Fossil SCM

Updating from trunk to keep up working in the branch with the most recent fossil code.

viriketo 2011-08-30 08:01 merge_renames merge

Commit 75aaf30aadc5cb32b00a5694e56a9216c07f09a5

Parent 281e5121e7e91bf…

83 files changed +38 -16 -3 +3 +9 +1 +2 +35 +1 +15 +73 +35 +314 +307 +1 +9 +144 +2 +80 +21 +3 +74 -21 +1 -1 +73 -36 +2 -2 +24 -8 +92 -15 +1 +17 -3 +5 -1 +1 +201 -52 +2 -2 +22 -16 +151 -36 +32 -6 +9 +4 -4 +2 -1 +61 -1 +11 -5 +9 -3 +16 -7 +32 -25 +3 -4 +10 -2 +1 -6 +5 +6 -2 +1 -1 +1 -1 +1 -1 +12 -2 +3777 -1278 +50 -23 +4 -2 +2 -1 +2 -1 +1 -1 +280 -29 +136 -20 +1 -1 +58 -2 +8 -20 +644 -9 +7 -1 +3 -3 +4 -4 +5 -5 +5 -5 -10 +63 -26 +1 +80 -16 +60 +4 +21 -15 +11 -6 +11 -1 +15 -7 +3 +6 +22

~ BUILD.txt - Makefile ~ Makefile.classic + Makefile.classic ~ Makefile.in ~ VERSION ~ auto.def ~ autosetup/LICENSE ~ autosetup/README.autosetup ~ autosetup/autosetup ~ autosetup/cc-lib.tcl ~ autosetup/cc-shared.tcl ~ autosetup/cc.tcl ~ autosetup/config.guess ~ autosetup/config.sub ~ autosetup/find-tclsh ~ autosetup/jimsh0.c ~ autosetup/local.tcl ~ autosetup/system.tcl ~ autosetup/test-tclsh ~ configure ~ src/add.c ~ src/blob.c ~ src/branch.c ~ src/browse.c ~ src/cgi.c ~ src/checkin.c ~ src/checkout.c ~ src/clone.c ~ src/config.h ~ src/configure.c ~ src/db.c ~ src/descendants.c ~ src/diffcmd.c ~ src/export.c ~ src/file.c ~ src/finfo.c ~ src/glob.c ~ src/http.c ~ src/http_ssl.c ~ src/info.c ~ src/main.c ~ src/main.mk ~ src/makemake.tcl ~ src/manifest.c ~ src/merge.c ~ src/mkversion.c + src/mkversion.c ~ src/popen.c ~ src/printf.c ~ src/rebuild.c ~ src/search.c ~ src/setup.c ~ src/sqlite3.c ~ src/sqlite3.h ~ src/stash.c ~ src/stat.c ~ src/style.c ~ src/tag.c ~ src/tar.c ~ src/timeline.c ~ src/tkt.c ~ src/update.c ~ src/wiki.c ~ src/winhttp.c ~ test/release-checklist.wiki ~ win/Makefile.PellesCGMake ~ win/Makefile.dmc ~ win/Makefile.mingw ~ win/Makefile.msc - win/version.c ~ www/build.wiki ~ www/changes.wiki ~ www/checkin.wiki ~ www/foss-cklist.wiki ~ www/index.wiki ~ www/makefile.wiki ~ www/mkindex.tcl ~ www/permutedindex.wiki ~ www/quickstart.wiki ~ www/server.wiki ~ www/settings.wiki ~ www/ssl.wiki

M BUILD.txt

+38 -16

		--- BUILD.txt
		+++ BUILD.txt
		@@ -1,41 +1,60 @@
1		-All of the source code for fossil is contained in the src/ subdirectory.
2		-But there is a lot of generated code, so you will probably want to
3		-use the Makefile. To do a complete build on unix, just type:
	1	+To do a complete build, just type:
	2	+
	3	+ ./configure; make
	4	+
	5	+The ./configure script builds Makefile from Makefile.in based on
	6	+your system and any options you select (run "./configure --help"
	7	+for a listing of the available options.)
	8	+
	9	+If you wish to use the original Makefile with no configuration, you can
	10	+instead use:
4	11
5		- make
	12	+ make -f Makefile.classic
6	13
7	14	On a windows box, use one of the Makefiles in the win/ subdirectory,
8		-according to your compiler and environment. For example:
	15	+according to your compiler and environment. If you have GCC and MSYS
	16	+installed on your system, then consider:
	17	+
	18	+ make -f win/Makefile.mingw
	19	+
	20	+If you have VC++ installed on your system, then consider:
9	21
10		- make -f win/Makefile.w32
	22	+ cd win; nmake /f Makefile.msc
11	23
12	24	If you have trouble, or you want to do something fancy, just look at
13		-top level makefile. There are 6 configuration options that are all well
14		-commented. Instead of editing the Makefile, consider copying the Makefile
15		-to an alternative name such as "GNUMakefile", "BSDMakefile", or "makefile"
16		-and editing the copy.
	25	+Makefile.classic. There are 6 configuration options that are all well
	26	+commented. Instead of editing the Makefile.classic, consider copying
	27	+Makefile.classic to an alternative name such as "GNUMakefile",
	28	+"BSDMakefile", or "makefile" and editing the copy.
17	29
18	30
19	31	BUILDING OUTSIDE THE SOURCE TREE
20	32
21	33	An out of source build is pretty easy:
22	34
23		- 1. Make a new directory to do the builds in.
24		- 2. Copy "Makefile" from the source into the build directory and
25		- modify the SRCDIR macro along the lines of:
	35	+ 1. Make and change to a new directory to do the builds in.
	36	+ 2. Run the "configure" script from this directory.
	37	+ 3. Type: "make"
	38	+
	39	+For example:
26	40
27		- SRCDIR=../src
28		-
29		- 3. type: "make"
	41	+ mkdir build
	42	+ cd build
	43	+ ../configure
	44	+ make
30	45
31	46	This will now keep all generates files seperate from the maintained
32	47	source code.
33	48
34	49	--------------------------------------------------------------------------
35	50
36	51	Here are some notes on what is happening behind the scenes:
	52	+
	53	+* The configure script (if used) examines the options given
	54	+ and runs various tests with the C compiler to create Makefile
	55	+ from the Makefile.in template as well as autoconfig.h
37	56
38	57	* The Makefile just sets up a few macros and then invokes the
39	58	real makefile in src/main.mk. The src/main.mk makefile is
40	59	automatically generated by a TCL script found at src/makemake.tcl.
41	60	Do not edit src/main.mk directly. Update src/makemake.tcl and
		@@ -51,5 +70,8 @@
51	70	A header comment in src/translate.c explains in detail what it does.
52	71
53	72	* The src/mkindex.c program generates some C code that implements
54	73	static lookup tables. See the header comment in the source code
55	74	for details on what it does.
	75	+
	76	+Additional information on the build process is available from
	77	+http://www.fossil-scm.org/fossil/doc/trunk/www/makefile.wiki
56	78
57	79	DELETED Makefile
58	80	ADDED Makefile.classic
59	81	ADDED Makefile.in
60	82	ADDED VERSION
61	83	ADDED auto.def
62	84	ADDED autosetup/LICENSE
63	85	ADDED autosetup/README.autosetup
64	86	ADDED autosetup/autosetup
65	87	ADDED autosetup/cc-lib.tcl
66	88	ADDED autosetup/cc-shared.tcl
67	89	ADDED autosetup/cc.tcl
68	90	ADDED autosetup/config.guess
69	91	ADDED autosetup/config.sub
70	92	ADDED autosetup/find-tclsh
71	93	ADDED autosetup/jimsh0.c
72	94	ADDED autosetup/local.tcl
73	95	ADDED autosetup/system.tcl
74	96	ADDED autosetup/test-tclsh
75	97	ADDED configure

	--- BUILD.txt
	+++ BUILD.txt
	@@ -1,41 +1,60 @@
1	All of the source code for fossil is contained in the src/ subdirectory.
2	But there is a lot of generated code, so you will probably want to
3	use the Makefile. To do a complete build on unix, just type:







4
5	make
6
7	On a windows box, use one of the Makefiles in the win/ subdirectory,
8	according to your compiler and environment. For example:





9
10	make -f win/Makefile.w32
11
12	If you have trouble, or you want to do something fancy, just look at
13	top level makefile. There are 6 configuration options that are all well
14	commented. Instead of editing the Makefile, consider copying the Makefile
15	to an alternative name such as "GNUMakefile", "BSDMakefile", or "makefile"
16	and editing the copy.
17
18
19	BUILDING OUTSIDE THE SOURCE TREE
20
21	An out of source build is pretty easy:
22
23	1. Make a new directory to do the builds in.
24	2. Copy "Makefile" from the source into the build directory and
25	modify the SRCDIR macro along the lines of:


26
27	SRCDIR=../src
28
29	3. type: "make"

30
31	This will now keep all generates files seperate from the maintained
32	source code.
33
34	--------------------------------------------------------------------------
35
36	Here are some notes on what is happening behind the scenes:




37
38	* The Makefile just sets up a few macros and then invokes the
39	real makefile in src/main.mk. The src/main.mk makefile is
40	automatically generated by a TCL script found at src/makemake.tcl.
41	Do not edit src/main.mk directly. Update src/makemake.tcl and
	@@ -51,5 +70,8 @@
51	A header comment in src/translate.c explains in detail what it does.
52
53	* The src/mkindex.c program generates some C code that implements
54	static lookup tables. See the header comment in the source code
55	for details on what it does.



56
57	ELETED Makefile
58	DDED Makefile.classic
59	DDED Makefile.in
60	DDED VERSION
61	DDED auto.def
62	DDED autosetup/LICENSE
63	DDED autosetup/README.autosetup
64	DDED autosetup/autosetup
65	DDED autosetup/cc-lib.tcl
66	DDED autosetup/cc-shared.tcl
67	DDED autosetup/cc.tcl
68	DDED autosetup/config.guess
69	DDED autosetup/config.sub
70	DDED autosetup/find-tclsh
71	DDED autosetup/jimsh0.c
72	DDED autosetup/local.tcl
73	DDED autosetup/system.tcl
74	DDED autosetup/test-tclsh
75	DDED configure

	--- BUILD.txt
	+++ BUILD.txt
	@@ -1,41 +1,60 @@
1	To do a complete build, just type:
2
3	./configure; make
4
5	The ./configure script builds Makefile from Makefile.in based on
6	your system and any options you select (run "./configure --help"
7	for a listing of the available options.)
8
9	If you wish to use the original Makefile with no configuration, you can
10	instead use:
11
12	make -f Makefile.classic
13
14	On a windows box, use one of the Makefiles in the win/ subdirectory,
15	according to your compiler and environment. If you have GCC and MSYS
16	installed on your system, then consider:
17
18	make -f win/Makefile.mingw
19
20	If you have VC++ installed on your system, then consider:
21
22	cd win; nmake /f Makefile.msc
23
24	If you have trouble, or you want to do something fancy, just look at
25	Makefile.classic. There are 6 configuration options that are all well
26	commented. Instead of editing the Makefile.classic, consider copying
27	Makefile.classic to an alternative name such as "GNUMakefile",
28	"BSDMakefile", or "makefile" and editing the copy.
29
30
31	BUILDING OUTSIDE THE SOURCE TREE
32
33	An out of source build is pretty easy:
34
35	1. Make and change to a new directory to do the builds in.
36	2. Run the "configure" script from this directory.
37	3. Type: "make"
38
39	For example:
40
41	mkdir build
42	cd build
43	../configure
44	make
45
46	This will now keep all generates files seperate from the maintained
47	source code.
48
49	--------------------------------------------------------------------------
50
51	Here are some notes on what is happening behind the scenes:
52
53	* The configure script (if used) examines the options given
54	and runs various tests with the C compiler to create Makefile
55	from the Makefile.in template as well as autoconfig.h
56
57	* The Makefile just sets up a few macros and then invokes the
58	real makefile in src/main.mk. The src/main.mk makefile is
59	automatically generated by a TCL script found at src/makemake.tcl.
60	Do not edit src/main.mk directly. Update src/makemake.tcl and
	@@ -51,5 +70,8 @@
70	A header comment in src/translate.c explains in detail what it does.
71
72	* The src/mkindex.c program generates some C code that implements
73	static lookup tables. See the header comment in the source code
74	for details on what it does.
75
76	Additional information on the build process is available from
77	http://www.fossil-scm.org/fossil/doc/trunk/www/makefile.wiki
78
79	ELETED Makefile
80	DDED Makefile.classic
81	DDED Makefile.in
82	DDED VERSION
83	DDED auto.def
84	DDED autosetup/LICENSE
85	DDED autosetup/README.autosetup
86	DDED autosetup/autosetup
87	DDED autosetup/cc-lib.tcl
88	DDED autosetup/cc-shared.tcl
89	DDED autosetup/cc.tcl
90	DDED autosetup/config.guess
91	DDED autosetup/config.sub
92	DDED autosetup/find-tclsh
93	DDED autosetup/jimsh0.c
94	DDED autosetup/local.tcl
95	DDED autosetup/system.tcl
96	DDED autosetup/test-tclsh
97	DDED configure

D Makefile

-3

		--- a/Makefile
		+++ b/Makefile
		@@ -1,3 +0,0 @@
1		-#!/usr/bin/make
2		-#
3		-# This is tg the build. This is used

	--- a/Makefile
	+++ b/Makefile
	@@ -1,3 +0,0 @@
1	#!/usr/bin/make
2	#
3	# This is tg the build. This is used

	--- a/Makefile
	+++ b/Makefile
	@@ -1,3 +0,0 @@

M Makefile.classic

No diff available

A Makefile.classic

		--- a/Makefile.classic
		+++ b/Makefile.classic
		@@ -0,0 +1,3 @@
	1	+#!/usr/bin/make
	2	+#
	3	+# This is tg the build. This is used

	--- a/Makefile.classic
	+++ b/Makefile.classic
	@@ -0,0 +1,3 @@

	--- a/Makefile.classic
	+++ b/Makefile.classic
	@@ -0,0 +1,3 @@
1	#!/usr/bin/make
2	#
3	# This is tg the build. This is used

M Makefile.in

		--- a/Makefile.in
		+++ b/Makefile.in
		@@ -0,0 +1,9 @@
	1	+ -D_HAVE_SQLITE_ssil-@FOSSIL_CI_PFX@P@1BE,N:fossil:@FOSSIL_CI_PFX-@FOSSIL_CI_PFX@P@1BE,N:fossil:@FOSSIL_CI_PFX -D_HAVE_SQLITE_ssil-@FOSSIL_CI_PFX@P@1BE,N:fossil:@FOSSIL_CI_PFX -fuzz -D_HAVE_SQLITE_ssil-@FOSSIL_CI_PFX@P@1BE,N:fossil:@FOSSIL_CI_PFX-@FOSSIL_CI_PFX@P@1BE,N:fossil:@FOSSIL_CI_PFX -D_HAVE_SQLITE_ssil-@FOSSIL_CI_PFX@P@1BE,N:fossil:@FOSSIL_CI_PFX -fuzzSIL_CI_PFX@P@1BE,N:fossil:@FOSSIL_CI_PFX-@FOSSIL_CI_PFX@P@1BE,N:fossil:@FOSSIL_CI_PFX -D_HAVE_SQLITE_ssitclsh -DFOSSIL_ENABLE_LEGACY_MV_RM=1@CFLAGS@@CFLAGS@Rdi make source
	2	+#
	3	+#
	4	+OBJDIR = ./@P@1BE,N:fossil:@FOSSIL_CI_PFX -fuzz -D_HAVE_SQLITE_ssil-@FOSSILendif
	5	+ifeq ($(findstriSSIL_CI_PFX@P@1BE,N:fossil:@FOSSIL_CI_PFX -fuzz -D_HAVE_SQLITE$(MAKE) reconfig
	6	+endif
	7	+@AUTOREMAKE@ && $(MAKE)
	8	+endif
	9	+FOSSIL_ENABLE_M +

	--- a/Makefile.in
	+++ b/Makefile.in
	@@ -0,0 +1,9 @@

	--- a/Makefile.in
	+++ b/Makefile.in
	@@ -0,0 +1,9 @@
1	-D_HAVE_SQLITE_ssil-@FOSSIL_CI_PFX@P@1BE,N:fossil:@FOSSIL_CI_PFX-@FOSSIL_CI_PFX@P@1BE,N:fossil:@FOSSIL_CI_PFX -D_HAVE_SQLITE_ssil-@FOSSIL_CI_PFX@P@1BE,N:fossil:@FOSSIL_CI_PFX -fuzz -D_HAVE_SQLITE_ssil-@FOSSIL_CI_PFX@P@1BE,N:fossil:@FOSSIL_CI_PFX-@FOSSIL_CI_PFX@P@1BE,N:fossil:@FOSSIL_CI_PFX -D_HAVE_SQLITE_ssil-@FOSSIL_CI_PFX@P@1BE,N:fossil:@FOSSIL_CI_PFX -fuzzSIL_CI_PFX@P@1BE,N:fossil:@FOSSIL_CI_PFX-@FOSSIL_CI_PFX@P@1BE,N:fossil:@FOSSIL_CI_PFX -D_HAVE_SQLITE_ssitclsh -DFOSSIL_ENABLE_LEGACY_MV_RM=1@CFLAGS@@CFLAGS@Rdi make source
2	#
3	#
4	OBJDIR = ./@P@1BE,N:fossil:@FOSSIL_CI_PFX -fuzz -D_HAVE_SQLITE_ssil-@FOSSILendif
5	ifeq ($(findstriSSIL_CI_PFX@P@1BE,N:fossil:@FOSSIL_CI_PFX -fuzz -D_HAVE_SQLITE$(MAKE) reconfig
6	endif
7	@AUTOREMAKE@ && $(MAKE)
8	endif
9	FOSSIL_ENABLE_M +

M VERSION

		--- a/VERSION
		+++ b/VERSION
		@@ -0,0 +1 @@
	1	+1.19

	--- a/VERSION
	+++ b/VERSION
	@@ -0,0 +1 @@

	--- a/VERSION
	+++ b/VERSION
	@@ -0,0 +1 @@
1	1.19

M auto.def

		--- a/auto.def
		+++ b/auto.def
		@@ -0,0 +1,2 @@
	1	+D markdown=0 => {Build with markdown enginemarkdown]} {
	2	+MARKDOWMARKDOWipv6=1 =>

	--- a/auto.def
	+++ b/auto.def
	@@ -0,0 +1,2 @@

	--- a/auto.def
	+++ b/auto.def
	@@ -0,0 +1,2 @@
1	D markdown=0 => {Build with markdown enginemarkdown]} {
2	MARKDOWMARKDOWipv6=1 =>

M autosetup/LICENSE

+35

		--- a/autosetup/LICENSE
		+++ b/autosetup/LICENSE
		@@ -0,0 +1,35 @@
	1	+Unless explicitly stated, all files which form part of autosetup
	2	+are released under the following license:
	3	+
	4	+---------------------------------------------------------------------
	5	+autosetup - A build environment "autoconfigurator"
	6	+
	7	+Copyright (c) 2010-2011, WorkWare Systems <http://workware.net.au/>
	8	+
	9	+Redistribution and use in source and binary forms, with or without
	10	+modification, are permitted provided that the following conditions
	11	+are met:
	12	+
	13	+1. Redistributions of source code must retain the above copyright
	14	+ notice, this list of conditions and the following disclaimer.
	15	+2. Redistributions in binary form must reproduce the above
	16	+ copyright notice, this list of conditions and the following
	17	+ disclaimer in the documentation and/or other materials
	18	+ provided with the distribution.
	19	+
	20	+THIS SOFTWARE IS PROVIDED BY THE WORKWARE SYSTEMS ``AS IS'' AND ANY
	21	+EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO,
	22	+THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A
	23	+PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL WORKWARE
	24	+SYSTEMS OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT,
	25	+INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES
	26	+(INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
	27	+OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
	28	+HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT,
	29	+STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
	30	+ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF
	31	+ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
	32	+
	33	+The views and conclusions contained in the software and documentation
	34	+are those of the authors and should not be interpreted as representing
	35	+official policies, either expressed or implied, of WorkWare Systems.

	--- a/autosetup/LICENSE
	+++ b/autosetup/LICENSE
	@@ -0,0 +1,35 @@

	--- a/autosetup/LICENSE
	+++ b/autosetup/LICENSE
	@@ -0,0 +1,35 @@
1	Unless explicitly stated, all files which form part of autosetup
2	are released under the following license:
3
4	---------------------------------------------------------------------
5	autosetup - A build environment "autoconfigurator"
6
7	Copyright (c) 2010-2011, WorkWare Systems <http://workware.net.au/>
8
9	Redistribution and use in source and binary forms, with or without
10	modification, are permitted provided that the following conditions
11	are met:
12
13	1. Redistributions of source code must retain the above copyright
14	notice, this list of conditions and the following disclaimer.
15	2. Redistributions in binary form must reproduce the above
16	copyright notice, this list of conditions and the following
17	disclaimer in the documentation and/or other materials
18	provided with the distribution.
19
20	THIS SOFTWARE IS PROVIDED BY THE WORKWARE SYSTEMS ``AS IS'' AND ANY
21	EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO,
22	THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A
23	PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL WORKWARE
24	SYSTEMS OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT,
25	INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES
26	(INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
27	OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
28	HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT,
29	STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
30	ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF
31	ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
32
33	The views and conclusions contained in the software and documentation
34	are those of the authors and should not be interpreted as representing
35	official policies, either expressed or implied, of WorkWare Systems.

M autosetup/README.autosetup

		--- a/autosetup/README.autosetup
		+++ b/autosetup/README.autosetup
		@@ -0,0 +1 @@
	1	+This is autosetup v0.6.2. S

	--- a/autosetup/README.autosetup
	+++ b/autosetup/README.autosetup
	@@ -0,0 +1 @@

	--- a/autosetup/README.autosetup
	+++ b/autosetup/README.autosetup
	@@ -0,0 +1 @@
1	This is autosetup v0.6.2. S

M autosetup/autosetup

+15

		--- a/autosetup/autosetup
		+++ b/autosetup/autosetup
		@@ -0,0 +1,15 @@
	1	+#!/bin/sh
	2	+4 => "create an initial 'confbool init]$source#!/bin/sh
	3	+parray define
	4	+ } else {
	5	+ return $alias
	6	+ }[error-location $msg]use cc cc-shareduplevel #0#!/bin/sh
	7	+${m}.tcl
	8	+uplevel #0 [list source $source]put2put3 "Looking for $source"![string m 2>/dev/null} {
	9	+ set create_configure 1Use --initset cre#!/bin/sh
	10	+4 => "creat # Jim uses system() for exec under mingw, so
	11	+ # we need to fetch the output ourselves
	12	+ set tmpfile auto[format %04x [rand 10000]].tmp
	13	+ set rc [catch [list exec {*}$readfile $tmpfile]
	14	+ file delete $tmpfile
	15	+ return -code $rc $result

	--- a/autosetup/autosetup
	+++ b/autosetup/autosetup
	@@ -0,0 +1,15 @@

	--- a/autosetup/autosetup
	+++ b/autosetup/autosetup
	@@ -0,0 +1,15 @@
1	#!/bin/sh
2	4 => "create an initial 'confbool init]$source#!/bin/sh
3	parray define
4	} else {
5	return $alias
6	}[error-location $msg]use cc cc-shareduplevel #0#!/bin/sh
7	${m}.tcl
8	uplevel #0 [list source $source]put2put3 "Looking for $source"![string m 2>/dev/null} {
9	set create_configure 1Use --initset cre#!/bin/sh
10	4 => "creat # Jim uses system() for exec under mingw, so
11	# we need to fetch the output ourselves
12	set tmpfile auto[format %04x [rand 10000]].tmp
13	set rc [catch [list exec {*}$readfile $tmpfile]
14	file delete $tmpfile
15	return -code $rc $result

M autosetup/cc-lib.tcl

+73

		--- a/autosetup/cc-lib.tcl
		+++ b/autosetup/cc-lib.tcl
		@@ -0,0 +1,73 @@
	1	+# Copyright (c) 2011 WorkWare Systems http://www.workware.net.au/
	2	+# All rights reserved
	3	+
	4	+# @synopsis:
	5	+#
	6	+# Provides a library of common tests on top ofmodule-options {}of the 'cc' module.
	7	+
	8	+use cc
	9	+
	10	+# @cc-check-lfsAC_SYS_LARGEFILE macro
	11	+#
	12	+# defines 'HAVE_LFS' if L S is available,
	13	+# and defines '_FILE_OFFSET_BITS=64' if necessary
	14	+#
	15	+# Returns 1 if 'LFS' is available or 0 otherwise
	16	+#
	17	+proc cc-check-lfs {} {
	18	+ cc-check-includes sys/types.h
	19	+ msg-checking "Checking if -D_FILE_OFFSET_BITS=64 is needed..."
	20	+ set lfs 1
	21	+ if {[msg-quiet cc-with {-includes sys/types.h} {cc-check-sizeof off_t}] == 8} {
	22	+ msg-result no
	23	+ } elseif {[msg-quiet cc-with {-includes sys/types.h -cflags -D_FILE_OFFSET_BITS=64} {cc-check-sizeof off_t}] == 8} {
	24	+ define _FILE_OFFSET_BITS 64
	25	+ msg-result yes
	26	+ } else {
	27	+ set lfs 0
	28	+ msg-result none
	29	+ }
	30	+ define-feature lfs $lfs
	31	+ return $lfs
	32	+}
	33	+
	34	+# @cc-check-endian
	35	+#
	36	+# The equivalent of th eAC_C_BIGENDIAN macroth e 'AC_C_BIGENDIAN' macro.
	37	+#
	38	+# defines 'HAVE_BIG_ENDIAN' if endian is known to be big,
	39	+# or 'HAVE_LITTLE_ENDIAN' if endian is known to be little.
	40	+#
	41	+# Returns 1 if determined, or 0 if not.
	42	+#
	43	+proc cc-check-endian {} {
	44	+ cc-check-includes sys/types.h sys/param.h
	45	+ set rc 0
	46	+ msg-checking "Checking endian..."
	47	+ cc-with {-includes {sys/types.h sys/param.h}} {
	48	+ if {[cctest -code {
	49	+ #if !defined(BIG_ENDIAN) \|\| !defined(BYTE_ORDER)
	50	+ #error unknown
	51	+ #elif BYTE_ORDER != BIG_ENDIAN
	52	+ #error little
	53	+ #endif
	54	+ }]} {
	55	+ define-feature big-endian
	56	+ msg-result "big"
	57	+ set rc 1
	58	+ } elseif {[cctest -code {
	59	+ #if !defined(LITTLE_ENDIAN) \|\| !defined(BYTE_ORDER)
	60	+ #error unknown
	61	+ #elif BYTE_ORDER != LITTLE_ENDIAN
	62	+ #error big
	63	+ #endif
	64	+ }]} {
	65	+ define-feature little-endian
	66	+ msg-result "little"
	67	+ set rc 1
	68	+ } else {
	69	+ msg-result "unknown"
	70	+ }
	71	+ }
	72	+ return $rc
	73	+}

	--- a/autosetup/cc-lib.tcl
	+++ b/autosetup/cc-lib.tcl
	@@ -0,0 +1,73 @@

	--- a/autosetup/cc-lib.tcl
	+++ b/autosetup/cc-lib.tcl
	@@ -0,0 +1,73 @@
1	# Copyright (c) 2011 WorkWare Systems http://www.workware.net.au/
2	# All rights reserved
3
4	# @synopsis:
5	#
6	# Provides a library of common tests on top ofmodule-options {}of the 'cc' module.
7
8	use cc
9
10	# @cc-check-lfsAC_SYS_LARGEFILE macro
11	#
12	# defines 'HAVE_LFS' if L S is available,
13	# and defines '_FILE_OFFSET_BITS=64' if necessary
14	#
15	# Returns 1 if 'LFS' is available or 0 otherwise
16	#
17	proc cc-check-lfs {} {
18	cc-check-includes sys/types.h
19	msg-checking "Checking if -D_FILE_OFFSET_BITS=64 is needed..."
20	set lfs 1
21	if {[msg-quiet cc-with {-includes sys/types.h} {cc-check-sizeof off_t}] == 8} {
22	msg-result no
23	} elseif {[msg-quiet cc-with {-includes sys/types.h -cflags -D_FILE_OFFSET_BITS=64} {cc-check-sizeof off_t}] == 8} {
24	define _FILE_OFFSET_BITS 64
25	msg-result yes
26	} else {
27	set lfs 0
28	msg-result none
29	}
30	define-feature lfs $lfs
31	return $lfs
32	}
33
34	# @cc-check-endian
35	#
36	# The equivalent of th eAC_C_BIGENDIAN macroth e 'AC_C_BIGENDIAN' macro.
37	#
38	# defines 'HAVE_BIG_ENDIAN' if endian is known to be big,
39	# or 'HAVE_LITTLE_ENDIAN' if endian is known to be little.
40	#
41	# Returns 1 if determined, or 0 if not.
42	#
43	proc cc-check-endian {} {
44	cc-check-includes sys/types.h sys/param.h
45	set rc 0
46	msg-checking "Checking endian..."
47	cc-with {-includes {sys/types.h sys/param.h}} {
48	if {[cctest -code {
49	#if !defined(BIG_ENDIAN) \|\| !defined(BYTE_ORDER)
50	#error unknown
51	#elif BYTE_ORDER != BIG_ENDIAN
52	#error little
53	#endif
54	}]} {
55	define-feature big-endian
56	msg-result "big"
57	set rc 1
58	} elseif {[cctest -code {
59	#if !defined(LITTLE_ENDIAN) \|\| !defined(BYTE_ORDER)
60	#error unknown
61	#elif BYTE_ORDER != LITTLE_ENDIAN
62	#error big
63	#endif
64	}]} {
65	define-feature little-endian
66	msg-result "little"
67	set rc 1
68	} else {
69	msg-result "unknown"
70	}
71	}
72	return $rc
73	}

M autosetup/cc-shared.tcl

+35

		--- a/autosetup/cc-shared.tcl
		+++ b/autosetup/cc-shared.tcl
		@@ -0,0 +1,35 @@
	1	+# Copyright (c) 2010 WorkWare Systems http://www.workware.net.au/
	2	+# All rights reserved
	3	+
	4	+# @synopsis:
	5	+#
	6	+# The 'cc-shared' module provides support for shared libraries and shared objects.
	7	+# It defines the following variables:
	8	+#
	9	+## SH_CFLAGS Flags to use compiling sources destined for a shared library
	10	+## SH_LDFLAGS Flags to use linking t the soname when cread libs - %s = version
	11	+## SHOBJ_CFLAGS Flags to use compiling sources destined for a shared object
	12	+## ll symbols table, %s = path
	13	+## SH_LINKFLAGS Flags to use linking an executable which will load shared objects
	14	+## LD_LIBRARY_PATH Environment variable which specifie
	15	+module-options {}
	16	+
	17	+foreach i {SH_LINKFLAGS SH_CFLAGS SH_LDFLAGS SHOBJ_CFLAGS SHOBJ_LLD_LIBRARY_PATH define SHOBJ_LDFLAGS_R -bobject
	18	+## SHOBJ_LDFLAGS Flags to are System# Copyright ( Flags to usDYLD_LIBRARY_PATH
	19	+ }
	20	+ --ming* {sharedOBJ_LDFLAGSshared}
	21	+ --solaris* {
	22	+# Copyright (c) 20mwed
	23	+## SHOBJ_LD-Kpic Flags to us-Kpic"-G"
	24	+define LD_LIBthe toolch-strip-unneeded
	25	+
	26	+# Note: Thiain
	27	+# http://sourceforge sparc* {
	28	+AGS SHOBJ_LDFLAGS_rdynamicference for identify-fPICshared Flags to us-fPIC}
	29	+ * {
	30	+ # Generic Unix settingsstems http://www.workwa# Copyrigference for identifying the toolchain
	31	+# http://sourceforge.net/apps/mediawiki/predef/ define SHOBJ_LDFLAGS_R -bobject
	32	+## SHOBJ_LDFLAGS Flags to us-fpshared Flags to us-fpicf/ define SHOBJ_LDFLAGS_R -bobject
	33	+## SHOBJ_LDFLAGS Flags to us"-shared -nostartfiles"
	34	+ }
	35	+}

	--- a/autosetup/cc-shared.tcl
	+++ b/autosetup/cc-shared.tcl
	@@ -0,0 +1,35 @@

	--- a/autosetup/cc-shared.tcl
	+++ b/autosetup/cc-shared.tcl
	@@ -0,0 +1,35 @@
1	# Copyright (c) 2010 WorkWare Systems http://www.workware.net.au/
2	# All rights reserved
3
4	# @synopsis:
5	#
6	# The 'cc-shared' module provides support for shared libraries and shared objects.
7	# It defines the following variables:
8	#
9	## SH_CFLAGS Flags to use compiling sources destined for a shared library
10	## SH_LDFLAGS Flags to use linking t the soname when cread libs - %s = version
11	## SHOBJ_CFLAGS Flags to use compiling sources destined for a shared object
12	## ll symbols table, %s = path
13	## SH_LINKFLAGS Flags to use linking an executable which will load shared objects
14	## LD_LIBRARY_PATH Environment variable which specifie
15	module-options {}
16
17	foreach i {SH_LINKFLAGS SH_CFLAGS SH_LDFLAGS SHOBJ_CFLAGS SHOBJ_LLD_LIBRARY_PATH define SHOBJ_LDFLAGS_R -bobject
18	## SHOBJ_LDFLAGS Flags to are System# Copyright ( Flags to usDYLD_LIBRARY_PATH
19	}
20	--ming* {sharedOBJ_LDFLAGSshared}
21	--solaris* {
22	# Copyright (c) 20mwed
23	## SHOBJ_LD-Kpic Flags to us-Kpic"-G"
24	define LD_LIBthe toolch-strip-unneeded
25
26	# Note: Thiain
27	# http://sourceforge sparc* {
28	AGS SHOBJ_LDFLAGS_rdynamicference for identify-fPICshared Flags to us-fPIC}
29	* {
30	# Generic Unix settingsstems http://www.workwa# Copyrigference for identifying the toolchain
31	# http://sourceforge.net/apps/mediawiki/predef/ define SHOBJ_LDFLAGS_R -bobject
32	## SHOBJ_LDFLAGS Flags to us-fpshared Flags to us-fpicf/ define SHOBJ_LDFLAGS_R -bobject
33	## SHOBJ_LDFLAGS Flags to us"-shared -nostartfiles"
34	}
35	}

M autosetup/cc.tcl

+314

		--- a/autosetup/cc.tcl
		+++ b/autosetup/cc.tcl
		@@ -0,0 +1,314 @@
	1	+# Copyright (c) 2010 WorkWare Systems http://www.workware.net.au/
	2	+# All rights reserved
	3	+
	4	+# @synopsis:
	5	+#
	6	+# The 'cc' module supports checking various 'features' of the C or C++
	7	+# compiler/linker environment. Common commands are ,
	8	+# cc-check-types, cc-check-functions, cc-with, make-autoconf-h and make-templatement variables are used if set:
	9	+#
	10	+## CC - C compiler
	11	+## CXX - C++ compiler
	12	+## CPP - C preprocessor
	13	+##CACHE - Set to "none" to disable automatic use of ccache
	14	+## Ciler flags
	15	+## CXXFLAGS - Additional C++ compiler flags
	16	+## LDFLAGS - Additional compiler flags during linking
	17	+## LINKFLAGS - ?How is this diffes to use (for all tests)
	18	+## CROSS - Tool prefix for cross compilation
	19	+#
	20	+# The following variables are defined from the corresponding
	21	+# environment variables if set.
	22	+#
	23	+## CC_FOR_BUILD
	24	+## LD
	25	+
	26	+use systPPFLAGS
	27	+## LINKFLAGSmodule-options {}
	28	+
	29	+# Checks for the existence of the giveNote that the return code is not meaningful
	30	+proc cc-check-something {name code} {
	31	+ uplevel 1 $codee given type/strunction {function} {
	32	+ cctest -link 1 -declare "extern void $function\(void);" -code "$function\();"
	33	+}
	34	+
	35	+# Checks for the existence of the given type by compiling
	36	+proc cctest_type {type} {
	37	+ cctest -code "$type _x;"
	38	+}
	39	+
	40	+# Checks for the existence of the given type/structure member.
	41	+# e.g. "struct stat.st_mtime"
	42	+proc cctest_member {struct_member} {
	43	+ # split at the first dot
	44	+ regexp {^([^.]+)[.](.*)$} $lassign me;\n#endif"
	45	+}
	46	+
	47	+# @cc-ch
	48	+# e.g. "struct stypeses a variable with the 4 8 1 2 16 32} {
	49	+ if {[cctest -code "static int _x\[sizeof($type) == $i ? 1 : -1\] = { 1 };"]} {
	50	+ set size $i
	51	+ break
	52	+ }
	53	+ }
	54	+ msg-result $size
	55	+ set define [feature-define-name $type SIZEOF_]
	56	+ define $define $size
	57	+ }
	58	+ # Return the last result
	59	+ get-define $define
	60	+}
	61	+
	62	+# Checks for each feature in $list by using the given script.
	63	+#
	64	+# When the script is evaluated, $each is set to the feature
	65	+# being checked, and $extra is set to any additional cctest args.
	66	+#
	67	+# Returns 1 if all features were found, or 0 otherwise.
	68	+proc cc-check-some-feature {list script} {
	69	+ set ret 1
	70	+ foreach each $list {
	71	+ if {![check-feature $each $script]} {
	72	+ set ret 0
	73	+ }
	74	+ }
	75	+ return $ret
	76	+}
	77	+
	78	+# @cc-check-includes includes ...
	79	+#
	80	+# Checks that the given include files can be used.
	81	+proc cc-check-includes {args} {
	82	+ c
	83	+procproc cctest_define {cc-check-some-xp {^([^.]+)[.](.*)$} $struct_member -> struct member
	84	+ cctest -code "static $struct _s; return sizeof(_s.$member);"
	85	+}
	86	+
	87	+# Checks for the existence of the given define by compiling
	88	+#
	89	+proc cctest_define {name} {
	90	+ cctest -code "#ifndef $name\n#error not defined\n#endif"
	91	+}
	92	+
	93	+# Checks forcks that the given include files can be used.
	94	+proc cc-check-includes {args} {
	95	+ cc-check-some-feature $args {
	96	+ set with {}
	97	+ if {[dict exists $::autosetup(cc-include-deps) $each]} {
	98	+ set deps! cc-check-include$exeecks tha]} {
	99	+ cc-with [list -includes $w }
	100	+ define $TOOL $exeed if set:
	101	+#
	102	+## CC
	103	+# @synopsis:
	104	+#
	105	+# The 'cc' module supports checking various 'features' of the C or C++
	106	+# compiler/linker environment. Common commands are ,
	107	+# cc-check-types, cc-check-functions, cc-with, make-autoconf-h and make-templatement variables are used if set:
	108	+#
	109	+## CC - C compiler
	110	+## CXX - C++ compiler
	111	+## CPP - C preprocessor
	112	+##CACHE - Set to "none" to disable automatic use of ccache
	113	+## Ciler flags
	114	+## CXXFLAGS - Additional C++ compiler flags
	115	+## LDFLAGS - Additional compiler flags during linking
	116	+## LINKFLAGS - ?How is this diffes to use (for all tests)
	117	+## CROSS - Tool prefix for cross compilation
	118	+#
	119	+# The following variables are defined from the corresponding
	120	+# environment variables if set.
	121	+#
	122	+## CC_FOR_BUILD
	123	+## LD
	124	+
	125	+use systPPFLAGS
	126	+## LINKFLAGSmodule-options {}
	127	+
	128	+# Checks for the existence of the giveNote that the return code is not meaningful
	129	+proc cc-check-something {name code} {
	130	+ uplevel 1 $codee given type/strunction {function} {
	131	+ cctest -link 1 -declare "extern void $function\(void);" -code "$function\();"
	132	+}
	133	+
	134	+# Checks for the existence of the given type by compiling
	135	+proc cctest_type {type} {
	136	+ cctest -code "$type _x;"
	137	+}
	138	+
	139	+# Checks for the existence of the given type/structure member.
	140	+# e.g. "struct stat.st_mtime"
	141	+proc cctest_member {struct_member} {
	142	+ # split at the first dot
	143	+ regexp {^([^.]+)[.](.*)$} $lassign [split $struct_member .]rn sizeof(_s.$member);"
	144	+}
	145	+
	146	+# Checks for the existence of the given define by compiling
	147	+#
	148	+proc cctest_define {name} {
	149	+ cctest -code "#ifndef $name\n#error not defined\n#endif"
	150	+}
	151	+
	152	+# Checks for the existence of the given name either as
	153	+# a macro (#define) or an rvalue (such as an enum)
	154	+#
	155	+proc cctest_decl {name} {
	156	+ cctest -code "#ifndef $name\n(void)$name;\n#endif"
	157	+}
	158	+
	159	+# @cc-check-sizeof type ...
	160	+#
	161	+# Checks the size of the given types (between 1 and 32, inclusive).
	162	+# Defines a variable with the size determined, or 'unknown'"unknown"ng', defines 'SIZEOF_LSIZEOF_LONG_LONGe.
	163	+#
	164	+proc cc-check-sizeof {args} {
	165	+ foreach type $args {
	166	+ msg-checking "Checking for sizeof $type..."
	167	+ set size unknown
	168	+ # Try the most common sizes first
	169	+ foreach i {4 8 1 2 16 32} {
	170	+ if {[cctest -code "static int _x\[sizeof($type) == $i ? 1 : -1\] = { 1 };"]} {
	171	+ set size $i
	172	+ break
	173	+ }
	174	+ }
	175	+ msg-result $size
	176	+ set define [feature-define-name $type SIZEOF_]
	177	+ define $define $size
	178	+ }
	179	+ # Return the last result
	180	+ get-define $define
	181	+}
	182	+
	183	+# Checks for each feature in $list by using the given script.
	184	+#
	185	+# When the script is evaluated, $each is set to the feature
	186	+# being checked, and $extra is set to any additional cctest args.
	187	+#
	188	+# Returns 1 if all features were found, or 0 otherwise.
	189	+proc cc-check-some-feature {list script} {
	190	+ set ret 1
	191	+ foreach each $list {
	192	+ if {![check-feature $each $script]} {
	193	+ set ret 0
	194	+ }
	195	+ }
	196	+ return $ret
	197	+}
	198	+
	199	+# @cc-check-includes includes ...
	200	+#
	201	+# Checks that the given include files can be used.
	202	+proc cc-check-includes {args} {
	203	+ c
	204	+procproc cctest_define {cc-check-some-xp {^([^.]+)[.](.*)$} $struct_member -> struct member
	205	+ cctest -code "static $struct _s; return sizeof(_s.$member);"
	206	+}
	207	+
	208	+# Checks for the existence of the given define by compiling
	209	+#
	210	+proc cctest_define {name} {
	211	+ cctest -code "#ifndef $name\n#error not defined\n#endif"
	212	+}
	213	+
	214	+# Checks for the existence of the given name either as
	215	+# a macro (#define) or an rvalue (such as an enum)
	216	+#
	217	+proc cctest_decl {name} {
	218	+ cctest -code "#ifndef $name\n(void)$name;\n#endif"
	219	+}
	220	+ given types (bet#includeddetermined, or 'unknown' otherwise.
	221	+# e.g. for type 'long long', defines 'SIZEOF_LONG_LONG'.
	222	+# Returns the size of the last type.
	223	+#
	224	+proc cc-check-sizeof {args} {
	225	+ foreach type $adfirst dot
	226	+ regexp {^([^.]+)[.](.*)$} $struct_member -> struct member
	227	+ cctest -code "static $struct _s; return sizeof(_s.$member);"
	228	+}
	229	+
	230	+# Checks for the existence of the given define by compiling
	231	+#
	232	+proc cctest_define FLAGS - ?How is this di# @synopsis:
	233	+#
	234	+# The 'cc' module supports checking various 'features' of the C or C++
	235	+# compiler/linker environment. Common commands are ,
	236	+# cc-check-types, cc-check-functions, cc-with, make-autoconf-h and make-templatement variables are used if set:
	237	+#
	238	+## CC - C compiler
	239	+## CXX - C++ compiler
	240	+## CPP - C preprocessor
	241	+##CACHE - Set to "none" to disfor a declsable automatic use of ccache
	242	+## Ciler flags
	243	+## C# Copyright (c) 2010 Work foreach type $args {
	244	+ msg-checking "Checking for sizeof $type..."
	245	+ set size unknown
	246	+ # Try the most common sizes first
	247	+ foreach i {4 8 1 2 16 32} {
	248	+ if {[cctest -code "static int _x\[sizeof($type) == $i ? 1 : -1\] = { 1 };"]} {
	249	+ set size $i
	250	+ break
	251	+ }
	252	+ }
	253	+ msg-result $size
	254	+ set define [feature-define-name $type SIZEOF_]
	255	+ define $define $size
	256	+ }
	257	+ # Return the last result
	258	+ get-define $define
	259	+}
	260	+
	261	+# Checks for each feature in $list by using the given script.
	262	+#
	263	+# When the script is evaluated, $each is set to the feature
	264	+# being checked, and $extra is set to any additional cctest args.
	265	+#
	266	+# Returns 1 if all features were found, or 0 otherwise.
	267	+proc cc-check-some-feature {list script} {
	268	+ set ret 1ch $script]} {
	269	+ set ret 0
	270	+ }
	271	+ }
	272	+ return $ret
	273	+}
	274	+
	275	+# @cc-check-includes includes ...
	276	+#
	277	+# Checks that the given include files can be used.
	278	+proc cc-check-includes {args} {
	279	+ cc-check-some-feature $args {
	280	+ set with {}
	281	+ if {[dict exists $::autosetup(cc-include-deps) $each]} {
	282	+ set deps [dict keys [dict get $::autosetup(cc-include-deps) $each]]
	283	+ msg-quiet cc-check-includes {*}$deps
	284	+ foreach i $deps {
	285	+ if {[have-feature $i]} {
	286	+ lappend with $i
	287	+ure $each $scriptargs {
	288	+ set new($name) $value
	289	+ } $i]} {
	290	+ lappend$newme\n(void)$name;\n#th]} {
	291	+ cc-with [list -includes $with] {
	292	+ cctest -includes $each
	293	+ }
	294	+ } else {
	295	+ cctest -includes $each
	296	+ }
	297	+ }
	298	+}
	299	+
	300	+# @cc-include-needs include required ...
	301	+#
	302	+# Ensures that when checking for '$include', a check is first
	303	+# made for eachtk is first
	304	+# made for each '$required' file, and if found, it is included with '#include'.
	305	+proc cc-include-needs {file args} {
	306	+ foreach depfile $args {
	307	+ dict set ::autosetup(cc-include-deps) $file $depfile 1
	308	+ }
	309	+}
	310	+
	311	+# C --darwin* {
	312	+ # Don't-gstabs
	313	+ }
	314	+ }$value$valuecc-check-tools ld

	--- a/autosetup/cc.tcl
	+++ b/autosetup/cc.tcl
	@@ -0,0 +1,314 @@

	--- a/autosetup/cc.tcl
	+++ b/autosetup/cc.tcl
	@@ -0,0 +1,314 @@
1	# Copyright (c) 2010 WorkWare Systems http://www.workware.net.au/
2	# All rights reserved
3
4	# @synopsis:
5	#
6	# The 'cc' module supports checking various 'features' of the C or C++
7	# compiler/linker environment. Common commands are ,
8	# cc-check-types, cc-check-functions, cc-with, make-autoconf-h and make-templatement variables are used if set:
9	#
10	## CC - C compiler
11	## CXX - C++ compiler
12	## CPP - C preprocessor
13	##CACHE - Set to "none" to disable automatic use of ccache
14	## Ciler flags
15	## CXXFLAGS - Additional C++ compiler flags
16	## LDFLAGS - Additional compiler flags during linking
17	## LINKFLAGS - ?How is this diffes to use (for all tests)
18	## CROSS - Tool prefix for cross compilation
19	#
20	# The following variables are defined from the corresponding
21	# environment variables if set.
22	#
23	## CC_FOR_BUILD
24	## LD
25
26	use systPPFLAGS
27	## LINKFLAGSmodule-options {}
28
29	# Checks for the existence of the giveNote that the return code is not meaningful
30	proc cc-check-something {name code} {
31	uplevel 1 $codee given type/strunction {function} {
32	cctest -link 1 -declare "extern void $function\(void);" -code "$function\();"
33	}
34
35	# Checks for the existence of the given type by compiling
36	proc cctest_type {type} {
37	cctest -code "$type _x;"
38	}
39
40	# Checks for the existence of the given type/structure member.
41	# e.g. "struct stat.st_mtime"
42	proc cctest_member {struct_member} {
43	# split at the first dot
44	regexp {^([^.]+)[.](.*)$} $lassign me;\n#endif"
45	}
46
47	# @cc-ch
48	# e.g. "struct stypeses a variable with the 4 8 1 2 16 32} {
49	if {[cctest -code "static int _x\[sizeof($type) == $i ? 1 : -1\] = { 1 };"]} {
50	set size $i
51	break
52	}
53	}
54	msg-result $size
55	set define [feature-define-name $type SIZEOF_]
56	define $define $size
57	}
58	# Return the last result
59	get-define $define
60	}
61
62	# Checks for each feature in $list by using the given script.
63	#
64	# When the script is evaluated, $each is set to the feature
65	# being checked, and $extra is set to any additional cctest args.
66	#
67	# Returns 1 if all features were found, or 0 otherwise.
68	proc cc-check-some-feature {list script} {
69	set ret 1
70	foreach each $list {
71	if {![check-feature $each $script]} {
72	set ret 0
73	}
74	}
75	return $ret
76	}
77
78	# @cc-check-includes includes ...
79	#
80	# Checks that the given include files can be used.
81	proc cc-check-includes {args} {
82	c
83	procproc cctest_define {cc-check-some-xp {^([^.]+)[.](.*)$} $struct_member -> struct member
84	cctest -code "static $struct _s; return sizeof(_s.$member);"
85	}
86
87	# Checks for the existence of the given define by compiling
88	#
89	proc cctest_define {name} {
90	cctest -code "#ifndef $name\n#error not defined\n#endif"
91	}
92
93	# Checks forcks that the given include files can be used.
94	proc cc-check-includes {args} {
95	cc-check-some-feature $args {
96	set with {}
97	if {[dict exists $::autosetup(cc-include-deps) $each]} {
98	set deps! cc-check-include$exeecks tha]} {
99	cc-with [list -includes $w }
100	define $TOOL $exeed if set:
101	#
102	## CC
103	# @synopsis:
104	#
105	# The 'cc' module supports checking various 'features' of the C or C++
106	# compiler/linker environment. Common commands are ,
107	# cc-check-types, cc-check-functions, cc-with, make-autoconf-h and make-templatement variables are used if set:
108	#
109	## CC - C compiler
110	## CXX - C++ compiler
111	## CPP - C preprocessor
112	##CACHE - Set to "none" to disable automatic use of ccache
113	## Ciler flags
114	## CXXFLAGS - Additional C++ compiler flags
115	## LDFLAGS - Additional compiler flags during linking
116	## LINKFLAGS - ?How is this diffes to use (for all tests)
117	## CROSS - Tool prefix for cross compilation
118	#
119	# The following variables are defined from the corresponding
120	# environment variables if set.
121	#
122	## CC_FOR_BUILD
123	## LD
124
125	use systPPFLAGS
126	## LINKFLAGSmodule-options {}
127
128	# Checks for the existence of the giveNote that the return code is not meaningful
129	proc cc-check-something {name code} {
130	uplevel 1 $codee given type/strunction {function} {
131	cctest -link 1 -declare "extern void $function\(void);" -code "$function\();"
132	}
133
134	# Checks for the existence of the given type by compiling
135	proc cctest_type {type} {
136	cctest -code "$type _x;"
137	}
138
139	# Checks for the existence of the given type/structure member.
140	# e.g. "struct stat.st_mtime"
141	proc cctest_member {struct_member} {
142	# split at the first dot
143	regexp {^([^.]+)[.](.*)$} $lassign [split $struct_member .]rn sizeof(_s.$member);"
144	}
145
146	# Checks for the existence of the given define by compiling
147	#
148	proc cctest_define {name} {
149	cctest -code "#ifndef $name\n#error not defined\n#endif"
150	}
151
152	# Checks for the existence of the given name either as
153	# a macro (#define) or an rvalue (such as an enum)
154	#
155	proc cctest_decl {name} {
156	cctest -code "#ifndef $name\n(void)$name;\n#endif"
157	}
158
159	# @cc-check-sizeof type ...
160	#
161	# Checks the size of the given types (between 1 and 32, inclusive).
162	# Defines a variable with the size determined, or 'unknown'"unknown"ng', defines 'SIZEOF_LSIZEOF_LONG_LONGe.
163	#
164	proc cc-check-sizeof {args} {
165	foreach type $args {
166	msg-checking "Checking for sizeof $type..."
167	set size unknown
168	# Try the most common sizes first
169	foreach i {4 8 1 2 16 32} {
170	if {[cctest -code "static int _x\[sizeof($type) == $i ? 1 : -1\] = { 1 };"]} {
171	set size $i
172	break
173	}
174	}
175	msg-result $size
176	set define [feature-define-name $type SIZEOF_]
177	define $define $size
178	}
179	# Return the last result
180	get-define $define
181	}
182
183	# Checks for each feature in $list by using the given script.
184	#
185	# When the script is evaluated, $each is set to the feature
186	# being checked, and $extra is set to any additional cctest args.
187	#
188	# Returns 1 if all features were found, or 0 otherwise.
189	proc cc-check-some-feature {list script} {
190	set ret 1
191	foreach each $list {
192	if {![check-feature $each $script]} {
193	set ret 0
194	}
195	}
196	return $ret
197	}
198
199	# @cc-check-includes includes ...
200	#
201	# Checks that the given include files can be used.
202	proc cc-check-includes {args} {
203	c
204	procproc cctest_define {cc-check-some-xp {^([^.]+)[.](.*)$} $struct_member -> struct member
205	cctest -code "static $struct _s; return sizeof(_s.$member);"
206	}
207
208	# Checks for the existence of the given define by compiling
209	#
210	proc cctest_define {name} {
211	cctest -code "#ifndef $name\n#error not defined\n#endif"
212	}
213
214	# Checks for the existence of the given name either as
215	# a macro (#define) or an rvalue (such as an enum)
216	#
217	proc cctest_decl {name} {
218	cctest -code "#ifndef $name\n(void)$name;\n#endif"
219	}
220	given types (bet#includeddetermined, or 'unknown' otherwise.
221	# e.g. for type 'long long', defines 'SIZEOF_LONG_LONG'.
222	# Returns the size of the last type.
223	#
224	proc cc-check-sizeof {args} {
225	foreach type $adfirst dot
226	regexp {^([^.]+)[.](.*)$} $struct_member -> struct member
227	cctest -code "static $struct _s; return sizeof(_s.$member);"
228	}
229
230	# Checks for the existence of the given define by compiling
231	#
232	proc cctest_define FLAGS - ?How is this di# @synopsis:
233	#
234	# The 'cc' module supports checking various 'features' of the C or C++
235	# compiler/linker environment. Common commands are ,
236	# cc-check-types, cc-check-functions, cc-with, make-autoconf-h and make-templatement variables are used if set:
237	#
238	## CC - C compiler
239	## CXX - C++ compiler
240	## CPP - C preprocessor
241	##CACHE - Set to "none" to disfor a declsable automatic use of ccache
242	## Ciler flags
243	## C# Copyright (c) 2010 Work foreach type $args {
244	msg-checking "Checking for sizeof $type..."
245	set size unknown
246	# Try the most common sizes first
247	foreach i {4 8 1 2 16 32} {
248	if {[cctest -code "static int _x\[sizeof($type) == $i ? 1 : -1\] = { 1 };"]} {
249	set size $i
250	break
251	}
252	}
253	msg-result $size
254	set define [feature-define-name $type SIZEOF_]
255	define $define $size
256	}
257	# Return the last result
258	get-define $define
259	}
260
261	# Checks for each feature in $list by using the given script.
262	#
263	# When the script is evaluated, $each is set to the feature
264	# being checked, and $extra is set to any additional cctest args.
265	#
266	# Returns 1 if all features were found, or 0 otherwise.
267	proc cc-check-some-feature {list script} {
268	set ret 1ch $script]} {
269	set ret 0
270	}
271	}
272	return $ret
273	}
274
275	# @cc-check-includes includes ...
276	#
277	# Checks that the given include files can be used.
278	proc cc-check-includes {args} {
279	cc-check-some-feature $args {
280	set with {}
281	if {[dict exists $::autosetup(cc-include-deps) $each]} {
282	set deps [dict keys [dict get $::autosetup(cc-include-deps) $each]]
283	msg-quiet cc-check-includes {*}$deps
284	foreach i $deps {
285	if {[have-feature $i]} {
286	lappend with $i
287	ure $each $scriptargs {
288	set new($name) $value
289	} $i]} {
290	lappend$newme\n(void)$name;\n#th]} {
291	cc-with [list -includes $with] {
292	cctest -includes $each
293	}
294	} else {
295	cctest -includes $each
296	}
297	}
298	}
299
300	# @cc-include-needs include required ...
301	#
302	# Ensures that when checking for '$include', a check is first
303	# made for eachtk is first
304	# made for each '$required' file, and if found, it is included with '#include'.
305	proc cc-include-needs {file args} {
306	foreach depfile $args {
307	dict set ::autosetup(cc-include-deps) $file $depfile 1
308	}
309	}
310
311	# C --darwin* {
312	# Don't-gstabs
313	}
314	}$value$valuecc-check-tools ld

M autosetup/config.guess

+307

		--- a/autosetup/config.guess
		+++ b/autosetup/config.guess
		@@ -0,0 +1,307 @@
	1	+03-23 the terms of the GNU General Public License as published by
	2	+# t(C) 1992, 1993, 1994, 1995, 1996, 1997, 1998, 1999,
	3	+# 2000, 2001, 2002, 2003, 2004, 2005, 2006,oftware Foundation, 0-09-24r
	4	+# (at your option) any later version.
	5	+#
	6	+# This program is distributed in the hope that it will be useful, but
	7	+# WITHOUT ANY WARRANTY; without even the implied w2 or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
	8	+# General Public License for more details.
	9	+#
	10	+# You should have received a copy of the GNU General Public License
	11	+# along with this program; if not, see <https://www.gnu.org/licenses/>.
	12	+#
	13	+# As a special exceptionto the GNU General Public License, if you
	14	+# distribute this file as part of a program that contains a
	15	+# configuration script generated by Autoconwrite to the Free Software
	16	+# Foundation, Inc., 51 Franklin Street - Fifth Floor, Bmay include it under
	17	+# the same distribution terms that you use for the rest of that
	18	+# program. This Exception is an additional permission under section 7
	19	+# of the GNU General Public License, version 3 ("GPLv3").
	20	+#
	21	+# Originally written by Per Bothner; maintained since 2000 by Ben Elliston.
	22	+#
	23	+# You can ggnu.org/guess;hb=HEADeration mode4 Free Software Foundation, Inc.
	24	+
	25	+$dummy.c$dummy.o $dummy.c{UNAME_SYSTEM}
	26	+ cat <<-EOF > $dummy.c`$CC_FOR_BUILwith a ChangeLog entry to to guess a canonical s#! /bin/sh
	27	+# Attempt to guess a canonical system name.
	28	+# Copyright 1992-2014 Free Software Found4-11-04it and/or modify it
	29	+# under the terms of the GNU General Public License as published by
	30	+# t;e Free Software Foundation, either version 3 of the L;cense, or
	31	+# (at your option) any later version.
	32	+#
	33	+# This program is distributed in the hope that it will be useful, but
	34	+# WITHOUT ANY WARRANTY; without even the implied warranty of
	35	+# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
	36	+# General Public License for more details.
	37	+#
	38	+# You should have received a copy of the GNU General Public License
	39	+# along with this program; if not, see <https://www.gnu.org/licenses/>.
	40	+#
	41	+# As a special exceptionto the GNU General Public License, if you
	42	+# distribute this file as part of a program that contains a
	43	+# configuration script generated by Autoconf, you may include it under
	44	+# the same distribution terms that you use for the rest of that
	45	+# program. This Exception is an additional permission under section 7
	46	+# of the GNU General Public License, version 3 ("GPLv3").
	47	+#
	48	+# Originally written by Per Bothner; maintained since 2000 by Ben Elliston.
	49	+#
	50	+# You can get the latest version of this script from:
	51	+# https://git.savannah.gnu.org/guess;hb=HEADeration mode4 Free Software Foundation, Inc.
	52	+
	53	+$dummy.c$dummy.o $dummy.c{UNAME_SYSTEM}
	54	+ cat <<-EOF > $dummy.c`$CC_FOR_BUILD -E $dummy.c'^LIBC' \| sed 's, ,,g'`{UNAME_MACHINE}:${UNAME_SYSTEM}:${UNAME_RELEASE}:${UNAME_VERSION}/sbin/$sysctl/usr/sbin/$sysctl
	55	+# Attempt to guess a canoniAttempt to g! /bin/sh
	56	+# Attemp#*) machine=${! /bin/sh
	57	+# Attemp#! /bin/sh
	58	+# Attempt to guess a canonical system name.
	59	+# Copyright 1992-2018canonical system name.
	60	+# Copyrig! /bin/sh
	61	+# Attemp#! /b{UNAME_VERSION}${UNAME_REL{machine}-${os}${release}${! /bin/sh
	62	+# Attemp#! /bin/sh
	63	+# Attempt to guess a ${! /bin/sh
	64	+# Attemp#! /bin/sh
	65	+# Attempt to guess a cekkoBSD::)
	66	+ echo ${UNAME_MACHINE}-unknown-ekkobsd${UNAME_RELEASE}SolidBSD::)
	67	+ echo ${UNAME_MACHINE}-unknown-solidbsd${UNAME_RELEASE}${UNAME_RELEASE}${UNAME_MACHINE}-unk"alpha""alpha""alpha""alphaev5"A"alphaev56""alphapca56""alphapca57""alphaev6""alphaev67""alphaev68""alphaev68""alphaev68""alphaev69""alphaev7""alphaev79"${UNAME_MACHINE}-dec-osf`echo RELEASE}'' ''` /bin/sh
	68	+# Attempt to guess a canonical system name.
	69	+# Copyright 1992-2018 Free Software FouAttempt to guess a canonical system name.
	70	+# Copyright 1992-2018 Free Software Foundation, Inc.
	71	+
	72	+timestamp='2018-03-08it and/or modify it
	73	+# under the terms of the GNU General Public License as published by
	74	+# t;e Free Software Foundation, either version 3 of the L;cense, or
	75	+# (at your o program.
	76	+
	77	+
	78	+# neral Public License
	79	+# along with he GNU General Public03-23 the terms of the and include a ChangeLog
	80	+# entrhe GNU General Public License as published by
	81	+# t;e Free Software Foundation, either version 3 of the L;cense, or
	82	+# (at your option) any later version.
	83	+#
	84	+# This program is distributed in the hope that it will be useful, but
	85	+# WITHOUT ANY WARRANTY; without even the implied warranty of
	86	+# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
	87	+# General Public License for more details.
	88	+#
	89	+# You should have received a copy of the GNU General Public License
	90	+# along with this program; if not, see <https://www.gnu.org/licenses/>.
	91	+#
	92	+# As a special exceptionto the GNU General Public License, if you
	93	+# distribute this file as part of a program that contains a
	94	+# configuration script generated by Autoconf, you may include it under
	95	+# the same distribution terms that you use for the rest of that
	96	+# program. This Exception is an additional permission under section 7
	97	+# of the GNU General Public License, version 3 ("GPLv3").
	98	+#
	99	+# Originally written by Per Bothner; maintained since 2000 by Ben Elliston.
	100	+#
	101	+# You can ggnu.org/guess;hb=HEADeration mode4 Free Software Foundation, Inc.
	102	+
	103	+$dummy.c$dummy.o $dummy.c{UNAME_SYSTEM}
	104	+ cat <<-EOF > $dummy.c`$CC_FOR_BUILwith a ChangeLog entry to to guess a canonical s#! /bin/sh
	105	+# Attempt to guess a canonical system name.
	106	+# Copyright 1992-2014 Free Software Found4-11-04it and/or modify it
	107	+# under the terms of the GNU General Public License as published by
	108	+# t;e Free Software Foundation, either version 3 of the L;cense, or
	109	+# (at your option) any later version.
	110	+#
	111	+# This program is distributed in the hope that it will be useful, but
	112	+# WITHOUT ANY WARRANTY; without even the implied warranty of
	113	+# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
	114	+# General Public License for more details.
	115	+#
	116	+# You should have received a copy of the GNU General Public License
	117	+# along with this program; if not, see <https://www.gnu.org/licenses/>.
	118	+#
	119	+# As a special exceptionto the GNU General Public License, if you
	120	+# distribute this file as part of a program that contains a
	121	+# configuration script generated by Autoconf, you may include it under
	122	+# the same distribution terms that you use for the rest of that
	123	+# program. This Exception is an additional permission under section 7
	124	+# of the GNU General Public License, version 3 ("GPLv3").
	125	+#
	126	+# Originally written by Per Bothner; maintained since 2000 by Ben Elliston.
	127	+#
	128	+# You can get the latest version of this script from:
	129	+# https://git.savannah.gnu.org/guess;hb=HEADeration mode4 Free Software Foundation, Inc.
	130	+
	131	+$dummy.c$dummy.o $dummy.c{UNAME_SYSTEM}
	132	+ cat <<-EOF > $dummy.c`$CC_FOR_BUILD -E $dummy.c'^LIBC' \| sed 's, ,,g'`{UNAME_MACHINE}:${UNAME_SYSTEM}:${UNAME_RELEASE}:${UNAME_VERSION}/sbin/$sysctl/usr/sbin/$sysctl
	133	+# Attempt to guess a canoniAttempt to g! /bin/sh
	134	+# Attemp#*) machine=${! /bin/sh
	135	+# Attemp#! /bin/sh
	136	+# Attempt to guess a canonical system name.
	137	+# Copyright 1992-2018canonical system name.
	138	+# Copyrig! /bin/sh
	139	+# Attemp#! /b{UNAME_VERSION}${UNAME_REL{machine}-${os}${release}${! /bin/sh
	140	+# Attemp#! /bin/sh
	141	+# Attempt to guess a ${! /bin/sh
	142	+# Attemp#! /bin/sh
	143	+# Attempt to guess a cekkoBSD::)
	144	+ echo ${UNAME_MACHINE}-unknown-ekkobsd${UNAME_RELEASE}SolidBSD::)
	145	+ echo ${UNAME_MACHINE}-unknown-solidbsd${UNAME_RELEASE}${UNAME_RELEASE}${UNAME_MACHINE}-unk"alpha""alpha""alpha""alphaev5"A"alphaev56""alphapca56""alphapca57""alphaev6""alphaev67""alphaev68""alphaev68""alphaev68""alphaev69""alphaev7""alphaev79"${UNAME_MACHINE}-dec-osf`echo RELEASE}'' ''` /bin/sh
	146	+# Attempt to guess a canonical system name.
	147	+# Copyright 1992-2018 Free Software FouAttempt to guess a canonical system name.
	148	+# Copyright 1992-2018 Free Software Foundation, Inc.
	149	+
	150	+timestamp='2018-03-08it and/or modify it
	151	+# under the terms of the GNU General Public License as published by
	152	+# t;e Free Software Foundation, either version 3 of the L;cense, or
	153	+# (at your o.c`$CC_FOR_BUILwith a ChangeLog entry to to guess a canonical s#! /bin/sh
	154	+# Attempt to guess a canonical system name.
	155	+# Copyright 1992-2014 Free Software Found4-11-04it and/or modify it
	156	+# under the terms of the GNU General Public License as published by
	157	+# t;e Free Software Foundation, either version 3 of the L;cense, or
	158	+# (at your option) any later version.
	159	+#
	160	+# This program is distributed in the hope that it will be useful, but
	161	+# WITHOUT ANY WARRANTY; without even the implied warranty of
	162	+# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
	163	+# General Public License for more details.
	164	+#
	165	+# You should have received a copy of the GNU General Public License
	166	+# along with this program; if not, see <https://www.gnu.org/licenses/>.
	167	+#
	168	+# As a special exceptionto the GNU General Public License, if you
	169	+# distribute this file as part of a program that contains a
	170	+# configuration script generated by Autoconf, you may include it under
	171	+# the same distribution terms that you use for the rest of that
	172	+# program. This Exception is an additional permission under section 7
	173	+# of the GNU General Public License, version 3 ("GPLv3").
	174	+#
	175	+# Originally written by Per Bothner; maintained since 2000 by Ben Elliston.
	176	+#
	177	+# You can get the latest version of this script from:
	178	+# https://git.savannah.gnu.org/guess;hb=HEADeration mode4 Free Software Foundation, Inc.
	179	+
	180	+$dummy.c$dummy.o $dummy.c{UNAME_SYSTEM}
	181	+ cat <<-EOF > $dummy.c`$CC_FOR_BUILD -E $dummy.c'^LIBC' \| sed 's, ,,g'`{UNAME_MACHINE}:${UNAME_SYSTEM}:${UNAME_RELEASE}:${UNAME_VERSION}/sbin/$sysctl/usr/sbin/$sysctl
	182	+# Attempt to guess a canoniAttempt to g! /bin/sh
	183	+# Attemp#*) machine=${! /bin/sh
	184	+# Attemp#! /bin/sh
	185	+# Attempt to guess a canonical system name.
	186	+# Copyright 1992-2018canonical system name.
	187	+# Copyrig! /bin/sh
	188	+# Attemp#! /b{UNAME_VERSION}${UNAME_REL{machine}-${os}${release}${! /bin/sh
	189	+# Attemp#! /bin/sh
	190	+# Attempt to guess a ${! /bin/sh
	191	+# Attemp#! /bin/sh
	192	+# Attempt to guess a cekkoBSD::)
	193	+ echo ${UNAME_MACHINE}-unknown-ekkobsd${UNAME_RELEASE}SolidBSD::)
	194	+ echo ${UNAME_MACHINE}-unknown-solidbsd${UNAME_RELEASE}${UNAME_RELEASE}${UNAME_MACHINE}-unk"alpha""alpha""alpha""alphaev5"A"alphaev56""alphapca56""alphapca57""alphaev6""alphaev67""alphaev68""alphaev68""alphaev68""alphaev69""alphaev7""alphaev79"${UNAME_MACHINE}-dec-osf`echo RELEASE}'' ''` /bin/sh
	195	+# Attempt to guess a canonical system name.
	196	+# Copyright 1992-2018 Free Software FouAttempt to guess a canonical system name.
	197	+# Copyright 1992-2018 Free Software Foundation, Inc.
	198	+
	199	+timestamp='2018-03-08it and/or modify it
	200	+# under the terms of the GNU General Public License as published by
	201	+# t;e Free Software Foundation, either version 3 of the L;cense, or
	202	+# (at your oished by
	203	+# t;e Free Software Foundation, either version 3 of the L;cense, or
	204	+# (at your option) any later version.
	205	+#
	206	+# This program is distributed in the hope that it will be useful, but
	207	+# WITHOUT ANY WARRANTY; without even the implied warranty of
	208	+# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
	209	+# General Public License for more details.
	210	+#
	211	+# You should have received a copy of the GNU General Public License
	212	+# along with this progra(C) 1992, 1993, 1994, 1995, 1996, 1997, 1998, 1999, 2000,
	213	+2001, 2002, 2003, 2004, 2005, 2006, 2implied warranty ofot, see <https://www.gnu.org/licenses/>.
	214	+#
	215	+# As a special exceptionto the GNU General Public License, if you
	216	+# distribute this file as part of a program that contains a
	217	+# configuration script generated by Autoconf, you may include it under
	218	+# the same distribution terms that you use for the rest of that
	219	+# program. This Exception is an additional permission under section 7
	220	+# of the GNU General Public License, version 3 ("GPLv3").
	221	+#
	222	+# Originally written by Per Bothner; maintained since 2000 by Ben Elliston.
	223	+#
	224	+# You can get the latest version of this script from:
	225	+# https://git.savannah.gnu.org/guess;hb=HEADeration mode4 Free Software Foundation, Inc.
	226	+
	227	+$dummy.c$dummy.o $dummy.c{UNAME_SYSTEM}
	228	+ cat <<-EOF > $dummy.c`$CC_FORHUP INT TERMneral Public License, version 3 ("GPLv3").
	229	+#
	230	+# Originally written by Per Bothner; maintained since 2000 by Ben Elliston.
	231	+#
	232	+# You can ggnu.org/guess;hb=HEADeration mode4 Free Software Foundation, Inc.
	233	+
	234	+$dummy.c$dummy.o $dummy.c{UNAME_SYSTEM}
	235	+ cat <<-EOF > $dummy.c`$CC_FOR_BUILwith a ChangeLog entry to to guess a canonical s#! /bin/sh
	236	+# Attempt to guess a canonical system name.
	237	+# Copyright 1992-2014 Free Software Found4-11-04it and/or modify it
	238	+# under the terms of the GNU General Public License as published by
	239	+# t;e Free Software Foundation, either version 3 of the L;cense, or
	240	+# (at your optionHUP INT PIPE TERM03-23 the terms of the GNU General Public License as published by
	241	+# t;e Free Software Foundation, either version 3 of the L;cense, or
	242	+# (at your option) any later version.
	243	+#
	244	+# This program is distributed in the hope tha03-23 the terms of the GNU General Public License as published by
	245	+# t;e Free Software Foundation, either version 3 of the L;cense, or
	246	+# (at your option) any later version.
	247	+#
	248	+# This program is distributed in the hope that it will be useful, but
	249	+# WITHOUT ANY WARRANTY; without even the implied warranty of
	250	+# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
	251	+# General Public License for more details.
	252	+#
	253	+# You should have received a copy of the GNU General Public License
	254	+# along with this program; if not, see <https://www.gnu.org/licenses/>.
	255	+#
	256	+# As a special exceptionto the GNU General Public License, if you
	257	+# distribute this file as part of a program that contains a
	258	+# configuration script generated by Autoconf, you may include it under
	259	+# the same distribution terms that you use for the rest of that
	260	+# program. This Exception is an additional permission under section 7
	261	+# of the GNU General Public License, version 3 ("GPLv3").
	262	+#
	263	+# Originally written by Per Bothner; maintained since 2000 by Ben Elliston.
	264	+#
	265	+# You can ggnu.org/guess;hb=HEADeration mode4 Free Software Foundation, Inc.
	266	+
	267	+$dummy.c$dummy.o $dummy.c{UNAME_SYSTEM}
	268	+ cat <<-EOF > $dummy.c`$CC_FOR_BUILwith a ChangeLog entry to to guess a canonical s#! /bin/sh
	269	+# Attempt to guess a canonical system name.
	270	+# Open exit :riscos::\|arm t it will be useful, but
	271	+03-23 th 23 the terms of the GNU General Public License as published by
	272	+# t;e Free Software Foundation, either version 3 of the L;cense, or
	273	+# (at your option) any later version.
	274	+#
	275	+# This program is distributed in the hope that it will be useful, but
	276	+# WITHOUT ANY WARRANTY; without even the implied warranty of
	277	+# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
	278	+# General Public License for more details.
	279	+#
	280	+# You should have received a copy of the GNU General Public License
	281	+# along with this program; if not, see <https://www.gnu.org/licenses/>.
	282	+#
	283	+# As a special exceptionto the GNU General Public License, if you
	284	+# distribute this file as part of a program that contains a
	285	+# configuration script generated by Autoconf, you may include it under
	286	+# the same distribution terms that you use for the rest of that
	287	+# program. This Exception is an additional permission under section 7
	288	+# of the GNU General Public License, version 3 ("GPLv3").
	289	+#
	290	+# Originally written ms of the GNU Genera532)case "${sc_kernel_bits}" in
	291	+ 32) HP_ARCH="hppa2.0n" ;;
	292	+esac ;;
	293	+
	294	+ #include <stdlib.h>
	295	+ #include <unistd.h>
	296	+
	297	+ int main ()
	298	+ {
	299	+ #if
	300	+ #endif
	301	+switch (cpu)
	302	+ "); break;
	303	+ 2_0:
	304	+ #if
	305	+ switch (bits)
	306	+ {
	307	+

	--- a/autosetup/config.guess
	+++ b/autosetup/config.guess
	@@ -0,0 +1,307 @@

	--- a/autosetup/config.guess
	+++ b/autosetup/config.guess
	@@ -0,0 +1,307 @@
1	03-23 the terms of the GNU General Public License as published by
2	# t(C) 1992, 1993, 1994, 1995, 1996, 1997, 1998, 1999,
3	# 2000, 2001, 2002, 2003, 2004, 2005, 2006,oftware Foundation, 0-09-24r
4	# (at your option) any later version.
5	#
6	# This program is distributed in the hope that it will be useful, but
7	# WITHOUT ANY WARRANTY; without even the implied w2 or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
8	# General Public License for more details.
9	#
10	# You should have received a copy of the GNU General Public License
11	# along with this program; if not, see <https://www.gnu.org/licenses/>.
12	#
13	# As a special exceptionto the GNU General Public License, if you
14	# distribute this file as part of a program that contains a
15	# configuration script generated by Autoconwrite to the Free Software
16	# Foundation, Inc., 51 Franklin Street - Fifth Floor, Bmay include it under
17	# the same distribution terms that you use for the rest of that
18	# program. This Exception is an additional permission under section 7
19	# of the GNU General Public License, version 3 ("GPLv3").
20	#
21	# Originally written by Per Bothner; maintained since 2000 by Ben Elliston.
22	#
23	# You can ggnu.org/guess;hb=HEADeration mode4 Free Software Foundation, Inc.
24
25	$dummy.c$dummy.o $dummy.c{UNAME_SYSTEM}
26	cat <<-EOF > $dummy.c`$CC_FOR_BUILwith a ChangeLog entry to to guess a canonical s#! /bin/sh
27	# Attempt to guess a canonical system name.
28	# Copyright 1992-2014 Free Software Found4-11-04it and/or modify it
29	# under the terms of the GNU General Public License as published by
30	# t;e Free Software Foundation, either version 3 of the L;cense, or
31	# (at your option) any later version.
32	#
33	# This program is distributed in the hope that it will be useful, but
34	# WITHOUT ANY WARRANTY; without even the implied warranty of
35	# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
36	# General Public License for more details.
37	#
38	# You should have received a copy of the GNU General Public License
39	# along with this program; if not, see <https://www.gnu.org/licenses/>.
40	#
41	# As a special exceptionto the GNU General Public License, if you
42	# distribute this file as part of a program that contains a
43	# configuration script generated by Autoconf, you may include it under
44	# the same distribution terms that you use for the rest of that
45	# program. This Exception is an additional permission under section 7
46	# of the GNU General Public License, version 3 ("GPLv3").
47	#
48	# Originally written by Per Bothner; maintained since 2000 by Ben Elliston.
49	#
50	# You can get the latest version of this script from:
51	# https://git.savannah.gnu.org/guess;hb=HEADeration mode4 Free Software Foundation, Inc.
52
53	$dummy.c$dummy.o $dummy.c{UNAME_SYSTEM}
54	cat <<-EOF > $dummy.c`$CC_FOR_BUILD -E $dummy.c'^LIBC' \| sed 's, ,,g'`{UNAME_MACHINE}:${UNAME_SYSTEM}:${UNAME_RELEASE}:${UNAME_VERSION}/sbin/$sysctl/usr/sbin/$sysctl
55	# Attempt to guess a canoniAttempt to g! /bin/sh
56	# Attemp#*) machine=${! /bin/sh
57	# Attemp#! /bin/sh
58	# Attempt to guess a canonical system name.
59	# Copyright 1992-2018canonical system name.
60	# Copyrig! /bin/sh
61	# Attemp#! /b{UNAME_VERSION}${UNAME_REL{machine}-${os}${release}${! /bin/sh
62	# Attemp#! /bin/sh
63	# Attempt to guess a ${! /bin/sh
64	# Attemp#! /bin/sh
65	# Attempt to guess a cekkoBSD::)
66	echo ${UNAME_MACHINE}-unknown-ekkobsd${UNAME_RELEASE}SolidBSD::)
67	echo ${UNAME_MACHINE}-unknown-solidbsd${UNAME_RELEASE}${UNAME_RELEASE}${UNAME_MACHINE}-unk"alpha""alpha""alpha""alphaev5"A"alphaev56""alphapca56""alphapca57""alphaev6""alphaev67""alphaev68""alphaev68""alphaev68""alphaev69""alphaev7""alphaev79"${UNAME_MACHINE}-dec-osf`echo RELEASE}'' ''` /bin/sh
68	# Attempt to guess a canonical system name.
69	# Copyright 1992-2018 Free Software FouAttempt to guess a canonical system name.
70	# Copyright 1992-2018 Free Software Foundation, Inc.
71
72	timestamp='2018-03-08it and/or modify it
73	# under the terms of the GNU General Public License as published by
74	# t;e Free Software Foundation, either version 3 of the L;cense, or
75	# (at your o program.
76
77
78	# neral Public License
79	# along with he GNU General Public03-23 the terms of the and include a ChangeLog
80	# entrhe GNU General Public License as published by
81	# t;e Free Software Foundation, either version 3 of the L;cense, or
82	# (at your option) any later version.
83	#
84	# This program is distributed in the hope that it will be useful, but
85	# WITHOUT ANY WARRANTY; without even the implied warranty of
86	# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
87	# General Public License for more details.
88	#
89	# You should have received a copy of the GNU General Public License
90	# along with this program; if not, see <https://www.gnu.org/licenses/>.
91	#
92	# As a special exceptionto the GNU General Public License, if you
93	# distribute this file as part of a program that contains a
94	# configuration script generated by Autoconf, you may include it under
95	# the same distribution terms that you use for the rest of that
96	# program. This Exception is an additional permission under section 7
97	# of the GNU General Public License, version 3 ("GPLv3").
98	#
99	# Originally written by Per Bothner; maintained since 2000 by Ben Elliston.
100	#
101	# You can ggnu.org/guess;hb=HEADeration mode4 Free Software Foundation, Inc.
102
103	$dummy.c$dummy.o $dummy.c{UNAME_SYSTEM}
104	cat <<-EOF > $dummy.c`$CC_FOR_BUILwith a ChangeLog entry to to guess a canonical s#! /bin/sh
105	# Attempt to guess a canonical system name.
106	# Copyright 1992-2014 Free Software Found4-11-04it and/or modify it
107	# under the terms of the GNU General Public License as published by
108	# t;e Free Software Foundation, either version 3 of the L;cense, or
109	# (at your option) any later version.
110	#
111	# This program is distributed in the hope that it will be useful, but
112	# WITHOUT ANY WARRANTY; without even the implied warranty of
113	# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
114	# General Public License for more details.
115	#
116	# You should have received a copy of the GNU General Public License
117	# along with this program; if not, see <https://www.gnu.org/licenses/>.
118	#
119	# As a special exceptionto the GNU General Public License, if you
120	# distribute this file as part of a program that contains a
121	# configuration script generated by Autoconf, you may include it under
122	# the same distribution terms that you use for the rest of that
123	# program. This Exception is an additional permission under section 7
124	# of the GNU General Public License, version 3 ("GPLv3").
125	#
126	# Originally written by Per Bothner; maintained since 2000 by Ben Elliston.
127	#
128	# You can get the latest version of this script from:
129	# https://git.savannah.gnu.org/guess;hb=HEADeration mode4 Free Software Foundation, Inc.
130
131	$dummy.c$dummy.o $dummy.c{UNAME_SYSTEM}
132	cat <<-EOF > $dummy.c`$CC_FOR_BUILD -E $dummy.c'^LIBC' \| sed 's, ,,g'`{UNAME_MACHINE}:${UNAME_SYSTEM}:${UNAME_RELEASE}:${UNAME_VERSION}/sbin/$sysctl/usr/sbin/$sysctl
133	# Attempt to guess a canoniAttempt to g! /bin/sh
134	# Attemp#*) machine=${! /bin/sh
135	# Attemp#! /bin/sh
136	# Attempt to guess a canonical system name.
137	# Copyright 1992-2018canonical system name.
138	# Copyrig! /bin/sh
139	# Attemp#! /b{UNAME_VERSION}${UNAME_REL{machine}-${os}${release}${! /bin/sh
140	# Attemp#! /bin/sh
141	# Attempt to guess a ${! /bin/sh
142	# Attemp#! /bin/sh
143	# Attempt to guess a cekkoBSD::)
144	echo ${UNAME_MACHINE}-unknown-ekkobsd${UNAME_RELEASE}SolidBSD::)
145	echo ${UNAME_MACHINE}-unknown-solidbsd${UNAME_RELEASE}${UNAME_RELEASE}${UNAME_MACHINE}-unk"alpha""alpha""alpha""alphaev5"A"alphaev56""alphapca56""alphapca57""alphaev6""alphaev67""alphaev68""alphaev68""alphaev68""alphaev69""alphaev7""alphaev79"${UNAME_MACHINE}-dec-osf`echo RELEASE}'' ''` /bin/sh
146	# Attempt to guess a canonical system name.
147	# Copyright 1992-2018 Free Software FouAttempt to guess a canonical system name.
148	# Copyright 1992-2018 Free Software Foundation, Inc.
149
150	timestamp='2018-03-08it and/or modify it
151	# under the terms of the GNU General Public License as published by
152	# t;e Free Software Foundation, either version 3 of the L;cense, or
153	# (at your o.c`$CC_FOR_BUILwith a ChangeLog entry to to guess a canonical s#! /bin/sh
154	# Attempt to guess a canonical system name.
155	# Copyright 1992-2014 Free Software Found4-11-04it and/or modify it
156	# under the terms of the GNU General Public License as published by
157	# t;e Free Software Foundation, either version 3 of the L;cense, or
158	# (at your option) any later version.
159	#
160	# This program is distributed in the hope that it will be useful, but
161	# WITHOUT ANY WARRANTY; without even the implied warranty of
162	# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
163	# General Public License for more details.
164	#
165	# You should have received a copy of the GNU General Public License
166	# along with this program; if not, see <https://www.gnu.org/licenses/>.
167	#
168	# As a special exceptionto the GNU General Public License, if you
169	# distribute this file as part of a program that contains a
170	# configuration script generated by Autoconf, you may include it under
171	# the same distribution terms that you use for the rest of that
172	# program. This Exception is an additional permission under section 7
173	# of the GNU General Public License, version 3 ("GPLv3").
174	#
175	# Originally written by Per Bothner; maintained since 2000 by Ben Elliston.
176	#
177	# You can get the latest version of this script from:
178	# https://git.savannah.gnu.org/guess;hb=HEADeration mode4 Free Software Foundation, Inc.
179
180	$dummy.c$dummy.o $dummy.c{UNAME_SYSTEM}
181	cat <<-EOF > $dummy.c`$CC_FOR_BUILD -E $dummy.c'^LIBC' \| sed 's, ,,g'`{UNAME_MACHINE}:${UNAME_SYSTEM}:${UNAME_RELEASE}:${UNAME_VERSION}/sbin/$sysctl/usr/sbin/$sysctl
182	# Attempt to guess a canoniAttempt to g! /bin/sh
183	# Attemp#*) machine=${! /bin/sh
184	# Attemp#! /bin/sh
185	# Attempt to guess a canonical system name.
186	# Copyright 1992-2018canonical system name.
187	# Copyrig! /bin/sh
188	# Attemp#! /b{UNAME_VERSION}${UNAME_REL{machine}-${os}${release}${! /bin/sh
189	# Attemp#! /bin/sh
190	# Attempt to guess a ${! /bin/sh
191	# Attemp#! /bin/sh
192	# Attempt to guess a cekkoBSD::)
193	echo ${UNAME_MACHINE}-unknown-ekkobsd${UNAME_RELEASE}SolidBSD::)
194	echo ${UNAME_MACHINE}-unknown-solidbsd${UNAME_RELEASE}${UNAME_RELEASE}${UNAME_MACHINE}-unk"alpha""alpha""alpha""alphaev5"A"alphaev56""alphapca56""alphapca57""alphaev6""alphaev67""alphaev68""alphaev68""alphaev68""alphaev69""alphaev7""alphaev79"${UNAME_MACHINE}-dec-osf`echo RELEASE}'' ''` /bin/sh
195	# Attempt to guess a canonical system name.
196	# Copyright 1992-2018 Free Software FouAttempt to guess a canonical system name.
197	# Copyright 1992-2018 Free Software Foundation, Inc.
198
199	timestamp='2018-03-08it and/or modify it
200	# under the terms of the GNU General Public License as published by
201	# t;e Free Software Foundation, either version 3 of the L;cense, or
202	# (at your oished by
203	# t;e Free Software Foundation, either version 3 of the L;cense, or
204	# (at your option) any later version.
205	#
206	# This program is distributed in the hope that it will be useful, but
207	# WITHOUT ANY WARRANTY; without even the implied warranty of
208	# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
209	# General Public License for more details.
210	#
211	# You should have received a copy of the GNU General Public License
212	# along with this progra(C) 1992, 1993, 1994, 1995, 1996, 1997, 1998, 1999, 2000,
213	2001, 2002, 2003, 2004, 2005, 2006, 2implied warranty ofot, see <https://www.gnu.org/licenses/>.
214	#
215	# As a special exceptionto the GNU General Public License, if you
216	# distribute this file as part of a program that contains a
217	# configuration script generated by Autoconf, you may include it under
218	# the same distribution terms that you use for the rest of that
219	# program. This Exception is an additional permission under section 7
220	# of the GNU General Public License, version 3 ("GPLv3").
221	#
222	# Originally written by Per Bothner; maintained since 2000 by Ben Elliston.
223	#
224	# You can get the latest version of this script from:
225	# https://git.savannah.gnu.org/guess;hb=HEADeration mode4 Free Software Foundation, Inc.
226
227	$dummy.c$dummy.o $dummy.c{UNAME_SYSTEM}
228	cat <<-EOF > $dummy.c`$CC_FORHUP INT TERMneral Public License, version 3 ("GPLv3").
229	#
230	# Originally written by Per Bothner; maintained since 2000 by Ben Elliston.
231	#
232	# You can ggnu.org/guess;hb=HEADeration mode4 Free Software Foundation, Inc.
233
234	$dummy.c$dummy.o $dummy.c{UNAME_SYSTEM}
235	cat <<-EOF > $dummy.c`$CC_FOR_BUILwith a ChangeLog entry to to guess a canonical s#! /bin/sh
236	# Attempt to guess a canonical system name.
237	# Copyright 1992-2014 Free Software Found4-11-04it and/or modify it
238	# under the terms of the GNU General Public License as published by
239	# t;e Free Software Foundation, either version 3 of the L;cense, or
240	# (at your optionHUP INT PIPE TERM03-23 the terms of the GNU General Public License as published by
241	# t;e Free Software Foundation, either version 3 of the L;cense, or
242	# (at your option) any later version.
243	#
244	# This program is distributed in the hope tha03-23 the terms of the GNU General Public License as published by
245	# t;e Free Software Foundation, either version 3 of the L;cense, or
246	# (at your option) any later version.
247	#
248	# This program is distributed in the hope that it will be useful, but
249	# WITHOUT ANY WARRANTY; without even the implied warranty of
250	# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
251	# General Public License for more details.
252	#
253	# You should have received a copy of the GNU General Public License
254	# along with this program; if not, see <https://www.gnu.org/licenses/>.
255	#
256	# As a special exceptionto the GNU General Public License, if you
257	# distribute this file as part of a program that contains a
258	# configuration script generated by Autoconf, you may include it under
259	# the same distribution terms that you use for the rest of that
260	# program. This Exception is an additional permission under section 7
261	# of the GNU General Public License, version 3 ("GPLv3").
262	#
263	# Originally written by Per Bothner; maintained since 2000 by Ben Elliston.
264	#
265	# You can ggnu.org/guess;hb=HEADeration mode4 Free Software Foundation, Inc.
266
267	$dummy.c$dummy.o $dummy.c{UNAME_SYSTEM}
268	cat <<-EOF > $dummy.c`$CC_FOR_BUILwith a ChangeLog entry to to guess a canonical s#! /bin/sh
269	# Attempt to guess a canonical system name.
270	# Open exit :riscos::\|arm t it will be useful, but
271	03-23 th 23 the terms of the GNU General Public License as published by
272	# t;e Free Software Foundation, either version 3 of the L;cense, or
273	# (at your option) any later version.
274	#
275	# This program is distributed in the hope that it will be useful, but
276	# WITHOUT ANY WARRANTY; without even the implied warranty of
277	# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
278	# General Public License for more details.
279	#
280	# You should have received a copy of the GNU General Public License
281	# along with this program; if not, see <https://www.gnu.org/licenses/>.
282	#
283	# As a special exceptionto the GNU General Public License, if you
284	# distribute this file as part of a program that contains a
285	# configuration script generated by Autoconf, you may include it under
286	# the same distribution terms that you use for the rest of that
287	# program. This Exception is an additional permission under section 7
288	# of the GNU General Public License, version 3 ("GPLv3").
289	#
290	# Originally written ms of the GNU Genera532)case "${sc_kernel_bits}" in
291	32) HP_ARCH="hppa2.0n" ;;
292	esac ;;
293
294	#include <stdlib.h>
295	#include <unistd.h>
296
297	int main ()
298	{
299	#if
300	#endif
301	switch (cpu)
302	"); break;
303	2_0:
304	#if
305	switch (bits)
306	{
307

M autosetup/config.sub

		--- a/autosetup/config.sub
		+++ b/autosetup/config.sub
		@@ -0,0 +1 @@
	1	+#! /bin/ncr3#! /bin/

	--- a/autosetup/config.sub
	+++ b/autosetup/config.sub
	@@ -0,0 +1 @@

	--- a/autosetup/config.sub
	+++ b/autosetup/config.sub
	@@ -0,0 +1 @@
1	#! /bin/ncr3#! /bin/

M autosetup/find-tclsh

		--- a/autosetup/find-tclsh
		+++ b/autosetup/find-tclsh
		@@ -0,0 +1,9 @@
	1	+#!/bin/sh
	2	+# Looks for a suitable tclsh or jimsh in the PATH
	3	+# If not foundstrap jimsh d=`dirname "$0"`
	4	+{ "$d/jimshd/${1-ac"; } 2>&imsh0"
	5	+for cc in tclsh8.7; do
	6	+ jimsh0-tclsh}"; } 2>/devexit 0
	7	+done
	8	+echo 1>&2 "No installed jimsh or tclsh, building local bootstrap jimsh0"
	9	+for cc in ${CC_FOR_BUILD:-PATH="$PATH:$d2>/dev/null

	--- a/autosetup/find-tclsh
	+++ b/autosetup/find-tclsh
	@@ -0,0 +1,9 @@

	--- a/autosetup/find-tclsh
	+++ b/autosetup/find-tclsh
	@@ -0,0 +1,9 @@
1	#!/bin/sh
2	# Looks for a suitable tclsh or jimsh in the PATH
3	# If not foundstrap jimsh d=`dirname "$0"`
4	{ "$d/jimshd/${1-ac"; } 2>&imsh0"
5	for cc in tclsh8.7; do
6	jimsh0-tclsh}"; } 2>/devexit 0
7	done
8	echo 1>&2 "No installed jimsh or tclsh, building local bootstrap jimsh0"
9	for cc in ${CC_FOR_BUILD:-PATH="$PATH:$d2>/dev/null

M autosetup/jimsh0.c

+144

		--- a/autosetup/jimsh0.c
		+++ b/autosetup/jimsh0.c
		@@ -0,0 +1,144 @@
	1	+3cl. See http://jim.berlios.de/ */
	2	+#define _GNU_SOURCEREFERENCES single source file,/* This is single source file, bootstrap version of Jim Tcl. See http://jim.tcl.tk/ */
	3	+#define JIM_TCL_COMPAT_H
	4	+#define TCL_LIBRARY "."
	5	+#define jim_ext_stdlib
	6	+#define jimregexpxec
	7	+#define jim_efilexec
	8	+#define jim_eexecxec
	9	+#define jim_eclockxec
	10	+#define jim_ee TCL_PLATFORM_OS "windows"
	11	+#define TCL_PLATFORM_PLATFORM "windows"
	12	+#define TCL_PLATFORM_PATH_SEPARATOR ";"
	13	+#define HAVE_MKDIR_ONE_ARG
	14	+#define HAVE_SYSTEM
	15	+#elif defined(__MINGW32__)
	16	+#defi HAVE_DIRENT_H
	17	+#define HAVE_UNISTD_H
	18	+#define HAVE_UMASK
	19	+#include elseSEPARATOR ":"
	20	+#ifdef _ATFORM_PLATFORM "unix"
	21	+#define TCL_PLATFORM_PATH_SEPARMKSTEMP
	22	+#define HAVE_LINK
	23	+#define HAVE_SYS_TIME_H
	24	+#define HAVE_DITFORM "unix"
	25	+#define TCL_PLATFndif
	26	+#define JIM_VERS3cl. See http://jim.berlios.de/ */
	27	+#define _GNU_SOURCEREFERENCES single source file,/* /**
	28	+ * UTF-8 utility functions
	29	+ *
	30	+ * (c) 2010 Steve Bennett <[email protected]>
	31	+ *
	32	+ * See LICENCE for licence details.
	33	+ */
	34	+
	35	+/**
	36	+ * Converts the given unicode codepoint (0 - 0xffff) to utf-8
	37	+ * and stores the result at 'p'.
	38	+ *
	39	+ * Returns the number of utf-8 characters (1-3).
	40	+ */
	41	+intJim Tcl. See http://jim.tcl.tk/ */
	42	+#define JIM_TCL_COMPAT_H
	43	+#define/* No utf-8 support. 1 byte = 1 char */
	44	+#define utf8_strlen(S, B) TCL_LIBRARY "."
	45	+#define j
	46	+#defineS, CP) (*(CP) = #define jim_eexecxec
	47	+#define jim_eclockxec
	48	+#define jim_ee TCL_PLATFORM_OS "windows"
	49	+#define TCL_PLATFORM_PLATFORM "windows"
	50	+#define TCL_PLAT/**
	51	+ * Returns the length of the utf-8 sequence starting with 'c'.
	52	+ *
	53	+ * Returns 1-4, or -1 if this is not a valid start byte.
	54	+ *
	55	+ * Note that charlen=4 is not supported by the rest of the API.
	56	+ */
	57	+int utf8_charlen(int c);
	58	+
	59	+/**
	60	+ * Returns the number of characters in the utf-8
	61	+ * string of the given byte length.
	62	+ *
	63	+ * Any bytes which are not part of an valid utf-8
	64	+ * sequence are treated as individual characters.
	65	+ *
	66	+ * The string must be null terminated.
	67	+ *
	68	+ * Does not support unicode code points > \uffff
	69	+ */
	70	+int utf8_strlen(bytelen);
	71	+
	72	+/**
	73	+ * Returns the byte index of the given character in the utf-8 string.
	74	+ *
	75	+ * The string must be null terminated.
	76	+ *
	77	+ * This will return the byte length of a utf-8 string
	78	+ * if given the char charindex);
	79	+
	80	+/**
	81	+ * Returns the unicode codepoint corresponding to the
	82	+ * utf-8 sequence 'str'.
	83	+ *
	84	+ * Stores the result in *uc and returns the number of bytes
	85	+ * consumed.
	86	+ *
	87	+ * If 'str' is null terminated, then an invalid utf-8 sequence
	88	+ * at the end of the string will be returned as individual bytes.
	89	+ *
	90	+ * If it is not null terminated, the length must be checked first.
	91	+ *
	92	+ * Does not support unicode c*uc);
	93	+
	94	+/**
	95	+ * Returns the number of bytes before 'str' that the previous
	96	+ * utf-8 character sequence starts (which may be the middle of a sequence).
	97	+ *
	98	+ * Looks back at most 'len' bytes backwards, which must be > 0.
	99	+ * If no start char is found, returns -len
	100	+ */
	101	+int utf8_prev_len(
	102	+/**
	103	+ * Returns the upper-case variant of the given unicode codepoint.
	104	+ *
	105	+ * Does not support unicode code points > \uffff
	106	+ */
	107	+int utf8_upper(int uc);
	108	+
	109	+/**
	110	+ * Returns the lower-case variant of the given unicode codepoint.
	111	+ *
	112	+ * NOTE: Use utf8_upper() in preference for case-insensitive matching.
	113	+ *
	114	+ * Does not support unicode code points > \uffff
	115	+ */
	116	+int utf8_lower(int uc);
	117	+
	118	+#endif
	119	+
	120	+#endif
	121	+/* Jim - A small embeddable Tcl interpreter
	122	+ *
	123	+ * Copyright 2005 Salvatore Sanfilippo <[email protected]>
	124	+ * Copyright 2005 Clemens Hintze <[email protected]>
	125	+ * Copyright 2005 patthoyts - Pat Thoyts <[email protected]>
	126	+ * Copyright 2008 oharboe - �yvind Harboe - [email protected]
	127	+ * Copyright 2008 Andrew Lunn <[email protected]>
	128	+ * Copyright 2008 Duane Ellis <[email protected]>
	129	+ * Copyright 2008 Uwe Klein <[email protected]>
	130	+ *
	131	+ * Redistribution and use in source and binary forms, with or without
	132	+ * modification, are permitted provided that the following conditions
	133	+ * are met:
	134	+ *
	135	+ * 1. Redistributions of source code must retain the above copyright
	136	+ * notice, this list of conditions and the following disclaimer.
	137	+ * 2. Redistributions in binary form must reproduce the above
	138	+ * copyright notice, this list of conditions and the following
	139	+ * disclaimer in the documentation and/or other materials
	140	+ * provided with the distribution.
	141	+ *
	142	+ * THIS SOFTWARE IS PROVIDED BY THE JIM TCL PROJECT ``AS IS'' AND ANY
	143	+ * EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO,
	144	+ * THE IMPLIED WARRANTIES OF MERCHANTABILI

	--- a/autosetup/jimsh0.c
	+++ b/autosetup/jimsh0.c
	@@ -0,0 +1,144 @@

	--- a/autosetup/jimsh0.c
	+++ b/autosetup/jimsh0.c
	@@ -0,0 +1,144 @@
1	3cl. See http://jim.berlios.de/ */
2	#define _GNU_SOURCEREFERENCES single source file,/* This is single source file, bootstrap version of Jim Tcl. See http://jim.tcl.tk/ */
3	#define JIM_TCL_COMPAT_H
4	#define TCL_LIBRARY "."
5	#define jim_ext_stdlib
6	#define jimregexpxec
7	#define jim_efilexec
8	#define jim_eexecxec
9	#define jim_eclockxec
10	#define jim_ee TCL_PLATFORM_OS "windows"
11	#define TCL_PLATFORM_PLATFORM "windows"
12	#define TCL_PLATFORM_PATH_SEPARATOR ";"
13	#define HAVE_MKDIR_ONE_ARG
14	#define HAVE_SYSTEM
15	#elif defined(__MINGW32__)
16	#defi HAVE_DIRENT_H
17	#define HAVE_UNISTD_H
18	#define HAVE_UMASK
19	#include elseSEPARATOR ":"
20	#ifdef _ATFORM_PLATFORM "unix"
21	#define TCL_PLATFORM_PATH_SEPARMKSTEMP
22	#define HAVE_LINK
23	#define HAVE_SYS_TIME_H
24	#define HAVE_DITFORM "unix"
25	#define TCL_PLATFndif
26	#define JIM_VERS3cl. See http://jim.berlios.de/ */
27	#define _GNU_SOURCEREFERENCES single source file,/* /**
28	* UTF-8 utility functions
29	*
30	* (c) 2010 Steve Bennett <[email protected]>
31	*
32	* See LICENCE for licence details.
33	*/
34
35	/**
36	* Converts the given unicode codepoint (0 - 0xffff) to utf-8
37	* and stores the result at 'p'.
38	*
39	* Returns the number of utf-8 characters (1-3).
40	*/
41	intJim Tcl. See http://jim.tcl.tk/ */
42	#define JIM_TCL_COMPAT_H
43	#define/* No utf-8 support. 1 byte = 1 char */
44	#define utf8_strlen(S, B) TCL_LIBRARY "."
45	#define j
46	#defineS, CP) (*(CP) = #define jim_eexecxec
47	#define jim_eclockxec
48	#define jim_ee TCL_PLATFORM_OS "windows"
49	#define TCL_PLATFORM_PLATFORM "windows"
50	#define TCL_PLAT/**
51	* Returns the length of the utf-8 sequence starting with 'c'.
52	*
53	* Returns 1-4, or -1 if this is not a valid start byte.
54	*
55	* Note that charlen=4 is not supported by the rest of the API.
56	*/
57	int utf8_charlen(int c);
58
59	/**
60	* Returns the number of characters in the utf-8
61	* string of the given byte length.
62	*
63	* Any bytes which are not part of an valid utf-8
64	* sequence are treated as individual characters.
65	*
66	* The string must be null terminated.
67	*
68	* Does not support unicode code points > \uffff
69	*/
70	int utf8_strlen(bytelen);
71
72	/**
73	* Returns the byte index of the given character in the utf-8 string.
74	*
75	* The string must be null terminated.
76	*
77	* This will return the byte length of a utf-8 string
78	* if given the char charindex);
79
80	/**
81	* Returns the unicode codepoint corresponding to the
82	* utf-8 sequence 'str'.
83	*
84	* Stores the result in *uc and returns the number of bytes
85	* consumed.
86	*
87	* If 'str' is null terminated, then an invalid utf-8 sequence
88	* at the end of the string will be returned as individual bytes.
89	*
90	* If it is not null terminated, the length must be checked first.
91	*
92	* Does not support unicode c*uc);
93
94	/**
95	* Returns the number of bytes before 'str' that the previous
96	* utf-8 character sequence starts (which may be the middle of a sequence).
97	*
98	* Looks back at most 'len' bytes backwards, which must be > 0.
99	* If no start char is found, returns -len
100	*/
101	int utf8_prev_len(
102	/**
103	* Returns the upper-case variant of the given unicode codepoint.
104	*
105	* Does not support unicode code points > \uffff
106	*/
107	int utf8_upper(int uc);
108
109	/**
110	* Returns the lower-case variant of the given unicode codepoint.
111	*
112	* NOTE: Use utf8_upper() in preference for case-insensitive matching.
113	*
114	* Does not support unicode code points > \uffff
115	*/
116	int utf8_lower(int uc);
117
118	#endif
119
120	#endif
121	/* Jim - A small embeddable Tcl interpreter
122	*
123	* Copyright 2005 Salvatore Sanfilippo <[email protected]>
124	* Copyright 2005 Clemens Hintze <[email protected]>
125	* Copyright 2005 patthoyts - Pat Thoyts <[email protected]>
126	* Copyright 2008 oharboe - �yvind Harboe - [email protected]
127	* Copyright 2008 Andrew Lunn <[email protected]>
128	* Copyright 2008 Duane Ellis <[email protected]>
129	* Copyright 2008 Uwe Klein <[email protected]>
130	*
131	* Redistribution and use in source and binary forms, with or without
132	* modification, are permitted provided that the following conditions
133	* are met:
134	*
135	* 1. Redistributions of source code must retain the above copyright
136	* notice, this list of conditions and the following disclaimer.
137	* 2. Redistributions in binary form must reproduce the above
138	* copyright notice, this list of conditions and the following
139	* disclaimer in the documentation and/or other materials
140	* provided with the distribution.
141	*
142	* THIS SOFTWARE IS PROVIDED BY THE JIM TCL PROJECT ``AS IS'' AND ANY
143	* EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO,
144	* THE IMPLIED WARRANTIES OF MERCHANTABILI

M autosetup/local.tcl

		--- a/autosetup/local.tcl
		+++ b/autosetup/local.tcl
		@@ -0,0 +1,2 @@
	1	+# For this project, d
	2	+set useropts(nopager)tosetup(optdefault) nopager 1

	--- a/autosetup/local.tcl
	+++ b/autosetup/local.tcl
	@@ -0,0 +1,2 @@

	--- a/autosetup/local.tcl
	+++ b/autosetup/local.tcl
	@@ -0,0 +1,2 @@
1	# For this project, d
2	set useropts(nopager)tosetup(optdefault) nopager 1

M autosetup/system.tcl

+80

		--- a/autosetup/system.tcl
		+++ b/autosetup/system.tcl
		@@ -0,0 +1,80 @@
	1	+# Copyright (c) 2010 WorkWare Systems http://www.workware.net.au/
	2	+# All rights reserved
	3	+
	4	+# @synopsis:
	5	+#
	6	+# This module supports common system interroga--host, --build, --prefix, and setting srcdir, builddir, and EXEXT.
	7	+#
	8	+# It naming convention, where searchisys/type.h defines HAVE_SYS_TYPES_H
	9	+#
	10	+# It defines the fol--prefix unless overridden by the module-optionsation and options
	11	+# such as '--host', '--build', '--prefix', and setting 'srcdir', 'builddir', and 'EXEEXT'.
	12	+#
	13	+# It also support the "feature" naming convention, where searching
	14	+# for a feature such as 'sys/type => ines the followins to '$defaultprefix overridden by the user:
	15	+#
	16	+## datadir
	17	+## sysconfdir
	18	+## sharedstatedir
	19	+## localstatedir
	20	+## infodir
	21	+## mandir
	22	+## includedir
	23	+#
	24	+# If '--prefix' is not /usr/localaults to '/usr/local' unless 'options-defaults { prefix ... }'prefix ... }' #
	25	+ program-transform-name:
	26	+}
	27	+
	28	+# @check-feature name { script }
	29	+#
	30	+# defines feature '$nReturns 1 if exists, or 0 if not
	31	+#ode} {
	32	+ msg-checking "Checking for $name..."
	33	+ set r [uplevel 1 $code
	34	+ define-feature $name $r
	35	+ if {$r} {
	36	+ msg-result "ok"
	37	+ } else {
	38	+ msg-result "not found"
	39	+ }
	40	+ return $r
	41	+}
	42	+
	43	+# @have-feature name ?default=0?
	44	+#
	45	+# Returns the value of feature '$name' if defined,the feature if de
	46	+# is translated thehave-feature {name {defathe 0}. Use @top_sr# Copyright (c) 2010 WorkWare Systems http://www.workware.net.au/
	47	+# All rights reserved
	48	+
	49	+# @synopsis:
	50	+#
	51	+# This module supports common system interroga--host, --build, --prefix, and setting srcdir, builddir, and EXEXT.
	52	+#
	53	+# It naming convention, where searchisys/type.h defines HAVE_SYS_TYPES_H
	54	+#
	55	+# It defines the fol--prefix unless overridden by the module-optionsation and options
	56	+# such as '--host', '--build', '--prefix', and setting 'srcdir', 'builddir', and 'EXEEXT'.
	57	+#
	58	+# It also support the "feature" naming convention, where searching
	59	+# for a feature such as 'sys/type => ines the followins to '$defaultprefix overridden by the user:
	60	+#
	61	+## datadir
	62	+## sysconfdir
	63	+## sharedstatedir
	64	+## localstatedir
	65	+## infodir
	66	+## mandir
	67	+## includedir
	68	+#
	69	+# If '--prefix' is not /usr/localaults to '/usr/local' unless 'options-defaults { prefix ... }'prefix ... }' #
	70	+ program-transform-name:
	71	+}
	72	+
	73	+# @check-feature set prefixdefine[get-env exec-prefix \${prefix}]]\${exec_prefix}/bin
	74	+ sbindir \${exec_prefix}/sbin
	75	+ libexecdir \${exec_prefix}/libexec
	76	+ libdir \${exec_prefix}/lib
	77	+ datadir \${prefix}/share
	78	+ sysconfdir \${prefix}/etc
	79	+ sharedstatedir \${prefix}\${prefix}/var
	80	+ infodir \${prefix}\${prefix}/share Copyright (c) ht (c) 2010 WorkWare Sys

	--- a/autosetup/system.tcl
	+++ b/autosetup/system.tcl
	@@ -0,0 +1,80 @@

	--- a/autosetup/system.tcl
	+++ b/autosetup/system.tcl
	@@ -0,0 +1,80 @@
1	# Copyright (c) 2010 WorkWare Systems http://www.workware.net.au/
2	# All rights reserved
3
4	# @synopsis:
5	#
6	# This module supports common system interroga--host, --build, --prefix, and setting srcdir, builddir, and EXEXT.
7	#
8	# It naming convention, where searchisys/type.h defines HAVE_SYS_TYPES_H
9	#
10	# It defines the fol--prefix unless overridden by the module-optionsation and options
11	# such as '--host', '--build', '--prefix', and setting 'srcdir', 'builddir', and 'EXEEXT'.
12	#
13	# It also support the "feature" naming convention, where searching
14	# for a feature such as 'sys/type => ines the followins to '$defaultprefix overridden by the user:
15	#
16	## datadir
17	## sysconfdir
18	## sharedstatedir
19	## localstatedir
20	## infodir
21	## mandir
22	## includedir
23	#
24	# If '--prefix' is not /usr/localaults to '/usr/local' unless 'options-defaults { prefix ... }'prefix ... }' #
25	program-transform-name:
26	}
27
28	# @check-feature name { script }
29	#
30	# defines feature '$nReturns 1 if exists, or 0 if not
31	#ode} {
32	msg-checking "Checking for $name..."
33	set r [uplevel 1 $code
34	define-feature $name $r
35	if {$r} {
36	msg-result "ok"
37	} else {
38	msg-result "not found"
39	}
40	return $r
41	}
42
43	# @have-feature name ?default=0?
44	#
45	# Returns the value of feature '$name' if defined,the feature if de
46	# is translated thehave-feature {name {defathe 0}. Use @top_sr# Copyright (c) 2010 WorkWare Systems http://www.workware.net.au/
47	# All rights reserved
48
49	# @synopsis:
50	#
51	# This module supports common system interroga--host, --build, --prefix, and setting srcdir, builddir, and EXEXT.
52	#
53	# It naming convention, where searchisys/type.h defines HAVE_SYS_TYPES_H
54	#
55	# It defines the fol--prefix unless overridden by the module-optionsation and options
56	# such as '--host', '--build', '--prefix', and setting 'srcdir', 'builddir', and 'EXEEXT'.
57	#
58	# It also support the "feature" naming convention, where searching
59	# for a feature such as 'sys/type => ines the followins to '$defaultprefix overridden by the user:
60	#
61	## datadir
62	## sysconfdir
63	## sharedstatedir
64	## localstatedir
65	## infodir
66	## mandir
67	## includedir
68	#
69	# If '--prefix' is not /usr/localaults to '/usr/local' unless 'options-defaults { prefix ... }'prefix ... }' #
70	program-transform-name:
71	}
72
73	# @check-feature set prefixdefine[get-env exec-prefix \${prefix}]]\${exec_prefix}/bin
74	sbindir \${exec_prefix}/sbin
75	libexecdir \${exec_prefix}/libexec
76	libdir \${exec_prefix}/lib
77	datadir \${prefix}/share
78	sysconfdir \${prefix}/etc
79	sharedstatedir \${prefix}\${prefix}/var
80	infodir \${prefix}\${prefix}/share Copyright (c) ht (c) 2010 WorkWare Sys

M autosetup/test-tclsh

+21

		--- a/autosetup/test-tclsh
		+++ b/autosetup/test-tclsh
		@@ -0,0 +1,21 @@
	1	+# A small Tcl script to verify that the chosen
	2	+# interpreter works. Sometimes we might e.g. pick up
	3	+# an interpreter for a different arch.
	4	+# Outputs the full path to the interpreter
	5	+
	6	+if {[catch {info version} version] == 0} {
	7	+ # This is Jim Tcl
	8	+ if {$version >= 0.70} {
	9	+ # Ensure that rege
	10	+ # Older versions of jimsh mayinfo patchlevel]]} {
	11	+ puts [info nameofexecutable]
	12	+ exit 0
	13	+ }
	14	+}
	15	+exit 1
	16	+puts [file join [p]
	17	+ exit 0
	18	+ }
	19	+} elseif {[catch {info tclversion} version] == 0} {
	20	+ if {$version >= 8.5 && ![string match 8.5a* [info patchlevel]]} {
	21	+ puts [info nameofexecut

	--- a/autosetup/test-tclsh
	+++ b/autosetup/test-tclsh
	@@ -0,0 +1,21 @@

	--- a/autosetup/test-tclsh
	+++ b/autosetup/test-tclsh
	@@ -0,0 +1,21 @@
1	# A small Tcl script to verify that the chosen
2	# interpreter works. Sometimes we might e.g. pick up
3	# an interpreter for a different arch.
4	# Outputs the full path to the interpreter
5
6	if {[catch {info version} version] == 0} {
7	# This is Jim Tcl
8	if {$version >= 0.70} {
9	# Ensure that rege
10	# Older versions of jimsh mayinfo patchlevel]]} {
11	puts [info nameofexecutable]
12	exit 0
13	}
14	}
15	exit 1
16	puts [file join [p]
17	exit 0
18	}
19	} elseif {[catch {info tclversion} version] == 0} {
20	if {$version >= 8.5 && ![string match 8.5a* [info patchlevel]]} {
21	puts [info nameofexecut

M configure

		--- a/configure
		+++ b/configure
		@@ -0,0 +1,3 @@
	1	+#!/bin/sh
	2	+dir=autosetup"
	3	+# exec "`$dir/find-tcl

	--- a/configure
	+++ b/configure
	@@ -0,0 +1,3 @@

	--- a/configure
	+++ b/configure
	@@ -0,0 +1,3 @@
1	#!/bin/sh
2	dir=autosetup"
3	# exec "`$dir/find-tcl

M src/add.c

+74 -21

		--- src/add.c
		+++ src/add.c
		@@ -92,66 +92,76 @@
92	92	**
93	93	** Omit any file whose name is pOmit.
94	94	*/
95	95	static int add_one_file(
96	96	const char zPath, / Tree-name of file to add. */
97		- int vid /* Add to this VFILE */
	97	+ int vid, /* Add to this VFILE */
	98	+ int caseSensitive /* True if filenames are case sensitive */
98	99	){
	100	+ const char *zCollate = caseSensitive ? "binary" : "nocase";
99	101	if( !file_is_simple_pathname(zPath) ){
100	102	fossil_fatal("filename contains illegal characters: %s", zPath);
101	103	}
102		-#if defined(_WIN32)
103	104	if( db_exists("SELECT 1 FROM vfile"
104		- " WHERE pathname=%Q COLLATE nocase", zPath) ){
	105	+ " WHERE pathname=%Q COLLATE %s", zPath, zCollate) ){
105	106	db_multi_exec("UPDATE vfile SET deleted=0"
106		- " WHERE pathname=%Q COLLATE nocase", zPath);
107		- }
108		-#else
109		- if( db_exists("SELECT 1 FROM vfile WHERE pathname=%Q", zPath) ){
110		- db_multi_exec("UPDATE vfile SET deleted=0 WHERE pathname=%Q", zPath);
111		- }
112		-#endif
113		- else{
	107	+ " WHERE pathname=%Q COLLATE %s", zPath, zCollate);
	108	+ }else{
114	109	char *zFullname = mprintf("%s%s", g.zLocalRoot, zPath);
115	110	db_multi_exec(
116	111	"INSERT INTO vfile(vid,deleted,rid,mrid,pathname,isexe)"
117	112	"VALUES(%d,0,0,0,%Q,%d)",
118	113	vid, zPath, file_isexe(zFullname));
119	114	fossil_free(zFullname);
120	115	}
121		- fossil_print("ADDED %s\n", zPath);
122		- return 1;
	116	+ if( db_changes() ){
	117	+ fossil_print("ADDED %s\n", zPath);
	118	+ return 1;
	119	+ }else{
	120	+ fossil_print("SKIP %s\n", zPath);
	121	+ return 0;
	122	+ }
123	123	}
124	124
125	125	/*
126	126	** Add all files in the sfile temp table.
127	127	**
128	128	** Automatically exclude the repository file.
129	129	*/
130		-static int add_files_in_sfile(int vid){
	130	+static int add_files_in_sfile(int vid, int caseSensitive){
131	131	const char zRepo; / Name of the repository database file */
132	132	int nAdd = 0; /* Number of files added */
133	133	int i; /* Loop counter */
134	134	const char zReserved; / Name of a reserved file */
135	135	Blob repoName; /* Treename of the repository */
136	136	Stmt loop; /* SQL to loop over all files to add */
	137	+ int (xCmp)(const char,const char*);
137	138
138	139	if( !file_tree_name(g.zRepositoryName, &repoName, 0) ){
139	140	blob_zero(&repoName);
140	141	zRepo = "";
141	142	}else{
142	143	zRepo = blob_str(&repoName);
	144	+ }
	145	+ if( caseSensitive ){
	146	+ xCmp = fossil_strcmp;
	147	+ }else{
	148	+ xCmp = fossil_stricmp;
	149	+ db_multi_exec(
	150	+ "CREATE INDEX IF NOT EXISTS vfile_nocase"
	151	+ " ON vfile(pathname COLLATE nocase)"
	152	+ );
143	153	}
144	154	db_prepare(&loop, "SELECT x FROM sfile ORDER BY x");
145	155	while( db_step(&loop)==SQLITE_ROW ){
146	156	const char *zToAdd = db_column_text(&loop, 0);
147	157	if( fossil_strcmp(zToAdd, zRepo)==0 ) continue;
148	158	for(i=0; (zReserved = fossil_reserved_name(i))!=0; i++){
149		- if( fossil_strcmp(zToAdd, zReserved)==0 ) break;
	159	+ if( xCmp(zToAdd, zReserved)==0 ) break;
150	160	}
151	161	if( zReserved ) continue;
152		- nAdd += add_one_file(zToAdd, vid);
	162	+ nAdd += add_one_file(zToAdd, vid, caseSensitive);
153	163	}
154	164	db_finalize(&loop);
155	165	blob_reset(&repoName);
156	166	return nAdd;
157	167	}
		@@ -180,14 +190,17 @@
180	190	int i; /* Loop counter */
181	191	int vid; /* Currently checked out version */
182	192	int nRoot; /* Full path characters in g.zLocalRoot */
183	193	const char zIgnoreFlag; / The --ignore option or ignore-glob setting */
184	194	Glob pIgnore; / Ignore everything matching this glob pattern */
	195	+ int caseSensitive; /* True if filenames are case sensitive */
185	196
186	197	zIgnoreFlag = find_option("ignore",0,1);
187	198	includeDotFiles = find_option("dotfiles",0,0)!=0;
	199	+ capture_case_sensitive_option();
188	200	db_must_be_within_tree();
	201	+ caseSensitive = filenames_are_case_sensitive();
189	202	if( zIgnoreFlag==0 ){
190	203	zIgnoreFlag = db_get("ignore-glob", 0);
191	204	}
192	205	vid = db_lget_int("checkout",0);
193	206	if( vid==0 ){
		@@ -229,11 +242,11 @@
229	242	}
230	243	blob_reset(&fullName);
231	244	}
232	245	glob_free(pIgnore);
233	246
234		- add_files_in_sfile(vid);
	247	+ add_files_in_sfile(vid, caseSensitive);
235	248	db_end_transaction(0);
236	249	}
237	250
238	251	/*
239	252	** COMMAND: rm
		@@ -289,10 +302,48 @@
289	302	"UPDATE vfile SET deleted=1 WHERE pathname IN sfile;"
290	303	"DELETE FROM vfile WHERE rid=0 AND deleted;"
291	304	);
292	305	db_end_transaction(0);
293	306	}
	307	+
	308	+/*
	309	+** Capture the command-line --case-sensitive option.
	310	+*/
	311	+static const char *zCaseSensitive = 0;
	312	+void capture_case_sensitive_option(void){
	313	+ if( zCaseSensitive==0 ){
	314	+ zCaseSensitive = find_option("case-sensitive",0,1);
	315	+ }
	316	+}
	317	+
	318	+/*
	319	+** This routine determines if files should be case-sensitive or not.
	320	+** In other words, this routine determines if two filenames that
	321	+** differ only in case should be considered the same name or not.
	322	+**
	323	+** The case-sensitive setting determines the default value. If
	324	+** the case-sensitive setting is undefined, then case sensitivity
	325	+** defaults on for Mac and Windows and off for all other unix.
	326	+**
	327	+** The --case-sensitive BOOLEAN command-line option overrides any
	328	+** setting.
	329	+*/
	330	+int filenames_are_case_sensitive(void){
	331	+ int caseSensitive;
	332	+
	333	+ if( zCaseSensitive ){
	334	+ caseSensitive = is_truth(zCaseSensitive);
	335	+ }else{
	336	+#if !defined(_WIN32) && !defined(__DARWIN__) && !defined(__APPLE__)
	337	+ caseSensitive = 1;
	338	+#else
	339	+ caseSensitive = 0;
	340	+#endif
	341	+ caseSensitive = db_get_boolean("case-sensitive",caseSensitive);
	342	+ }
	343	+ return caseSensitive;
	344	+}
294	345
295	346	/*
296	347	** COMMAND: addremove
297	348	**
298	349	** Usage: %fossil addremove ?--dotfiles? ?--ignore GLOBPATTERN? ?--test?
		@@ -319,27 +370,29 @@
319	370	**
320	371	** The --test option shows what would happen without actually doing anything.
321	372	**
322	373	** This command can be used to track third party software.
323	374	**
324		-**
325	375	** SUMMARY: fossil addremove
326		-** Options: ?--dotfiles? ?--ignore GLOBPATTERN? ?--test?
	376	+** Options: ?--dotfiles? ?--ignore GLOB? ?--test? ?--case-sensitive BOOL?
327	377	*/
328		-void import_cmd(void){
	378	+void addremove_cmd(void){
329	379	Blob path;
330	380	const char *zIgnoreFlag = find_option("ignore",0,1);
331	381	int allFlag = find_option("dotfiles",0,0)!=0;
332	382	int isTest = find_option("test",0,0)!=0;
	383	+ int caseSensitive;
333	384	int n;
334	385	Stmt q;
335	386	int vid;
336	387	int nAdd = 0;
337	388	int nDelete = 0;
338	389	Glob *pIgnore;
339	390
	391	+ capture_case_sensitive_option();
340	392	db_must_be_within_tree();
	393	+ caseSensitive = filenames_are_case_sensitive();
341	394	if( zIgnoreFlag==0 ){
342	395	zIgnoreFlag = db_get("ignore-glob", 0);
343	396	}
344	397	vid = db_lget_int("checkout",0);
345	398	if( vid==0 ){
		@@ -358,11 +411,11 @@
358	411	blob_init(&path, g.zLocalRoot, n-1);
359	412	/* now we read the complete file structure into a temp table */
360	413	pIgnore = glob_create(zIgnoreFlag);
361	414	vfile_scan(&path, blob_size(&path), allFlag, pIgnore);
362	415	glob_free(pIgnore);
363		- nAdd = add_files_in_sfile(vid);
	416	+ nAdd = add_files_in_sfile(vid, caseSensitive);
364	417
365	418	/* step 2: search for missing files */
366	419	db_prepare(&q,
367	420	"SELECT pathname, %Q \|\| pathname, deleted FROM vfile"
368	421	" WHERE NOT deleted"
369	422

	--- src/add.c
	+++ src/add.c
	@@ -92,66 +92,76 @@
92	**
93	** Omit any file whose name is pOmit.
94	*/
95	static int add_one_file(
96	const char zPath, / Tree-name of file to add. */
97	int vid /* Add to this VFILE */

98	){

99	if( !file_is_simple_pathname(zPath) ){
100	fossil_fatal("filename contains illegal characters: %s", zPath);
101	}
102	#if defined(_WIN32)
103	if( db_exists("SELECT 1 FROM vfile"
104	" WHERE pathname=%Q COLLATE nocase", zPath) ){
105	db_multi_exec("UPDATE vfile SET deleted=0"
106	" WHERE pathname=%Q COLLATE nocase", zPath);
107	}
108	#else
109	if( db_exists("SELECT 1 FROM vfile WHERE pathname=%Q", zPath) ){
110	db_multi_exec("UPDATE vfile SET deleted=0 WHERE pathname=%Q", zPath);
111	}
112	#endif
113	else{
114	char *zFullname = mprintf("%s%s", g.zLocalRoot, zPath);
115	db_multi_exec(
116	"INSERT INTO vfile(vid,deleted,rid,mrid,pathname,isexe)"
117	"VALUES(%d,0,0,0,%Q,%d)",
118	vid, zPath, file_isexe(zFullname));
119	fossil_free(zFullname);
120	}
121	fossil_print("ADDED %s\n", zPath);
122	return 1;





123	}
124
125	/*
126	** Add all files in the sfile temp table.
127	**
128	** Automatically exclude the repository file.
129	*/
130	static int add_files_in_sfile(int vid){
131	const char zRepo; / Name of the repository database file */
132	int nAdd = 0; /* Number of files added */
133	int i; /* Loop counter */
134	const char zReserved; / Name of a reserved file */
135	Blob repoName; /* Treename of the repository */
136	Stmt loop; /* SQL to loop over all files to add */

137
138	if( !file_tree_name(g.zRepositoryName, &repoName, 0) ){
139	blob_zero(&repoName);
140	zRepo = "";
141	}else{
142	zRepo = blob_str(&repoName);









143	}
144	db_prepare(&loop, "SELECT x FROM sfile ORDER BY x");
145	while( db_step(&loop)==SQLITE_ROW ){
146	const char *zToAdd = db_column_text(&loop, 0);
147	if( fossil_strcmp(zToAdd, zRepo)==0 ) continue;
148	for(i=0; (zReserved = fossil_reserved_name(i))!=0; i++){
149	if( fossil_strcmp(zToAdd, zReserved)==0 ) break;
150	}
151	if( zReserved ) continue;
152	nAdd += add_one_file(zToAdd, vid);
153	}
154	db_finalize(&loop);
155	blob_reset(&repoName);
156	return nAdd;
157	}
	@@ -180,14 +190,17 @@
180	int i; /* Loop counter */
181	int vid; /* Currently checked out version */
182	int nRoot; /* Full path characters in g.zLocalRoot */
183	const char zIgnoreFlag; / The --ignore option or ignore-glob setting */
184	Glob pIgnore; / Ignore everything matching this glob pattern */

185
186	zIgnoreFlag = find_option("ignore",0,1);
187	includeDotFiles = find_option("dotfiles",0,0)!=0;

188	db_must_be_within_tree();

189	if( zIgnoreFlag==0 ){
190	zIgnoreFlag = db_get("ignore-glob", 0);
191	}
192	vid = db_lget_int("checkout",0);
193	if( vid==0 ){
	@@ -229,11 +242,11 @@
229	}
230	blob_reset(&fullName);
231	}
232	glob_free(pIgnore);
233
234	add_files_in_sfile(vid);
235	db_end_transaction(0);
236	}
237
238	/*
239	** COMMAND: rm
	@@ -289,10 +302,48 @@
289	"UPDATE vfile SET deleted=1 WHERE pathname IN sfile;"
290	"DELETE FROM vfile WHERE rid=0 AND deleted;"
291	);
292	db_end_transaction(0);
293	}






































294
295	/*
296	** COMMAND: addremove
297	**
298	** Usage: %fossil addremove ?--dotfiles? ?--ignore GLOBPATTERN? ?--test?
	@@ -319,27 +370,29 @@
319	**
320	** The --test option shows what would happen without actually doing anything.
321	**
322	** This command can be used to track third party software.
323	**
324	**
325	** SUMMARY: fossil addremove
326	** Options: ?--dotfiles? ?--ignore GLOBPATTERN? ?--test?
327	*/
328	void import_cmd(void){
329	Blob path;
330	const char *zIgnoreFlag = find_option("ignore",0,1);
331	int allFlag = find_option("dotfiles",0,0)!=0;
332	int isTest = find_option("test",0,0)!=0;

333	int n;
334	Stmt q;
335	int vid;
336	int nAdd = 0;
337	int nDelete = 0;
338	Glob *pIgnore;
339

340	db_must_be_within_tree();

341	if( zIgnoreFlag==0 ){
342	zIgnoreFlag = db_get("ignore-glob", 0);
343	}
344	vid = db_lget_int("checkout",0);
345	if( vid==0 ){
	@@ -358,11 +411,11 @@
358	blob_init(&path, g.zLocalRoot, n-1);
359	/* now we read the complete file structure into a temp table */
360	pIgnore = glob_create(zIgnoreFlag);
361	vfile_scan(&path, blob_size(&path), allFlag, pIgnore);
362	glob_free(pIgnore);
363	nAdd = add_files_in_sfile(vid);
364
365	/* step 2: search for missing files */
366	db_prepare(&q,
367	"SELECT pathname, %Q \|\| pathname, deleted FROM vfile"
368	" WHERE NOT deleted"
369

	--- src/add.c
	+++ src/add.c
	@@ -92,66 +92,76 @@
92	**
93	** Omit any file whose name is pOmit.
94	*/
95	static int add_one_file(
96	const char zPath, / Tree-name of file to add. */
97	int vid, /* Add to this VFILE */
98	int caseSensitive /* True if filenames are case sensitive */
99	){
100	const char *zCollate = caseSensitive ? "binary" : "nocase";
101	if( !file_is_simple_pathname(zPath) ){
102	fossil_fatal("filename contains illegal characters: %s", zPath);
103	}

104	if( db_exists("SELECT 1 FROM vfile"
105	" WHERE pathname=%Q COLLATE %s", zPath, zCollate) ){
106	db_multi_exec("UPDATE vfile SET deleted=0"
107	" WHERE pathname=%Q COLLATE %s", zPath, zCollate);
108	}else{






109	char *zFullname = mprintf("%s%s", g.zLocalRoot, zPath);
110	db_multi_exec(
111	"INSERT INTO vfile(vid,deleted,rid,mrid,pathname,isexe)"
112	"VALUES(%d,0,0,0,%Q,%d)",
113	vid, zPath, file_isexe(zFullname));
114	fossil_free(zFullname);
115	}
116	if( db_changes() ){
117	fossil_print("ADDED %s\n", zPath);
118	return 1;
119	}else{
120	fossil_print("SKIP %s\n", zPath);
121	return 0;
122	}
123	}
124
125	/*
126	** Add all files in the sfile temp table.
127	**
128	** Automatically exclude the repository file.
129	*/
130	static int add_files_in_sfile(int vid, int caseSensitive){
131	const char zRepo; / Name of the repository database file */
132	int nAdd = 0; /* Number of files added */
133	int i; /* Loop counter */
134	const char zReserved; / Name of a reserved file */
135	Blob repoName; /* Treename of the repository */
136	Stmt loop; /* SQL to loop over all files to add */
137	int (xCmp)(const char,const char*);
138
139	if( !file_tree_name(g.zRepositoryName, &repoName, 0) ){
140	blob_zero(&repoName);
141	zRepo = "";
142	}else{
143	zRepo = blob_str(&repoName);
144	}
145	if( caseSensitive ){
146	xCmp = fossil_strcmp;
147	}else{
148	xCmp = fossil_stricmp;
149	db_multi_exec(
150	"CREATE INDEX IF NOT EXISTS vfile_nocase"
151	" ON vfile(pathname COLLATE nocase)"
152	);
153	}
154	db_prepare(&loop, "SELECT x FROM sfile ORDER BY x");
155	while( db_step(&loop)==SQLITE_ROW ){
156	const char *zToAdd = db_column_text(&loop, 0);
157	if( fossil_strcmp(zToAdd, zRepo)==0 ) continue;
158	for(i=0; (zReserved = fossil_reserved_name(i))!=0; i++){
159	if( xCmp(zToAdd, zReserved)==0 ) break;
160	}
161	if( zReserved ) continue;
162	nAdd += add_one_file(zToAdd, vid, caseSensitive);
163	}
164	db_finalize(&loop);
165	blob_reset(&repoName);
166	return nAdd;
167	}
	@@ -180,14 +190,17 @@
190	int i; /* Loop counter */
191	int vid; /* Currently checked out version */
192	int nRoot; /* Full path characters in g.zLocalRoot */
193	const char zIgnoreFlag; / The --ignore option or ignore-glob setting */
194	Glob pIgnore; / Ignore everything matching this glob pattern */
195	int caseSensitive; /* True if filenames are case sensitive */
196
197	zIgnoreFlag = find_option("ignore",0,1);
198	includeDotFiles = find_option("dotfiles",0,0)!=0;
199	capture_case_sensitive_option();
200	db_must_be_within_tree();
201	caseSensitive = filenames_are_case_sensitive();
202	if( zIgnoreFlag==0 ){
203	zIgnoreFlag = db_get("ignore-glob", 0);
204	}
205	vid = db_lget_int("checkout",0);
206	if( vid==0 ){
	@@ -229,11 +242,11 @@
242	}
243	blob_reset(&fullName);
244	}
245	glob_free(pIgnore);
246
247	add_files_in_sfile(vid, caseSensitive);
248	db_end_transaction(0);
249	}
250
251	/*
252	** COMMAND: rm
	@@ -289,10 +302,48 @@
302	"UPDATE vfile SET deleted=1 WHERE pathname IN sfile;"
303	"DELETE FROM vfile WHERE rid=0 AND deleted;"
304	);
305	db_end_transaction(0);
306	}
307
308	/*
309	** Capture the command-line --case-sensitive option.
310	*/
311	static const char *zCaseSensitive = 0;
312	void capture_case_sensitive_option(void){
313	if( zCaseSensitive==0 ){
314	zCaseSensitive = find_option("case-sensitive",0,1);
315	}
316	}
317
318	/*
319	** This routine determines if files should be case-sensitive or not.
320	** In other words, this routine determines if two filenames that
321	** differ only in case should be considered the same name or not.
322	**
323	** The case-sensitive setting determines the default value. If
324	** the case-sensitive setting is undefined, then case sensitivity
325	** defaults on for Mac and Windows and off for all other unix.
326	**
327	** The --case-sensitive BOOLEAN command-line option overrides any
328	** setting.
329	*/
330	int filenames_are_case_sensitive(void){
331	int caseSensitive;
332
333	if( zCaseSensitive ){
334	caseSensitive = is_truth(zCaseSensitive);
335	}else{
336	#if !defined(_WIN32) && !defined(__DARWIN__) && !defined(__APPLE__)
337	caseSensitive = 1;
338	#else
339	caseSensitive = 0;
340	#endif
341	caseSensitive = db_get_boolean("case-sensitive",caseSensitive);
342	}
343	return caseSensitive;
344	}
345
346	/*
347	** COMMAND: addremove
348	**
349	** Usage: %fossil addremove ?--dotfiles? ?--ignore GLOBPATTERN? ?--test?
	@@ -319,27 +370,29 @@
370	**
371	** The --test option shows what would happen without actually doing anything.
372	**
373	** This command can be used to track third party software.
374	**

375	** SUMMARY: fossil addremove
376	** Options: ?--dotfiles? ?--ignore GLOB? ?--test? ?--case-sensitive BOOL?
377	*/
378	void addremove_cmd(void){
379	Blob path;
380	const char *zIgnoreFlag = find_option("ignore",0,1);
381	int allFlag = find_option("dotfiles",0,0)!=0;
382	int isTest = find_option("test",0,0)!=0;
383	int caseSensitive;
384	int n;
385	Stmt q;
386	int vid;
387	int nAdd = 0;
388	int nDelete = 0;
389	Glob *pIgnore;
390
391	capture_case_sensitive_option();
392	db_must_be_within_tree();
393	caseSensitive = filenames_are_case_sensitive();
394	if( zIgnoreFlag==0 ){
395	zIgnoreFlag = db_get("ignore-glob", 0);
396	}
397	vid = db_lget_int("checkout",0);
398	if( vid==0 ){
	@@ -358,11 +411,11 @@
411	blob_init(&path, g.zLocalRoot, n-1);
412	/* now we read the complete file structure into a temp table */
413	pIgnore = glob_create(zIgnoreFlag);
414	vfile_scan(&path, blob_size(&path), allFlag, pIgnore);
415	glob_free(pIgnore);
416	nAdd = add_files_in_sfile(vid, caseSensitive);
417
418	/* step 2: search for missing files */
419	db_prepare(&q,
420	"SELECT pathname, %Q \|\| pathname, deleted FROM vfile"
421	" WHERE NOT deleted"
422

M src/blob.c

+1 -1

		--- src/blob.c
		+++ src/blob.c
		@@ -120,11 +120,11 @@
120	120	** This routine is called if a blob operation fails because we
121	121	** have run out of memory.
122	122	*/
123	123	static void blob_panic(void){
124	124	static const char zErrMsg[] = "out of memory\n";
125		- write(2, zErrMsg, sizeof(zErrMsg)-1);
	125	+ fputs(zErrMsg, stderr);
126	126	fossil_exit(1);
127	127	}
128	128
129	129	/*
130	130	** A reallocation function that assumes that aData came from malloc().
131	131

	--- src/blob.c
	+++ src/blob.c
	@@ -120,11 +120,11 @@
120	** This routine is called if a blob operation fails because we
121	** have run out of memory.
122	*/
123	static void blob_panic(void){
124	static const char zErrMsg[] = "out of memory\n";
125	write(2, zErrMsg, sizeof(zErrMsg)-1);
126	fossil_exit(1);
127	}
128
129	/*
130	** A reallocation function that assumes that aData came from malloc().
131

	--- src/blob.c
	+++ src/blob.c
	@@ -120,11 +120,11 @@
120	** This routine is called if a blob operation fails because we
121	** have run out of memory.
122	*/
123	static void blob_panic(void){
124	static const char zErrMsg[] = "out of memory\n";
125	fputs(zErrMsg, stderr);
126	fossil_exit(1);
127	}
128
129	/*
130	** A reallocation function that assumes that aData came from malloc().
131

M src/branch.c

+73 -36

		--- src/branch.c
		+++ src/branch.c
		@@ -174,12 +174,49 @@
174	174
175	175	/* Commit */
176	176	db_end_transaction(0);
177	177
178	178	/* Do an autosync push, if requested */
179		- autosync(AUTOSYNC_PUSH);
	179	+ if( !isPrivate ) autosync(AUTOSYNC_PUSH);
	180	+}
	181	+
	182	+/*
	183	+** Prepare a query that will list all branches.
	184	+*/
	185	+static void prepareBranchQuery(Stmt *pQuery, int showAll, int showClosed){
	186	+ if( showClosed ){
	187	+ db_prepare(pQuery,
	188	+ "SELECT value FROM tagxref"
	189	+ " WHERE tagid=%d AND value NOT NULL "
	190	+ "EXCEPT "
	191	+ "SELECT value FROM tagxref"
	192	+ " WHERE tagid=%d"
	193	+ " AND rid IN leaf"
	194	+ " AND NOT %z"
	195	+ " ORDER BY value COLLATE nocase /sort/",
	196	+ TAG_BRANCH, TAG_BRANCH, leaf_is_closed_sql("tagxref.rid")
	197	+ );
	198	+ }else if( showAll ){
	199	+ db_prepare(pQuery,
	200	+ "SELECT DISTINCT value FROM tagxref"
	201	+ " WHERE tagid=%d AND value NOT NULL"
	202	+ " AND rid IN leaf"
	203	+ " ORDER BY value COLLATE nocase /sort/",
	204	+ TAG_BRANCH
	205	+ );
	206	+ }else{
	207	+ db_prepare(pQuery,
	208	+ "SELECT DISTINCT value FROM tagxref"
	209	+ " WHERE tagid=%d AND value NOT NULL"
	210	+ " AND rid IN leaf"
	211	+ " AND NOT %z"
	212	+ " ORDER BY value COLLATE nocase /sort/",
	213	+ TAG_BRANCH, leaf_is_closed_sql("tagxref.rid")
	214	+ );
	215	+ }
180	216	}
	217	+
181	218
182	219	/*
183	220	** COMMAND: branch
184	221	**
185	222	** Usage: %fossil branch SUBCOMMAND ... ?-R\|--repository FILE?
		@@ -194,11 +231,12 @@
194	231	** --private option makes the branch private.
195	232	**
196	233	** %fossil branch list
197	234	** %fossil branch ls
198	235	**
199		-** List all branches
	236	+** List all branches. Use --all or --closed to list all branches
	237	+** or closed branches. The default is to show only open branches.
200	238	**
201	239	*/
202	240	void branch_cmd(void){
203	241	int n;
204	242	const char *zCmd = "list";
		@@ -212,24 +250,19 @@
212	250	branch_new();
213	251	}else if( (strncmp(zCmd,"list",n)==0)\|\|(strncmp(zCmd, "ls", n)==0) ){
214	252	Stmt q;
215	253	int vid;
216	254	char *zCurrent = 0;
	255	+ int showAll = find_option("all",0,0)!=0;
	256	+ int showClosed = find_option("closed",0,0)!=0;
217	257
218	258	if( g.localOpen ){
219	259	vid = db_lget_int("checkout", 0);
220	260	zCurrent = db_text(0, "SELECT value FROM tagxref"
221	261	" WHERE rid=%d AND tagid=%d", vid, TAG_BRANCH);
222	262	}
223		- db_prepare(&q,
224		- "SELECT DISTINCT value FROM tagxref"
225		- " WHERE tagid=%d AND value NOT NULL"
226		- " AND rid IN leaf"
227		- " AND NOT %z"
228		- " ORDER BY value /sort/",
229		- TAG_BRANCH, leaf_is_closed_sql("tagxref.rid")
230		- );
	263	+ prepareBranchQuery(&q, showAll, showClosed);
231	264	while( db_step(&q)==SQLITE_ROW ){
232	265	const char *zBr = db_column_text(&q, 0);
233	266	int isCur = zCurrent!=0 && fossil_strcmp(zCurrent,zBr)==0;
234	267	fossil_print("%s%s\n", (isCur ? "* " : " "), zBr);
235	268	}
		@@ -247,21 +280,38 @@
247	280	*/
248	281	void brlist_page(void){
249	282	Stmt q;
250	283	int cnt;
251	284	int showClosed = P("closed")!=0;
	285	+ int showAll = P("all")!=0;
	286	+ int colorTest = P("colortest")!=0;
252	287
253	288	login_check_credentials();
254	289	if( !g.okRead ){ login_needed(); return; }
	290	+ if( colorTest ){
	291	+ showClosed = 0;
	292	+ showAll = 1;
	293	+ }
255	294
256		- style_header(showClosed ? "Closed Branches" : "Open Branches");
	295	+ style_header(showClosed ? "Closed Branches" :
	296	+ showAll ? "All Branches" : "Open Branches");
257	297	style_submenu_element("Timeline", "Timeline", "brtimeline");
258	298	if( showClosed ){
	299	+ style_submenu_element("All", "All", "brlist?all");
	300	+ style_submenu_element("Open","Open","brlist");
	301	+ }else if( showAll ){
	302	+ style_submenu_element("Closed", "Closed", "brlist?closed");
259	303	style_submenu_element("Open","Open","brlist");
260	304	}else{
	305	+ style_submenu_element("All", "All", "brlist?all");
261	306	style_submenu_element("Closed","Closed","brlist?closed");
262	307	}
	308	+ if( !colorTest ){
	309	+ style_submenu_element("Color-Test", "Color-Test", "brlist?colortest");
	310	+ }else{
	311	+ style_submenu_element("All", "All", "brlist?all");
	312	+ }
263	313	login_anonymous_available();
264	314	style_sidebox_begin("Nomenclature:", "33%");
265	315	@ <ol>
266	316	@ <li> An <div class="sideboxDescribed"><a href="brlist">
267	317	@ open branch</a></div> is a branch that has one or
		@@ -275,46 +325,33 @@
275	325	@ Closed branches are fixed and do not change (unless they are first
276	326	@ reopened)</li>
277	327	@ </ol>
278	328	style_sidebox_end();
279	329
	330	+ prepareBranchQuery(&q, showAll, showClosed);
280	331	cnt = 0;
281		- if( showClosed ){
282		- db_prepare(&q,
283		- "SELECT value FROM tagxref"
284		- " WHERE tagid=%d AND value NOT NULL "
285		- "EXCEPT "
286		- "SELECT value FROM tagxref"
287		- " WHERE tagid=%d"
288		- " AND rid IN leaf"
289		- " AND NOT %z"
290		- " ORDER BY value /sort/",
291		- TAG_BRANCH, TAG_BRANCH, leaf_is_closed_sql("tagxref.rid")
292		- );
293		- }else{
294		- db_prepare(&q,
295		- "SELECT DISTINCT value FROM tagxref"
296		- " WHERE tagid=%d AND value NOT NULL"
297		- " AND rid IN leaf"
298		- " AND NOT %z"
299		- " ORDER BY value /sort/",
300		- TAG_BRANCH, leaf_is_closed_sql("tagxref.rid")
301		- );
302		- }
303	332	while( db_step(&q)==SQLITE_ROW ){
304	333	const char *zBr = db_column_text(&q, 0);
305	334	if( cnt==0 ){
306		- if( showClosed ){
	335	+ if( colorTest ){
	336	+ @ <h2>Default background colors for all branches:</h2>
	337	+ }else if( showAll ){
	338	+ @ <h2>All Branches:</h2>
	339	+ }else if( showClosed ){
307	340	@ <h2>Closed Branches:</h2>
308	341	}else{
309	342	@ <h2>Open Branches:</h2>
310	343	}
311	344	@ <ul>
312	345	cnt++;
313	346	}
314		- if( g.okHistory ){
315		- @ <li><a href="%s(g.zTop)/timeline?r=%T(zBr)">%h(zBr)</a></li>
	347	+ if( colorTest ){
	348	+ const char *zColor = hash_color(zBr);
	349	+ @ <li><span style="background-color: %s(zColor)">
	350	+ @ %h(zBr) → %s(zColor)</span></li>
	351	+ }else if( g.okHistory ){
	352	+ @ <li><a href="%s(g.zTop)/timeline?r=%T(zBr)")>%h(zBr)</a></li>
316	353	}else{
317	354	@ <li><b>%h(zBr)</b></li>
318	355	}
319	356	}
320	357	if( cnt ){
321	358

	--- src/branch.c
	+++ src/branch.c
	@@ -174,12 +174,49 @@
174
175	/* Commit */
176	db_end_transaction(0);
177
178	/* Do an autosync push, if requested */
179	autosync(AUTOSYNC_PUSH);




































180	}

181
182	/*
183	** COMMAND: branch
184	**
185	** Usage: %fossil branch SUBCOMMAND ... ?-R\|--repository FILE?
	@@ -194,11 +231,12 @@
194	** --private option makes the branch private.
195	**
196	** %fossil branch list
197	** %fossil branch ls
198	**
199	** List all branches

200	**
201	*/
202	void branch_cmd(void){
203	int n;
204	const char *zCmd = "list";
	@@ -212,24 +250,19 @@
212	branch_new();
213	}else if( (strncmp(zCmd,"list",n)==0)\|\|(strncmp(zCmd, "ls", n)==0) ){
214	Stmt q;
215	int vid;
216	char *zCurrent = 0;


217
218	if( g.localOpen ){
219	vid = db_lget_int("checkout", 0);
220	zCurrent = db_text(0, "SELECT value FROM tagxref"
221	" WHERE rid=%d AND tagid=%d", vid, TAG_BRANCH);
222	}
223	db_prepare(&q,
224	"SELECT DISTINCT value FROM tagxref"
225	" WHERE tagid=%d AND value NOT NULL"
226	" AND rid IN leaf"
227	" AND NOT %z"
228	" ORDER BY value /sort/",
229	TAG_BRANCH, leaf_is_closed_sql("tagxref.rid")
230	);
231	while( db_step(&q)==SQLITE_ROW ){
232	const char *zBr = db_column_text(&q, 0);
233	int isCur = zCurrent!=0 && fossil_strcmp(zCurrent,zBr)==0;
234	fossil_print("%s%s\n", (isCur ? "* " : " "), zBr);
235	}
	@@ -247,21 +280,38 @@
247	*/
248	void brlist_page(void){
249	Stmt q;
250	int cnt;
251	int showClosed = P("closed")!=0;


252
253	login_check_credentials();
254	if( !g.okRead ){ login_needed(); return; }




255
256	style_header(showClosed ? "Closed Branches" : "Open Branches");

257	style_submenu_element("Timeline", "Timeline", "brtimeline");
258	if( showClosed ){




259	style_submenu_element("Open","Open","brlist");
260	}else{

261	style_submenu_element("Closed","Closed","brlist?closed");
262	}





263	login_anonymous_available();
264	style_sidebox_begin("Nomenclature:", "33%");
265	@ <ol>
266	@ <li> An <div class="sideboxDescribed"><a href="brlist">
267	@ open branch</a></div> is a branch that has one or
	@@ -275,46 +325,33 @@
275	@ Closed branches are fixed and do not change (unless they are first
276	@ reopened)</li>
277	@ </ol>
278	style_sidebox_end();
279

280	cnt = 0;
281	if( showClosed ){
282	db_prepare(&q,
283	"SELECT value FROM tagxref"
284	" WHERE tagid=%d AND value NOT NULL "
285	"EXCEPT "
286	"SELECT value FROM tagxref"
287	" WHERE tagid=%d"
288	" AND rid IN leaf"
289	" AND NOT %z"
290	" ORDER BY value /sort/",
291	TAG_BRANCH, TAG_BRANCH, leaf_is_closed_sql("tagxref.rid")
292	);
293	}else{
294	db_prepare(&q,
295	"SELECT DISTINCT value FROM tagxref"
296	" WHERE tagid=%d AND value NOT NULL"
297	" AND rid IN leaf"
298	" AND NOT %z"
299	" ORDER BY value /sort/",
300	TAG_BRANCH, leaf_is_closed_sql("tagxref.rid")
301	);
302	}
303	while( db_step(&q)==SQLITE_ROW ){
304	const char *zBr = db_column_text(&q, 0);
305	if( cnt==0 ){
306	if( showClosed ){




307	@ <h2>Closed Branches:</h2>
308	}else{
309	@ <h2>Open Branches:</h2>
310	}
311	@ <ul>
312	cnt++;
313	}
314	if( g.okHistory ){
315	@ <li><a href="%s(g.zTop)/timeline?r=%T(zBr)">%h(zBr)</a></li>




316	}else{
317	@ <li><b>%h(zBr)</b></li>
318	}
319	}
320	if( cnt ){
321

	--- src/branch.c
	+++ src/branch.c
	@@ -174,12 +174,49 @@
174
175	/* Commit */
176	db_end_transaction(0);
177
178	/* Do an autosync push, if requested */
179	if( !isPrivate ) autosync(AUTOSYNC_PUSH);
180	}
181
182	/*
183	** Prepare a query that will list all branches.
184	*/
185	static void prepareBranchQuery(Stmt *pQuery, int showAll, int showClosed){
186	if( showClosed ){
187	db_prepare(pQuery,
188	"SELECT value FROM tagxref"
189	" WHERE tagid=%d AND value NOT NULL "
190	"EXCEPT "
191	"SELECT value FROM tagxref"
192	" WHERE tagid=%d"
193	" AND rid IN leaf"
194	" AND NOT %z"
195	" ORDER BY value COLLATE nocase /sort/",
196	TAG_BRANCH, TAG_BRANCH, leaf_is_closed_sql("tagxref.rid")
197	);
198	}else if( showAll ){
199	db_prepare(pQuery,
200	"SELECT DISTINCT value FROM tagxref"
201	" WHERE tagid=%d AND value NOT NULL"
202	" AND rid IN leaf"
203	" ORDER BY value COLLATE nocase /sort/",
204	TAG_BRANCH
205	);
206	}else{
207	db_prepare(pQuery,
208	"SELECT DISTINCT value FROM tagxref"
209	" WHERE tagid=%d AND value NOT NULL"
210	" AND rid IN leaf"
211	" AND NOT %z"
212	" ORDER BY value COLLATE nocase /sort/",
213	TAG_BRANCH, leaf_is_closed_sql("tagxref.rid")
214	);
215	}
216	}
217
218
219	/*
220	** COMMAND: branch
221	**
222	** Usage: %fossil branch SUBCOMMAND ... ?-R\|--repository FILE?
	@@ -194,11 +231,12 @@
231	** --private option makes the branch private.
232	**
233	** %fossil branch list
234	** %fossil branch ls
235	**
236	** List all branches. Use --all or --closed to list all branches
237	** or closed branches. The default is to show only open branches.
238	**
239	*/
240	void branch_cmd(void){
241	int n;
242	const char *zCmd = "list";
	@@ -212,24 +250,19 @@
250	branch_new();
251	}else if( (strncmp(zCmd,"list",n)==0)\|\|(strncmp(zCmd, "ls", n)==0) ){
252	Stmt q;
253	int vid;
254	char *zCurrent = 0;
255	int showAll = find_option("all",0,0)!=0;
256	int showClosed = find_option("closed",0,0)!=0;
257
258	if( g.localOpen ){
259	vid = db_lget_int("checkout", 0);
260	zCurrent = db_text(0, "SELECT value FROM tagxref"
261	" WHERE rid=%d AND tagid=%d", vid, TAG_BRANCH);
262	}
263	prepareBranchQuery(&q, showAll, showClosed);







264	while( db_step(&q)==SQLITE_ROW ){
265	const char *zBr = db_column_text(&q, 0);
266	int isCur = zCurrent!=0 && fossil_strcmp(zCurrent,zBr)==0;
267	fossil_print("%s%s\n", (isCur ? "* " : " "), zBr);
268	}
	@@ -247,21 +280,38 @@
280	*/
281	void brlist_page(void){
282	Stmt q;
283	int cnt;
284	int showClosed = P("closed")!=0;
285	int showAll = P("all")!=0;
286	int colorTest = P("colortest")!=0;
287
288	login_check_credentials();
289	if( !g.okRead ){ login_needed(); return; }
290	if( colorTest ){
291	showClosed = 0;
292	showAll = 1;
293	}
294
295	style_header(showClosed ? "Closed Branches" :
296	showAll ? "All Branches" : "Open Branches");
297	style_submenu_element("Timeline", "Timeline", "brtimeline");
298	if( showClosed ){
299	style_submenu_element("All", "All", "brlist?all");
300	style_submenu_element("Open","Open","brlist");
301	}else if( showAll ){
302	style_submenu_element("Closed", "Closed", "brlist?closed");
303	style_submenu_element("Open","Open","brlist");
304	}else{
305	style_submenu_element("All", "All", "brlist?all");
306	style_submenu_element("Closed","Closed","brlist?closed");
307	}
308	if( !colorTest ){
309	style_submenu_element("Color-Test", "Color-Test", "brlist?colortest");
310	}else{
311	style_submenu_element("All", "All", "brlist?all");
312	}
313	login_anonymous_available();
314	style_sidebox_begin("Nomenclature:", "33%");
315	@ <ol>
316	@ <li> An <div class="sideboxDescribed"><a href="brlist">
317	@ open branch</a></div> is a branch that has one or
	@@ -275,46 +325,33 @@
325	@ Closed branches are fixed and do not change (unless they are first
326	@ reopened)</li>
327	@ </ol>
328	style_sidebox_end();
329
330	prepareBranchQuery(&q, showAll, showClosed);
331	cnt = 0;






















332	while( db_step(&q)==SQLITE_ROW ){
333	const char *zBr = db_column_text(&q, 0);
334	if( cnt==0 ){
335	if( colorTest ){
336	@ <h2>Default background colors for all branches:</h2>
337	}else if( showAll ){
338	@ <h2>All Branches:</h2>
339	}else if( showClosed ){
340	@ <h2>Closed Branches:</h2>
341	}else{
342	@ <h2>Open Branches:</h2>
343	}
344	@ <ul>
345	cnt++;
346	}
347	if( colorTest ){
348	const char *zColor = hash_color(zBr);
349	@ <li><span style="background-color: %s(zColor)">
350	@ %h(zBr) → %s(zColor)</span></li>
351	}else if( g.okHistory ){
352	@ <li><a href="%s(g.zTop)/timeline?r=%T(zBr)")>%h(zBr)</a></li>
353	}else{
354	@ <li><b>%h(zBr)</b></li>
355	}
356	}
357	if( cnt ){
358

M src/browse.c

+2 -2

		--- src/browse.c
		+++ src/browse.c
		@@ -146,11 +146,11 @@
146	146	/* Compute the title of the page */
147	147	blob_zero(&dirname);
148	148	if( zD ){
149	149	blob_append(&dirname, "in directory ", -1);
150	150	hyperlinked_path(zD, &dirname, zCI);
151		- zPrefix = mprintf("%h/", zD);
	151	+ zPrefix = mprintf("%s/", zD);
152	152	}else{
153	153	blob_append(&dirname, "in the top-level directory", -1);
154	154	zPrefix = "";
155	155	}
156	156	if( zCI ){
		@@ -267,11 +267,11 @@
267	267	if( zFN[0]=='/' ){
268	268	zFN++;
269	269	@ <li><a href="%s(zSubdirLink)%T(zFN)">%h(zFN)/</a></li>
270	270	}else if( zCI ){
271	271	const char *zUuid = db_column_text(&q, 1);
272		- @ <li><a href="%s(g.zTop)/artifact?name=%s(zUuid)">%h(zFN)</a></li>
	272	+ @ <li><a href="%s(g.zTop)/artifact/%s(zUuid)">%h(zFN)</a></li>
273	273	}else{
274	274	@ <li><a href="%s(g.zTop)/finfo?name=%T(zPrefix)%T(zFN)">%h(zFN)
275	275	@ </a></li>
276	276	}
277	277	}
278	278

	--- src/browse.c
	+++ src/browse.c
	@@ -146,11 +146,11 @@
146	/* Compute the title of the page */
147	blob_zero(&dirname);
148	if( zD ){
149	blob_append(&dirname, "in directory ", -1);
150	hyperlinked_path(zD, &dirname, zCI);
151	zPrefix = mprintf("%h/", zD);
152	}else{
153	blob_append(&dirname, "in the top-level directory", -1);
154	zPrefix = "";
155	}
156	if( zCI ){
	@@ -267,11 +267,11 @@
267	if( zFN[0]=='/' ){
268	zFN++;
269	@ <li><a href="%s(zSubdirLink)%T(zFN)">%h(zFN)/</a></li>
270	}else if( zCI ){
271	const char *zUuid = db_column_text(&q, 1);
272	@ <li><a href="%s(g.zTop)/artifact?name=%s(zUuid)">%h(zFN)</a></li>
273	}else{
274	@ <li><a href="%s(g.zTop)/finfo?name=%T(zPrefix)%T(zFN)">%h(zFN)
275	@ </a></li>
276	}
277	}
278

	--- src/browse.c
	+++ src/browse.c
	@@ -146,11 +146,11 @@
146	/* Compute the title of the page */
147	blob_zero(&dirname);
148	if( zD ){
149	blob_append(&dirname, "in directory ", -1);
150	hyperlinked_path(zD, &dirname, zCI);
151	zPrefix = mprintf("%s/", zD);
152	}else{
153	blob_append(&dirname, "in the top-level directory", -1);
154	zPrefix = "";
155	}
156	if( zCI ){
	@@ -267,11 +267,11 @@
267	if( zFN[0]=='/' ){
268	zFN++;
269	@ <li><a href="%s(zSubdirLink)%T(zFN)">%h(zFN)/</a></li>
270	}else if( zCI ){
271	const char *zUuid = db_column_text(&q, 1);
272	@ <li><a href="%s(g.zTop)/artifact/%s(zUuid)">%h(zFN)</a></li>
273	}else{
274	@ <li><a href="%s(g.zTop)/finfo?name=%T(zPrefix)%T(zFN)">%h(zFN)
275	@ </a></li>
276	}
277	}
278

M src/cgi.c

+24 -8

		--- src/cgi.c
		+++ src/cgi.c
		@@ -295,11 +295,11 @@
295	295	if( blob_size(&extraHeader)>0 ){
296	296	fprintf(g.httpOut, "%s", blob_buffer(&extraHeader));
297	297	}
298	298
299	299	/* Add headers to turn on useful security options in browsers. */
300		- fprintf(g.httpOut, "X-Frame-Options: DENY\r\n");
	300	+ fprintf(g.httpOut, "X-Frame-Options: SAMEORIGIN\r\n");
301	301	/* This stops fossil pages appearing in frames or iframes, preventing
302	302	** click-jacking attacks on supporting browsers.
303	303	**
304	304	** Other good headers would be
305	305	** Strict-Transport-Security: max-age=62208000
		@@ -353,10 +353,11 @@
353	353	if( size>0 ){
354	354	fwrite(blob_buffer(&cgiContent[i]), 1, size, g.httpOut);
355	355	}
356	356	}
357	357	}
	358	+ fflush(g.httpOut);
358	359	CGIDEBUG(("DONE\n"));
359	360	}
360	361
361	362	/*
362	363	** Do a redirect request to the URL given in the argument.
		@@ -887,17 +888,26 @@
887	888	}
888	889
889	890	/*
890	891	** Print all query parameters on standard output. Format the
891	892	** parameters as HTML. This is used for testing and debugging.
	893	+** Release builds omit the values of the cookies to avoid defeating
	894	+** the purpose of setting HttpOnly cookies.
892	895	*/
893	896	void cgi_print_all(void){
894	897	int i;
	898	+ int showAll = 0;
	899	+#ifdef FOSSIL_DEBUG
	900	+ /* Show the values of cookies in debug mode. */
	901	+ showAll = 1;
	902	+#endif
895	903	cgi_parameter("",""); /* Force the parameters into sorted order */
896	904	for(i=0; i<nUsedQP; i++){
897		- cgi_printf("%s = %s <br />\n",
898		- htmlize(aParamQP[i].zName, -1), htmlize(aParamQP[i].zValue, -1));
	905	+ if( showAll \|\| (fossil_stricmp("HTTP_COOKIE",aParamQP[i].zName)!=0 && fossil_strnicmp("fossil-",aParamQP[i].zName,7)!=0) ){
	906	+ cgi_printf("%s = %s <br />\n",
	907	+ htmlize(aParamQP[i].zName, -1), htmlize(aParamQP[i].zValue, -1));
	908	+ }
899	909	}
900	910	}
901	911
902	912	/*
903	913	** This routine works like "printf" except that it has the
		@@ -1143,11 +1153,13 @@
1143	1153	fossil_print("Listening for HTTP requests on TCP port %d\n", iPort);
1144	1154	fflush(stdout);
1145	1155	}
1146	1156	if( zBrowser ){
1147	1157	zBrowser = mprintf(zBrowser, iPort);
1148		- system(zBrowser);
	1158	+ if( system(zBrowser)<0 ){
	1159	+ fossil_warning("cannot start browser: %s\n", zBrowser);
	1160	+ }
1149	1161	}
1150	1162	while( 1 ){
1151	1163	if( nchildren>MAX_PARALLEL ){
1152	1164	/* Slow down if connections are arriving too fast */
1153	1165	sleep( nchildren-MAX_PARALLEL );
		@@ -1164,20 +1176,24 @@
1164	1176	child = fork();
1165	1177	if( child!=0 ){
1166	1178	if( child>0 ) nchildren++;
1167	1179	close(connection);
1168	1180	}else{
	1181	+ int nErr = 0, fd;
1169	1182	close(0);
1170		- dup(connection);
	1183	+ fd = dup(connection);
	1184	+ if( fd!=0 ) nErr++;
1171	1185	close(1);
1172		- dup(connection);
	1186	+ fd = dup(connection);
	1187	+ if( fd!=1 ) nErr++;
1173	1188	if( !g.fHttpTrace && !g.fSqlTrace ){
1174	1189	close(2);
1175		- dup(connection);
	1190	+ fd = dup(connection);
	1191	+ if( fd!=2 ) nErr++;
1176	1192	}
1177	1193	close(connection);
1178		- return 0;
	1194	+ return nErr;
1179	1195	}
1180	1196	}
1181	1197	}
1182	1198	/* Bury dead children */
1183	1199	while( waitpid(0, 0, WNOHANG)>0 ){
1184	1200

	--- src/cgi.c
	+++ src/cgi.c
	@@ -295,11 +295,11 @@
295	if( blob_size(&extraHeader)>0 ){
296	fprintf(g.httpOut, "%s", blob_buffer(&extraHeader));
297	}
298
299	/* Add headers to turn on useful security options in browsers. */
300	fprintf(g.httpOut, "X-Frame-Options: DENY\r\n");
301	/* This stops fossil pages appearing in frames or iframes, preventing
302	** click-jacking attacks on supporting browsers.
303	**
304	** Other good headers would be
305	** Strict-Transport-Security: max-age=62208000
	@@ -353,10 +353,11 @@
353	if( size>0 ){
354	fwrite(blob_buffer(&cgiContent[i]), 1, size, g.httpOut);
355	}
356	}
357	}

358	CGIDEBUG(("DONE\n"));
359	}
360
361	/*
362	** Do a redirect request to the URL given in the argument.
	@@ -887,17 +888,26 @@
887	}
888
889	/*
890	** Print all query parameters on standard output. Format the
891	** parameters as HTML. This is used for testing and debugging.


892	*/
893	void cgi_print_all(void){
894	int i;





895	cgi_parameter("",""); /* Force the parameters into sorted order */
896	for(i=0; i<nUsedQP; i++){
897	cgi_printf("%s = %s <br />\n",
898	htmlize(aParamQP[i].zName, -1), htmlize(aParamQP[i].zValue, -1));


899	}
900	}
901
902	/*
903	** This routine works like "printf" except that it has the
	@@ -1143,11 +1153,13 @@
1143	fossil_print("Listening for HTTP requests on TCP port %d\n", iPort);
1144	fflush(stdout);
1145	}
1146	if( zBrowser ){
1147	zBrowser = mprintf(zBrowser, iPort);
1148	system(zBrowser);


1149	}
1150	while( 1 ){
1151	if( nchildren>MAX_PARALLEL ){
1152	/* Slow down if connections are arriving too fast */
1153	sleep( nchildren-MAX_PARALLEL );
	@@ -1164,20 +1176,24 @@
1164	child = fork();
1165	if( child!=0 ){
1166	if( child>0 ) nchildren++;
1167	close(connection);
1168	}else{

1169	close(0);
1170	dup(connection);

1171	close(1);
1172	dup(connection);

1173	if( !g.fHttpTrace && !g.fSqlTrace ){
1174	close(2);
1175	dup(connection);

1176	}
1177	close(connection);
1178	return 0;
1179	}
1180	}
1181	}
1182	/* Bury dead children */
1183	while( waitpid(0, 0, WNOHANG)>0 ){
1184

	--- src/cgi.c
	+++ src/cgi.c
	@@ -295,11 +295,11 @@
295	if( blob_size(&extraHeader)>0 ){
296	fprintf(g.httpOut, "%s", blob_buffer(&extraHeader));
297	}
298
299	/* Add headers to turn on useful security options in browsers. */
300	fprintf(g.httpOut, "X-Frame-Options: SAMEORIGIN\r\n");
301	/* This stops fossil pages appearing in frames or iframes, preventing
302	** click-jacking attacks on supporting browsers.
303	**
304	** Other good headers would be
305	** Strict-Transport-Security: max-age=62208000
	@@ -353,10 +353,11 @@
353	if( size>0 ){
354	fwrite(blob_buffer(&cgiContent[i]), 1, size, g.httpOut);
355	}
356	}
357	}
358	fflush(g.httpOut);
359	CGIDEBUG(("DONE\n"));
360	}
361
362	/*
363	** Do a redirect request to the URL given in the argument.
	@@ -887,17 +888,26 @@
888	}
889
890	/*
891	** Print all query parameters on standard output. Format the
892	** parameters as HTML. This is used for testing and debugging.
893	** Release builds omit the values of the cookies to avoid defeating
894	** the purpose of setting HttpOnly cookies.
895	*/
896	void cgi_print_all(void){
897	int i;
898	int showAll = 0;
899	#ifdef FOSSIL_DEBUG
900	/* Show the values of cookies in debug mode. */
901	showAll = 1;
902	#endif
903	cgi_parameter("",""); /* Force the parameters into sorted order */
904	for(i=0; i<nUsedQP; i++){
905	if( showAll \|\| (fossil_stricmp("HTTP_COOKIE",aParamQP[i].zName)!=0 && fossil_strnicmp("fossil-",aParamQP[i].zName,7)!=0) ){
906	cgi_printf("%s = %s <br />\n",
907	htmlize(aParamQP[i].zName, -1), htmlize(aParamQP[i].zValue, -1));
908	}
909	}
910	}
911
912	/*
913	** This routine works like "printf" except that it has the
	@@ -1143,11 +1153,13 @@
1153	fossil_print("Listening for HTTP requests on TCP port %d\n", iPort);
1154	fflush(stdout);
1155	}
1156	if( zBrowser ){
1157	zBrowser = mprintf(zBrowser, iPort);
1158	if( system(zBrowser)<0 ){
1159	fossil_warning("cannot start browser: %s\n", zBrowser);
1160	}
1161	}
1162	while( 1 ){
1163	if( nchildren>MAX_PARALLEL ){
1164	/* Slow down if connections are arriving too fast */
1165	sleep( nchildren-MAX_PARALLEL );
	@@ -1164,20 +1176,24 @@
1176	child = fork();
1177	if( child!=0 ){
1178	if( child>0 ) nchildren++;
1179	close(connection);
1180	}else{
1181	int nErr = 0, fd;
1182	close(0);
1183	fd = dup(connection);
1184	if( fd!=0 ) nErr++;
1185	close(1);
1186	fd = dup(connection);
1187	if( fd!=1 ) nErr++;
1188	if( !g.fHttpTrace && !g.fSqlTrace ){
1189	close(2);
1190	fd = dup(connection);
1191	if( fd!=2 ) nErr++;
1192	}
1193	close(connection);
1194	return nErr;
1195	}
1196	}
1197	}
1198	/* Bury dead children */
1199	while( waitpid(0, 0, WNOHANG)>0 ){
1200

M src/checkin.c

+92 -15

		--- src/checkin.c
		+++ src/checkin.c
		@@ -32,60 +32,72 @@
32	32	** are not true files results in a fatal error.
33	33	*/
34	34	static void status_report(
35	35	Blob report, / Append the status report here */
36	36	const char zPrefix, / Prefix on each line of the report */
37		- int missingIsFatal /* MISSING and NOT_A_FILE are fatal errors */
	37	+ int missingIsFatal, /* MISSING and NOT_A_FILE are fatal errors */
	38	+ int cwdRelative /* Report relative to the current working dir */
38	39	){
39	40	Stmt q;
40	41	int nPrefix = strlen(zPrefix);
41	42	int nErr = 0;
	43	+ Blob rewrittenPathname;
42	44	db_prepare(&q,
43	45	"SELECT pathname, deleted, chnged, rid, coalesce(origname!=pathname,0)"
44	46	" FROM vfile "
45	47	" WHERE file_is_selected(id)"
46	48	" AND (chnged OR deleted OR rid=0 OR pathname!=origname) ORDER BY 1"
47	49	);
	50	+ blob_zero(&rewrittenPathname);
48	51	while( db_step(&q)==SQLITE_ROW ){
49	52	const char *zPathname = db_column_text(&q,0);
	53	+ const char *zDisplayName = zPathname;
50	54	int isDeleted = db_column_int(&q, 1);
51	55	int isChnged = db_column_int(&q,2);
52	56	int isNew = db_column_int(&q,3)==0;
53	57	int isRenamed = db_column_int(&q,4);
54	58	char *zFullName = mprintf("%s%s", g.zLocalRoot, zPathname);
	59	+ if( cwdRelative ){
	60	+ file_relative_name(zFullName, &rewrittenPathname);
	61	+ zDisplayName = blob_str(&rewrittenPathname);
	62	+ if( zDisplayName[0]=='.' && zDisplayName[1]=='/' ){
	63	+ zDisplayName += 2; /* no unnecessary ./ prefix */
	64	+ }
	65	+ }
55	66	blob_append(report, zPrefix, nPrefix);
56	67	if( isDeleted ){
57		- blob_appendf(report, "DELETED %s\n", zPathname);
	68	+ blob_appendf(report, "DELETED %s\n", zDisplayName);
58	69	}else if( !file_isfile(zFullName) ){
59	70	if( file_access(zFullName, 0)==0 ){
60		- blob_appendf(report, "NOT_A_FILE %s\n", zPathname);
	71	+ blob_appendf(report, "NOT_A_FILE %s\n", zDisplayName);
61	72	if( missingIsFatal ){
62		- fossil_warning("not a file: %s", zPathname);
	73	+ fossil_warning("not a file: %s", zDisplayName);
63	74	nErr++;
64	75	}
65	76	}else{
66		- blob_appendf(report, "MISSING %s\n", zPathname);
	77	+ blob_appendf(report, "MISSING %s\n", zDisplayName);
67	78	if( missingIsFatal ){
68		- fossil_warning("missing file: %s", zPathname);
	79	+ fossil_warning("missing file: %s", zDisplayName);
69	80	nErr++;
70	81	}
71	82	}
72	83	}else if( isNew ){
73		- blob_appendf(report, "ADDED %s\n", zPathname);
	84	+ blob_appendf(report, "ADDED %s\n", zDisplayName);
74	85	}else if( isDeleted ){
75		- blob_appendf(report, "DELETED %s\n", zPathname);
	86	+ blob_appendf(report, "DELETED %s\n", zDisplayName);
76	87	}else if( isChnged==2 ){
77		- blob_appendf(report, "UPDATED_BY_MERGE %s\n", zPathname);
	88	+ blob_appendf(report, "UPDATED_BY_MERGE %s\n", zDisplayName);
78	89	}else if( isChnged==3 ){
79		- blob_appendf(report, "ADDED_BY_MERGE %s\n", zPathname);
	90	+ blob_appendf(report, "ADDED_BY_MERGE %s\n", zDisplayName);
80	91	}else if( isChnged==1 ){
81		- blob_appendf(report, "EDITED %s\n", zPathname);
	92	+ blob_appendf(report, "EDITED %s\n", zDisplayName);
82	93	}else if( isRenamed ){
83		- blob_appendf(report, "RENAMED %s\n", zPathname);
	94	+ blob_appendf(report, "RENAMED %s\n", zDisplayName);
84	95	}
85	96	free(zFullName);
86	97	}
	98	+ blob_reset(&rewrittenPathname);
87	99	db_finalize(&q);
88	100	db_prepare(&q, "SELECT uuid FROM vmerge JOIN blob ON merge=rid"
89	101	" WHERE id=0");
90	102	while( db_step(&q)==SQLITE_ROW ){
91	103	blob_append(report, zPrefix, nPrefix);
		@@ -94,33 +106,59 @@
94	106	db_finalize(&q);
95	107	if( nErr ){
96	108	fossil_fatal("aborting due to prior errors");
97	109	}
98	110	}
	111	+
	112	+/*
	113	+** Use the "relative-paths" setting and the --abs-paths and
	114	+** --rel-paths command line options to determine whether the
	115	+** status report should be shown relative to the current
	116	+** working directory.
	117	+*/
	118	+static int determine_cwd_relative_option()
	119	+{
	120	+ int relativePaths = db_get_boolean("relative-paths", 1);
	121	+ int absPathOption = find_option("abs-paths", 0, 0)!=0;
	122	+ int relPathOption = find_option("rel-paths", 0, 0)!=0;
	123	+ if( absPathOption ){ relativePaths = 0; }
	124	+ if( relPathOption ){ relativePaths = 1; }
	125	+ return relativePaths;
	126	+}
99	127
100	128	/*
101	129	** COMMAND: changes
102	130	**
103	131	** Usage: %fossil changes
104	132	**
105	133	** Report on the edit status of all files in the current checkout.
106	134	** See also the "status" and "extra" commands.
107	135	**
	136	+** Pathnames are displayed according to the "relative-paths" setting,
	137	+** unless overridden by the --abs-paths or --rel-paths options.
	138	+**
108	139	** Options:
109	140	**
110	141	** --sha1sum Verify file status using SHA1 hashing rather
111	142	** than relying on file mtimes.
	143	+**
	144	+** --abs-paths Display absolute pathnames.
	145	+**
	146	+** --rel-paths Display pathnames relative to the current working
	147	+** directory.
112	148	*/
113	149	void changes_cmd(void){
114	150	Blob report;
115	151	int vid;
116	152	int useSha1sum = find_option("sha1sum", 0, 0)!=0;
	153	+ int cwdRelative = 0;
117	154	db_must_be_within_tree();
	155	+ cwdRelative = determine_cwd_relative_option();
118	156	blob_zero(&report);
119	157	vid = db_lget_int("checkout", 0);
120	158	vfile_check_signature(vid, 0, useSha1sum);
121		- status_report(&report, "", 0);
	159	+ status_report(&report, "", 0, cwdRelative);
122	160	blob_write_to_file(&report, "-");
123	161	}
124	162
125	163	/*
126	164	** COMMAND: status
		@@ -127,14 +165,22 @@
127	165	**
128	166	** Usage: %fossil status
129	167	**
130	168	** Report on the status of the current checkout.
131	169	**
	170	+** Pathnames are displayed according to the "relative-paths" setting,
	171	+** unless overridden by the --abs-paths or --rel-paths options.
	172	+**
132	173	** Options:
133	174	**
134	175	** --sha1sum Verify file status using SHA1 hashing rather
135	176	** than relying on file mtimes.
	177	+**
	178	+** --abs-paths Display absolute pathnames.
	179	+**
	180	+** --rel-paths Display pathnames relative to the current working
	181	+** directory.
136	182	*/
137	183	void status_cmd(void){
138	184	int vid;
139	185	db_must_be_within_tree();
140	186	/* 012345678901234 */
		@@ -212,22 +258,41 @@
212	258	** ignored but can be included by adding the --dotfiles option.
213	259	**
214	260	** The GLOBPATTERN is a comma-separated list of GLOB expressions for
215	261	** files that are ignored. The GLOBPATTERN specified by the "ignore-glob"
216	262	** is used if the --ignore option is omitted.
	263	+**
	264	+** Pathnames are displayed according to the "relative-paths" setting,
	265	+** unless overridden by the --abs-paths or --rel-paths options.
	266	+**
	267	+** Options:
	268	+**
	269	+** --dotfiles Include files with names beginning with "."
	270	+**
	271	+** --ignore GLOBPATTERN
	272	+** Override the "ignore-glob" setting.
	273	+**
	274	+** --abs-paths Display absolute pathnames.
	275	+**
	276	+** --rel-paths Display pathnames relative to the current working
	277	+** directory.
217	278	*/
218	279	void extra_cmd(void){
219	280	Blob path;
220	281	Blob repo;
221	282	Stmt q;
222	283	int n;
223	284	const char *zIgnoreFlag = find_option("ignore",0,1);
224	285	int allFlag = find_option("dotfiles",0,0)!=0;
	286	+ int cwdRelative = 0;
225	287	int outputManifest;
226	288	Glob *pIgnore;
	289	+ Blob rewrittenPathname;
	290	+ const char zPathname, zDisplayName;
227	291
228	292	db_must_be_within_tree();
	293	+ cwdRelative = determine_cwd_relative_option();
229	294	outputManifest = db_get_boolean("manifest",0);
230	295	db_multi_exec("CREATE TEMP TABLE sfile(x TEXT PRIMARY KEY)");
231	296	n = strlen(g.zLocalRoot);
232	297	blob_init(&path, g.zLocalRoot, n-1);
233	298	if( zIgnoreFlag==0 ){
		@@ -243,13 +308,25 @@
243	308	fossil_all_reserved_names()
244	309	);
245	310	if( file_tree_name(g.zRepositoryName, &repo, 0) ){
246	311	db_multi_exec("DELETE FROM sfile WHERE x=%B", &repo);
247	312	}
	313	+ blob_zero(&rewrittenPathname);
248	314	while( db_step(&q)==SQLITE_ROW ){
249		- fossil_print("%s\n", db_column_text(&q, 0));
	315	+ zDisplayName = zPathname = db_column_text(&q, 0);
	316	+ if( cwdRelative ) {
	317	+ char *zFullName = mprintf("%s%s", g.zLocalRoot, zPathname);
	318	+ file_relative_name(zFullName, &rewrittenPathname);
	319	+ free(zFullName);
	320	+ zDisplayName = blob_str(&rewrittenPathname);
	321	+ if( zDisplayName[0]=='.' && zDisplayName[1]=='/' ){
	322	+ zDisplayName += 2; /* no unnecessary ./ prefix */
	323	+ }
	324	+ }
	325	+ fossil_print("%s\n", zDisplayName);
250	326	}
	327	+ blob_reset(&rewrittenPathname);
251	328	db_finalize(&q);
252	329	}
253	330
254	331	/*
255	332	** COMMAND: clean
		@@ -367,11 +444,11 @@
367	444	"# PRIVATE BRANCH: This check-in will be private and will not sync to\n"
368	445	"# repositories.\n"
369	446	"#\n", -1
370	447	);
371	448	}
372		- status_report(&text, "# ", 1);
	449	+ status_report(&text, "# ", 1, 0);
373	450	zEditor = db_get("editor", 0);
374	451	if( zEditor==0 ){
375	452	zEditor = getenv("VISUAL");
376	453	}
377	454	if( zEditor==0 ){
378	455

	--- src/checkin.c
	+++ src/checkin.c
	@@ -32,60 +32,72 @@
32	** are not true files results in a fatal error.
33	*/
34	static void status_report(
35	Blob report, / Append the status report here */
36	const char zPrefix, / Prefix on each line of the report */
37	int missingIsFatal /* MISSING and NOT_A_FILE are fatal errors */

38	){
39	Stmt q;
40	int nPrefix = strlen(zPrefix);
41	int nErr = 0;

42	db_prepare(&q,
43	"SELECT pathname, deleted, chnged, rid, coalesce(origname!=pathname,0)"
44	" FROM vfile "
45	" WHERE file_is_selected(id)"
46	" AND (chnged OR deleted OR rid=0 OR pathname!=origname) ORDER BY 1"
47	);

48	while( db_step(&q)==SQLITE_ROW ){
49	const char *zPathname = db_column_text(&q,0);

50	int isDeleted = db_column_int(&q, 1);
51	int isChnged = db_column_int(&q,2);
52	int isNew = db_column_int(&q,3)==0;
53	int isRenamed = db_column_int(&q,4);
54	char *zFullName = mprintf("%s%s", g.zLocalRoot, zPathname);







55	blob_append(report, zPrefix, nPrefix);
56	if( isDeleted ){
57	blob_appendf(report, "DELETED %s\n", zPathname);
58	}else if( !file_isfile(zFullName) ){
59	if( file_access(zFullName, 0)==0 ){
60	blob_appendf(report, "NOT_A_FILE %s\n", zPathname);
61	if( missingIsFatal ){
62	fossil_warning("not a file: %s", zPathname);
63	nErr++;
64	}
65	}else{
66	blob_appendf(report, "MISSING %s\n", zPathname);
67	if( missingIsFatal ){
68	fossil_warning("missing file: %s", zPathname);
69	nErr++;
70	}
71	}
72	}else if( isNew ){
73	blob_appendf(report, "ADDED %s\n", zPathname);
74	}else if( isDeleted ){
75	blob_appendf(report, "DELETED %s\n", zPathname);
76	}else if( isChnged==2 ){
77	blob_appendf(report, "UPDATED_BY_MERGE %s\n", zPathname);
78	}else if( isChnged==3 ){
79	blob_appendf(report, "ADDED_BY_MERGE %s\n", zPathname);
80	}else if( isChnged==1 ){
81	blob_appendf(report, "EDITED %s\n", zPathname);
82	}else if( isRenamed ){
83	blob_appendf(report, "RENAMED %s\n", zPathname);
84	}
85	free(zFullName);
86	}

87	db_finalize(&q);
88	db_prepare(&q, "SELECT uuid FROM vmerge JOIN blob ON merge=rid"
89	" WHERE id=0");
90	while( db_step(&q)==SQLITE_ROW ){
91	blob_append(report, zPrefix, nPrefix);
	@@ -94,33 +106,59 @@
94	db_finalize(&q);
95	if( nErr ){
96	fossil_fatal("aborting due to prior errors");
97	}
98	}
















99
100	/*
101	** COMMAND: changes
102	**
103	** Usage: %fossil changes
104	**
105	** Report on the edit status of all files in the current checkout.
106	** See also the "status" and "extra" commands.
107	**



108	** Options:
109	**
110	** --sha1sum Verify file status using SHA1 hashing rather
111	** than relying on file mtimes.





112	*/
113	void changes_cmd(void){
114	Blob report;
115	int vid;
116	int useSha1sum = find_option("sha1sum", 0, 0)!=0;

117	db_must_be_within_tree();

118	blob_zero(&report);
119	vid = db_lget_int("checkout", 0);
120	vfile_check_signature(vid, 0, useSha1sum);
121	status_report(&report, "", 0);
122	blob_write_to_file(&report, "-");
123	}
124
125	/*
126	** COMMAND: status
	@@ -127,14 +165,22 @@
127	**
128	** Usage: %fossil status
129	**
130	** Report on the status of the current checkout.
131	**



132	** Options:
133	**
134	** --sha1sum Verify file status using SHA1 hashing rather
135	** than relying on file mtimes.





136	*/
137	void status_cmd(void){
138	int vid;
139	db_must_be_within_tree();
140	/* 012345678901234 */
	@@ -212,22 +258,41 @@
212	** ignored but can be included by adding the --dotfiles option.
213	**
214	** The GLOBPATTERN is a comma-separated list of GLOB expressions for
215	** files that are ignored. The GLOBPATTERN specified by the "ignore-glob"
216	** is used if the --ignore option is omitted.















217	*/
218	void extra_cmd(void){
219	Blob path;
220	Blob repo;
221	Stmt q;
222	int n;
223	const char *zIgnoreFlag = find_option("ignore",0,1);
224	int allFlag = find_option("dotfiles",0,0)!=0;

225	int outputManifest;
226	Glob *pIgnore;


227
228	db_must_be_within_tree();

229	outputManifest = db_get_boolean("manifest",0);
230	db_multi_exec("CREATE TEMP TABLE sfile(x TEXT PRIMARY KEY)");
231	n = strlen(g.zLocalRoot);
232	blob_init(&path, g.zLocalRoot, n-1);
233	if( zIgnoreFlag==0 ){
	@@ -243,13 +308,25 @@
243	fossil_all_reserved_names()
244	);
245	if( file_tree_name(g.zRepositoryName, &repo, 0) ){
246	db_multi_exec("DELETE FROM sfile WHERE x=%B", &repo);
247	}

248	while( db_step(&q)==SQLITE_ROW ){
249	fossil_print("%s\n", db_column_text(&q, 0));










250	}

251	db_finalize(&q);
252	}
253
254	/*
255	** COMMAND: clean
	@@ -367,11 +444,11 @@
367	"# PRIVATE BRANCH: This check-in will be private and will not sync to\n"
368	"# repositories.\n"
369	"#\n", -1
370	);
371	}
372	status_report(&text, "# ", 1);
373	zEditor = db_get("editor", 0);
374	if( zEditor==0 ){
375	zEditor = getenv("VISUAL");
376	}
377	if( zEditor==0 ){
378

	--- src/checkin.c
	+++ src/checkin.c
	@@ -32,60 +32,72 @@
32	** are not true files results in a fatal error.
33	*/
34	static void status_report(
35	Blob report, / Append the status report here */
36	const char zPrefix, / Prefix on each line of the report */
37	int missingIsFatal, /* MISSING and NOT_A_FILE are fatal errors */
38	int cwdRelative /* Report relative to the current working dir */
39	){
40	Stmt q;
41	int nPrefix = strlen(zPrefix);
42	int nErr = 0;
43	Blob rewrittenPathname;
44	db_prepare(&q,
45	"SELECT pathname, deleted, chnged, rid, coalesce(origname!=pathname,0)"
46	" FROM vfile "
47	" WHERE file_is_selected(id)"
48	" AND (chnged OR deleted OR rid=0 OR pathname!=origname) ORDER BY 1"
49	);
50	blob_zero(&rewrittenPathname);
51	while( db_step(&q)==SQLITE_ROW ){
52	const char *zPathname = db_column_text(&q,0);
53	const char *zDisplayName = zPathname;
54	int isDeleted = db_column_int(&q, 1);
55	int isChnged = db_column_int(&q,2);
56	int isNew = db_column_int(&q,3)==0;
57	int isRenamed = db_column_int(&q,4);
58	char *zFullName = mprintf("%s%s", g.zLocalRoot, zPathname);
59	if( cwdRelative ){
60	file_relative_name(zFullName, &rewrittenPathname);
61	zDisplayName = blob_str(&rewrittenPathname);
62	if( zDisplayName[0]=='.' && zDisplayName[1]=='/' ){
63	zDisplayName += 2; /* no unnecessary ./ prefix */
64	}
65	}
66	blob_append(report, zPrefix, nPrefix);
67	if( isDeleted ){
68	blob_appendf(report, "DELETED %s\n", zDisplayName);
69	}else if( !file_isfile(zFullName) ){
70	if( file_access(zFullName, 0)==0 ){
71	blob_appendf(report, "NOT_A_FILE %s\n", zDisplayName);
72	if( missingIsFatal ){
73	fossil_warning("not a file: %s", zDisplayName);
74	nErr++;
75	}
76	}else{
77	blob_appendf(report, "MISSING %s\n", zDisplayName);
78	if( missingIsFatal ){
79	fossil_warning("missing file: %s", zDisplayName);
80	nErr++;
81	}
82	}
83	}else if( isNew ){
84	blob_appendf(report, "ADDED %s\n", zDisplayName);
85	}else if( isDeleted ){
86	blob_appendf(report, "DELETED %s\n", zDisplayName);
87	}else if( isChnged==2 ){
88	blob_appendf(report, "UPDATED_BY_MERGE %s\n", zDisplayName);
89	}else if( isChnged==3 ){
90	blob_appendf(report, "ADDED_BY_MERGE %s\n", zDisplayName);
91	}else if( isChnged==1 ){
92	blob_appendf(report, "EDITED %s\n", zDisplayName);
93	}else if( isRenamed ){
94	blob_appendf(report, "RENAMED %s\n", zDisplayName);
95	}
96	free(zFullName);
97	}
98	blob_reset(&rewrittenPathname);
99	db_finalize(&q);
100	db_prepare(&q, "SELECT uuid FROM vmerge JOIN blob ON merge=rid"
101	" WHERE id=0");
102	while( db_step(&q)==SQLITE_ROW ){
103	blob_append(report, zPrefix, nPrefix);
	@@ -94,33 +106,59 @@
106	db_finalize(&q);
107	if( nErr ){
108	fossil_fatal("aborting due to prior errors");
109	}
110	}
111
112	/*
113	** Use the "relative-paths" setting and the --abs-paths and
114	** --rel-paths command line options to determine whether the
115	** status report should be shown relative to the current
116	** working directory.
117	*/
118	static int determine_cwd_relative_option()
119	{
120	int relativePaths = db_get_boolean("relative-paths", 1);
121	int absPathOption = find_option("abs-paths", 0, 0)!=0;
122	int relPathOption = find_option("rel-paths", 0, 0)!=0;
123	if( absPathOption ){ relativePaths = 0; }
124	if( relPathOption ){ relativePaths = 1; }
125	return relativePaths;
126	}
127
128	/*
129	** COMMAND: changes
130	**
131	** Usage: %fossil changes
132	**
133	** Report on the edit status of all files in the current checkout.
134	** See also the "status" and "extra" commands.
135	**
136	** Pathnames are displayed according to the "relative-paths" setting,
137	** unless overridden by the --abs-paths or --rel-paths options.
138	**
139	** Options:
140	**
141	** --sha1sum Verify file status using SHA1 hashing rather
142	** than relying on file mtimes.
143	**
144	** --abs-paths Display absolute pathnames.
145	**
146	** --rel-paths Display pathnames relative to the current working
147	** directory.
148	*/
149	void changes_cmd(void){
150	Blob report;
151	int vid;
152	int useSha1sum = find_option("sha1sum", 0, 0)!=0;
153	int cwdRelative = 0;
154	db_must_be_within_tree();
155	cwdRelative = determine_cwd_relative_option();
156	blob_zero(&report);
157	vid = db_lget_int("checkout", 0);
158	vfile_check_signature(vid, 0, useSha1sum);
159	status_report(&report, "", 0, cwdRelative);
160	blob_write_to_file(&report, "-");
161	}
162
163	/*
164	** COMMAND: status
	@@ -127,14 +165,22 @@
165	**
166	** Usage: %fossil status
167	**
168	** Report on the status of the current checkout.
169	**
170	** Pathnames are displayed according to the "relative-paths" setting,
171	** unless overridden by the --abs-paths or --rel-paths options.
172	**
173	** Options:
174	**
175	** --sha1sum Verify file status using SHA1 hashing rather
176	** than relying on file mtimes.
177	**
178	** --abs-paths Display absolute pathnames.
179	**
180	** --rel-paths Display pathnames relative to the current working
181	** directory.
182	*/
183	void status_cmd(void){
184	int vid;
185	db_must_be_within_tree();
186	/* 012345678901234 */
	@@ -212,22 +258,41 @@
258	** ignored but can be included by adding the --dotfiles option.
259	**
260	** The GLOBPATTERN is a comma-separated list of GLOB expressions for
261	** files that are ignored. The GLOBPATTERN specified by the "ignore-glob"
262	** is used if the --ignore option is omitted.
263	**
264	** Pathnames are displayed according to the "relative-paths" setting,
265	** unless overridden by the --abs-paths or --rel-paths options.
266	**
267	** Options:
268	**
269	** --dotfiles Include files with names beginning with "."
270	**
271	** --ignore GLOBPATTERN
272	** Override the "ignore-glob" setting.
273	**
274	** --abs-paths Display absolute pathnames.
275	**
276	** --rel-paths Display pathnames relative to the current working
277	** directory.
278	*/
279	void extra_cmd(void){
280	Blob path;
281	Blob repo;
282	Stmt q;
283	int n;
284	const char *zIgnoreFlag = find_option("ignore",0,1);
285	int allFlag = find_option("dotfiles",0,0)!=0;
286	int cwdRelative = 0;
287	int outputManifest;
288	Glob *pIgnore;
289	Blob rewrittenPathname;
290	const char zPathname, zDisplayName;
291
292	db_must_be_within_tree();
293	cwdRelative = determine_cwd_relative_option();
294	outputManifest = db_get_boolean("manifest",0);
295	db_multi_exec("CREATE TEMP TABLE sfile(x TEXT PRIMARY KEY)");
296	n = strlen(g.zLocalRoot);
297	blob_init(&path, g.zLocalRoot, n-1);
298	if( zIgnoreFlag==0 ){
	@@ -243,13 +308,25 @@
308	fossil_all_reserved_names()
309	);
310	if( file_tree_name(g.zRepositoryName, &repo, 0) ){
311	db_multi_exec("DELETE FROM sfile WHERE x=%B", &repo);
312	}
313	blob_zero(&rewrittenPathname);
314	while( db_step(&q)==SQLITE_ROW ){
315	zDisplayName = zPathname = db_column_text(&q, 0);
316	if( cwdRelative ) {
317	char *zFullName = mprintf("%s%s", g.zLocalRoot, zPathname);
318	file_relative_name(zFullName, &rewrittenPathname);
319	free(zFullName);
320	zDisplayName = blob_str(&rewrittenPathname);
321	if( zDisplayName[0]=='.' && zDisplayName[1]=='/' ){
322	zDisplayName += 2; /* no unnecessary ./ prefix */
323	}
324	}
325	fossil_print("%s\n", zDisplayName);
326	}
327	blob_reset(&rewrittenPathname);
328	db_finalize(&q);
329	}
330
331	/*
332	** COMMAND: clean
	@@ -367,11 +444,11 @@
444	"# PRIVATE BRANCH: This check-in will be private and will not sync to\n"
445	"# repositories.\n"
446	"#\n", -1
447	);
448	}
449	status_report(&text, "# ", 1, 0);
450	zEditor = db_get("editor", 0);
451	if( zEditor==0 ){
452	zEditor = getenv("VISUAL");
453	}
454	if( zEditor==0 ){
455

M src/checkout.c

		--- src/checkout.c
		+++ src/checkout.c
		@@ -232,10 +232,11 @@
232	232	if( !keepFlag ){
233	233	vfile_to_disk(vid, 0, 1, promptFlag);
234	234	}
235	235	checkout_set_all_exe(vid);
236	236	manifest_to_disk(vid);
	237	+ ensure_empty_dirs_created();
237	238	db_lset_int("checkout", vid);
238	239	undo_reset();
239	240	db_multi_exec("DELETE FROM vmerge");
240	241	if( !keepFlag && db_get_boolean("repo-cksum",1) ){
241	242	vfile_aggregate_checksum_manifest(vid, &cksum1, &cksum1b);
242	243

	--- src/checkout.c
	+++ src/checkout.c
	@@ -232,10 +232,11 @@
232	if( !keepFlag ){
233	vfile_to_disk(vid, 0, 1, promptFlag);
234	}
235	checkout_set_all_exe(vid);
236	manifest_to_disk(vid);

237	db_lset_int("checkout", vid);
238	undo_reset();
239	db_multi_exec("DELETE FROM vmerge");
240	if( !keepFlag && db_get_boolean("repo-cksum",1) ){
241	vfile_aggregate_checksum_manifest(vid, &cksum1, &cksum1b);
242

	--- src/checkout.c
	+++ src/checkout.c
	@@ -232,10 +232,11 @@
232	if( !keepFlag ){
233	vfile_to_disk(vid, 0, 1, promptFlag);
234	}
235	checkout_set_all_exe(vid);
236	manifest_to_disk(vid);
237	ensure_empty_dirs_created();
238	db_lset_int("checkout", vid);
239	undo_reset();
240	db_multi_exec("DELETE FROM vmerge");
241	if( !keepFlag && db_get_boolean("repo-cksum",1) ){
242	vfile_aggregate_checksum_manifest(vid, &cksum1, &cksum1b);
243

M src/clone.c

+17 -3

		--- src/clone.c
		+++ src/clone.c
		@@ -37,15 +37,17 @@
37	37	**
38	38	** Options:
39	39	**
40	40	** --admin-user\|-A USERNAME Make USERNAME the administrator
41	41	** --private Also clone private branches
	42	+** --ssl-identity=filename Use the SSL identity if requested by the server
42	43	**
43	44	*/
44	45	void clone_cmd(void){
45	46	char *zPassword;
46	47	const char zDefaultUser; / Optional name of the default user */
	48	+ const char zPw; / The user clone password */
47	49	int nErr = 0;
48	50	int bPrivate; /* Also clone private branches */
49	51
50	52	bPrivate = find_option("private",0,0)!=0;
51	53	url_proxy_options();
		@@ -76,13 +78,15 @@
76	78	"DELETE FROM blob WHERE rid IN private;"
77	79	"DELETE FROM delta wHERE rid IN private;"
78	80	"DELETE FROM private;"
79	81	);
80	82	shun_artifacts();
81		- g.zLogin = db_text(0, "SELECT login FROM user WHERE cap LIKE '%%s%%'");
82		- if( g.zLogin==0 ){
83		- db_create_default_users(1,zDefaultUser);
	83	+ db_create_default_users(1, zDefaultUser);
	84	+ if( zDefaultUser ){
	85	+ g.zLogin = zDefaultUser;
	86	+ }else{
	87	+ g.zLogin = db_text(0, "SELECT login FROM user WHERE cap LIKE '%%s%%'");
84	88	}
85	89	fossil_print("Repository cloned into %s\n", g.argv[3]);
86	90	}else{
87	91	db_create_repository(g.argv[3]);
88	92	db_open_repository(g.argv[3]);
		@@ -91,10 +95,18 @@
91	95	db_initial_setup(0, zDefaultUser, 0);
92	96	user_select();
93	97	db_set("content-schema", CONTENT_SCHEMA, 0);
94	98	db_set("aux-schema", AUX_SCHEMA, 0);
95	99	db_set("last-sync-url", g.argv[2], 0);
	100	+ if( g.zSSLIdentity!=0 ){
	101	+ /* If the --ssl-identity option was specified, store it as a setting */
	102	+ Blob fn;
	103	+ blob_zero(&fn);
	104	+ file_canonical_name(g.zSSLIdentity, &fn);
	105	+ db_set("ssl-identity", blob_str(&fn), 0);
	106	+ blob_reset(&fn);
	107	+ }
96	108	db_multi_exec(
97	109	"REPLACE INTO config(name,value,mtime)"
98	110	" VALUES('server-code', lower(hex(randomblob(20))), now());"
99	111	);
100	112	url_enable_proxy(0);
		@@ -116,7 +128,9 @@
116	128	rebuild_db(0, 1, 0);
117	129	fossil_print("project-id: %s\n", db_get("project-code", 0));
118	130	fossil_print("server-id: %s\n", db_get("server-code", 0));
119	131	zPassword = db_text(0, "SELECT pw FROM user WHERE login=%Q", g.zLogin);
120	132	fossil_print("admin-user: %s (password is \"%s\")\n", g.zLogin, zPassword);
	133	+ zPw = g.urlPasswd;
	134	+ if( !g.dontKeepUrl && zPw) db_set("last-sync-pw", obscure(zPw), 0);
121	135	db_end_transaction(0);
122	136	}
123	137

	--- src/clone.c
	+++ src/clone.c
	@@ -37,15 +37,17 @@
37	**
38	** Options:
39	**
40	** --admin-user\|-A USERNAME Make USERNAME the administrator
41	** --private Also clone private branches

42	**
43	*/
44	void clone_cmd(void){
45	char *zPassword;
46	const char zDefaultUser; / Optional name of the default user */

47	int nErr = 0;
48	int bPrivate; /* Also clone private branches */
49
50	bPrivate = find_option("private",0,0)!=0;
51	url_proxy_options();
	@@ -76,13 +78,15 @@
76	"DELETE FROM blob WHERE rid IN private;"
77	"DELETE FROM delta wHERE rid IN private;"
78	"DELETE FROM private;"
79	);
80	shun_artifacts();
81	g.zLogin = db_text(0, "SELECT login FROM user WHERE cap LIKE '%%s%%'");
82	if( g.zLogin==0 ){
83	db_create_default_users(1,zDefaultUser);


84	}
85	fossil_print("Repository cloned into %s\n", g.argv[3]);
86	}else{
87	db_create_repository(g.argv[3]);
88	db_open_repository(g.argv[3]);
	@@ -91,10 +95,18 @@
91	db_initial_setup(0, zDefaultUser, 0);
92	user_select();
93	db_set("content-schema", CONTENT_SCHEMA, 0);
94	db_set("aux-schema", AUX_SCHEMA, 0);
95	db_set("last-sync-url", g.argv[2], 0);








96	db_multi_exec(
97	"REPLACE INTO config(name,value,mtime)"
98	" VALUES('server-code', lower(hex(randomblob(20))), now());"
99	);
100	url_enable_proxy(0);
	@@ -116,7 +128,9 @@
116	rebuild_db(0, 1, 0);
117	fossil_print("project-id: %s\n", db_get("project-code", 0));
118	fossil_print("server-id: %s\n", db_get("server-code", 0));
119	zPassword = db_text(0, "SELECT pw FROM user WHERE login=%Q", g.zLogin);
120	fossil_print("admin-user: %s (password is \"%s\")\n", g.zLogin, zPassword);


121	db_end_transaction(0);
122	}
123

	--- src/clone.c
	+++ src/clone.c
	@@ -37,15 +37,17 @@
37	**
38	** Options:
39	**
40	** --admin-user\|-A USERNAME Make USERNAME the administrator
41	** --private Also clone private branches
42	** --ssl-identity=filename Use the SSL identity if requested by the server
43	**
44	*/
45	void clone_cmd(void){
46	char *zPassword;
47	const char zDefaultUser; / Optional name of the default user */
48	const char zPw; / The user clone password */
49	int nErr = 0;
50	int bPrivate; /* Also clone private branches */
51
52	bPrivate = find_option("private",0,0)!=0;
53	url_proxy_options();
	@@ -76,13 +78,15 @@
78	"DELETE FROM blob WHERE rid IN private;"
79	"DELETE FROM delta wHERE rid IN private;"
80	"DELETE FROM private;"
81	);
82	shun_artifacts();
83	db_create_default_users(1, zDefaultUser);
84	if( zDefaultUser ){
85	g.zLogin = zDefaultUser;
86	}else{
87	g.zLogin = db_text(0, "SELECT login FROM user WHERE cap LIKE '%%s%%'");
88	}
89	fossil_print("Repository cloned into %s\n", g.argv[3]);
90	}else{
91	db_create_repository(g.argv[3]);
92	db_open_repository(g.argv[3]);
	@@ -91,10 +95,18 @@
95	db_initial_setup(0, zDefaultUser, 0);
96	user_select();
97	db_set("content-schema", CONTENT_SCHEMA, 0);
98	db_set("aux-schema", AUX_SCHEMA, 0);
99	db_set("last-sync-url", g.argv[2], 0);
100	if( g.zSSLIdentity!=0 ){
101	/* If the --ssl-identity option was specified, store it as a setting */
102	Blob fn;
103	blob_zero(&fn);
104	file_canonical_name(g.zSSLIdentity, &fn);
105	db_set("ssl-identity", blob_str(&fn), 0);
106	blob_reset(&fn);
107	}
108	db_multi_exec(
109	"REPLACE INTO config(name,value,mtime)"
110	" VALUES('server-code', lower(hex(randomblob(20))), now());"
111	);
112	url_enable_proxy(0);
	@@ -116,7 +128,9 @@
128	rebuild_db(0, 1, 0);
129	fossil_print("project-id: %s\n", db_get("project-code", 0));
130	fossil_print("server-id: %s\n", db_get("server-code", 0));
131	zPassword = db_text(0, "SELECT pw FROM user WHERE login=%Q", g.zLogin);
132	fossil_print("admin-user: %s (password is \"%s\")\n", g.zLogin, zPassword);
133	zPw = g.urlPasswd;
134	if( !g.dontKeepUrl && zPw) db_set("last-sync-pw", obscure(zPw), 0);
135	db_end_transaction(0);
136	}
137

M src/config.h

+5 -1

		--- src/config.h
		+++ src/config.h
		@@ -24,10 +24,14 @@
24	24	#define _LARGE_FILE 1
25	25	#ifndef _FILE_OFFSET_BITS
26	26	# define _FILE_OFFSET_BITS 64
27	27	#endif
28	28	#define _LARGEFILE_SOURCE 1
	29	+
	30	+#ifdef HAVE_AUTOCONFIG_H
	31	+#include "autoconfig.h"
	32	+#endif
29	33
30	34	#ifndef _RC_COMPILE_
31	35
32	36	/*
33	37	** System header files used by all modules
		@@ -85,11 +89,11 @@
85	89	#include "sqlite3.h"
86	90
87	91	/*
88	92	** On Solaris, getpass() will only return up to 8 characters. getpassphrase() returns up to 257.
89	93	*/
90		-#if defined(__sun__) \|\| defined(sun)
	94	+#if HAVE_GETPASSPHRASE
91	95	#define getpass getpassphrase
92	96	#endif
93	97
94	98	/*
95	99	** Typedef for a 64-bit integer
96	100

	--- src/config.h
	+++ src/config.h
	@@ -24,10 +24,14 @@
24	#define _LARGE_FILE 1
25	#ifndef _FILE_OFFSET_BITS
26	# define _FILE_OFFSET_BITS 64
27	#endif
28	#define _LARGEFILE_SOURCE 1




29
30	#ifndef _RC_COMPILE_
31
32	/*
33	** System header files used by all modules
	@@ -85,11 +89,11 @@
85	#include "sqlite3.h"
86
87	/*
88	** On Solaris, getpass() will only return up to 8 characters. getpassphrase() returns up to 257.
89	*/
90	#if defined(__sun__) \|\| defined(sun)
91	#define getpass getpassphrase
92	#endif
93
94	/*
95	** Typedef for a 64-bit integer
96

	--- src/config.h
	+++ src/config.h
	@@ -24,10 +24,14 @@
24	#define _LARGE_FILE 1
25	#ifndef _FILE_OFFSET_BITS
26	# define _FILE_OFFSET_BITS 64
27	#endif
28	#define _LARGEFILE_SOURCE 1
29
30	#ifdef HAVE_AUTOCONFIG_H
31	#include "autoconfig.h"
32	#endif
33
34	#ifndef _RC_COMPILE_
35
36	/*
37	** System header files used by all modules
	@@ -85,11 +89,11 @@
89	#include "sqlite3.h"
90
91	/*
92	** On Solaris, getpass() will only return up to 8 characters. getpassphrase() returns up to 257.
93	*/
94	#if HAVE_GETPASSPHRASE
95	#define getpass getpassphrase
96	#endif
97
98	/*
99	** Typedef for a 64-bit integer
100

M src/configure.c

		--- src/configure.c
		+++ src/configure.c
		@@ -81,10 +81,11 @@
81	81	{ "project-name", CONFIGSET_PROJ },
82	82	{ "project-description", CONFIGSET_PROJ },
83	83	{ "manifest", CONFIGSET_PROJ },
84	84	{ "ignore-glob", CONFIGSET_PROJ },
85	85	{ "crnl-glob", CONFIGSET_PROJ },
	86	+ { "empty-dirs", CONFIGSET_PROJ },
86	87	{ "index-page", CONFIGSET_SKIN },
87	88	{ "timeline-block-markup", CONFIGSET_SKIN },
88	89	{ "timeline-max-comment", CONFIGSET_SKIN },
89	90	{ "ticket-table", CONFIGSET_TKT },
90	91	{ "ticket-common", CONFIGSET_TKT },
91	92

	--- src/configure.c
	+++ src/configure.c
	@@ -81,10 +81,11 @@
81	{ "project-name", CONFIGSET_PROJ },
82	{ "project-description", CONFIGSET_PROJ },
83	{ "manifest", CONFIGSET_PROJ },
84	{ "ignore-glob", CONFIGSET_PROJ },
85	{ "crnl-glob", CONFIGSET_PROJ },

86	{ "index-page", CONFIGSET_SKIN },
87	{ "timeline-block-markup", CONFIGSET_SKIN },
88	{ "timeline-max-comment", CONFIGSET_SKIN },
89	{ "ticket-table", CONFIGSET_TKT },
90	{ "ticket-common", CONFIGSET_TKT },
91

	--- src/configure.c
	+++ src/configure.c
	@@ -81,10 +81,11 @@
81	{ "project-name", CONFIGSET_PROJ },
82	{ "project-description", CONFIGSET_PROJ },
83	{ "manifest", CONFIGSET_PROJ },
84	{ "ignore-glob", CONFIGSET_PROJ },
85	{ "crnl-glob", CONFIGSET_PROJ },
86	{ "empty-dirs", CONFIGSET_PROJ },
87	{ "index-page", CONFIGSET_SKIN },
88	{ "timeline-block-markup", CONFIGSET_SKIN },
89	{ "timeline-max-comment", CONFIGSET_SKIN },
90	{ "ticket-table", CONFIGSET_TKT },
91	{ "ticket-common", CONFIGSET_TKT },
92

M src/db.c

+201 -52

		--- src/db.c
		+++ src/db.c
		@@ -153,13 +153,17 @@
153	153	** Force a rollback and shutdown the database
154	154	*/
155	155	void db_force_rollback(void){
156	156	int i;
157	157	static int busy = 0;
	158	+ sqlite3_stmt *pStmt = 0;
158	159	if( busy \|\| g.db==0 ) return;
159	160	busy = 1;
160	161	undo_rollback();
	162	+ while( (pStmt = sqlite3_next_stmt(g.db,pStmt))!=0 ){
	163	+ sqlite3_reset(pStmt);
	164	+ }
161	165	while( pAllStmt ){
162	166	db_finalize(pAllStmt);
163	167	}
164	168	if( nBegin ){
165	169	sqlite3_exec(g.db, "ROLLBACK", 0, 0, 0);
		@@ -809,10 +813,11 @@
809	813	static const char *aDbName[] = { "/_FOSSIL_", "/.fos" };
810	814
811	815	if( g.localOpen) return 1;
812	816	file_getcwd(zPwd, sizeof(zPwd)-20);
813	817	n = strlen(zPwd);
	818	+ if( n==1 && zPwd[0]=='/' ) zPwd[0] = '.';
814	819	while( n>0 ){
815	820	if( file_access(zPwd, W_OK) ) break;
816	821	for(i=0; i<sizeof(aDbName)/sizeof(aDbName[0]); i++){
817	822	sqlite3_snprintf(sizeof(zPwd)-n, &zPwd[n], "%s", aDbName[i]);
818	823	if( isValidLocalDb(zPwd) ){
		@@ -1058,14 +1063,13 @@
1058	1063
1059	1064	/*
1060	1065	** Create the default user accounts in the USER table.
1061	1066	*/
1062	1067	void db_create_default_users(int setupUserOnly, const char *zDefaultUser){
1063		- const char *zUser;
1064		- zUser = db_get("default-user", 0);
	1068	+ const char *zUser = zDefaultUser;
1065	1069	if( zUser==0 ){
1066		- zUser = zDefaultUser;
	1070	+ zUser = db_get("default-user", 0);
1067	1071	}
1068	1072	if( zUser==0 ){
1069	1073	#if defined(_WIN32)
1070	1074	zUser = getenv("USERNAME");
1071	1075	#else
		@@ -1074,12 +1078,15 @@
1074	1078	}
1075	1079	if( zUser==0 ){
1076	1080	zUser = "root";
1077	1081	}
1078	1082	db_multi_exec(
1079		- "INSERT INTO user(login, pw, cap, info)"
1080		- "VALUES(%Q,lower(hex(randomblob(3))),'s','')", zUser
	1083	+ "INSERT OR IGNORE INTO user(login, info) VALUES(%Q,'')", zUser
	1084	+ );
	1085	+ db_multi_exec(
	1086	+ "UPDATE user SET cap='s', pw=lower(hex(randomblob(3)))"
	1087	+ " WHERE login=%Q", zUser
1081	1088	);
1082	1089	if( !setupUserOnly ){
1083	1090	db_multi_exec(
1084	1091	"INSERT INTO user(login,pw,cap,info)"
1085	1092	" VALUES('anonymous',hex(randomblob(8)),'hmncz','Anon');"
		@@ -1363,19 +1370,19 @@
1363	1370	*/
1364	1371	int is_truth(const char *zVal){
1365	1372	static const char *azOn[] = { "on", "yes", "true", "1" };
1366	1373	int i;
1367	1374	for(i=0; i<sizeof(azOn)/sizeof(azOn[0]); i++){
1368		- if( fossil_strcmp(zVal,azOn[i])==0 ) return 1;
	1375	+ if( fossil_stricmp(zVal,azOn[i])==0 ) return 1;
1369	1376	}
1370	1377	return 0;
1371	1378	}
1372	1379	int is_false(const char *zVal){
1373	1380	static const char *azOff[] = { "off", "no", "false", "0" };
1374	1381	int i;
1375	1382	for(i=0; i<sizeof(azOff)/sizeof(azOff[0]); i++){
1376		- if( fossil_strcmp(zVal,azOff[i])==0 ) return 1;
	1383	+ if( fossil_stricmp(zVal,azOff[i])==0 ) return 1;
1377	1384	}
1378	1385	return 0;
1379	1386	}
1380	1387
1381	1388	/*
		@@ -1392,24 +1399,114 @@
1392	1399	sqlite3 *dbTemp = g.db;
1393	1400	g.db = g.dbConfig;
1394	1401	g.dbConfig = dbTemp;
1395	1402	}
1396	1403	}
	1404	+
	1405	+/*
	1406	+** Logic for reading potentially versioned settings from
	1407	+** .fossil-settings/<name> , and emits warnings if necessary.
	1408	+** Returns the non-versioned value without modification if there is no
	1409	+** versioned value.
	1410	+*/
	1411	+static char db_get_do_versionable(const char zName, char *zNonVersionedSetting){
	1412	+ char *zVersionedSetting = 0;
	1413	+ int noWarn = 0;
	1414	+ struct _cacheEntry {
	1415	+ struct _cacheEntry *next;
	1416	+ const char zName, zValue;
	1417	+ } *cacheEntry = 0;
	1418	+ static struct _cacheEntry *cache = 0;
	1419	+
	1420	+ /* Look up name in cache */
	1421	+ cacheEntry = cache;
	1422	+ while( cacheEntry!=0 ){
	1423	+ if( fossil_strcmp(cacheEntry->zName, zName)==0 ){
	1424	+ zVersionedSetting = fossil_strdup(cacheEntry->zValue);
	1425	+ break;
	1426	+ }
	1427	+ cacheEntry = cacheEntry->next;
	1428	+ }
	1429	+ /* Attempt to read value from file in checkout if there wasn't a cache hit
	1430	+ ** and a checkout is open. */
	1431	+ if( cacheEntry==0 ){
	1432	+ Blob versionedPathname;
	1433	+ char *zVersionedPathname;
	1434	+ blob_zero(&versionedPathname);
	1435	+ blob_appendf(&versionedPathname, "%s/.fossil-settings/%s",
	1436	+ g.zLocalRoot, zName);
	1437	+ zVersionedPathname = blob_str(&versionedPathname);
	1438	+ if( file_size(zVersionedPathname)>=0 ){
	1439	+ /* File exists, and contains the value for this setting. Load from
	1440	+ ** the file. */
	1441	+ Blob setting;
	1442	+ blob_zero(&setting);
	1443	+ if( blob_read_from_file(&setting, zVersionedPathname) >= 0 ){
	1444	+ blob_trim(&setting); /* Avoid non-obvious problems with line endings
	1445	+ ** on boolean properties */
	1446	+ zVersionedSetting = strdup(blob_str(&setting));
	1447	+ }
	1448	+ blob_reset(&setting);
	1449	+ /* See if there's a no-warn flag */
	1450	+ blob_append(&versionedPathname, ".no-warn", -1);
	1451	+ if( file_size(blob_str(&versionedPathname))>=0 ){
	1452	+ noWarn = 1;
	1453	+ }
	1454	+ }
	1455	+ blob_reset(&versionedPathname);
	1456	+ /* Store result in cache, which can be the value or 0 if not found */
	1457	+ cacheEntry = (struct _cacheEntry*)fossil_malloc(sizeof(struct _cacheEntry));
	1458	+ cacheEntry->next = cache;
	1459	+ cacheEntry->zName = zName;
	1460	+ cacheEntry->zValue = fossil_strdup(zVersionedSetting);
	1461	+ cache = cacheEntry;
	1462	+ }
	1463	+ /* Display a warning? */
	1464	+ if( zVersionedSetting!=0 && zNonVersionedSetting!=0
	1465	+ && zNonVersionedSetting[0]!='\0' && !noWarn
	1466	+ ){
	1467	+ /* There's a versioned setting, and a non-versioned setting. Tell
	1468	+ ** the user about the conflict */
	1469	+ fossil_warning(
	1470	+ "setting %s has both versioned and non-versioned values: using "
	1471	+ "versioned value from file .fossil-settings/%s (to silence this "
	1472	+ "warning, either create an empty file named "
	1473	+ ".fossil-settings/%s.no-warn or delete the non-versioned setting "
	1474	+ " with \"fossil unset %s\")", zName, zName, zName, zName
	1475	+ );
	1476	+ }
	1477	+ /* Prefer the versioned setting */
	1478	+ return ( zVersionedSetting!=0 ) ? zVersionedSetting : zNonVersionedSetting;
	1479	+}
	1480	+
1397	1481
1398	1482	/*
1399	1483	** Get and set values from the CONFIG, GLOBAL_CONFIG and VVAR table in the
1400	1484	** repository and local databases.
1401	1485	*/
1402	1486	char db_get(const char zName, char *zDefault){
1403	1487	char *z = 0;
	1488	+ int i;
	1489	+ const struct stControlSettings *ctrlSetting = 0;
	1490	+ /* Is this a setting? */
	1491	+ for(i=0; ctrlSettings[i].name; i++){
	1492	+ if( strcmp(ctrlSettings[i].name, zName)==0 ){
	1493	+ ctrlSetting = &(ctrlSettings[i]);
	1494	+ break;
	1495	+ }
	1496	+ }
1404	1497	if( g.repositoryOpen ){
1405	1498	z = db_text(0, "SELECT value FROM config WHERE name=%Q", zName);
1406	1499	}
1407	1500	if( z==0 && g.configOpen ){
1408	1501	db_swap_connections();
1409	1502	z = db_text(0, "SELECT value FROM global_config WHERE name=%Q", zName);
1410	1503	db_swap_connections();
	1504	+ }
	1505	+ if( ctrlSetting!=0 && ctrlSetting->versionable && g.localOpen ){
	1506	+ /* This is a versionable setting, try and get the info from a checked out file */
	1507	+ z = db_get_do_versionable(zName, z);
1411	1508	}
1412	1509	if( z==0 ){
1413	1510	z = zDefault;
1414	1511	}
1415	1512	return z;
		@@ -1596,30 +1693,39 @@
1596	1693	}
1597	1694
1598	1695	/*
1599	1696	** Print the value of a setting named zName
1600	1697	*/
1601		-static void print_setting(const char *zName){
	1698	+static void print_setting(const struct stControlSettings *ctrlSetting, int localOpen){
1602	1699	Stmt q;
1603	1700	if( g.repositoryOpen ){
1604	1701	db_prepare(&q,
1605	1702	"SELECT '(local)', value FROM config WHERE name=%Q"
1606	1703	" UNION ALL "
1607	1704	"SELECT '(global)', value FROM global_config WHERE name=%Q",
1608		- zName, zName
	1705	+ ctrlSetting->name, ctrlSetting->name
1609	1706	);
1610	1707	}else{
1611	1708	db_prepare(&q,
1612	1709	"SELECT '(global)', value FROM global_config WHERE name=%Q",
1613		- zName
	1710	+ ctrlSetting->name
1614	1711	);
1615	1712	}
1616	1713	if( db_step(&q)==SQLITE_ROW ){
1617		- fossil_print("%-20s %-8s %s\n", zName, db_column_text(&q, 0),
	1714	+ fossil_print("%-20s %-8s %s\n", ctrlSetting->name, db_column_text(&q, 0),
1618	1715	db_column_text(&q, 1));
1619	1716	}else{
1620		- fossil_print("%-20s\n", zName);
	1717	+ fossil_print("%-20s\n", ctrlSetting->name);
	1718	+ }
	1719	+ if( ctrlSetting->versionable && localOpen ){
	1720	+ /* Check to see if this is overridden by a versionable settings file */
	1721	+ Blob versionedPathname;
	1722	+ blob_zero(&versionedPathname);
	1723	+ blob_appendf(&versionedPathname, "%s/.fossil-settings/%s", g.zLocalRoot, ctrlSetting->name);
	1724	+ if( file_size(blob_str(&versionedPathname))>=0 ){
	1725	+ fossil_print(" (overridden by contents of file .fossil-settings/%s)\n", ctrlSetting->name);
	1726	+ }
1621	1727	}
1622	1728	db_finalize(&q);
1623	1729	}
1624	1730
1625	1731
		@@ -1635,42 +1741,49 @@
1635	1741	#if INTERFACE
1636	1742	struct stControlSettings {
1637	1743	char const name; / Name of the setting */
1638	1744	char const var; / Internal variable name used by db_set() */
1639	1745	int width; /* Width of display. 0 for boolean values */
	1746	+ int versionable; /* Is this setting versionable? */
1640	1747	char const def; / Default value */
1641	1748	};
1642	1749	#endif /* INTERFACE */
1643	1750	struct stControlSettings const ctrlSettings[] = {
1644		- { "access-log", 0, 0, "off" },
1645		- { "auto-captcha", "autocaptcha", 0, "on" },
1646		- { "auto-shun", 0, 0, "on" },
1647		- { "autosync", 0, 0, "on" },
1648		- { "binary-glob", 0, 32, "" },
1649		- { "clearsign", 0, 0, "off" },
1650		- { "crnl-glob", 0, 16, "" },
1651		- { "default-perms", 0, 16, "u" },
1652		- { "diff-command", 0, 16, "" },
1653		- { "dont-push", 0, 0, "off" },
1654		- { "editor", 0, 16, "" },
1655		- { "gdiff-command", 0, 16, "gdiff" },
1656		- { "gmerge-command",0, 40, "" },
1657		- { "https-login", 0, 0, "off" },
1658		- { "ignore-glob", 0, 40, "" },
1659		- { "http-port", 0, 16, "8080" },
1660		- { "localauth", 0, 0, "off" },
1661		- { "main-branch", 0, 40, "trunk" },
1662		- { "manifest", 0, 0, "off" },
1663		- { "max-upload", 0, 25, "250000" },
1664		- { "mtime-changes", 0, 0, "on" },
1665		- { "pgp-command", 0, 32, "gpg --clearsign -o " },
1666		- { "proxy", 0, 32, "off" },
1667		- { "repo-cksum", 0, 0, "on" },
1668		- { "self-register", 0, 0, "off" },
1669		- { "ssh-command", 0, 32, "" },
1670		- { "web-browser", 0, 32, "" },
1671		- { 0,0,0,0 }
	1751	+ { "access-log", 0, 0, 0, "off" },
	1752	+ { "auto-captcha", "autocaptcha", 0, 0, "on" },
	1753	+ { "auto-shun", 0, 0, 0, "on" },
	1754	+ { "autosync", 0, 0, 0, "on" },
	1755	+ { "binary-glob", 0, 32, 1, "" },
	1756	+ { "clearsign", 0, 0, 0, "off" },
	1757	+ { "case-sensitive",0, 0, 0, "on" },
	1758	+ { "crnl-glob", 0, 16, 1, "" },
	1759	+ { "default-perms", 0, 16, 0, "u" },
	1760	+ { "diff-command", 0, 16, 0, "" },
	1761	+ { "dont-push", 0, 0, 0, "off" },
	1762	+ { "editor", 0, 16, 0, "" },
	1763	+ { "gdiff-command", 0, 16, 0, "gdiff" },
	1764	+ { "gmerge-command",0, 40, 0, "" },
	1765	+ { "https-login", 0, 0, 0, "off" },
	1766	+ { "ignore-glob", 0, 40, 1, "" },
	1767	+ { "empty-dirs", 0, 40, 1, "" },
	1768	+ { "http-port", 0, 16, 0, "8080" },
	1769	+ { "localauth", 0, 0, 0, "off" },
	1770	+ { "main-branch", 0, 40, 0, "trunk" },
	1771	+ { "manifest", 0, 0, 1, "off" },
	1772	+ { "max-upload", 0, 25, 0, "250000" },
	1773	+ { "mtime-changes", 0, 0, 0, "on" },
	1774	+ { "pgp-command", 0, 32, 0, "gpg --clearsign -o " },
	1775	+ { "proxy", 0, 32, 0, "off" },
	1776	+ { "relative-paths",0, 0, 0, "on" },
	1777	+ { "repo-cksum", 0, 0, 0, "on" },
	1778	+ { "self-register", 0, 0, 0, "off" },
	1779	+ { "ssl-ca-location",0, 40, 0, "" },
	1780	+ { "ssl-identity", 0, 40, 0, "" },
	1781	+ { "ssh-command", 0, 32, 0, "" },
	1782	+ { "web-browser", 0, 32, 0, "" },
	1783	+ { "white-foreground", 0, 0, 0, "off" },
	1784	+ { 0,0,0,0,0 }
1672	1785	};
1673	1786
1674	1787	/*
1675	1788	** COMMAND: settings
1676	1789	** COMMAND: unset
		@@ -1679,10 +1792,14 @@
1679	1792	** %fossil unset PROPERTY ?-global?
1680	1793	**
1681	1794	** The "settings" command with no arguments lists all properties and their
1682	1795	** values. With just a property name it shows the value of that property.
1683	1796	** With a value argument it changes the property for the current repository.
	1797	+**
	1798	+** Settings marked as versionable are overridden by the contents of the
	1799	+** file named .fossil-settings/PROPERTY in the checked out files, if that
	1800	+** file exists.
1684	1801	**
1685	1802	** The "unset" command clears a property setting.
1686	1803	**
1687	1804	**
1688	1805	** auto-captcha If enabled, the Login page provides a button to
		@@ -1696,20 +1813,25 @@
1696	1813	** or update and automatically push after commit or
1697	1814	** tag or branch creation. If the value is "pullonly"
1698	1815	** then only pull operations occur automatically.
1699	1816	** Default: on
1700	1817	**
1701		-** binary-glob The VALUE is a comma-separated list of GLOB patterns
1702		-** that should be treated as binary files for merging
1703		-** purposes. Example: *.xml
	1818	+** binary-glob The VALUE is a comma or newline-separated list of
	1819	+** (versionable) GLOB patterns that should be treated as binary files
	1820	+** for merging purposes. Example: *.xml
	1821	+**
	1822	+** case-sensitive If TRUE, the files whose names differ only in case
	1823	+** care considered distinct. If FALSE files whose names
	1824	+** differ only in case are the same file. Defaults to
	1825	+** TRUE for unix and FALSE for windows and mac.
1704	1826	**
1705	1827	** clearsign When enabled, fossil will attempt to sign all commits
1706	1828	** with gpg. When disabled (the default), commits will
1707	1829	** be unsigned. Default: off
1708	1830	**
1709		-** crnl-glob A comma-separated list of GLOB patterns for text files
1710		-** in which it is ok to have CR+NL line endings.
	1831	+** crnl-glob A comma or newline-separated list of GLOB patterns for
	1832	+** (versionable) text files in which it is ok to have CR+NL line endings.
1711	1833	** Set to "*" to disable CR+NL checking.
1712	1834	**
1713	1835	** default-perms Permissions given automatically to new users. For more
1714	1836	** information on permissions see Users page in Server
1715	1837	** Administration of the HTTP UI. Default: u.
		@@ -1717,10 +1839,15 @@
1717	1839	** diff-command External command to run when performing a diff.
1718	1840	** If undefined, the internal text diff will be used.
1719	1841	**
1720	1842	** dont-push Prevent this repository from pushing from client to
1721	1843	** server. Useful when setting up a private branch.
	1844	+**
	1845	+** empty-dirs A comma or newline-separated list of pathnames. On
	1846	+** (versionable) update and checkout commands, if no file or directory
	1847	+** exists with that name, an empty directory will be
	1848	+** created.
1722	1849	**
1723	1850	** editor Text editor command used for check-in comments.
1724	1851	**
1725	1852	** gdiff-command External command to run when performing a graphical
1726	1853	** diff. If undefined, text diff will be used.
		@@ -1735,23 +1862,23 @@
1735	1862	** and "ui" commands. Default: 8080
1736	1863	**
1737	1864	** https-login Send login creditials using HTTPS instead of HTTP
1738	1865	** even if the login page request came via HTTP.
1739	1866	**
1740		-** ignore-glob The VALUE is a comma-separated list of GLOB patterns
1741		-** specifying files that the "extra" command will ignore.
1742		-** Example: .o,.obj,*.exe
	1867	+** ignore-glob The VALUE is a comma or newline-separated list of GLOB
	1868	+** (versionable) patterns specifying files that the "extra" command will
	1869	+** ignore. Example: .o,.obj,*.exe
1743	1870	**
1744	1871	** localauth If enabled, require that HTTP connections from
1745	1872	** 127.0.0.1 be authenticated by password. If
1746	1873	** false, all HTTP requests from localhost have
1747	1874	** unrestricted access to the repository.
1748	1875	**
1749	1876	** main-branch The primary branch for the project. Default: trunk
1750	1877	**
1751	1878	** manifest If enabled, automatically create files "manifest" and
1752		-** "manifest.uuid" in every checkout. The SQLite and
	1879	+** (versionable) "manifest.uuid" in every checkout. The SQLite and
1753	1880	** Fossil repositories both require this. Default: off.
1754	1881	**
1755	1882	** max-upload A limit on the size of uplink HTTP requests. The
1756	1883	** default is 250000 bytes.
1757	1884	**
		@@ -1763,10 +1890,13 @@
1763	1890	**
1764	1891	** proxy URL of the HTTP proxy. If undefined or "off" then
1765	1892	** the "http_proxy" environment variable is consulted.
1766	1893	** If the http_proxy environment variable is undefined
1767	1894	** then a direct HTTP connection is used.
	1895	+**
	1896	+** relative-paths When showing changes and extras, report paths relative
	1897	+** to the current working directory. Default: "on"
1768	1898	**
1769	1899	** repo-cksum Compute checksums over all files in each checkout
1770	1900	** as a double-check of correctness. Defaults to "on".
1771	1901	** Disable on large repositories for a performance
1772	1902	** improvement.
		@@ -1773,10 +1903,28 @@
1773	1903	**
1774	1904	** self-register Allow users to register themselves through the HTTP UI.
1775	1905	** This is useful if you want to see other names than
1776	1906	** "Anonymous" in e.g. ticketing system. On the other hand
1777	1907	** users can not be deleted. Default: off.
	1908	+**
	1909	+** ssl-ca-location The full pathname to a file containing PEM encoded
	1910	+** CA root certificates, or a directory of certificates
	1911	+** with filenames formed from the certificate hashes as
	1912	+** required by OpenSSL.
	1913	+** If set, this will override the OS default list of
	1914	+** OpenSSL CAs. If unset, the default list will be used.
	1915	+** Some platforms may add additional certificates.
	1916	+** Check your platform behaviour is as required if the
	1917	+** exact contents of the CA root is critical for your
	1918	+** application.
	1919	+**
	1920	+** ssl-identity The full pathname to a file containing a certificate
	1921	+** and private key in PEM format. Create by concatenating
	1922	+** the certificate and private key files.
	1923	+** This identity will be presented to SSL servers to
	1924	+** authenticate this client, in addition to the normal
	1925	+** password authentication.
1778	1926	**
1779	1927	** ssh-command Command used to talk to a remote machine with
1780	1928	** the "ssh://" protocol.
1781	1929	**
1782	1930	** web-browser A shell command used to launch your preferred
		@@ -1797,12 +1945,13 @@
1797	1945	}
1798	1946	if( unsetFlag && g.argc!=3 ){
1799	1947	usage("PROPERTY ?-global?");
1800	1948	}
1801	1949	if( g.argc==2 ){
	1950	+ int openLocal = db_open_local();
1802	1951	for(i=0; ctrlSettings[i].name; i++){
1803		- print_setting(ctrlSettings[i].name);
	1952	+ print_setting(&ctrlSettings[i], openLocal);
1804	1953	}
1805	1954	}else if( g.argc==3 \|\| g.argc==4 ){
1806	1955	const char *zName = g.argv[2];
1807	1956	int isManifest;
1808	1957	int n = strlen(zName);
		@@ -1820,13 +1969,13 @@
1820	1969	db_unset(ctrlSettings[i].name, globalFlag);
1821	1970	}else if( g.argc==4 ){
1822	1971	db_set(ctrlSettings[i].name, g.argv[3], globalFlag);
1823	1972	}else{
1824	1973	isManifest = 0;
1825		- print_setting(ctrlSettings[i].name);
	1974	+ print_setting(&ctrlSettings[i], db_open_local());
1826	1975	}
1827		- if( isManifest ){
	1976	+ if( isManifest && g.localOpen ){
1828	1977	manifest_to_disk(db_lget_int("checkout", 0));
1829	1978	}
1830	1979	}else{
1831	1980	usage("?PROPERTY? ?VALUE?");
1832	1981	}
1833	1982

	--- src/db.c
	+++ src/db.c
	@@ -153,13 +153,17 @@
153	** Force a rollback and shutdown the database
154	*/
155	void db_force_rollback(void){
156	int i;
157	static int busy = 0;

158	if( busy \|\| g.db==0 ) return;
159	busy = 1;
160	undo_rollback();



161	while( pAllStmt ){
162	db_finalize(pAllStmt);
163	}
164	if( nBegin ){
165	sqlite3_exec(g.db, "ROLLBACK", 0, 0, 0);
	@@ -809,10 +813,11 @@
809	static const char *aDbName[] = { "/_FOSSIL_", "/.fos" };
810
811	if( g.localOpen) return 1;
812	file_getcwd(zPwd, sizeof(zPwd)-20);
813	n = strlen(zPwd);

814	while( n>0 ){
815	if( file_access(zPwd, W_OK) ) break;
816	for(i=0; i<sizeof(aDbName)/sizeof(aDbName[0]); i++){
817	sqlite3_snprintf(sizeof(zPwd)-n, &zPwd[n], "%s", aDbName[i]);
818	if( isValidLocalDb(zPwd) ){
	@@ -1058,14 +1063,13 @@
1058
1059	/*
1060	** Create the default user accounts in the USER table.
1061	*/
1062	void db_create_default_users(int setupUserOnly, const char *zDefaultUser){
1063	const char *zUser;
1064	zUser = db_get("default-user", 0);
1065	if( zUser==0 ){
1066	zUser = zDefaultUser;
1067	}
1068	if( zUser==0 ){
1069	#if defined(_WIN32)
1070	zUser = getenv("USERNAME");
1071	#else
	@@ -1074,12 +1078,15 @@
1074	}
1075	if( zUser==0 ){
1076	zUser = "root";
1077	}
1078	db_multi_exec(
1079	"INSERT INTO user(login, pw, cap, info)"
1080	"VALUES(%Q,lower(hex(randomblob(3))),'s','')", zUser



1081	);
1082	if( !setupUserOnly ){
1083	db_multi_exec(
1084	"INSERT INTO user(login,pw,cap,info)"
1085	" VALUES('anonymous',hex(randomblob(8)),'hmncz','Anon');"
	@@ -1363,19 +1370,19 @@
1363	*/
1364	int is_truth(const char *zVal){
1365	static const char *azOn[] = { "on", "yes", "true", "1" };
1366	int i;
1367	for(i=0; i<sizeof(azOn)/sizeof(azOn[0]); i++){
1368	if( fossil_strcmp(zVal,azOn[i])==0 ) return 1;
1369	}
1370	return 0;
1371	}
1372	int is_false(const char *zVal){
1373	static const char *azOff[] = { "off", "no", "false", "0" };
1374	int i;
1375	for(i=0; i<sizeof(azOff)/sizeof(azOff[0]); i++){
1376	if( fossil_strcmp(zVal,azOff[i])==0 ) return 1;
1377	}
1378	return 0;
1379	}
1380
1381	/*
	@@ -1392,24 +1399,114 @@
1392	sqlite3 *dbTemp = g.db;
1393	g.db = g.dbConfig;
1394	g.dbConfig = dbTemp;
1395	}
1396	}













































































1397
1398	/*
1399	** Get and set values from the CONFIG, GLOBAL_CONFIG and VVAR table in the
1400	** repository and local databases.
1401	*/
1402	char db_get(const char zName, char *zDefault){
1403	char *z = 0;









1404	if( g.repositoryOpen ){
1405	z = db_text(0, "SELECT value FROM config WHERE name=%Q", zName);
1406	}
1407	if( z==0 && g.configOpen ){
1408	db_swap_connections();
1409	z = db_text(0, "SELECT value FROM global_config WHERE name=%Q", zName);
1410	db_swap_connections();




1411	}
1412	if( z==0 ){
1413	z = zDefault;
1414	}
1415	return z;
	@@ -1596,30 +1693,39 @@
1596	}
1597
1598	/*
1599	** Print the value of a setting named zName
1600	*/
1601	static void print_setting(const char *zName){
1602	Stmt q;
1603	if( g.repositoryOpen ){
1604	db_prepare(&q,
1605	"SELECT '(local)', value FROM config WHERE name=%Q"
1606	" UNION ALL "
1607	"SELECT '(global)', value FROM global_config WHERE name=%Q",
1608	zName, zName
1609	);
1610	}else{
1611	db_prepare(&q,
1612	"SELECT '(global)', value FROM global_config WHERE name=%Q",
1613	zName
1614	);
1615	}
1616	if( db_step(&q)==SQLITE_ROW ){
1617	fossil_print("%-20s %-8s %s\n", zName, db_column_text(&q, 0),
1618	db_column_text(&q, 1));
1619	}else{
1620	fossil_print("%-20s\n", zName);









1621	}
1622	db_finalize(&q);
1623	}
1624
1625
	@@ -1635,42 +1741,49 @@
1635	#if INTERFACE
1636	struct stControlSettings {
1637	char const name; / Name of the setting */
1638	char const var; / Internal variable name used by db_set() */
1639	int width; /* Width of display. 0 for boolean values */

1640	char const def; / Default value */
1641	};
1642	#endif /* INTERFACE */
1643	struct stControlSettings const ctrlSettings[] = {
1644	{ "access-log", 0, 0, "off" },
1645	{ "auto-captcha", "autocaptcha", 0, "on" },
1646	{ "auto-shun", 0, 0, "on" },
1647	{ "autosync", 0, 0, "on" },
1648	{ "binary-glob", 0, 32, "" },
1649	{ "clearsign", 0, 0, "off" },
1650	{ "crnl-glob", 0, 16, "" },
1651	{ "default-perms", 0, 16, "u" },
1652	{ "diff-command", 0, 16, "" },
1653	{ "dont-push", 0, 0, "off" },
1654	{ "editor", 0, 16, "" },
1655	{ "gdiff-command", 0, 16, "gdiff" },
1656	{ "gmerge-command",0, 40, "" },
1657	{ "https-login", 0, 0, "off" },
1658	{ "ignore-glob", 0, 40, "" },
1659	{ "http-port", 0, 16, "8080" },
1660	{ "localauth", 0, 0, "off" },
1661	{ "main-branch", 0, 40, "trunk" },
1662	{ "manifest", 0, 0, "off" },
1663	{ "max-upload", 0, 25, "250000" },
1664	{ "mtime-changes", 0, 0, "on" },
1665	{ "pgp-command", 0, 32, "gpg --clearsign -o " },
1666	{ "proxy", 0, 32, "off" },
1667	{ "repo-cksum", 0, 0, "on" },
1668	{ "self-register", 0, 0, "off" },
1669	{ "ssh-command", 0, 32, "" },
1670	{ "web-browser", 0, 32, "" },
1671	{ 0,0,0,0 }






1672	};
1673
1674	/*
1675	** COMMAND: settings
1676	** COMMAND: unset
	@@ -1679,10 +1792,14 @@
1679	** %fossil unset PROPERTY ?-global?
1680	**
1681	** The "settings" command with no arguments lists all properties and their
1682	** values. With just a property name it shows the value of that property.
1683	** With a value argument it changes the property for the current repository.




1684	**
1685	** The "unset" command clears a property setting.
1686	**
1687	**
1688	** auto-captcha If enabled, the Login page provides a button to
	@@ -1696,20 +1813,25 @@
1696	** or update and automatically push after commit or
1697	** tag or branch creation. If the value is "pullonly"
1698	** then only pull operations occur automatically.
1699	** Default: on
1700	**
1701	** binary-glob The VALUE is a comma-separated list of GLOB patterns
1702	** that should be treated as binary files for merging
1703	** purposes. Example: *.xml





1704	**
1705	** clearsign When enabled, fossil will attempt to sign all commits
1706	** with gpg. When disabled (the default), commits will
1707	** be unsigned. Default: off
1708	**
1709	** crnl-glob A comma-separated list of GLOB patterns for text files
1710	** in which it is ok to have CR+NL line endings.
1711	** Set to "*" to disable CR+NL checking.
1712	**
1713	** default-perms Permissions given automatically to new users. For more
1714	** information on permissions see Users page in Server
1715	** Administration of the HTTP UI. Default: u.
	@@ -1717,10 +1839,15 @@
1717	** diff-command External command to run when performing a diff.
1718	** If undefined, the internal text diff will be used.
1719	**
1720	** dont-push Prevent this repository from pushing from client to
1721	** server. Useful when setting up a private branch.





1722	**
1723	** editor Text editor command used for check-in comments.
1724	**
1725	** gdiff-command External command to run when performing a graphical
1726	** diff. If undefined, text diff will be used.
	@@ -1735,23 +1862,23 @@
1735	** and "ui" commands. Default: 8080
1736	**
1737	** https-login Send login creditials using HTTPS instead of HTTP
1738	** even if the login page request came via HTTP.
1739	**
1740	** ignore-glob The VALUE is a comma-separated list of GLOB patterns
1741	** specifying files that the "extra" command will ignore.
1742	** Example: .o,.obj,*.exe
1743	**
1744	** localauth If enabled, require that HTTP connections from
1745	** 127.0.0.1 be authenticated by password. If
1746	** false, all HTTP requests from localhost have
1747	** unrestricted access to the repository.
1748	**
1749	** main-branch The primary branch for the project. Default: trunk
1750	**
1751	** manifest If enabled, automatically create files "manifest" and
1752	** "manifest.uuid" in every checkout. The SQLite and
1753	** Fossil repositories both require this. Default: off.
1754	**
1755	** max-upload A limit on the size of uplink HTTP requests. The
1756	** default is 250000 bytes.
1757	**
	@@ -1763,10 +1890,13 @@
1763	**
1764	** proxy URL of the HTTP proxy. If undefined or "off" then
1765	** the "http_proxy" environment variable is consulted.
1766	** If the http_proxy environment variable is undefined
1767	** then a direct HTTP connection is used.



1768	**
1769	** repo-cksum Compute checksums over all files in each checkout
1770	** as a double-check of correctness. Defaults to "on".
1771	** Disable on large repositories for a performance
1772	** improvement.
	@@ -1773,10 +1903,28 @@
1773	**
1774	** self-register Allow users to register themselves through the HTTP UI.
1775	** This is useful if you want to see other names than
1776	** "Anonymous" in e.g. ticketing system. On the other hand
1777	** users can not be deleted. Default: off.


















1778	**
1779	** ssh-command Command used to talk to a remote machine with
1780	** the "ssh://" protocol.
1781	**
1782	** web-browser A shell command used to launch your preferred
	@@ -1797,12 +1945,13 @@
1797	}
1798	if( unsetFlag && g.argc!=3 ){
1799	usage("PROPERTY ?-global?");
1800	}
1801	if( g.argc==2 ){

1802	for(i=0; ctrlSettings[i].name; i++){
1803	print_setting(ctrlSettings[i].name);
1804	}
1805	}else if( g.argc==3 \|\| g.argc==4 ){
1806	const char *zName = g.argv[2];
1807	int isManifest;
1808	int n = strlen(zName);
	@@ -1820,13 +1969,13 @@
1820	db_unset(ctrlSettings[i].name, globalFlag);
1821	}else if( g.argc==4 ){
1822	db_set(ctrlSettings[i].name, g.argv[3], globalFlag);
1823	}else{
1824	isManifest = 0;
1825	print_setting(ctrlSettings[i].name);
1826	}
1827	if( isManifest ){
1828	manifest_to_disk(db_lget_int("checkout", 0));
1829	}
1830	}else{
1831	usage("?PROPERTY? ?VALUE?");
1832	}
1833

	--- src/db.c
	+++ src/db.c
	@@ -153,13 +153,17 @@
153	** Force a rollback and shutdown the database
154	*/
155	void db_force_rollback(void){
156	int i;
157	static int busy = 0;
158	sqlite3_stmt *pStmt = 0;
159	if( busy \|\| g.db==0 ) return;
160	busy = 1;
161	undo_rollback();
162	while( (pStmt = sqlite3_next_stmt(g.db,pStmt))!=0 ){
163	sqlite3_reset(pStmt);
164	}
165	while( pAllStmt ){
166	db_finalize(pAllStmt);
167	}
168	if( nBegin ){
169	sqlite3_exec(g.db, "ROLLBACK", 0, 0, 0);
	@@ -809,10 +813,11 @@
813	static const char *aDbName[] = { "/_FOSSIL_", "/.fos" };
814
815	if( g.localOpen) return 1;
816	file_getcwd(zPwd, sizeof(zPwd)-20);
817	n = strlen(zPwd);
818	if( n==1 && zPwd[0]=='/' ) zPwd[0] = '.';
819	while( n>0 ){
820	if( file_access(zPwd, W_OK) ) break;
821	for(i=0; i<sizeof(aDbName)/sizeof(aDbName[0]); i++){
822	sqlite3_snprintf(sizeof(zPwd)-n, &zPwd[n], "%s", aDbName[i]);
823	if( isValidLocalDb(zPwd) ){
	@@ -1058,14 +1063,13 @@
1063
1064	/*
1065	** Create the default user accounts in the USER table.
1066	*/
1067	void db_create_default_users(int setupUserOnly, const char *zDefaultUser){
1068	const char *zUser = zDefaultUser;

1069	if( zUser==0 ){
1070	zUser = db_get("default-user", 0);
1071	}
1072	if( zUser==0 ){
1073	#if defined(_WIN32)
1074	zUser = getenv("USERNAME");
1075	#else
	@@ -1074,12 +1078,15 @@
1078	}
1079	if( zUser==0 ){
1080	zUser = "root";
1081	}
1082	db_multi_exec(
1083	"INSERT OR IGNORE INTO user(login, info) VALUES(%Q,'')", zUser
1084	);
1085	db_multi_exec(
1086	"UPDATE user SET cap='s', pw=lower(hex(randomblob(3)))"
1087	" WHERE login=%Q", zUser
1088	);
1089	if( !setupUserOnly ){
1090	db_multi_exec(
1091	"INSERT INTO user(login,pw,cap,info)"
1092	" VALUES('anonymous',hex(randomblob(8)),'hmncz','Anon');"
	@@ -1363,19 +1370,19 @@
1370	*/
1371	int is_truth(const char *zVal){
1372	static const char *azOn[] = { "on", "yes", "true", "1" };
1373	int i;
1374	for(i=0; i<sizeof(azOn)/sizeof(azOn[0]); i++){
1375	if( fossil_stricmp(zVal,azOn[i])==0 ) return 1;
1376	}
1377	return 0;
1378	}
1379	int is_false(const char *zVal){
1380	static const char *azOff[] = { "off", "no", "false", "0" };
1381	int i;
1382	for(i=0; i<sizeof(azOff)/sizeof(azOff[0]); i++){
1383	if( fossil_stricmp(zVal,azOff[i])==0 ) return 1;
1384	}
1385	return 0;
1386	}
1387
1388	/*
	@@ -1392,24 +1399,114 @@
1399	sqlite3 *dbTemp = g.db;
1400	g.db = g.dbConfig;
1401	g.dbConfig = dbTemp;
1402	}
1403	}
1404
1405	/*
1406	** Logic for reading potentially versioned settings from
1407	** .fossil-settings/<name> , and emits warnings if necessary.
1408	** Returns the non-versioned value without modification if there is no
1409	** versioned value.
1410	*/
1411	static char db_get_do_versionable(const char zName, char *zNonVersionedSetting){
1412	char *zVersionedSetting = 0;
1413	int noWarn = 0;
1414	struct _cacheEntry {
1415	struct _cacheEntry *next;
1416	const char zName, zValue;
1417	} *cacheEntry = 0;
1418	static struct _cacheEntry *cache = 0;
1419
1420	/* Look up name in cache */
1421	cacheEntry = cache;
1422	while( cacheEntry!=0 ){
1423	if( fossil_strcmp(cacheEntry->zName, zName)==0 ){
1424	zVersionedSetting = fossil_strdup(cacheEntry->zValue);
1425	break;
1426	}
1427	cacheEntry = cacheEntry->next;
1428	}
1429	/* Attempt to read value from file in checkout if there wasn't a cache hit
1430	** and a checkout is open. */
1431	if( cacheEntry==0 ){
1432	Blob versionedPathname;
1433	char *zVersionedPathname;
1434	blob_zero(&versionedPathname);
1435	blob_appendf(&versionedPathname, "%s/.fossil-settings/%s",
1436	g.zLocalRoot, zName);
1437	zVersionedPathname = blob_str(&versionedPathname);
1438	if( file_size(zVersionedPathname)>=0 ){
1439	/* File exists, and contains the value for this setting. Load from
1440	** the file. */
1441	Blob setting;
1442	blob_zero(&setting);
1443	if( blob_read_from_file(&setting, zVersionedPathname) >= 0 ){
1444	blob_trim(&setting); /* Avoid non-obvious problems with line endings
1445	** on boolean properties */
1446	zVersionedSetting = strdup(blob_str(&setting));
1447	}
1448	blob_reset(&setting);
1449	/* See if there's a no-warn flag */
1450	blob_append(&versionedPathname, ".no-warn", -1);
1451	if( file_size(blob_str(&versionedPathname))>=0 ){
1452	noWarn = 1;
1453	}
1454	}
1455	blob_reset(&versionedPathname);
1456	/* Store result in cache, which can be the value or 0 if not found */
1457	cacheEntry = (struct _cacheEntry*)fossil_malloc(sizeof(struct _cacheEntry));
1458	cacheEntry->next = cache;
1459	cacheEntry->zName = zName;
1460	cacheEntry->zValue = fossil_strdup(zVersionedSetting);
1461	cache = cacheEntry;
1462	}
1463	/* Display a warning? */
1464	if( zVersionedSetting!=0 && zNonVersionedSetting!=0
1465	&& zNonVersionedSetting[0]!='\0' && !noWarn
1466	){
1467	/* There's a versioned setting, and a non-versioned setting. Tell
1468	** the user about the conflict */
1469	fossil_warning(
1470	"setting %s has both versioned and non-versioned values: using "
1471	"versioned value from file .fossil-settings/%s (to silence this "
1472	"warning, either create an empty file named "
1473	".fossil-settings/%s.no-warn or delete the non-versioned setting "
1474	" with \"fossil unset %s\")", zName, zName, zName, zName
1475	);
1476	}
1477	/* Prefer the versioned setting */
1478	return ( zVersionedSetting!=0 ) ? zVersionedSetting : zNonVersionedSetting;
1479	}
1480
1481
1482	/*
1483	** Get and set values from the CONFIG, GLOBAL_CONFIG and VVAR table in the
1484	** repository and local databases.
1485	*/
1486	char db_get(const char zName, char *zDefault){
1487	char *z = 0;
1488	int i;
1489	const struct stControlSettings *ctrlSetting = 0;
1490	/* Is this a setting? */
1491	for(i=0; ctrlSettings[i].name; i++){
1492	if( strcmp(ctrlSettings[i].name, zName)==0 ){
1493	ctrlSetting = &(ctrlSettings[i]);
1494	break;
1495	}
1496	}
1497	if( g.repositoryOpen ){
1498	z = db_text(0, "SELECT value FROM config WHERE name=%Q", zName);
1499	}
1500	if( z==0 && g.configOpen ){
1501	db_swap_connections();
1502	z = db_text(0, "SELECT value FROM global_config WHERE name=%Q", zName);
1503	db_swap_connections();
1504	}
1505	if( ctrlSetting!=0 && ctrlSetting->versionable && g.localOpen ){
1506	/* This is a versionable setting, try and get the info from a checked out file */
1507	z = db_get_do_versionable(zName, z);
1508	}
1509	if( z==0 ){
1510	z = zDefault;
1511	}
1512	return z;
	@@ -1596,30 +1693,39 @@
1693	}
1694
1695	/*
1696	** Print the value of a setting named zName
1697	*/
1698	static void print_setting(const struct stControlSettings *ctrlSetting, int localOpen){
1699	Stmt q;
1700	if( g.repositoryOpen ){
1701	db_prepare(&q,
1702	"SELECT '(local)', value FROM config WHERE name=%Q"
1703	" UNION ALL "
1704	"SELECT '(global)', value FROM global_config WHERE name=%Q",
1705	ctrlSetting->name, ctrlSetting->name
1706	);
1707	}else{
1708	db_prepare(&q,
1709	"SELECT '(global)', value FROM global_config WHERE name=%Q",
1710	ctrlSetting->name
1711	);
1712	}
1713	if( db_step(&q)==SQLITE_ROW ){
1714	fossil_print("%-20s %-8s %s\n", ctrlSetting->name, db_column_text(&q, 0),
1715	db_column_text(&q, 1));
1716	}else{
1717	fossil_print("%-20s\n", ctrlSetting->name);
1718	}
1719	if( ctrlSetting->versionable && localOpen ){
1720	/* Check to see if this is overridden by a versionable settings file */
1721	Blob versionedPathname;
1722	blob_zero(&versionedPathname);
1723	blob_appendf(&versionedPathname, "%s/.fossil-settings/%s", g.zLocalRoot, ctrlSetting->name);
1724	if( file_size(blob_str(&versionedPathname))>=0 ){
1725	fossil_print(" (overridden by contents of file .fossil-settings/%s)\n", ctrlSetting->name);
1726	}
1727	}
1728	db_finalize(&q);
1729	}
1730
1731
	@@ -1635,42 +1741,49 @@
1741	#if INTERFACE
1742	struct stControlSettings {
1743	char const name; / Name of the setting */
1744	char const var; / Internal variable name used by db_set() */
1745	int width; /* Width of display. 0 for boolean values */
1746	int versionable; /* Is this setting versionable? */
1747	char const def; / Default value */
1748	};
1749	#endif /* INTERFACE */
1750	struct stControlSettings const ctrlSettings[] = {
1751	{ "access-log", 0, 0, 0, "off" },
1752	{ "auto-captcha", "autocaptcha", 0, 0, "on" },
1753	{ "auto-shun", 0, 0, 0, "on" },
1754	{ "autosync", 0, 0, 0, "on" },
1755	{ "binary-glob", 0, 32, 1, "" },
1756	{ "clearsign", 0, 0, 0, "off" },
1757	{ "case-sensitive",0, 0, 0, "on" },
1758	{ "crnl-glob", 0, 16, 1, "" },
1759	{ "default-perms", 0, 16, 0, "u" },
1760	{ "diff-command", 0, 16, 0, "" },
1761	{ "dont-push", 0, 0, 0, "off" },
1762	{ "editor", 0, 16, 0, "" },
1763	{ "gdiff-command", 0, 16, 0, "gdiff" },
1764	{ "gmerge-command",0, 40, 0, "" },
1765	{ "https-login", 0, 0, 0, "off" },
1766	{ "ignore-glob", 0, 40, 1, "" },
1767	{ "empty-dirs", 0, 40, 1, "" },
1768	{ "http-port", 0, 16, 0, "8080" },
1769	{ "localauth", 0, 0, 0, "off" },
1770	{ "main-branch", 0, 40, 0, "trunk" },
1771	{ "manifest", 0, 0, 1, "off" },
1772	{ "max-upload", 0, 25, 0, "250000" },
1773	{ "mtime-changes", 0, 0, 0, "on" },
1774	{ "pgp-command", 0, 32, 0, "gpg --clearsign -o " },
1775	{ "proxy", 0, 32, 0, "off" },
1776	{ "relative-paths",0, 0, 0, "on" },
1777	{ "repo-cksum", 0, 0, 0, "on" },
1778	{ "self-register", 0, 0, 0, "off" },
1779	{ "ssl-ca-location",0, 40, 0, "" },
1780	{ "ssl-identity", 0, 40, 0, "" },
1781	{ "ssh-command", 0, 32, 0, "" },
1782	{ "web-browser", 0, 32, 0, "" },
1783	{ "white-foreground", 0, 0, 0, "off" },
1784	{ 0,0,0,0,0 }
1785	};
1786
1787	/*
1788	** COMMAND: settings
1789	** COMMAND: unset
	@@ -1679,10 +1792,14 @@
1792	** %fossil unset PROPERTY ?-global?
1793	**
1794	** The "settings" command with no arguments lists all properties and their
1795	** values. With just a property name it shows the value of that property.
1796	** With a value argument it changes the property for the current repository.
1797	**
1798	** Settings marked as versionable are overridden by the contents of the
1799	** file named .fossil-settings/PROPERTY in the checked out files, if that
1800	** file exists.
1801	**
1802	** The "unset" command clears a property setting.
1803	**
1804	**
1805	** auto-captcha If enabled, the Login page provides a button to
	@@ -1696,20 +1813,25 @@
1813	** or update and automatically push after commit or
1814	** tag or branch creation. If the value is "pullonly"
1815	** then only pull operations occur automatically.
1816	** Default: on
1817	**
1818	** binary-glob The VALUE is a comma or newline-separated list of
1819	** (versionable) GLOB patterns that should be treated as binary files
1820	** for merging purposes. Example: *.xml
1821	**
1822	** case-sensitive If TRUE, the files whose names differ only in case
1823	** care considered distinct. If FALSE files whose names
1824	** differ only in case are the same file. Defaults to
1825	** TRUE for unix and FALSE for windows and mac.
1826	**
1827	** clearsign When enabled, fossil will attempt to sign all commits
1828	** with gpg. When disabled (the default), commits will
1829	** be unsigned. Default: off
1830	**
1831	** crnl-glob A comma or newline-separated list of GLOB patterns for
1832	** (versionable) text files in which it is ok to have CR+NL line endings.
1833	** Set to "*" to disable CR+NL checking.
1834	**
1835	** default-perms Permissions given automatically to new users. For more
1836	** information on permissions see Users page in Server
1837	** Administration of the HTTP UI. Default: u.
	@@ -1717,10 +1839,15 @@
1839	** diff-command External command to run when performing a diff.
1840	** If undefined, the internal text diff will be used.
1841	**
1842	** dont-push Prevent this repository from pushing from client to
1843	** server. Useful when setting up a private branch.
1844	**
1845	** empty-dirs A comma or newline-separated list of pathnames. On
1846	** (versionable) update and checkout commands, if no file or directory
1847	** exists with that name, an empty directory will be
1848	** created.
1849	**
1850	** editor Text editor command used for check-in comments.
1851	**
1852	** gdiff-command External command to run when performing a graphical
1853	** diff. If undefined, text diff will be used.
	@@ -1735,23 +1862,23 @@
1862	** and "ui" commands. Default: 8080
1863	**
1864	** https-login Send login creditials using HTTPS instead of HTTP
1865	** even if the login page request came via HTTP.
1866	**
1867	** ignore-glob The VALUE is a comma or newline-separated list of GLOB
1868	** (versionable) patterns specifying files that the "extra" command will
1869	** ignore. Example: .o,.obj,*.exe
1870	**
1871	** localauth If enabled, require that HTTP connections from
1872	** 127.0.0.1 be authenticated by password. If
1873	** false, all HTTP requests from localhost have
1874	** unrestricted access to the repository.
1875	**
1876	** main-branch The primary branch for the project. Default: trunk
1877	**
1878	** manifest If enabled, automatically create files "manifest" and
1879	** (versionable) "manifest.uuid" in every checkout. The SQLite and
1880	** Fossil repositories both require this. Default: off.
1881	**
1882	** max-upload A limit on the size of uplink HTTP requests. The
1883	** default is 250000 bytes.
1884	**
	@@ -1763,10 +1890,13 @@
1890	**
1891	** proxy URL of the HTTP proxy. If undefined or "off" then
1892	** the "http_proxy" environment variable is consulted.
1893	** If the http_proxy environment variable is undefined
1894	** then a direct HTTP connection is used.
1895	**
1896	** relative-paths When showing changes and extras, report paths relative
1897	** to the current working directory. Default: "on"
1898	**
1899	** repo-cksum Compute checksums over all files in each checkout
1900	** as a double-check of correctness. Defaults to "on".
1901	** Disable on large repositories for a performance
1902	** improvement.
	@@ -1773,10 +1903,28 @@
1903	**
1904	** self-register Allow users to register themselves through the HTTP UI.
1905	** This is useful if you want to see other names than
1906	** "Anonymous" in e.g. ticketing system. On the other hand
1907	** users can not be deleted. Default: off.
1908	**
1909	** ssl-ca-location The full pathname to a file containing PEM encoded
1910	** CA root certificates, or a directory of certificates
1911	** with filenames formed from the certificate hashes as
1912	** required by OpenSSL.
1913	** If set, this will override the OS default list of
1914	** OpenSSL CAs. If unset, the default list will be used.
1915	** Some platforms may add additional certificates.
1916	** Check your platform behaviour is as required if the
1917	** exact contents of the CA root is critical for your
1918	** application.
1919	**
1920	** ssl-identity The full pathname to a file containing a certificate
1921	** and private key in PEM format. Create by concatenating
1922	** the certificate and private key files.
1923	** This identity will be presented to SSL servers to
1924	** authenticate this client, in addition to the normal
1925	** password authentication.
1926	**
1927	** ssh-command Command used to talk to a remote machine with
1928	** the "ssh://" protocol.
1929	**
1930	** web-browser A shell command used to launch your preferred
	@@ -1797,12 +1945,13 @@
1945	}
1946	if( unsetFlag && g.argc!=3 ){
1947	usage("PROPERTY ?-global?");
1948	}
1949	if( g.argc==2 ){
1950	int openLocal = db_open_local();
1951	for(i=0; ctrlSettings[i].name; i++){
1952	print_setting(&ctrlSettings[i], openLocal);
1953	}
1954	}else if( g.argc==3 \|\| g.argc==4 ){
1955	const char *zName = g.argv[2];
1956	int isManifest;
1957	int n = strlen(zName);
	@@ -1820,13 +1969,13 @@
1969	db_unset(ctrlSettings[i].name, globalFlag);
1970	}else if( g.argc==4 ){
1971	db_set(ctrlSettings[i].name, g.argv[3], globalFlag);
1972	}else{
1973	isManifest = 0;
1974	print_setting(&ctrlSettings[i], db_open_local());
1975	}
1976	if( isManifest && g.localOpen ){
1977	manifest_to_disk(db_lget_int("checkout", 0));
1978	}
1979	}else{
1980	usage("?PROPERTY? ?VALUE?");
1981	}
1982

M src/descendants.c

+2 -2

		--- src/descendants.c
		+++ src/descendants.c
		@@ -287,11 +287,11 @@
287	287	"%s"
288	288	" AND event.objid IN (SELECT rid FROM leaves)"
289	289	" ORDER BY event.mtime DESC",
290	290	timeline_query_for_tty()
291	291	);
292		- print_timeline(&q, 20);
	292	+ print_timeline(&q, 20, 0);
293	293	db_finalize(&q);
294	294	}
295	295
296	296	/*
297	297	** COMMAND: leaves
		@@ -322,11 +322,11 @@
322	322	}else if( !showAll ){
323	323	blob_appendf(&sql," AND NOT %z", leaf_is_closed_sql("blob.rid"));
324	324	}
325	325	db_prepare(&q, "%s ORDER BY event.mtime DESC", blob_str(&sql));
326	326	blob_reset(&sql);
327		- print_timeline(&q, 2000);
	327	+ print_timeline(&q, 2000, 0);
328	328	db_finalize(&q);
329	329	}
330	330
331	331	/*
332	332	** WEBPAGE: leaves
333	333

	--- src/descendants.c
	+++ src/descendants.c
	@@ -287,11 +287,11 @@
287	"%s"
288	" AND event.objid IN (SELECT rid FROM leaves)"
289	" ORDER BY event.mtime DESC",
290	timeline_query_for_tty()
291	);
292	print_timeline(&q, 20);
293	db_finalize(&q);
294	}
295
296	/*
297	** COMMAND: leaves
	@@ -322,11 +322,11 @@
322	}else if( !showAll ){
323	blob_appendf(&sql," AND NOT %z", leaf_is_closed_sql("blob.rid"));
324	}
325	db_prepare(&q, "%s ORDER BY event.mtime DESC", blob_str(&sql));
326	blob_reset(&sql);
327	print_timeline(&q, 2000);
328	db_finalize(&q);
329	}
330
331	/*
332	** WEBPAGE: leaves
333

	--- src/descendants.c
	+++ src/descendants.c
	@@ -287,11 +287,11 @@
287	"%s"
288	" AND event.objid IN (SELECT rid FROM leaves)"
289	" ORDER BY event.mtime DESC",
290	timeline_query_for_tty()
291	);
292	print_timeline(&q, 20, 0);
293	db_finalize(&q);
294	}
295
296	/*
297	** COMMAND: leaves
	@@ -322,11 +322,11 @@
322	}else if( !showAll ){
323	blob_appendf(&sql," AND NOT %z", leaf_is_closed_sql("blob.rid"));
324	}
325	db_prepare(&q, "%s ORDER BY event.mtime DESC", blob_str(&sql));
326	blob_reset(&sql);
327	print_timeline(&q, 2000, 0);
328	db_finalize(&q);
329	}
330
331	/*
332	** WEBPAGE: leaves
333

M src/diffcmd.c

+22 -16

		--- src/diffcmd.c
		+++ src/diffcmd.c
		@@ -175,23 +175,24 @@
175	175	blob_reset(&cmd);
176	176	}
177	177	}
178	178
179	179	/*
180		-** Do a diff against a single file named in g.argv[2] from version zFrom
	180	+** Do a diff against a single file named in zFileTreeName from version zFrom
181	181	** against the same file on disk.
182	182	*/
183	183	static void diff_one_against_disk(
184	184	const char zFrom, / Name of file */
185	185	const char zDiffCmd, / Use this "diff" command */
186		- int ignoreEolWs /* Ignore whitespace changes at end of lines */
	186	+ int ignoreEolWs, /* Ignore whitespace changes at end of lines */
	187	+ const char *zFileTreeName
187	188	){
188	189	Blob fname;
189	190	Blob content;
190		- file_tree_name(g.argv[2], &fname, 1);
	191	+ file_tree_name(zFileTreeName, &fname, 1);
191	192	historical_version_of_file(zFrom, blob_str(&fname), &content, 0, 0);
192		- diff_file(&content, g.argv[2], g.argv[2], zDiffCmd, ignoreEolWs);
	193	+ diff_file(&content, zFileTreeName, zFileTreeName, zDiffCmd, ignoreEolWs);
193	194	blob_reset(&content);
194	195	blob_reset(&fname);
195	196	}
196	197
197	198	/*
		@@ -295,22 +296,22 @@
295	296	}
296	297
297	298	/*
298	299	** Output the differences between two versions of a single file.
299	300	** zFrom and zTo are the check-ins containing the two file versions.
300		-** The filename is contained in g.argv[2].
301	301	*/
302	302	static void diff_one_two_versions(
303	303	const char *zFrom,
304	304	const char *zTo,
305	305	const char *zDiffCmd,
306		- int ignoreEolWs
	306	+ int ignoreEolWs,
	307	+ const char *zFileTreeName
307	308	){
308	309	char *zName;
309	310	Blob fname;
310	311	Blob v1, v2;
311		- file_tree_name(g.argv[2], &fname, 1);
	312	+ file_tree_name(zFileTreeName, &fname, 1);
312	313	zName = blob_str(&fname);
313	314	historical_version_of_file(zFrom, zName, &v1, 0, 0);
314	315	historical_version_of_file(zTo, zName, &v2, 0, 0);
315	316	diff_file_mem(&v1, &v2, zName, zDiffCmd, ignoreEolWs);
316	317	blob_reset(&v1);
		@@ -408,16 +409,16 @@
408	409
409	410	/*
410	411	** COMMAND: diff
411	412	** COMMAND: gdiff
412	413	**
413		-** Usage: %fossil diff\|gdiff ?options? ?FILE?
	414	+** Usage: %fossil diff\|gdiff ?options? ?FILE1? ?FILE2 ...?
414	415	**
415		-** Show the difference between the current version of FILE (as it
416		-** exists on disk) and that same file as it was checked out. Or
417		-** if the FILE argument is omitted, show the unsaved changed currently
418		-** in the working check-out.
	416	+** Show the difference between the current version of each of the FILEs
	417	+** specified (as they exist on disk) and that same file as it was checked
	418	+** out. Or if the FILE arguments are omitted, show the unsaved changed
	419	+** currently in the working check-out.
419	420	**
420	421	** If the "--from VERSION" or "-r VERSION" option is used it specifies
421	422	** the source check-in for the diff operation. If not specified, the
422	423	** source check-in is the base check-in for the current check-out.
423	424	**
		@@ -440,10 +441,11 @@
440	441	int hasNFlag; /* True if -N or --new-file flag is used */
441	442	const char zFrom; / Source version number */
442	443	const char zTo; / Target version number */
443	444	const char zDiffCmd = 0; / External diff command. NULL for internal diff */
444	445	int diffFlags = 0; /* Flags to control the DIFF */
	446	+ int f;
445	447
446	448	isGDiff = g.argv[1][0]=='g';
447	449	isInternDiff = find_option("internal","i",0)!=0;
448	450	zFrom = find_option("from", "r", 1);
449	451	zTo = find_option("to", 0, 1);
		@@ -455,12 +457,14 @@
455	457	db_must_be_within_tree();
456	458	verify_all_options();
457	459	if( !isInternDiff ){
458	460	zDiffCmd = db_get(isGDiff ? "gdiff-command" : "diff-command", 0);
459	461	}
460		- if( g.argc==3 ){
461		- diff_one_against_disk(zFrom, zDiffCmd, 0);
	462	+ if( g.argc>=3 ){
	463	+ for(f=2; f<g.argc; ++f){
	464	+ diff_one_against_disk(zFrom, zDiffCmd, 0, g.argv[f]);
	465	+ }
462	466	}else{
463	467	diff_all_against_disk(zFrom, zDiffCmd, diffFlags);
464	468	}
465	469	}else if( zFrom==0 ){
466	470	fossil_fatal("must use --from if --to is present");
		@@ -468,12 +472,14 @@
468	472	db_find_and_open_repository(0, 0);
469	473	verify_all_options();
470	474	if( !isInternDiff ){
471	475	zDiffCmd = db_get(isGDiff ? "gdiff-command" : "diff-command", 0);
472	476	}
473		- if( g.argc==3 ){
474		- diff_one_two_versions(zFrom, zTo, zDiffCmd, 0);
	477	+ if( g.argc>=3 ){
	478	+ for(f=2; f<g.argc; ++f){
	479	+ diff_one_two_versions(zFrom, zTo, zDiffCmd, 0, g.argv[f]);
	480	+ }
475	481	}else{
476	482	diff_all_two_versions(zFrom, zTo, zDiffCmd, diffFlags);
477	483	}
478	484	}
479	485	}
480	486

	--- src/diffcmd.c
	+++ src/diffcmd.c
	@@ -175,23 +175,24 @@
175	blob_reset(&cmd);
176	}
177	}
178
179	/*
180	** Do a diff against a single file named in g.argv[2] from version zFrom
181	** against the same file on disk.
182	*/
183	static void diff_one_against_disk(
184	const char zFrom, / Name of file */
185	const char zDiffCmd, / Use this "diff" command */
186	int ignoreEolWs /* Ignore whitespace changes at end of lines */

187	){
188	Blob fname;
189	Blob content;
190	file_tree_name(g.argv[2], &fname, 1);
191	historical_version_of_file(zFrom, blob_str(&fname), &content, 0, 0);
192	diff_file(&content, g.argv[2], g.argv[2], zDiffCmd, ignoreEolWs);
193	blob_reset(&content);
194	blob_reset(&fname);
195	}
196
197	/*
	@@ -295,22 +296,22 @@
295	}
296
297	/*
298	** Output the differences between two versions of a single file.
299	** zFrom and zTo are the check-ins containing the two file versions.
300	** The filename is contained in g.argv[2].
301	*/
302	static void diff_one_two_versions(
303	const char *zFrom,
304	const char *zTo,
305	const char *zDiffCmd,
306	int ignoreEolWs

307	){
308	char *zName;
309	Blob fname;
310	Blob v1, v2;
311	file_tree_name(g.argv[2], &fname, 1);
312	zName = blob_str(&fname);
313	historical_version_of_file(zFrom, zName, &v1, 0, 0);
314	historical_version_of_file(zTo, zName, &v2, 0, 0);
315	diff_file_mem(&v1, &v2, zName, zDiffCmd, ignoreEolWs);
316	blob_reset(&v1);
	@@ -408,16 +409,16 @@
408
409	/*
410	** COMMAND: diff
411	** COMMAND: gdiff
412	**
413	** Usage: %fossil diff\|gdiff ?options? ?FILE?
414	**
415	** Show the difference between the current version of FILE (as it
416	** exists on disk) and that same file as it was checked out. Or
417	** if the FILE argument is omitted, show the unsaved changed currently
418	** in the working check-out.
419	**
420	** If the "--from VERSION" or "-r VERSION" option is used it specifies
421	** the source check-in for the diff operation. If not specified, the
422	** source check-in is the base check-in for the current check-out.
423	**
	@@ -440,10 +441,11 @@
440	int hasNFlag; /* True if -N or --new-file flag is used */
441	const char zFrom; / Source version number */
442	const char zTo; / Target version number */
443	const char zDiffCmd = 0; / External diff command. NULL for internal diff */
444	int diffFlags = 0; /* Flags to control the DIFF */

445
446	isGDiff = g.argv[1][0]=='g';
447	isInternDiff = find_option("internal","i",0)!=0;
448	zFrom = find_option("from", "r", 1);
449	zTo = find_option("to", 0, 1);
	@@ -455,12 +457,14 @@
455	db_must_be_within_tree();
456	verify_all_options();
457	if( !isInternDiff ){
458	zDiffCmd = db_get(isGDiff ? "gdiff-command" : "diff-command", 0);
459	}
460	if( g.argc==3 ){
461	diff_one_against_disk(zFrom, zDiffCmd, 0);


462	}else{
463	diff_all_against_disk(zFrom, zDiffCmd, diffFlags);
464	}
465	}else if( zFrom==0 ){
466	fossil_fatal("must use --from if --to is present");
	@@ -468,12 +472,14 @@
468	db_find_and_open_repository(0, 0);
469	verify_all_options();
470	if( !isInternDiff ){
471	zDiffCmd = db_get(isGDiff ? "gdiff-command" : "diff-command", 0);
472	}
473	if( g.argc==3 ){
474	diff_one_two_versions(zFrom, zTo, zDiffCmd, 0);


475	}else{
476	diff_all_two_versions(zFrom, zTo, zDiffCmd, diffFlags);
477	}
478	}
479	}
480

	--- src/diffcmd.c
	+++ src/diffcmd.c
	@@ -175,23 +175,24 @@
175	blob_reset(&cmd);
176	}
177	}
178
179	/*
180	** Do a diff against a single file named in zFileTreeName from version zFrom
181	** against the same file on disk.
182	*/
183	static void diff_one_against_disk(
184	const char zFrom, / Name of file */
185	const char zDiffCmd, / Use this "diff" command */
186	int ignoreEolWs, /* Ignore whitespace changes at end of lines */
187	const char *zFileTreeName
188	){
189	Blob fname;
190	Blob content;
191	file_tree_name(zFileTreeName, &fname, 1);
192	historical_version_of_file(zFrom, blob_str(&fname), &content, 0, 0);
193	diff_file(&content, zFileTreeName, zFileTreeName, zDiffCmd, ignoreEolWs);
194	blob_reset(&content);
195	blob_reset(&fname);
196	}
197
198	/*
	@@ -295,22 +296,22 @@
296	}
297
298	/*
299	** Output the differences between two versions of a single file.
300	** zFrom and zTo are the check-ins containing the two file versions.

301	*/
302	static void diff_one_two_versions(
303	const char *zFrom,
304	const char *zTo,
305	const char *zDiffCmd,
306	int ignoreEolWs,
307	const char *zFileTreeName
308	){
309	char *zName;
310	Blob fname;
311	Blob v1, v2;
312	file_tree_name(zFileTreeName, &fname, 1);
313	zName = blob_str(&fname);
314	historical_version_of_file(zFrom, zName, &v1, 0, 0);
315	historical_version_of_file(zTo, zName, &v2, 0, 0);
316	diff_file_mem(&v1, &v2, zName, zDiffCmd, ignoreEolWs);
317	blob_reset(&v1);
	@@ -408,16 +409,16 @@
409
410	/*
411	** COMMAND: diff
412	** COMMAND: gdiff
413	**
414	** Usage: %fossil diff\|gdiff ?options? ?FILE1? ?FILE2 ...?
415	**
416	** Show the difference between the current version of each of the FILEs
417	** specified (as they exist on disk) and that same file as it was checked
418	** out. Or if the FILE arguments are omitted, show the unsaved changed
419	** currently in the working check-out.
420	**
421	** If the "--from VERSION" or "-r VERSION" option is used it specifies
422	** the source check-in for the diff operation. If not specified, the
423	** source check-in is the base check-in for the current check-out.
424	**
	@@ -440,10 +441,11 @@
441	int hasNFlag; /* True if -N or --new-file flag is used */
442	const char zFrom; / Source version number */
443	const char zTo; / Target version number */
444	const char zDiffCmd = 0; / External diff command. NULL for internal diff */
445	int diffFlags = 0; /* Flags to control the DIFF */
446	int f;
447
448	isGDiff = g.argv[1][0]=='g';
449	isInternDiff = find_option("internal","i",0)!=0;
450	zFrom = find_option("from", "r", 1);
451	zTo = find_option("to", 0, 1);
	@@ -455,12 +457,14 @@
457	db_must_be_within_tree();
458	verify_all_options();
459	if( !isInternDiff ){
460	zDiffCmd = db_get(isGDiff ? "gdiff-command" : "diff-command", 0);
461	}
462	if( g.argc>=3 ){
463	for(f=2; f<g.argc; ++f){
464	diff_one_against_disk(zFrom, zDiffCmd, 0, g.argv[f]);
465	}
466	}else{
467	diff_all_against_disk(zFrom, zDiffCmd, diffFlags);
468	}
469	}else if( zFrom==0 ){
470	fossil_fatal("must use --from if --to is present");
	@@ -468,12 +472,14 @@
472	db_find_and_open_repository(0, 0);
473	verify_all_options();
474	if( !isInternDiff ){
475	zDiffCmd = db_get(isGDiff ? "gdiff-command" : "diff-command", 0);
476	}
477	if( g.argc>=3 ){
478	for(f=2; f<g.argc; ++f){
479	diff_one_two_versions(zFrom, zTo, zDiffCmd, 0, g.argv[f]);
480	}
481	}else{
482	diff_all_two_versions(zFrom, zTo, zDiffCmd, diffFlags);
483	}
484	}
485	}
486

M src/export.c

+151 -36

		--- src/export.c
		+++ src/export.c
		@@ -78,15 +78,17 @@
78	78	free(zName);
79	79	free(zEmail);
80	80	db_reset(&q);
81	81	}
82	82
	83	+#define BLOBMARK(rid) ((rid) * 2)
	84	+#define COMMITMARK(rid) ((rid) * 2 + 1)
83	85
84	86	/*
85	87	** COMMAND: export
86	88	**
87		-** Usage: %fossil export --git ?REPOSITORY?
	89	+** Usage: %fossil export --git ?options? ?REPOSITORY?
88	90	**
89	91	** Write an export of all check-ins to standard output. The export is
90	92	** written in the git-fast-export file format assuming the --git option is
91	93	** provided. The git-fast-export format is currently the only VCS
92	94	** interchange format supported, though other formats may be added in
		@@ -95,98 +97,190 @@
95	97	** Run this command within a checkout. Or use the -R or --repository
96	98	** option to specify a Fossil repository to be exported.
97	99	**
98	100	** Only check-ins are exported using --git. Git does not support tickets
99	101	** or wiki or events or attachments, so none of those are exported.
	102	+**
	103	+** If the "--import-marks FILE" option is used, it contains a list of
	104	+** rids to skip.
	105	+**
	106	+** If the "--export-marks FILE" option is used, the rid of all commits and
	107	+** blobs written on exit for use with "--import-marks" on the next run.
100	108	*/
101	109	void export_cmd(void){
102		- Stmt q;
	110	+ Stmt q, q2, q3;
103	111	int i;
104		- int firstCkin; /* Integer offset to check-in marks */
105	112	Bag blobs, vers;
	113	+ const char *markfile_in;
	114	+ const char *markfile_out;
	115	+
106	116	bag_init(&blobs);
107	117	bag_init(&vers);
108	118
109	119	find_option("git", 0, 0); /* Ignore the --git option for now */
	120	+ markfile_in = find_option("import-marks", 0, 1);
	121	+ markfile_out = find_option("export-marks", 0, 1);
	122	+
110	123	db_find_and_open_repository(0, 2);
111	124	verify_all_options();
112	125	if( g.argc!=2 && g.argc!=3 ){ usage("--git ?REPOSITORY?"); }
	126	+
	127	+ db_multi_exec("CREATE TEMPORARY TABLE oldblob(rid INTEGER PRIMARY KEY)");
	128	+ db_multi_exec("CREATE TEMPORARY TABLE oldcommit(rid INTEGER PRIMARY KEY)");
	129	+ if( markfile_in!=0 ){
	130	+ Stmt qb,qc;
	131	+ char line[100];
	132	+ FILE *f;
	133	+
	134	+ f = fopen(markfile_in, "r");
	135	+ if( f==0 ){
	136	+ fossil_panic("cannot open %s for reading", markfile_in);
	137	+ }
	138	+ db_prepare(&qb, "INSERT OR IGNORE INTO oldblob VALUES (:rid)");
	139	+ db_prepare(&qc, "INSERT OR IGNORE INTO oldcommit VALUES (:rid)");
	140	+ while( fgets(line, sizeof(line), f)!=0 ){
	141	+ if( *line == 'b' ){
	142	+ db_bind_text(&qb, ":rid", line + 1);
	143	+ db_step(&qb);
	144	+ db_reset(&qb);
	145	+ bag_insert(&blobs, atoi(line + 1));
	146	+ }else if( *line == 'c' ){
	147	+ db_bind_text(&qc, ":rid", line + 1);
	148	+ db_step(&qc);
	149	+ db_reset(&qc);
	150	+ bag_insert(&vers, atoi(line + 1));
	151	+ }else{
	152	+ fossil_panic("bad input from %s: %s", markfile_in, line);
	153	+ }
	154	+ }
	155	+ db_finalize(&qb);
	156	+ db_finalize(&qc);
	157	+ fclose(f);
	158	+ }
113	159
114	160	/* Step 1: Generate "blob" records for every artifact that is part
115	161	** of a check-in
116	162	*/
117	163	fossil_binary_mode(stdout);
118		- db_prepare(&q, "SELECT DISTINCT fid FROM mlink WHERE fid>0");
	164	+ db_multi_exec("CREATE TEMPORARY TABLE newblob(rid INTEGER KEY, srcid INTEGER)");
	165	+ db_multi_exec("CREATE INDEX newblob_src ON newblob(srcid)");
	166	+ db_multi_exec(
	167	+ "INSERT INTO newblob"
	168	+ " SELECT DISTINCT fid,"
	169	+ " CASE WHEN EXISTS(SELECT 1 FROM delta WHERE rid=fid AND NOT EXISTS(SELECT 1 FROM oldblob WHERE srcid=fid))"
	170	+ " THEN (SELECT srcid FROM delta WHERE rid=fid)"
	171	+ " ELSE 0"
	172	+ " END"
	173	+ " FROM mlink"
	174	+ " WHERE fid>0 AND NOT EXISTS(SELECT 1 FROM oldblob WHERE rid=fid)");
	175	+ db_prepare(&q,
	176	+ "SELECT DISTINCT fid FROM mlink"
	177	+ " WHERE fid>0 AND NOT EXISTS(SELECT 1 FROM oldblob WHERE rid=fid)");
	178	+ db_prepare(&q2, "INSERT INTO oldblob VALUES (:rid)");
	179	+ db_prepare(&q3, "SELECT rid FROM newblob WHERE srcid= (:srcid)");
119	180	while( db_step(&q)==SQLITE_ROW ){
120	181	int rid = db_column_int(&q, 0);
121	182	Blob content;
122		- content_get(rid, &content);
123		- printf("blob\nmark :%d\ndata %d\n", rid, blob_size(&content));
124		- bag_insert(&blobs, rid);
125		- fwrite(blob_buffer(&content), 1, blob_size(&content), stdout);
126		- printf("\n");
127		- blob_reset(&content);
	183	+
	184	+ while( !bag_find(&blobs, rid) ){
	185	+ content_get(rid, &content);
	186	+ db_bind_int(&q2, ":rid", rid);
	187	+ db_step(&q2);
	188	+ db_reset(&q2);
	189	+ printf("blob\nmark :%d\ndata %d\n", BLOBMARK(rid), blob_size(&content));
	190	+ bag_insert(&blobs, rid);
	191	+ fwrite(blob_buffer(&content), 1, blob_size(&content), stdout);
	192	+ printf("\n");
	193	+ blob_reset(&content);
	194	+
	195	+ db_bind_int(&q3, ":srcid", rid);
	196	+ if( db_step(&q3) != SQLITE_ROW ){
	197	+ db_reset(&q3);
	198	+ break;
	199	+ }
	200	+ rid = db_column_int(&q3, 0);
	201	+ db_reset(&q3);
	202	+ }
128	203	}
129	204	db_finalize(&q);
	205	+ db_finalize(&q2);
	206	+ db_finalize(&q3);
130	207
131	208	/* Output the commit records.
132	209	*/
133		- firstCkin = db_int(0, "SELECT max(rid) FROM blob")+1;
134	210	db_prepare(&q,
135	211	"SELECT strftime('%%s',mtime), objid, coalesce(comment,ecomment),"
136	212	" coalesce(user,euser),"
137	213	" (SELECT value FROM tagxref WHERE rid=objid AND tagid=%d)"
138	214	" FROM event"
139		- " WHERE type='ci'"
	215	+ " WHERE type='ci' AND NOT EXISTS (SELECT 1 FROM oldcommit WHERE objid=rid)"
140	216	" ORDER BY mtime ASC",
141	217	TAG_BRANCH
142	218	);
	219	+ db_prepare(&q2, "INSERT INTO oldcommit VALUES (:rid)");
143	220	while( db_step(&q)==SQLITE_ROW ){
	221	+ Stmt q4;
144	222	const char *zSecondsSince1970 = db_column_text(&q, 0);
145	223	int ckinId = db_column_int(&q, 1);
146	224	const char *zComment = db_column_text(&q, 2);
147	225	const char *zUser = db_column_text(&q, 3);
148	226	const char *zBranch = db_column_text(&q, 4);
149	227	char *zBr;
150		- Manifest *p;
151		- ManifestFile *pFile;
152		- const char *zFromType;
153	228
154	229	bag_insert(&vers, ckinId);
	230	+ db_bind_int(&q2, ":rid", ckinId);
	231	+ db_step(&q2);
	232	+ db_reset(&q2);
155	233	if( zBranch==0 ) zBranch = "trunk";
156	234	zBr = mprintf("%s", zBranch);
157	235	for(i=0; zBr[i]; i++){
158	236	if( !fossil_isalnum(zBr[i]) ) zBr[i] = '_';
159	237	}
160		- printf("commit refs/heads/%s\nmark :%d\n", zBr, ckinId+firstCkin);
	238	+ printf("commit refs/heads/%s\nmark :%d\n", zBr, COMMITMARK(ckinId));
161	239	free(zBr);
162	240	printf("committer");
163	241	print_person(zUser);
164	242	printf(" %s +0000\n", zSecondsSince1970);
165	243	if( zComment==0 ) zComment = "null comment";
166	244	printf("data %d\n%s\n", (int)strlen(zComment), zComment);
167		- p = manifest_get(ckinId, CFTYPE_ANY);
168		- zFromType = "from";
169		- for(i=0; i<p->nParent; i++){
170		- int pid = fast_uuid_to_rid(p->azParent[i]);
171		- if( pid==0 \|\| !bag_find(&vers, pid) ) continue;
172		- printf("%s :%d\n", zFromType, fast_uuid_to_rid(p->azParent[i])+firstCkin);
173		- zFromType = "merge";
174		- }
175		- printf("deleteall\n");
176		- manifest_file_rewind(p);
177		- while( (pFile=manifest_file_next(p, 0))!=0 ){
178		- int fid = fast_uuid_to_rid(pFile->zUuid);
179		- const char *zPerm = "100644";
180		- if( fid==0 ) continue;
181		- if( pFile->zPerm && strstr(pFile->zPerm,"x") ) zPerm = "100755";
182		- if( !bag_find(&blobs, fid) ) continue;
183		- printf("M %s :%d %s\n", zPerm, fid, pFile->zName);
184		- }
185		- manifest_cache_insert(p);
	245	+ db_prepare(&q3, "SELECT pid FROM plink WHERE cid=%d AND isprim", ckinId);
	246	+ if( db_step(&q3) == SQLITE_ROW ){
	247	+ printf("from :%d\n", COMMITMARK(db_column_int(&q3, 0)));
	248	+ db_prepare(&q4,
	249	+ "SELECT pid FROM plink"
	250	+ " WHERE cid=%d AND NOT isprim"
	251	+ " AND NOT EXISTS(SELECT 1 FROM phantom WHERE rid=pid)"
	252	+ " ORDER BY pid",
	253	+ ckinId);
	254	+ while( db_step(&q4)==SQLITE_ROW ){
	255	+ printf("merge :%d\n", COMMITMARK(db_column_int(&q4,0)));
	256	+ }
	257	+ db_finalize(&q4);
	258	+ }else{
	259	+ printf("deleteall\n");
	260	+ }
	261	+
	262	+ db_prepare(&q4,
	263	+ "SELECT filename.name, mlink.fid, mlink.mperm FROM mlink"
	264	+ " JOIN filename ON filename.fnid=mlink.fnid"
	265	+ " WHERE mlink.mid=%d",
	266	+ ckinId
	267	+ );
	268	+ while( db_step(&q4)==SQLITE_ROW ){
	269	+ const char *zName = db_column_text(&q4,0);
	270	+ int zNew = db_column_int(&q4,1);
	271	+ int mPerm = db_column_int(&q4,2);
	272	+ if( zNew==0)
	273	+ printf("D %s\n", zName);
	274	+ else if( bag_find(&blobs, zNew) )
	275	+ printf("M %s :%d %s\n", mPerm ? "100755" : "100644", BLOBMARK(zNew), zName);
	276	+ }
	277	+ db_finalize(&q4);
	278	+ db_finalize(&q3);
186	279	printf("\n");
187	280	}
	281	+ db_finalize(&q2);
188	282	db_finalize(&q);
189	283	bag_clear(&blobs);
190	284	manifest_cache_clear();
191	285
192	286
		@@ -207,13 +301,34 @@
207	301	zEncoded = mprintf("%s", zTagname);
208	302	for(i=0; zEncoded[i]; i++){
209	303	if( !fossil_isalnum(zEncoded[i]) ) zEncoded[i] = '_';
210	304	}
211	305	printf("tag %s\n", zEncoded);
212		- printf("from :%d\n", rid+firstCkin);
	306	+ printf("from :%d\n", COMMITMARK(rid));
213	307	printf("tagger <tagger> %s +0000\n", zSecSince1970);
214	308	printf("data 0\n");
215	309	fossil_free(zEncoded);
216	310	}
217	311	db_finalize(&q);
218	312	bag_clear(&vers);
	313	+
	314	+ if( markfile_out!=0 ){
	315	+ FILE *f;
	316	+ f = fopen(markfile_out, "w");
	317	+ if( f == 0 ){
	318	+ fossil_panic("cannot open %s for writing", markfile_out);
	319	+ }
	320	+ db_prepare(&q, "SELECT rid FROM oldblob");
	321	+ while( db_step(&q)==SQLITE_ROW ){
	322	+ fprintf(f, "b%d\n", db_column_int(&q, 0));
	323	+ }
	324	+ db_finalize(&q);
	325	+ db_prepare(&q, "SELECT rid FROM oldcommit");
	326	+ while( db_step(&q)==SQLITE_ROW ){
	327	+ fprintf(f, "c%d\n", db_column_int(&q, 0));
	328	+ }
	329	+ db_finalize(&q);
	330	+ if( ferror(f)!=0 \|\| fclose(f)!=0 ) {
	331	+ fossil_panic("error while writing %s", markfile_out);
	332	+ }
	333	+ }
219	334	}
220	335

	--- src/export.c
	+++ src/export.c
	@@ -78,15 +78,17 @@
78	free(zName);
79	free(zEmail);
80	db_reset(&q);
81	}
82


83
84	/*
85	** COMMAND: export
86	**
87	** Usage: %fossil export --git ?REPOSITORY?
88	**
89	** Write an export of all check-ins to standard output. The export is
90	** written in the git-fast-export file format assuming the --git option is
91	** provided. The git-fast-export format is currently the only VCS
92	** interchange format supported, though other formats may be added in
	@@ -95,98 +97,190 @@
95	** Run this command within a checkout. Or use the -R or --repository
96	** option to specify a Fossil repository to be exported.
97	**
98	** Only check-ins are exported using --git. Git does not support tickets
99	** or wiki or events or attachments, so none of those are exported.






100	*/
101	void export_cmd(void){
102	Stmt q;
103	int i;
104	int firstCkin; /* Integer offset to check-in marks */
105	Bag blobs, vers;



106	bag_init(&blobs);
107	bag_init(&vers);
108
109	find_option("git", 0, 0); /* Ignore the --git option for now */



110	db_find_and_open_repository(0, 2);
111	verify_all_options();
112	if( g.argc!=2 && g.argc!=3 ){ usage("--git ?REPOSITORY?"); }

































113
114	/* Step 1: Generate "blob" records for every artifact that is part
115	** of a check-in
116	*/
117	fossil_binary_mode(stdout);
118	db_prepare(&q, "SELECT DISTINCT fid FROM mlink WHERE fid>0");















119	while( db_step(&q)==SQLITE_ROW ){
120	int rid = db_column_int(&q, 0);
121	Blob content;
122	content_get(rid, &content);
123	printf("blob\nmark :%d\ndata %d\n", rid, blob_size(&content));
124	bag_insert(&blobs, rid);
125	fwrite(blob_buffer(&content), 1, blob_size(&content), stdout);
126	printf("\n");
127	blob_reset(&content);














128	}
129	db_finalize(&q);


130
131	/* Output the commit records.
132	*/
133	firstCkin = db_int(0, "SELECT max(rid) FROM blob")+1;
134	db_prepare(&q,
135	"SELECT strftime('%%s',mtime), objid, coalesce(comment,ecomment),"
136	" coalesce(user,euser),"
137	" (SELECT value FROM tagxref WHERE rid=objid AND tagid=%d)"
138	" FROM event"
139	" WHERE type='ci'"
140	" ORDER BY mtime ASC",
141	TAG_BRANCH
142	);

143	while( db_step(&q)==SQLITE_ROW ){

144	const char *zSecondsSince1970 = db_column_text(&q, 0);
145	int ckinId = db_column_int(&q, 1);
146	const char *zComment = db_column_text(&q, 2);
147	const char *zUser = db_column_text(&q, 3);
148	const char *zBranch = db_column_text(&q, 4);
149	char *zBr;
150	Manifest *p;
151	ManifestFile *pFile;
152	const char *zFromType;
153
154	bag_insert(&vers, ckinId);



155	if( zBranch==0 ) zBranch = "trunk";
156	zBr = mprintf("%s", zBranch);
157	for(i=0; zBr[i]; i++){
158	if( !fossil_isalnum(zBr[i]) ) zBr[i] = '_';
159	}
160	printf("commit refs/heads/%s\nmark :%d\n", zBr, ckinId+firstCkin);
161	free(zBr);
162	printf("committer");
163	print_person(zUser);
164	printf(" %s +0000\n", zSecondsSince1970);
165	if( zComment==0 ) zComment = "null comment";
166	printf("data %d\n%s\n", (int)strlen(zComment), zComment);
167	p = manifest_get(ckinId, CFTYPE_ANY);
168	zFromType = "from";
169	for(i=0; i<p->nParent; i++){
170	int pid = fast_uuid_to_rid(p->azParent[i]);
171	if( pid==0 \|\| !bag_find(&vers, pid) ) continue;
172	printf("%s :%d\n", zFromType, fast_uuid_to_rid(p->azParent[i])+firstCkin);
173	zFromType = "merge";
174	}
175	printf("deleteall\n");
176	manifest_file_rewind(p);
177	while( (pFile=manifest_file_next(p, 0))!=0 ){
178	int fid = fast_uuid_to_rid(pFile->zUuid);
179	const char *zPerm = "100644";
180	if( fid==0 ) continue;
181	if( pFile->zPerm && strstr(pFile->zPerm,"x") ) zPerm = "100755";
182	if( !bag_find(&blobs, fid) ) continue;
183	printf("M %s :%d %s\n", zPerm, fid, pFile->zName);
184	}
185	manifest_cache_insert(p);















186	printf("\n");
187	}

188	db_finalize(&q);
189	bag_clear(&blobs);
190	manifest_cache_clear();
191
192
	@@ -207,13 +301,34 @@
207	zEncoded = mprintf("%s", zTagname);
208	for(i=0; zEncoded[i]; i++){
209	if( !fossil_isalnum(zEncoded[i]) ) zEncoded[i] = '_';
210	}
211	printf("tag %s\n", zEncoded);
212	printf("from :%d\n", rid+firstCkin);
213	printf("tagger <tagger> %s +0000\n", zSecSince1970);
214	printf("data 0\n");
215	fossil_free(zEncoded);
216	}
217	db_finalize(&q);
218	bag_clear(&vers);





















219	}
220

	--- src/export.c
	+++ src/export.c
	@@ -78,15 +78,17 @@
78	free(zName);
79	free(zEmail);
80	db_reset(&q);
81	}
82
83	#define BLOBMARK(rid) ((rid) * 2)
84	#define COMMITMARK(rid) ((rid) * 2 + 1)
85
86	/*
87	** COMMAND: export
88	**
89	** Usage: %fossil export --git ?options? ?REPOSITORY?
90	**
91	** Write an export of all check-ins to standard output. The export is
92	** written in the git-fast-export file format assuming the --git option is
93	** provided. The git-fast-export format is currently the only VCS
94	** interchange format supported, though other formats may be added in
	@@ -95,98 +97,190 @@
97	** Run this command within a checkout. Or use the -R or --repository
98	** option to specify a Fossil repository to be exported.
99	**
100	** Only check-ins are exported using --git. Git does not support tickets
101	** or wiki or events or attachments, so none of those are exported.
102	**
103	** If the "--import-marks FILE" option is used, it contains a list of
104	** rids to skip.
105	**
106	** If the "--export-marks FILE" option is used, the rid of all commits and
107	** blobs written on exit for use with "--import-marks" on the next run.
108	*/
109	void export_cmd(void){
110	Stmt q, q2, q3;
111	int i;

112	Bag blobs, vers;
113	const char *markfile_in;
114	const char *markfile_out;
115
116	bag_init(&blobs);
117	bag_init(&vers);
118
119	find_option("git", 0, 0); /* Ignore the --git option for now */
120	markfile_in = find_option("import-marks", 0, 1);
121	markfile_out = find_option("export-marks", 0, 1);
122
123	db_find_and_open_repository(0, 2);
124	verify_all_options();
125	if( g.argc!=2 && g.argc!=3 ){ usage("--git ?REPOSITORY?"); }
126
127	db_multi_exec("CREATE TEMPORARY TABLE oldblob(rid INTEGER PRIMARY KEY)");
128	db_multi_exec("CREATE TEMPORARY TABLE oldcommit(rid INTEGER PRIMARY KEY)");
129	if( markfile_in!=0 ){
130	Stmt qb,qc;
131	char line[100];
132	FILE *f;
133
134	f = fopen(markfile_in, "r");
135	if( f==0 ){
136	fossil_panic("cannot open %s for reading", markfile_in);
137	}
138	db_prepare(&qb, "INSERT OR IGNORE INTO oldblob VALUES (:rid)");
139	db_prepare(&qc, "INSERT OR IGNORE INTO oldcommit VALUES (:rid)");
140	while( fgets(line, sizeof(line), f)!=0 ){
141	if( *line == 'b' ){
142	db_bind_text(&qb, ":rid", line + 1);
143	db_step(&qb);
144	db_reset(&qb);
145	bag_insert(&blobs, atoi(line + 1));
146	}else if( *line == 'c' ){
147	db_bind_text(&qc, ":rid", line + 1);
148	db_step(&qc);
149	db_reset(&qc);
150	bag_insert(&vers, atoi(line + 1));
151	}else{
152	fossil_panic("bad input from %s: %s", markfile_in, line);
153	}
154	}
155	db_finalize(&qb);
156	db_finalize(&qc);
157	fclose(f);
158	}
159
160	/* Step 1: Generate "blob" records for every artifact that is part
161	** of a check-in
162	*/
163	fossil_binary_mode(stdout);
164	db_multi_exec("CREATE TEMPORARY TABLE newblob(rid INTEGER KEY, srcid INTEGER)");
165	db_multi_exec("CREATE INDEX newblob_src ON newblob(srcid)");
166	db_multi_exec(
167	"INSERT INTO newblob"
168	" SELECT DISTINCT fid,"
169	" CASE WHEN EXISTS(SELECT 1 FROM delta WHERE rid=fid AND NOT EXISTS(SELECT 1 FROM oldblob WHERE srcid=fid))"
170	" THEN (SELECT srcid FROM delta WHERE rid=fid)"
171	" ELSE 0"
172	" END"
173	" FROM mlink"
174	" WHERE fid>0 AND NOT EXISTS(SELECT 1 FROM oldblob WHERE rid=fid)");
175	db_prepare(&q,
176	"SELECT DISTINCT fid FROM mlink"
177	" WHERE fid>0 AND NOT EXISTS(SELECT 1 FROM oldblob WHERE rid=fid)");
178	db_prepare(&q2, "INSERT INTO oldblob VALUES (:rid)");
179	db_prepare(&q3, "SELECT rid FROM newblob WHERE srcid= (:srcid)");
180	while( db_step(&q)==SQLITE_ROW ){
181	int rid = db_column_int(&q, 0);
182	Blob content;
183
184	while( !bag_find(&blobs, rid) ){
185	content_get(rid, &content);
186	db_bind_int(&q2, ":rid", rid);
187	db_step(&q2);
188	db_reset(&q2);
189	printf("blob\nmark :%d\ndata %d\n", BLOBMARK(rid), blob_size(&content));
190	bag_insert(&blobs, rid);
191	fwrite(blob_buffer(&content), 1, blob_size(&content), stdout);
192	printf("\n");
193	blob_reset(&content);
194
195	db_bind_int(&q3, ":srcid", rid);
196	if( db_step(&q3) != SQLITE_ROW ){
197	db_reset(&q3);
198	break;
199	}
200	rid = db_column_int(&q3, 0);
201	db_reset(&q3);
202	}
203	}
204	db_finalize(&q);
205	db_finalize(&q2);
206	db_finalize(&q3);
207
208	/* Output the commit records.
209	*/

210	db_prepare(&q,
211	"SELECT strftime('%%s',mtime), objid, coalesce(comment,ecomment),"
212	" coalesce(user,euser),"
213	" (SELECT value FROM tagxref WHERE rid=objid AND tagid=%d)"
214	" FROM event"
215	" WHERE type='ci' AND NOT EXISTS (SELECT 1 FROM oldcommit WHERE objid=rid)"
216	" ORDER BY mtime ASC",
217	TAG_BRANCH
218	);
219	db_prepare(&q2, "INSERT INTO oldcommit VALUES (:rid)");
220	while( db_step(&q)==SQLITE_ROW ){
221	Stmt q4;
222	const char *zSecondsSince1970 = db_column_text(&q, 0);
223	int ckinId = db_column_int(&q, 1);
224	const char *zComment = db_column_text(&q, 2);
225	const char *zUser = db_column_text(&q, 3);
226	const char *zBranch = db_column_text(&q, 4);
227	char *zBr;



228
229	bag_insert(&vers, ckinId);
230	db_bind_int(&q2, ":rid", ckinId);
231	db_step(&q2);
232	db_reset(&q2);
233	if( zBranch==0 ) zBranch = "trunk";
234	zBr = mprintf("%s", zBranch);
235	for(i=0; zBr[i]; i++){
236	if( !fossil_isalnum(zBr[i]) ) zBr[i] = '_';
237	}
238	printf("commit refs/heads/%s\nmark :%d\n", zBr, COMMITMARK(ckinId));
239	free(zBr);
240	printf("committer");
241	print_person(zUser);
242	printf(" %s +0000\n", zSecondsSince1970);
243	if( zComment==0 ) zComment = "null comment";
244	printf("data %d\n%s\n", (int)strlen(zComment), zComment);
245	db_prepare(&q3, "SELECT pid FROM plink WHERE cid=%d AND isprim", ckinId);
246	if( db_step(&q3) == SQLITE_ROW ){
247	printf("from :%d\n", COMMITMARK(db_column_int(&q3, 0)));
248	db_prepare(&q4,
249	"SELECT pid FROM plink"
250	" WHERE cid=%d AND NOT isprim"
251	" AND NOT EXISTS(SELECT 1 FROM phantom WHERE rid=pid)"
252	" ORDER BY pid",
253	ckinId);
254	while( db_step(&q4)==SQLITE_ROW ){
255	printf("merge :%d\n", COMMITMARK(db_column_int(&q4,0)));
256	}
257	db_finalize(&q4);
258	}else{
259	printf("deleteall\n");
260	}
261
262	db_prepare(&q4,
263	"SELECT filename.name, mlink.fid, mlink.mperm FROM mlink"
264	" JOIN filename ON filename.fnid=mlink.fnid"
265	" WHERE mlink.mid=%d",
266	ckinId
267	);
268	while( db_step(&q4)==SQLITE_ROW ){
269	const char *zName = db_column_text(&q4,0);
270	int zNew = db_column_int(&q4,1);
271	int mPerm = db_column_int(&q4,2);
272	if( zNew==0)
273	printf("D %s\n", zName);
274	else if( bag_find(&blobs, zNew) )
275	printf("M %s :%d %s\n", mPerm ? "100755" : "100644", BLOBMARK(zNew), zName);
276	}
277	db_finalize(&q4);
278	db_finalize(&q3);
279	printf("\n");
280	}
281	db_finalize(&q2);
282	db_finalize(&q);
283	bag_clear(&blobs);
284	manifest_cache_clear();
285
286
	@@ -207,13 +301,34 @@
301	zEncoded = mprintf("%s", zTagname);
302	for(i=0; zEncoded[i]; i++){
303	if( !fossil_isalnum(zEncoded[i]) ) zEncoded[i] = '_';
304	}
305	printf("tag %s\n", zEncoded);
306	printf("from :%d\n", COMMITMARK(rid));
307	printf("tagger <tagger> %s +0000\n", zSecSince1970);
308	printf("data 0\n");
309	fossil_free(zEncoded);
310	}
311	db_finalize(&q);
312	bag_clear(&vers);
313
314	if( markfile_out!=0 ){
315	FILE *f;
316	f = fopen(markfile_out, "w");
317	if( f == 0 ){
318	fossil_panic("cannot open %s for writing", markfile_out);
319	}
320	db_prepare(&q, "SELECT rid FROM oldblob");
321	while( db_step(&q)==SQLITE_ROW ){
322	fprintf(f, "b%d\n", db_column_int(&q, 0));
323	}
324	db_finalize(&q);
325	db_prepare(&q, "SELECT rid FROM oldcommit");
326	while( db_step(&q)==SQLITE_ROW ){
327	fprintf(f, "c%d\n", db_column_int(&q, 0));
328	}
329	db_finalize(&q);
330	if( ferror(f)!=0 \|\| fclose(f)!=0 ) {
331	fossil_panic("error while writing %s", markfile_out);
332	}
333	}
334	}
335

M src/file.c

+32 -6

		--- src/file.c
		+++ src/file.c
		@@ -19,10 +19,12 @@
19	19	*/
20	20	#include "config.h"
21	21	#include <sys/types.h>
22	22	#include <sys/stat.h>
23	23	#include <unistd.h>
	24	+#include <string.h>
	25	+#include <errno.h>
24	26	#include "file.h"
25	27
26	28	/*
27	29	** The file status information from the most recent stat() call.
28	30	**
		@@ -397,11 +399,11 @@
397	399	char *zPwdUtf8;
398	400	int nPwd;
399	401	int i;
400	402	char zPwd[2000];
401	403	if( getcwd(zPwd, sizeof(zPwd)-1)==0 ){
402		- fossil_fatal("pwd too big: max %d\n", (int)sizeof(zPwd)-1);
	404	+ fossil_fatal("cannot find the current working directory.");
403	405	}
404	406	zPwdUtf8 = fossil_mbcs_to_utf8(zPwd);
405	407	nPwd = strlen(zPwdUtf8);
406	408	if( nPwd > nBuf-1 ){
407	409	fossil_fatal("pwd too big: max %d\n", nBuf-1);
		@@ -409,11 +411,16 @@
409	411	for(i=0; zPwdUtf8[i]; i++) if( zPwdUtf8[i]=='\\' ) zPwdUtf8[i] = '/';
410	412	memcpy(zBuf, zPwdUtf8, nPwd+1);
411	413	fossil_mbcs_free(zPwdUtf8);
412	414	#else
413	415	if( getcwd(zBuf, nBuf-1)==0 ){
414		- fossil_fatal("pwd too big: max %d\n", nBuf-1);
	416	+ if( errno==ERANGE ){
	417	+ fossil_fatal("pwd too big: max %d\n", nBuf-1);
	418	+ }else{
	419	+ fossil_fatal("cannot find current working directory; %s",
	420	+ strerror(errno));
	421	+ }
415	422	}
416	423	#endif
417	424	}
418	425
419	426	/*
		@@ -492,26 +499,45 @@
492	499	}
493	500	}
494	501	}
495	502	return 1;
496	503	}
	504	+
	505	+/*
	506	+** Return a pointer to the first character in a pathname past the
	507	+** drive letter. This routine is a no-op on unix.
	508	+*/
	509	+char file_without_drive_letter(char zIn){
	510	+#ifdef _WIN32
	511	+ if( fossil_isalpha(zIn[0]) && zIn[1]==':' ) zIn += 2;
	512	+#endif
	513	+ return zIn;
	514	+}
497	515
498	516	/*
499	517	** Compute a pathname for a file or directory that is relative
500	518	** to the current directory.
501	519	*/
502	520	void file_relative_name(const char zOrigName, Blob pOut){
503	521	char *zPath;
504	522	blob_set(pOut, zOrigName);
505	523	blob_resize(pOut, file_simplify_name(blob_buffer(pOut), blob_size(pOut)));
506		- zPath = blob_buffer(pOut);
	524	+ zPath = file_without_drive_letter(blob_buffer(pOut));
507	525	if( zPath[0]=='/' ){
508	526	int i, j;
509	527	Blob tmp;
510		- char zPwd[2000];
511		- file_getcwd(zPwd, sizeof(zPwd)-20);
512		- for(i=1; zPath[i] && zPwd[i]==zPath[i]; i++){}
	528	+ char *zPwd;
	529	+ char zBuf[2000];
	530	+ zPwd = zBuf;
	531	+ file_getcwd(zBuf, sizeof(zBuf)-20);
	532	+ zPwd = file_without_drive_letter(zBuf);
	533	+ i = 1;
	534	+#ifdef _WIN32
	535	+ while( zPath[i] && fossil_tolower(zPwd[i])==fossil_tolower(zPath[i]) ) i++;
	536	+#else
	537	+ while( zPath[i] && zPwd[i]==zPath[i] ) i++;
	538	+#endif
513	539	if( zPath[i]==0 ){
514	540	blob_reset(pOut);
515	541	if( zPwd[i]==0 ){
516	542	blob_append(pOut, ".", 1);
517	543	}else{
518	544

	--- src/file.c
	+++ src/file.c
	@@ -19,10 +19,12 @@
19	*/
20	#include "config.h"
21	#include <sys/types.h>
22	#include <sys/stat.h>
23	#include <unistd.h>


24	#include "file.h"
25
26	/*
27	** The file status information from the most recent stat() call.
28	**
	@@ -397,11 +399,11 @@
397	char *zPwdUtf8;
398	int nPwd;
399	int i;
400	char zPwd[2000];
401	if( getcwd(zPwd, sizeof(zPwd)-1)==0 ){
402	fossil_fatal("pwd too big: max %d\n", (int)sizeof(zPwd)-1);
403	}
404	zPwdUtf8 = fossil_mbcs_to_utf8(zPwd);
405	nPwd = strlen(zPwdUtf8);
406	if( nPwd > nBuf-1 ){
407	fossil_fatal("pwd too big: max %d\n", nBuf-1);
	@@ -409,11 +411,16 @@
409	for(i=0; zPwdUtf8[i]; i++) if( zPwdUtf8[i]=='\\' ) zPwdUtf8[i] = '/';
410	memcpy(zBuf, zPwdUtf8, nPwd+1);
411	fossil_mbcs_free(zPwdUtf8);
412	#else
413	if( getcwd(zBuf, nBuf-1)==0 ){
414	fossil_fatal("pwd too big: max %d\n", nBuf-1);





415	}
416	#endif
417	}
418
419	/*
	@@ -492,26 +499,45 @@
492	}
493	}
494	}
495	return 1;
496	}











497
498	/*
499	** Compute a pathname for a file or directory that is relative
500	** to the current directory.
501	*/
502	void file_relative_name(const char zOrigName, Blob pOut){
503	char *zPath;
504	blob_set(pOut, zOrigName);
505	blob_resize(pOut, file_simplify_name(blob_buffer(pOut), blob_size(pOut)));
506	zPath = blob_buffer(pOut);
507	if( zPath[0]=='/' ){
508	int i, j;
509	Blob tmp;
510	char zPwd[2000];
511	file_getcwd(zPwd, sizeof(zPwd)-20);
512	for(i=1; zPath[i] && zPwd[i]==zPath[i]; i++){}








513	if( zPath[i]==0 ){
514	blob_reset(pOut);
515	if( zPwd[i]==0 ){
516	blob_append(pOut, ".", 1);
517	}else{
518

	--- src/file.c
	+++ src/file.c
	@@ -19,10 +19,12 @@
19	*/
20	#include "config.h"
21	#include <sys/types.h>
22	#include <sys/stat.h>
23	#include <unistd.h>
24	#include <string.h>
25	#include <errno.h>
26	#include "file.h"
27
28	/*
29	** The file status information from the most recent stat() call.
30	**
	@@ -397,11 +399,11 @@
399	char *zPwdUtf8;
400	int nPwd;
401	int i;
402	char zPwd[2000];
403	if( getcwd(zPwd, sizeof(zPwd)-1)==0 ){
404	fossil_fatal("cannot find the current working directory.");
405	}
406	zPwdUtf8 = fossil_mbcs_to_utf8(zPwd);
407	nPwd = strlen(zPwdUtf8);
408	if( nPwd > nBuf-1 ){
409	fossil_fatal("pwd too big: max %d\n", nBuf-1);
	@@ -409,11 +411,16 @@
411	for(i=0; zPwdUtf8[i]; i++) if( zPwdUtf8[i]=='\\' ) zPwdUtf8[i] = '/';
412	memcpy(zBuf, zPwdUtf8, nPwd+1);
413	fossil_mbcs_free(zPwdUtf8);
414	#else
415	if( getcwd(zBuf, nBuf-1)==0 ){
416	if( errno==ERANGE ){
417	fossil_fatal("pwd too big: max %d\n", nBuf-1);
418	}else{
419	fossil_fatal("cannot find current working directory; %s",
420	strerror(errno));
421	}
422	}
423	#endif
424	}
425
426	/*
	@@ -492,26 +499,45 @@
499	}
500	}
501	}
502	return 1;
503	}
504
505	/*
506	** Return a pointer to the first character in a pathname past the
507	** drive letter. This routine is a no-op on unix.
508	*/
509	char file_without_drive_letter(char zIn){
510	#ifdef _WIN32
511	if( fossil_isalpha(zIn[0]) && zIn[1]==':' ) zIn += 2;
512	#endif
513	return zIn;
514	}
515
516	/*
517	** Compute a pathname for a file or directory that is relative
518	** to the current directory.
519	*/
520	void file_relative_name(const char zOrigName, Blob pOut){
521	char *zPath;
522	blob_set(pOut, zOrigName);
523	blob_resize(pOut, file_simplify_name(blob_buffer(pOut), blob_size(pOut)));
524	zPath = file_without_drive_letter(blob_buffer(pOut));
525	if( zPath[0]=='/' ){
526	int i, j;
527	Blob tmp;
528	char *zPwd;
529	char zBuf[2000];
530	zPwd = zBuf;
531	file_getcwd(zBuf, sizeof(zBuf)-20);
532	zPwd = file_without_drive_letter(zBuf);
533	i = 1;
534	#ifdef _WIN32
535	while( zPath[i] && fossil_tolower(zPwd[i])==fossil_tolower(zPath[i]) ) i++;
536	#else
537	while( zPath[i] && zPwd[i]==zPath[i] ) i++;
538	#endif
539	if( zPath[i]==0 ){
540	blob_reset(pOut);
541	if( zPwd[i]==0 ){
542	blob_append(pOut, ".", 1);
543	}else{
544

M src/finfo.c

		--- src/finfo.c
		+++ src/finfo.c
		@@ -193,10 +193,12 @@
193	193	** Additional query parameters:
194	194	**
195	195	** a=DATE Only show changes after DATE
196	196	** b=DATE Only show changes before DATE
197	197	** n=NUM Show the first NUM changes only
	198	+** brbg Background color by branch name
	199	+** ubg Background color by user name
198	200	*/
199	201	void finfo_page(void){
200	202	Stmt q;
201	203	const char *zFilename;
202	204	char zPrevDate[20];
		@@ -204,10 +206,12 @@
204	206	const char *zB;
205	207	int n;
206	208	Blob title;
207	209	Blob sql;
208	210	GraphContext *pGraph;
	211	+ int brBg = P("brbg")!=0;
	212	+ int uBg = P("ubg")!=0;
209	213
210	214	login_check_credentials();
211	215	if( !g.okRead ){ login_needed(); return; }
212	216	style_header("File History");
213	217	login_anonymous_available();
		@@ -268,10 +272,15 @@
268	272	int gidx;
269	273	char zTime[10];
270	274	char zShort[20];
271	275	char zShortCkin[20];
272	276	if( zBr==0 ) zBr = "trunk";
	277	+ if( uBg ){
	278	+ zBgClr = hash_color(zUser);
	279	+ }else if( brBg \|\| zBgClr==0 \|\| zBgClr[0]==0 ){
	280	+ zBgClr = strcmp(zBr,"trunk")==0 ? "white" : hash_color(zBr);
	281	+ }
273	282	gidx = graph_add_row(pGraph, frid, fpid>0 ? 1 : 0, &fpid, zBr, zBgClr, 0);
274	283	if( memcmp(zDate, zPrevDate, 10) ){
275	284	sqlite3_snprintf(sizeof(zPrevDate), zPrevDate, "%.10s", zDate);
276	285	@ <tr><td>
277	286	@ <div class="divider">%s(zPrevDate)</div>
278	287

	--- src/finfo.c
	+++ src/finfo.c
	@@ -193,10 +193,12 @@
193	** Additional query parameters:
194	**
195	** a=DATE Only show changes after DATE
196	** b=DATE Only show changes before DATE
197	** n=NUM Show the first NUM changes only


198	*/
199	void finfo_page(void){
200	Stmt q;
201	const char *zFilename;
202	char zPrevDate[20];
	@@ -204,10 +206,12 @@
204	const char *zB;
205	int n;
206	Blob title;
207	Blob sql;
208	GraphContext *pGraph;


209
210	login_check_credentials();
211	if( !g.okRead ){ login_needed(); return; }
212	style_header("File History");
213	login_anonymous_available();
	@@ -268,10 +272,15 @@
268	int gidx;
269	char zTime[10];
270	char zShort[20];
271	char zShortCkin[20];
272	if( zBr==0 ) zBr = "trunk";





273	gidx = graph_add_row(pGraph, frid, fpid>0 ? 1 : 0, &fpid, zBr, zBgClr, 0);
274	if( memcmp(zDate, zPrevDate, 10) ){
275	sqlite3_snprintf(sizeof(zPrevDate), zPrevDate, "%.10s", zDate);
276	@ <tr><td>
277	@ <div class="divider">%s(zPrevDate)</div>
278

	--- src/finfo.c
	+++ src/finfo.c
	@@ -193,10 +193,12 @@
193	** Additional query parameters:
194	**
195	** a=DATE Only show changes after DATE
196	** b=DATE Only show changes before DATE
197	** n=NUM Show the first NUM changes only
198	** brbg Background color by branch name
199	** ubg Background color by user name
200	*/
201	void finfo_page(void){
202	Stmt q;
203	const char *zFilename;
204	char zPrevDate[20];
	@@ -204,10 +206,12 @@
206	const char *zB;
207	int n;
208	Blob title;
209	Blob sql;
210	GraphContext *pGraph;
211	int brBg = P("brbg")!=0;
212	int uBg = P("ubg")!=0;
213
214	login_check_credentials();
215	if( !g.okRead ){ login_needed(); return; }
216	style_header("File History");
217	login_anonymous_available();
	@@ -268,10 +272,15 @@
272	int gidx;
273	char zTime[10];
274	char zShort[20];
275	char zShortCkin[20];
276	if( zBr==0 ) zBr = "trunk";
277	if( uBg ){
278	zBgClr = hash_color(zUser);
279	}else if( brBg \|\| zBgClr==0 \|\| zBgClr[0]==0 ){
280	zBgClr = strcmp(zBr,"trunk")==0 ? "white" : hash_color(zBr);
281	}
282	gidx = graph_add_row(pGraph, frid, fpid>0 ? 1 : 0, &fpid, zBr, zBgClr, 0);
283	if( memcmp(zDate, zPrevDate, 10) ){
284	sqlite3_snprintf(sizeof(zPrevDate), zPrevDate, "%.10s", zDate);
285	@ <tr><td>
286	@ <div class="divider">%s(zPrevDate)</div>
287

M src/glob.c

+4 -4

		--- src/glob.c
		+++ src/glob.c
		@@ -110,24 +110,24 @@
110	110	p = fossil_malloc( sizeof(*p) + nList+1 );
111	111	memset(p, 0, sizeof(*p));
112	112	z = (char*)&p[1];
113	113	memcpy(z, zPatternList, nList+1);
114	114	while( z[0] ){
115		- while( z[0]==',' \|\| z[0]==' ' ) z++; /* Skip leading spaces */
	115	+ while( z[0]==',' \|\| z[0]==' ' \|\| z[0]=='\n' \|\| z[0]=='\r' ) z++; /* Skip leading spaces and newlines */
116	116	if( z[0]=='\'' \|\| z[0]=='"' ){
117	117	delimiter = z[0];
118	118	z++;
119	119	}else{
120	120	delimiter = ',';
121	121	}
122	122	if( z[0]==0 ) break;
123	123	p->azPattern = fossil_realloc(p->azPattern, (p->nPattern+1)sizeof(char) );
124	124	p->azPattern[p->nPattern++] = z;
125		- for(i=0; z[i] && z[i]!=delimiter; i++){}
	125	+ for(i=0; z[i] && z[i]!=delimiter && z[i]!='\n' && z[i]!='\r'; i++){}
126	126	if( delimiter==',' ){
127		- /* Remove trailing spaces on a comma-delimited pattern */
128		- for(j=i; j>1 && z[j-1]==' '; j--){}
	127	+ /* Remove trailing spaces / newlines on a comma-delimited pattern */
	128	+ for(j=i; j>1 && (z[j-1]==' ' \|\| z[j-1]=='\n' \|\| z[j-1]=='\r'); j--){}
129	129	if( j<i ) z[j] = 0;
130	130	}
131	131	if( z[i]==0 ) break;
132	132	z[i] = 0;
133	133	z += i+1;
134	134

	--- src/glob.c
	+++ src/glob.c
	@@ -110,24 +110,24 @@
110	p = fossil_malloc( sizeof(*p) + nList+1 );
111	memset(p, 0, sizeof(*p));
112	z = (char*)&p[1];
113	memcpy(z, zPatternList, nList+1);
114	while( z[0] ){
115	while( z[0]==',' \|\| z[0]==' ' ) z++; /* Skip leading spaces */
116	if( z[0]=='\'' \|\| z[0]=='"' ){
117	delimiter = z[0];
118	z++;
119	}else{
120	delimiter = ',';
121	}
122	if( z[0]==0 ) break;
123	p->azPattern = fossil_realloc(p->azPattern, (p->nPattern+1)sizeof(char) );
124	p->azPattern[p->nPattern++] = z;
125	for(i=0; z[i] && z[i]!=delimiter; i++){}
126	if( delimiter==',' ){
127	/* Remove trailing spaces on a comma-delimited pattern */
128	for(j=i; j>1 && z[j-1]==' '; j--){}
129	if( j<i ) z[j] = 0;
130	}
131	if( z[i]==0 ) break;
132	z[i] = 0;
133	z += i+1;
134

	--- src/glob.c
	+++ src/glob.c
	@@ -110,24 +110,24 @@
110	p = fossil_malloc( sizeof(*p) + nList+1 );
111	memset(p, 0, sizeof(*p));
112	z = (char*)&p[1];
113	memcpy(z, zPatternList, nList+1);
114	while( z[0] ){
115	while( z[0]==',' \|\| z[0]==' ' \|\| z[0]=='\n' \|\| z[0]=='\r' ) z++; /* Skip leading spaces and newlines */
116	if( z[0]=='\'' \|\| z[0]=='"' ){
117	delimiter = z[0];
118	z++;
119	}else{
120	delimiter = ',';
121	}
122	if( z[0]==0 ) break;
123	p->azPattern = fossil_realloc(p->azPattern, (p->nPattern+1)sizeof(char) );
124	p->azPattern[p->nPattern++] = z;
125	for(i=0; z[i] && z[i]!=delimiter && z[i]!='\n' && z[i]!='\r'; i++){}
126	if( delimiter==',' ){
127	/* Remove trailing spaces / newlines on a comma-delimited pattern */
128	for(j=i; j>1 && (z[j-1]==' ' \|\| z[j-1]=='\n' \|\| z[j-1]=='\r'); j--){}
129	if( j<i ) z[j] = 0;
130	}
131	if( z[i]==0 ) break;
132	z[i] = 0;
133	z += i+1;
134

M src/http.c

+2 -1

		--- src/http.c
		+++ src/http.c
		@@ -110,11 +110,12 @@
110	110	blob_appendf(pHdr, "Authorization: Basic %s\r\n", zEncoded);
111	111	fossil_free(zEncoded);
112	112	fossil_free(zCredentials);
113	113	}
114	114	blob_appendf(pHdr, "Host: %s\r\n", g.urlHostname);
115		- blob_appendf(pHdr, "User-Agent: Fossil/" MANIFEST_VERSION "\r\n");
	115	+ blob_appendf(pHdr, "User-Agent: Fossil/" RELEASE_VERSION
	116	+ "-" MANIFEST_VERSION "\r\n");
116	117	if( g.fHttpTrace ){
117	118	blob_appendf(pHdr, "Content-Type: application/x-fossil-debug\r\n");
118	119	}else{
119	120	blob_appendf(pHdr, "Content-Type: application/x-fossil\r\n");
120	121	}
121	122

	--- src/http.c
	+++ src/http.c
	@@ -110,11 +110,12 @@
110	blob_appendf(pHdr, "Authorization: Basic %s\r\n", zEncoded);
111	fossil_free(zEncoded);
112	fossil_free(zCredentials);
113	}
114	blob_appendf(pHdr, "Host: %s\r\n", g.urlHostname);
115	blob_appendf(pHdr, "User-Agent: Fossil/" MANIFEST_VERSION "\r\n");

116	if( g.fHttpTrace ){
117	blob_appendf(pHdr, "Content-Type: application/x-fossil-debug\r\n");
118	}else{
119	blob_appendf(pHdr, "Content-Type: application/x-fossil\r\n");
120	}
121

	--- src/http.c
	+++ src/http.c
	@@ -110,11 +110,12 @@
110	blob_appendf(pHdr, "Authorization: Basic %s\r\n", zEncoded);
111	fossil_free(zEncoded);
112	fossil_free(zCredentials);
113	}
114	blob_appendf(pHdr, "Host: %s\r\n", g.urlHostname);
115	blob_appendf(pHdr, "User-Agent: Fossil/" RELEASE_VERSION
116	"-" MANIFEST_VERSION "\r\n");
117	if( g.fHttpTrace ){
118	blob_appendf(pHdr, "Content-Type: application/x-fossil-debug\r\n");
119	}else{
120	blob_appendf(pHdr, "Content-Type: application/x-fossil\r\n");
121	}
122

M src/http_ssl.c

+61 -1

		--- src/http_ssl.c
		+++ src/http_ssl.c
		@@ -78,23 +78,76 @@
78	78	** Return the current SSL error message
79	79	*/
80	80	const char *ssl_errmsg(void){
81	81	return sslErrMsg;
82	82	}
	83	+
	84	+/*
	85	+** When a server requests a client certificate that hasn't been provided,
	86	+** display a warning message explaining what to do next.
	87	+*/
	88	+static int ssl_client_cert_callback(SSL ssl, X509 x509, EVP_PKEY *pkey){
	89	+ fossil_warning("The remote server requested a client certificate for "
	90	+ "authentication. Specify the pathname to a file containing the PEM "
	91	+ "encoded certificate and private key with the --ssl-identity option "
	92	+ "or the ssl-identity setting.");
	93	+ return 0; /* no cert available */
	94	+}
83	95
84	96	/*
85	97	** Call this routine once before any other use of the SSL interface.
86	98	** This routine does initial configuration of the SSL module.
87	99	*/
88	100	void ssl_global_init(void){
	101	+ const char zCaSetting = 0, zCaFile = 0, *zCaDirectory = 0;
	102	+
89	103	if( sslIsInit==0 ){
90	104	SSL_library_init();
91	105	SSL_load_error_strings();
92	106	ERR_load_BIO_strings();
93	107	OpenSSL_add_all_algorithms();
94	108	sslCtx = SSL_CTX_new(SSLv23_client_method());
95		- X509_STORE_set_default_paths(SSL_CTX_get_cert_store(sslCtx));
	109	+
	110	+ /* Set up acceptable CA root certificates */
	111	+ zCaSetting = db_get("ssl-ca-location", 0);
	112	+ if( zCaSetting==0 \|\| zCaSetting[0]=='\0' ){
	113	+ /* CA location not specified, use platform's default certificate store */
	114	+ X509_STORE_set_default_paths(SSL_CTX_get_cert_store(sslCtx));
	115	+ }else{
	116	+ /* User has specified a CA location, make sure it exists and use it */
	117	+ switch( file_isdir(zCaSetting) ){
	118	+ case 0: { /* doesn't exist */
	119	+ fossil_fatal("ssl-ca-location is set to '%s', "
	120	+ "but is not a file or directory", zCaSetting);
	121	+ break;
	122	+ }
	123	+ case 1: { /* directory */
	124	+ zCaDirectory = zCaSetting;
	125	+ break;
	126	+ }
	127	+ case 2: { /* file */
	128	+ zCaFile = zCaSetting;
	129	+ break;
	130	+ }
	131	+ }
	132	+ if( SSL_CTX_load_verify_locations(sslCtx, zCaFile, zCaDirectory)==0 ){
	133	+ fossil_fatal("Failed to use CA root certificates from "
	134	+ "ssl-ca-location '%s'", zCaSetting);
	135	+ }
	136	+ }
	137	+
	138	+ /* Load client SSL identity, preferring the filename specified on the command line */
	139	+ const char *identityFile = ( g.zSSLIdentity!= 0) ? g.zSSLIdentity : db_get("ssl-identity", 0);
	140	+ if( identityFile!=0 && identityFile[0]!='\0' ){
	141	+ if( SSL_CTX_use_certificate_file(sslCtx, identityFile, SSL_FILETYPE_PEM)!= 1
	142	+ \|\| SSL_CTX_use_PrivateKey_file(sslCtx, identityFile, SSL_FILETYPE_PEM)!=1 ){
	143	+ fossil_fatal("Could not load SSL identity from %s", identityFile);
	144	+ }
	145	+ }
	146	+ /* Register a callback to tell the user what to do when the server asks for a cert */
	147	+ SSL_CTX_set_client_cert_cb(sslCtx, ssl_client_cert_callback);
	148	+
96	149	sslIsInit = 1;
97	150	}
98	151	}
99	152
100	153	/*
		@@ -204,10 +257,17 @@
204	257	if( hasSavedCertificate ){
205	258	warning = "WARNING: Certificate doesn't match the "
206	259	"saved certificate for this host!";
207	260	}
208	261	prompt = mprintf("\nUnknown SSL certificate:\n\n%s\n\n%s\n"
	262	+ "Either:\n"
	263	+ " * verify the certificate is correct using the "
	264	+ "SHA1 fingerprint above\n"
	265	+ " * use the global ssl-ca-location setting to specify your CA root\n"
	266	+ " certificates list\n\n"
	267	+ "If you are not expecting this message, answer no and "
	268	+ "contact your server\nadministrator.\n\n"
209	269	"Accept certificate [a=always/y/N]? ", desc, warning);
210	270	BIO_free(mem);
211	271
212	272	prompt_user(prompt, &ans);
213	273	free(prompt);
214	274

	--- src/http_ssl.c
	+++ src/http_ssl.c
	@@ -78,23 +78,76 @@
78	** Return the current SSL error message
79	*/
80	const char *ssl_errmsg(void){
81	return sslErrMsg;
82	}












83
84	/*
85	** Call this routine once before any other use of the SSL interface.
86	** This routine does initial configuration of the SSL module.
87	*/
88	void ssl_global_init(void){


89	if( sslIsInit==0 ){
90	SSL_library_init();
91	SSL_load_error_strings();
92	ERR_load_BIO_strings();
93	OpenSSL_add_all_algorithms();
94	sslCtx = SSL_CTX_new(SSLv23_client_method());
95	X509_STORE_set_default_paths(SSL_CTX_get_cert_store(sslCtx));







































96	sslIsInit = 1;
97	}
98	}
99
100	/*
	@@ -204,10 +257,17 @@
204	if( hasSavedCertificate ){
205	warning = "WARNING: Certificate doesn't match the "
206	"saved certificate for this host!";
207	}
208	prompt = mprintf("\nUnknown SSL certificate:\n\n%s\n\n%s\n"







209	"Accept certificate [a=always/y/N]? ", desc, warning);
210	BIO_free(mem);
211
212	prompt_user(prompt, &ans);
213	free(prompt);
214

	--- src/http_ssl.c
	+++ src/http_ssl.c
	@@ -78,23 +78,76 @@
78	** Return the current SSL error message
79	*/
80	const char *ssl_errmsg(void){
81	return sslErrMsg;
82	}
83
84	/*
85	** When a server requests a client certificate that hasn't been provided,
86	** display a warning message explaining what to do next.
87	*/
88	static int ssl_client_cert_callback(SSL ssl, X509 x509, EVP_PKEY *pkey){
89	fossil_warning("The remote server requested a client certificate for "
90	"authentication. Specify the pathname to a file containing the PEM "
91	"encoded certificate and private key with the --ssl-identity option "
92	"or the ssl-identity setting.");
93	return 0; /* no cert available */
94	}
95
96	/*
97	** Call this routine once before any other use of the SSL interface.
98	** This routine does initial configuration of the SSL module.
99	*/
100	void ssl_global_init(void){
101	const char zCaSetting = 0, zCaFile = 0, *zCaDirectory = 0;
102
103	if( sslIsInit==0 ){
104	SSL_library_init();
105	SSL_load_error_strings();
106	ERR_load_BIO_strings();
107	OpenSSL_add_all_algorithms();
108	sslCtx = SSL_CTX_new(SSLv23_client_method());
109
110	/* Set up acceptable CA root certificates */
111	zCaSetting = db_get("ssl-ca-location", 0);
112	if( zCaSetting==0 \|\| zCaSetting[0]=='\0' ){
113	/* CA location not specified, use platform's default certificate store */
114	X509_STORE_set_default_paths(SSL_CTX_get_cert_store(sslCtx));
115	}else{
116	/* User has specified a CA location, make sure it exists and use it */
117	switch( file_isdir(zCaSetting) ){
118	case 0: { /* doesn't exist */
119	fossil_fatal("ssl-ca-location is set to '%s', "
120	"but is not a file or directory", zCaSetting);
121	break;
122	}
123	case 1: { /* directory */
124	zCaDirectory = zCaSetting;
125	break;
126	}
127	case 2: { /* file */
128	zCaFile = zCaSetting;
129	break;
130	}
131	}
132	if( SSL_CTX_load_verify_locations(sslCtx, zCaFile, zCaDirectory)==0 ){
133	fossil_fatal("Failed to use CA root certificates from "
134	"ssl-ca-location '%s'", zCaSetting);
135	}
136	}
137
138	/* Load client SSL identity, preferring the filename specified on the command line */
139	const char *identityFile = ( g.zSSLIdentity!= 0) ? g.zSSLIdentity : db_get("ssl-identity", 0);
140	if( identityFile!=0 && identityFile[0]!='\0' ){
141	if( SSL_CTX_use_certificate_file(sslCtx, identityFile, SSL_FILETYPE_PEM)!= 1
142	\|\| SSL_CTX_use_PrivateKey_file(sslCtx, identityFile, SSL_FILETYPE_PEM)!=1 ){
143	fossil_fatal("Could not load SSL identity from %s", identityFile);
144	}
145	}
146	/* Register a callback to tell the user what to do when the server asks for a cert */
147	SSL_CTX_set_client_cert_cb(sslCtx, ssl_client_cert_callback);
148
149	sslIsInit = 1;
150	}
151	}
152
153	/*
	@@ -204,10 +257,17 @@
257	if( hasSavedCertificate ){
258	warning = "WARNING: Certificate doesn't match the "
259	"saved certificate for this host!";
260	}
261	prompt = mprintf("\nUnknown SSL certificate:\n\n%s\n\n%s\n"
262	"Either:\n"
263	" * verify the certificate is correct using the "
264	"SHA1 fingerprint above\n"
265	" * use the global ssl-ca-location setting to specify your CA root\n"
266	" certificates list\n\n"
267	"If you are not expecting this message, answer no and "
268	"contact your server\nadministrator.\n\n"
269	"Accept certificate [a=always/y/N]? ", desc, warning);
270	BIO_free(mem);
271
272	prompt_user(prompt, &ans);
273	free(prompt);
274

M src/info.c

+11 -5

		--- src/info.c
		+++ src/info.c
		@@ -1213,12 +1213,18 @@
1213	1213	}
1214	1214
1215	1215
1216	1216	/*
1217	1217	** WEBPAGE: artifact
1218		-** URL: /artifact?name=ARTIFACTID
	1218	+** URL: /artifact/ARTIFACTID
1219	1219	** URL: /artifact?ci=CHECKIN&filename=PATH
	1220	+**
	1221	+** Additional query parameters:
	1222	+**
	1223	+** ln - show line numbers
	1224	+** ln=N - highlight line number N
	1225	+** ln=M-N - highlight lines M through N inclusive
1220	1226	**
1221	1227	** Show the complete content of a file identified by ARTIFACTID
1222	1228	** as preformatted text.
1223	1229	*/
1224	1230	void artifact_page(void){
		@@ -1260,24 +1266,24 @@
1260	1266	zMime = mimetype_from_name(blob_str(&downloadName));
1261	1267	if( zMime ){
1262	1268	if( fossil_strcmp(zMime, "text/html")==0 ){
1263	1269	if( P("txt") ){
1264	1270	style_submenu_element("Html", "Html",
1265		- "%s/artifact?name=%s", g.zTop, zUuid);
	1271	+ "%s/artifact/%s", g.zTop, zUuid);
1266	1272	}else{
1267	1273	renderAsHtml = 1;
1268	1274	style_submenu_element("Text", "Text",
1269		- "%s/artifact?name=%s&txt=1", g.zTop, zUuid);
	1275	+ "%s/artifact/%s?txt=1", g.zTop, zUuid);
1270	1276	}
1271	1277	}else if( fossil_strcmp(zMime, "application/x-fossil-wiki")==0 ){
1272	1278	if( P("txt") ){
1273	1279	style_submenu_element("Wiki", "Wiki",
1274		- "%s/artifact?name=%s", g.zTop, zUuid);
	1280	+ "%s/artifact/%s", g.zTop, zUuid);
1275	1281	}else{
1276	1282	renderAsWiki = 1;
1277	1283	style_submenu_element("Text", "Text",
1278		- "%s/artifact?name=%s&txt=1", g.zTop, zUuid);
	1284	+ "%s/artifact/%s?txt=1", g.zTop, zUuid);
1279	1285	}
1280	1286	}
1281	1287	}
1282	1288	@ </p></blockquote>
1283	1289	@ <hr />
1284	1290

	--- src/info.c
	+++ src/info.c
	@@ -1213,12 +1213,18 @@
1213	}
1214
1215
1216	/*
1217	** WEBPAGE: artifact
1218	** URL: /artifact?name=ARTIFACTID
1219	** URL: /artifact?ci=CHECKIN&filename=PATH






1220	**
1221	** Show the complete content of a file identified by ARTIFACTID
1222	** as preformatted text.
1223	*/
1224	void artifact_page(void){
	@@ -1260,24 +1266,24 @@
1260	zMime = mimetype_from_name(blob_str(&downloadName));
1261	if( zMime ){
1262	if( fossil_strcmp(zMime, "text/html")==0 ){
1263	if( P("txt") ){
1264	style_submenu_element("Html", "Html",
1265	"%s/artifact?name=%s", g.zTop, zUuid);
1266	}else{
1267	renderAsHtml = 1;
1268	style_submenu_element("Text", "Text",
1269	"%s/artifact?name=%s&txt=1", g.zTop, zUuid);
1270	}
1271	}else if( fossil_strcmp(zMime, "application/x-fossil-wiki")==0 ){
1272	if( P("txt") ){
1273	style_submenu_element("Wiki", "Wiki",
1274	"%s/artifact?name=%s", g.zTop, zUuid);
1275	}else{
1276	renderAsWiki = 1;
1277	style_submenu_element("Text", "Text",
1278	"%s/artifact?name=%s&txt=1", g.zTop, zUuid);
1279	}
1280	}
1281	}
1282	@ </p></blockquote>
1283	@ <hr />
1284

	--- src/info.c
	+++ src/info.c
	@@ -1213,12 +1213,18 @@
1213	}
1214
1215
1216	/*
1217	** WEBPAGE: artifact
1218	** URL: /artifact/ARTIFACTID
1219	** URL: /artifact?ci=CHECKIN&filename=PATH
1220	**
1221	** Additional query parameters:
1222	**
1223	** ln - show line numbers
1224	** ln=N - highlight line number N
1225	** ln=M-N - highlight lines M through N inclusive
1226	**
1227	** Show the complete content of a file identified by ARTIFACTID
1228	** as preformatted text.
1229	*/
1230	void artifact_page(void){
	@@ -1260,24 +1266,24 @@
1266	zMime = mimetype_from_name(blob_str(&downloadName));
1267	if( zMime ){
1268	if( fossil_strcmp(zMime, "text/html")==0 ){
1269	if( P("txt") ){
1270	style_submenu_element("Html", "Html",
1271	"%s/artifact/%s", g.zTop, zUuid);
1272	}else{
1273	renderAsHtml = 1;
1274	style_submenu_element("Text", "Text",
1275	"%s/artifact/%s?txt=1", g.zTop, zUuid);
1276	}
1277	}else if( fossil_strcmp(zMime, "application/x-fossil-wiki")==0 ){
1278	if( P("txt") ){
1279	style_submenu_element("Wiki", "Wiki",
1280	"%s/artifact/%s", g.zTop, zUuid);
1281	}else{
1282	renderAsWiki = 1;
1283	style_submenu_element("Text", "Text",
1284	"%s/artifact/%s?txt=1", g.zTop, zUuid);
1285	}
1286	}
1287	}
1288	@ </p></blockquote>
1289	@ <hr />
1290

M src/main.c

+9 -3

		--- src/main.c
		+++ src/main.c
		@@ -105,10 +105,11 @@
105	105	char urlProxyAuth; / Proxy-Authorizer: string */
106	106	char urlFossil; / The path of the ?fossil=path suffix on ssh: */
107	107	int dontKeepUrl; /* Do not persist the URL */
108	108
109	109	const char zLogin; / Login name. "" if not logged in. */
	110	+ const char zSSLIdentity; / Value of --ssl-identity option, filename of SSL client identity */
110	111	int useLocalauth; /* No login required if from 127.0.0.1 */
111	112	int noPswd; /* Logged in without password (on 127.0.0.1) */
112	113	int userUid; /* Integer user id */
113	114
114	115	/* Information used to populate the RCVFROM table */
		@@ -251,10 +252,11 @@
251	252	g.fSystemTrace = find_option("systemtrace", 0, 0)!=0;
252	253	if( g.fSqlTrace ) g.fSqlStats = 1;
253	254	g.fSqlPrint = find_option("sqlprint", 0, 0)!=0;
254	255	g.fHttpTrace = find_option("httptrace", 0, 0)!=0;
255	256	g.zLogin = find_option("user", "U", 1);
	257	+ g.zSSLIdentity = find_option("ssl-identity", 0, 1);
256	258	if( find_option("help",0,0)!=0 ){
257	259	/* --help anywhere on the command line is translated into
258	260	** "fossil help argv[1] argv[2]..." */
259	261	int i;
260	262	char *zNewArgv = fossil_malloc( sizeof(char)*(g.argc+2) );
		@@ -644,11 +646,11 @@
644	646	** Usage: %fossil version
645	647	**
646	648	** Print the source code version number for the fossil executable.
647	649	*/
648	650	void version_cmd(void){
649		- fossil_print("This is fossil version "
	651	+ fossil_print("This is fossil version " RELEASE_VERSION " "
650	652	MANIFEST_VERSION " " MANIFEST_DATE " UTC\n");
651	653	}
652	654
653	655
654	656	/*
		@@ -1102,11 +1104,12 @@
1102	1104	if( blob_eq(&key, "HOME:") && blob_token(&line, &value) ){
1103	1105	cgi_setenv("HOME", blob_str(&value));
1104	1106	blob_reset(&value);
1105	1107	continue;
1106	1108	}
1107		- if( blob_eq(&key, "repository:") && blob_token(&line, &value) ){
	1109	+ if( blob_eq(&key, "repository:") && blob_tail(&line, &value) ){
	1110	+ blob_trim(&value);
1108	1111	db_open_repository(blob_str(&value));
1109	1112	blob_reset(&value);
1110	1113	continue;
1111	1114	}
1112	1115	if( blob_eq(&key, "directory:") && blob_token(&line, &value) ){
		@@ -1437,11 +1440,14 @@
1437	1440	if( isUiCmd ){
1438	1441	zBrowser = db_get("web-browser", "start");
1439	1442	zBrowserCmd = mprintf("%s http://127.0.0.1:%%d/", zBrowser);
1440	1443	}
1441	1444	db_close(1);
1442		- win32_http_server(iPort, mxPort, zBrowserCmd, zStopperFile, zNotFound, flags);
	1445	+ if( win32_http_service(iPort, zNotFound, flags) ){
	1446	+ win32_http_server(iPort, mxPort, zBrowserCmd,
	1447	+ zStopperFile, zNotFound, flags);
	1448	+ }
1443	1449	#endif
1444	1450	}
1445	1451
1446	1452	/*
1447	1453	** COMMAND: test-echo
1448	1454

	--- src/main.c
	+++ src/main.c
	@@ -105,10 +105,11 @@
105	char urlProxyAuth; / Proxy-Authorizer: string */
106	char urlFossil; / The path of the ?fossil=path suffix on ssh: */
107	int dontKeepUrl; /* Do not persist the URL */
108
109	const char zLogin; / Login name. "" if not logged in. */

110	int useLocalauth; /* No login required if from 127.0.0.1 */
111	int noPswd; /* Logged in without password (on 127.0.0.1) */
112	int userUid; /* Integer user id */
113
114	/* Information used to populate the RCVFROM table */
	@@ -251,10 +252,11 @@
251	g.fSystemTrace = find_option("systemtrace", 0, 0)!=0;
252	if( g.fSqlTrace ) g.fSqlStats = 1;
253	g.fSqlPrint = find_option("sqlprint", 0, 0)!=0;
254	g.fHttpTrace = find_option("httptrace", 0, 0)!=0;
255	g.zLogin = find_option("user", "U", 1);

256	if( find_option("help",0,0)!=0 ){
257	/* --help anywhere on the command line is translated into
258	** "fossil help argv[1] argv[2]..." */
259	int i;
260	char *zNewArgv = fossil_malloc( sizeof(char)*(g.argc+2) );
	@@ -644,11 +646,11 @@
644	** Usage: %fossil version
645	**
646	** Print the source code version number for the fossil executable.
647	*/
648	void version_cmd(void){
649	fossil_print("This is fossil version "
650	MANIFEST_VERSION " " MANIFEST_DATE " UTC\n");
651	}
652
653
654	/*
	@@ -1102,11 +1104,12 @@
1102	if( blob_eq(&key, "HOME:") && blob_token(&line, &value) ){
1103	cgi_setenv("HOME", blob_str(&value));
1104	blob_reset(&value);
1105	continue;
1106	}
1107	if( blob_eq(&key, "repository:") && blob_token(&line, &value) ){

1108	db_open_repository(blob_str(&value));
1109	blob_reset(&value);
1110	continue;
1111	}
1112	if( blob_eq(&key, "directory:") && blob_token(&line, &value) ){
	@@ -1437,11 +1440,14 @@
1437	if( isUiCmd ){
1438	zBrowser = db_get("web-browser", "start");
1439	zBrowserCmd = mprintf("%s http://127.0.0.1:%%d/", zBrowser);
1440	}
1441	db_close(1);
1442	win32_http_server(iPort, mxPort, zBrowserCmd, zStopperFile, zNotFound, flags);



1443	#endif
1444	}
1445
1446	/*
1447	** COMMAND: test-echo
1448

	--- src/main.c
	+++ src/main.c
	@@ -105,10 +105,11 @@
105	char urlProxyAuth; / Proxy-Authorizer: string */
106	char urlFossil; / The path of the ?fossil=path suffix on ssh: */
107	int dontKeepUrl; /* Do not persist the URL */
108
109	const char zLogin; / Login name. "" if not logged in. */
110	const char zSSLIdentity; / Value of --ssl-identity option, filename of SSL client identity */
111	int useLocalauth; /* No login required if from 127.0.0.1 */
112	int noPswd; /* Logged in without password (on 127.0.0.1) */
113	int userUid; /* Integer user id */
114
115	/* Information used to populate the RCVFROM table */
	@@ -251,10 +252,11 @@
252	g.fSystemTrace = find_option("systemtrace", 0, 0)!=0;
253	if( g.fSqlTrace ) g.fSqlStats = 1;
254	g.fSqlPrint = find_option("sqlprint", 0, 0)!=0;
255	g.fHttpTrace = find_option("httptrace", 0, 0)!=0;
256	g.zLogin = find_option("user", "U", 1);
257	g.zSSLIdentity = find_option("ssl-identity", 0, 1);
258	if( find_option("help",0,0)!=0 ){
259	/* --help anywhere on the command line is translated into
260	** "fossil help argv[1] argv[2]..." */
261	int i;
262	char *zNewArgv = fossil_malloc( sizeof(char)*(g.argc+2) );
	@@ -644,11 +646,11 @@
646	** Usage: %fossil version
647	**
648	** Print the source code version number for the fossil executable.
649	*/
650	void version_cmd(void){
651	fossil_print("This is fossil version " RELEASE_VERSION " "
652	MANIFEST_VERSION " " MANIFEST_DATE " UTC\n");
653	}
654
655
656	/*
	@@ -1102,11 +1104,12 @@
1104	if( blob_eq(&key, "HOME:") && blob_token(&line, &value) ){
1105	cgi_setenv("HOME", blob_str(&value));
1106	blob_reset(&value);
1107	continue;
1108	}
1109	if( blob_eq(&key, "repository:") && blob_tail(&line, &value) ){
1110	blob_trim(&value);
1111	db_open_repository(blob_str(&value));
1112	blob_reset(&value);
1113	continue;
1114	}
1115	if( blob_eq(&key, "directory:") && blob_token(&line, &value) ){
	@@ -1437,11 +1440,14 @@
1440	if( isUiCmd ){
1441	zBrowser = db_get("web-browser", "start");
1442	zBrowserCmd = mprintf("%s http://127.0.0.1:%%d/", zBrowser);
1443	}
1444	db_close(1);
1445	if( win32_http_service(iPort, zNotFound, flags) ){
1446	win32_http_server(iPort, mxPort, zBrowserCmd,
1447	zStopperFile, zNotFound, flags);
1448	}
1449	#endif
1450	}
1451
1452	/*
1453	** COMMAND: test-echo
1454

M src/main.mk

+16 -7

		--- src/main.mk
		+++ src/main.mk
		@@ -281,22 +281,31 @@
281	281	$(BCC) -o $(OBJDIR)/makeheaders $(SRCDIR)/makeheaders.c
282	282
283	283	$(OBJDIR)/mkindex: $(SRCDIR)/mkindex.c
284	284	$(BCC) -o $(OBJDIR)/mkindex $(SRCDIR)/mkindex.c
285	285
	286	+$(OBJDIR)/mkversion: $(SRCDIR)/mkversion.c
	287	+ $(BCC) -o $(OBJDIR)/mkversion $(SRCDIR)/mkversion.c
	288	+
286	289	# WARNING. DANGER. Running the testsuite modifies the repository the
287	290	# build is done from, i.e. the checkout belongs to. Do not sync/push
288	291	# the repository after running the tests.
289	292	test: $(APPNAME)
290	293	$(TCLSH) test/tester.tcl $(APPNAME)
291	294
292		-$(OBJDIR)/VERSION.h: $(SRCDIR)/../manifest.uuid $(SRCDIR)/../manifest
293		- awk '{ printf "#define MANIFEST_UUID \"%s\"\n", $$1}' $(SRCDIR)/../manifest.uuid >$(OBJDIR)/VERSION.h
294		- awk '{ printf "#define MANIFEST_VERSION \"[%.10s]\"\n", $$1}' $(SRCDIR)/../manifest.uuid >>$(OBJDIR)/VERSION.h
295		- awk '$$1=="D"{printf "#define MANIFEST_DATE \"%s %s\"\n", substr($$2,1,10),substr($$2,12,8)}' $(SRCDIR)/../manifest >>$(OBJDIR)/VERSION.h
	295	+$(OBJDIR)/VERSION.h: $(SRCDIR)/../manifest.uuid $(SRCDIR)/../manifest $(SRCDIR)/../VERSION $(OBJDIR)/mkversion
	296	+ $(OBJDIR)/mkversion $(SRCDIR)/../manifest.uuid $(SRCDIR)/../manifest $(SRCDIR)/../VERSION >$(OBJDIR)/VERSION.h
296	297
297		-EXTRAOBJ = $(OBJDIR)/sqlite3.o $(OBJDIR)/shell.o $(OBJDIR)/th.o $(OBJDIR)/th_lang.o
	298	+# The USE_SYSTEM_SQLITE variable may be undefined, set to 0, or set
	299	+# to 1. If it is set to 1, then there is no need to build or link
	300	+# the sqlite3.o object. Instead, the system sqlite will be linked
	301	+# using -lsqlite3.
	302	+SQLITE3_OBJ.1 =
	303	+SQLITE3_OBJ.0 = $(OBJDIR)/sqlite3.o
	304	+SQLITE3_OBJ. = $(SQLITE3_OBJ.0)
	305	+
	306	+EXTRAOBJ = $(SQLITE3_OBJ.$(USE_SYSTEM_SQLITE)) $(OBJDIR)/shell.o $(OBJDIR)/th.o $(OBJDIR)/th_lang.o
298	307
299	308	$(APPNAME): $(OBJDIR)/headers $(OBJ) $(EXTRAOBJ)
300	309	$(TCC) -o $(APPNAME) $(OBJ) $(EXTRAOBJ) $(LIB)
301	310
302	311	# This rule prevents make from using its default rules to try build
		@@ -889,16 +898,16 @@
889	898	$(OBJDIR)/zip.o: $(OBJDIR)/zip_.c $(OBJDIR)/zip.h $(SRCDIR)/config.h
890	899	$(XTCC) -o $(OBJDIR)/zip.o -c $(OBJDIR)/zip_.c
891	900
892	901	$(OBJDIR)/zip.h: $(OBJDIR)/headers
893	902	$(OBJDIR)/sqlite3.o: $(SRCDIR)/sqlite3.c
894		- $(XTCC) -DSQLITE_OMIT_LOAD_EXTENSION=1 -DSQLITE_THREADSAFE=0 -DSQLITE_DEFAULT_FILE_FORMAT=4 -DSQLITE_ENABLE_STAT2 -Dlocaltime=fossil_localtime -DSQLITE_ENABLE_LOCKING_STYLE=0 -c $(SRCDIR)/sqlite3.c -o $(OBJDIR)/sqlite3.o
	903	+ $(XTCC) -DSQLITE_OMIT_LOAD_EXTENSION=1 -DSQLITE_THREADSAFE=0 -DSQLITE_DEFAULT_FILE_FORMAT=4 -DSQLITE_ENABLE_STAT3 -Dlocaltime=fossil_localtime -DSQLITE_ENABLE_LOCKING_STYLE=0 -c $(SRCDIR)/sqlite3.c -o $(OBJDIR)/sqlite3.o
895	904
896		-$(OBJDIR)/shell.o: $(SRCDIR)/shell.c
	905	+$(OBJDIR)/shell.o: $(SRCDIR)/shell.c $(SRCDIR)/sqlite3.h
897	906	$(XTCC) -Dmain=sqlite3_shell -DSQLITE_OMIT_LOAD_EXTENSION=1 -c $(SRCDIR)/shell.c -o $(OBJDIR)/shell.o
898	907
899	908	$(OBJDIR)/th.o: $(SRCDIR)/th.c
900	909	$(XTCC) -I$(SRCDIR) -c $(SRCDIR)/th.c -o $(OBJDIR)/th.o
901	910
902	911	$(OBJDIR)/th_lang.o: $(SRCDIR)/th_lang.c
903	912	$(XTCC) -I$(SRCDIR) -c $(SRCDIR)/th_lang.c -o $(OBJDIR)/th_lang.o
904	913
905	914

	--- src/main.mk
	+++ src/main.mk
	@@ -281,22 +281,31 @@
281	$(BCC) -o $(OBJDIR)/makeheaders $(SRCDIR)/makeheaders.c
282
283	$(OBJDIR)/mkindex: $(SRCDIR)/mkindex.c
284	$(BCC) -o $(OBJDIR)/mkindex $(SRCDIR)/mkindex.c
285



286	# WARNING. DANGER. Running the testsuite modifies the repository the
287	# build is done from, i.e. the checkout belongs to. Do not sync/push
288	# the repository after running the tests.
289	test: $(APPNAME)
290	$(TCLSH) test/tester.tcl $(APPNAME)
291
292	$(OBJDIR)/VERSION.h: $(SRCDIR)/../manifest.uuid $(SRCDIR)/../manifest
293	awk '{ printf "#define MANIFEST_UUID \"%s\"\n", $$1}' $(SRCDIR)/../manifest.uuid >$(OBJDIR)/VERSION.h
294	awk '{ printf "#define MANIFEST_VERSION \"[%.10s]\"\n", $$1}' $(SRCDIR)/../manifest.uuid >>$(OBJDIR)/VERSION.h
295	awk '$$1=="D"{printf "#define MANIFEST_DATE \"%s %s\"\n", substr($$2,1,10),substr($$2,12,8)}' $(SRCDIR)/../manifest >>$(OBJDIR)/VERSION.h
296
297	EXTRAOBJ = $(OBJDIR)/sqlite3.o $(OBJDIR)/shell.o $(OBJDIR)/th.o $(OBJDIR)/th_lang.o








298
299	$(APPNAME): $(OBJDIR)/headers $(OBJ) $(EXTRAOBJ)
300	$(TCC) -o $(APPNAME) $(OBJ) $(EXTRAOBJ) $(LIB)
301
302	# This rule prevents make from using its default rules to try build
	@@ -889,16 +898,16 @@
889	$(OBJDIR)/zip.o: $(OBJDIR)/zip_.c $(OBJDIR)/zip.h $(SRCDIR)/config.h
890	$(XTCC) -o $(OBJDIR)/zip.o -c $(OBJDIR)/zip_.c
891
892	$(OBJDIR)/zip.h: $(OBJDIR)/headers
893	$(OBJDIR)/sqlite3.o: $(SRCDIR)/sqlite3.c
894	$(XTCC) -DSQLITE_OMIT_LOAD_EXTENSION=1 -DSQLITE_THREADSAFE=0 -DSQLITE_DEFAULT_FILE_FORMAT=4 -DSQLITE_ENABLE_STAT2 -Dlocaltime=fossil_localtime -DSQLITE_ENABLE_LOCKING_STYLE=0 -c $(SRCDIR)/sqlite3.c -o $(OBJDIR)/sqlite3.o
895
896	$(OBJDIR)/shell.o: $(SRCDIR)/shell.c
897	$(XTCC) -Dmain=sqlite3_shell -DSQLITE_OMIT_LOAD_EXTENSION=1 -c $(SRCDIR)/shell.c -o $(OBJDIR)/shell.o
898
899	$(OBJDIR)/th.o: $(SRCDIR)/th.c
900	$(XTCC) -I$(SRCDIR) -c $(SRCDIR)/th.c -o $(OBJDIR)/th.o
901
902	$(OBJDIR)/th_lang.o: $(SRCDIR)/th_lang.c
903	$(XTCC) -I$(SRCDIR) -c $(SRCDIR)/th_lang.c -o $(OBJDIR)/th_lang.o
904
905

	--- src/main.mk
	+++ src/main.mk
	@@ -281,22 +281,31 @@
281	$(BCC) -o $(OBJDIR)/makeheaders $(SRCDIR)/makeheaders.c
282
283	$(OBJDIR)/mkindex: $(SRCDIR)/mkindex.c
284	$(BCC) -o $(OBJDIR)/mkindex $(SRCDIR)/mkindex.c
285
286	$(OBJDIR)/mkversion: $(SRCDIR)/mkversion.c
287	$(BCC) -o $(OBJDIR)/mkversion $(SRCDIR)/mkversion.c
288
289	# WARNING. DANGER. Running the testsuite modifies the repository the
290	# build is done from, i.e. the checkout belongs to. Do not sync/push
291	# the repository after running the tests.
292	test: $(APPNAME)
293	$(TCLSH) test/tester.tcl $(APPNAME)
294
295	$(OBJDIR)/VERSION.h: $(SRCDIR)/../manifest.uuid $(SRCDIR)/../manifest $(SRCDIR)/../VERSION $(OBJDIR)/mkversion
296	$(OBJDIR)/mkversion $(SRCDIR)/../manifest.uuid $(SRCDIR)/../manifest $(SRCDIR)/../VERSION >$(OBJDIR)/VERSION.h


297
298	# The USE_SYSTEM_SQLITE variable may be undefined, set to 0, or set
299	# to 1. If it is set to 1, then there is no need to build or link
300	# the sqlite3.o object. Instead, the system sqlite will be linked
301	# using -lsqlite3.
302	SQLITE3_OBJ.1 =
303	SQLITE3_OBJ.0 = $(OBJDIR)/sqlite3.o
304	SQLITE3_OBJ. = $(SQLITE3_OBJ.0)
305
306	EXTRAOBJ = $(SQLITE3_OBJ.$(USE_SYSTEM_SQLITE)) $(OBJDIR)/shell.o $(OBJDIR)/th.o $(OBJDIR)/th_lang.o
307
308	$(APPNAME): $(OBJDIR)/headers $(OBJ) $(EXTRAOBJ)
309	$(TCC) -o $(APPNAME) $(OBJ) $(EXTRAOBJ) $(LIB)
310
311	# This rule prevents make from using its default rules to try build
	@@ -889,16 +898,16 @@
898	$(OBJDIR)/zip.o: $(OBJDIR)/zip_.c $(OBJDIR)/zip.h $(SRCDIR)/config.h
899	$(XTCC) -o $(OBJDIR)/zip.o -c $(OBJDIR)/zip_.c
900
901	$(OBJDIR)/zip.h: $(OBJDIR)/headers
902	$(OBJDIR)/sqlite3.o: $(SRCDIR)/sqlite3.c
903	$(XTCC) -DSQLITE_OMIT_LOAD_EXTENSION=1 -DSQLITE_THREADSAFE=0 -DSQLITE_DEFAULT_FILE_FORMAT=4 -DSQLITE_ENABLE_STAT3 -Dlocaltime=fossil_localtime -DSQLITE_ENABLE_LOCKING_STYLE=0 -c $(SRCDIR)/sqlite3.c -o $(OBJDIR)/sqlite3.o
904
905	$(OBJDIR)/shell.o: $(SRCDIR)/shell.c $(SRCDIR)/sqlite3.h
906	$(XTCC) -Dmain=sqlite3_shell -DSQLITE_OMIT_LOAD_EXTENSION=1 -c $(SRCDIR)/shell.c -o $(OBJDIR)/shell.o
907
908	$(OBJDIR)/th.o: $(SRCDIR)/th.c
909	$(XTCC) -I$(SRCDIR) -c $(SRCDIR)/th.c -o $(OBJDIR)/th.o
910
911	$(OBJDIR)/th_lang.o: $(SRCDIR)/th_lang.c
912	$(XTCC) -I$(SRCDIR) -c $(SRCDIR)/th_lang.c -o $(OBJDIR)/th_lang.o
913
914

M src/makemake.tcl

+32 -25

		--- src/makemake.tcl
		+++ src/makemake.tcl
		@@ -176,27 +176,34 @@
176	176	$(BCC) -o $(OBJDIR)/makeheaders $(SRCDIR)/makeheaders.c
177	177
178	178	$(OBJDIR)/mkindex: $(SRCDIR)/mkindex.c
179	179	$(BCC) -o $(OBJDIR)/mkindex $(SRCDIR)/mkindex.c
180	180
	181	+$(OBJDIR)/mkversion: $(SRCDIR)/mkversion.c
	182	+ $(BCC) -o $(OBJDIR)/mkversion $(SRCDIR)/mkversion.c
	183	+
181	184	# WARNING. DANGER. Running the testsuite modifies the repository the
182	185	# build is done from, i.e. the checkout belongs to. Do not sync/push
183	186	# the repository after running the tests.
184	187	test: $(APPNAME)
185	188	$(TCLSH) test/tester.tcl $(APPNAME)
186	189
187		-$(OBJDIR)/VERSION.h: $(SRCDIR)/../manifest.uuid $(SRCDIR)/../manifest
188		- awk '{ printf "#define MANIFEST_UUID \"%s\"\n", $$1}' \
189		- $(SRCDIR)/../manifest.uuid >$(OBJDIR)/VERSION.h
190		- awk '{ printf "#define MANIFEST_VERSION \"[%.10s]\"\n", $$1}' \
191		- $(SRCDIR)/../manifest.uuid >>$(OBJDIR)/VERSION.h
192		- awk '$$1=="D"{printf "#define MANIFEST_DATE \"%s %s\"\n",\
193		- substr($$2,1,10),substr($$2,12,8)}' \
194		- $(SRCDIR)/../manifest >>$(OBJDIR)/VERSION.h
	190	+$(OBJDIR)/VERSION.h: $(SRCDIR)/../manifest.uuid $(SRCDIR)/../manifest $(SRCDIR)/../VERSION $(OBJDIR)/mkversion
	191	+ $(OBJDIR)/mkversion $(SRCDIR)/../manifest.uuid \
	192	+ $(SRCDIR)/../manifest \
	193	+ $(SRCDIR)/../VERSION >$(OBJDIR)/VERSION.h
	194	+
	195	+# The USE_SYSTEM_SQLITE variable may be undefined, set to 0, or set
	196	+# to 1. If it is set to 1, then there is no need to build or link
	197	+# the sqlite3.o object. Instead, the system sqlite will be linked
	198	+# using -lsqlite3.
	199	+SQLITE3_OBJ.1 =
	200	+SQLITE3_OBJ.0 = $(OBJDIR)/sqlite3.o
	201	+SQLITE3_OBJ. = $(SQLITE3_OBJ.0)
195	202
196	203	EXTRAOBJ = \
197		- $(OBJDIR)/sqlite3.o \
	204	+ $(SQLITE3_OBJ.$(USE_SYSTEM_SQLITE)) \
198	205	$(OBJDIR)/shell.o \
199	206	$(OBJDIR)/th.o \
200	207	$(OBJDIR)/th_lang.o
201	208
202	209	$(APPNAME): $(OBJDIR)/headers $(OBJ) $(EXTRAOBJ)
		@@ -241,17 +248,17 @@
241	248
242	249	writeln "\$(OBJDIR)/sqlite3.o:\t\$(SRCDIR)/sqlite3.c"
243	250	set opt {-DSQLITE_OMIT_LOAD_EXTENSION=1}
244	251	append opt " -DSQLITE_THREADSAFE=0 -DSQLITE_DEFAULT_FILE_FORMAT=4"
245	252	#append opt " -DSQLITE_ENABLE_FTS3=1"
246		-append opt " -DSQLITE_ENABLE_STAT2"
	253	+append opt " -DSQLITE_ENABLE_STAT3"
247	254	append opt " -Dlocaltime=fossil_localtime"
248	255	append opt " -DSQLITE_ENABLE_LOCKING_STYLE=0"
249	256	set SQLITE_OPTIONS $opt
250	257	writeln "\t\$(XTCC) $opt -c \$(SRCDIR)/sqlite3.c -o \$(OBJDIR)/sqlite3.o\n"
251	258
252		-writeln "\$(OBJDIR)/shell.o:\t\$(SRCDIR)/shell.c"
	259	+writeln "\$(OBJDIR)/shell.o:\t\$(SRCDIR)/shell.c \$(SRCDIR)/sqlite3.h"
253	260	set opt {-Dmain=sqlite3_shell}
254	261	append opt " -DSQLITE_OMIT_LOAD_EXTENSION=1"
255	262	writeln "\t\$(XTCC) $opt -c \$(SRCDIR)/shell.c -o \$(OBJDIR)/shell.o\n"
256	263
257	264	writeln "\$(OBJDIR)/th.o:\t\$(SRCDIR)/th.c"
		@@ -389,21 +396,21 @@
389	396	$(BCC) -o $(OBJDIR)/makeheaders $(SRCDIR)/makeheaders.c
390	397
391	398	$(OBJDIR)/mkindex: $(SRCDIR)/mkindex.c
392	399	$(BCC) -o $(OBJDIR)/mkindex $(SRCDIR)/mkindex.c
393	400
394		-$(VERSION): $(SRCDIR)/../win/version.c
395		- $(BCC) -o $(OBJDIR)/version $(SRCDIR)/../win/version.c
	401	+$(VERSION): $(SRCDIR)/mkversion.c
	402	+ $(BCC) -o $(OBJDIR)/version $(SRCDIR)/mkversion.c
396	403
397	404	# WARNING. DANGER. Running the testsuite modifies the repository the
398	405	# build is done from, i.e. the checkout belongs to. Do not sync/push
399	406	# the repository after running the tests.
400	407	test: $(APPNAME)
401	408	$(TCLSH) test/tester.tcl $(APPNAME)
402	409
403	410	$(OBJDIR)/VERSION.h: $(SRCDIR)/../manifest.uuid $(SRCDIR)/../manifest $(VERSION)
404		- $(VERSION) $(SRCDIR)/../manifest.uuid $(SRCDIR)/../manifest >$(OBJDIR)/VERSION.h
	411	+ $(VERSION) $(SRCDIR)/../manifest.uuid $(SRCDIR)/../manifest $(SRCDIR)/../VERSION >$(OBJDIR)/VERSION.h
405	412
406	413	EXTRAOBJ = \
407	414	$(OBJDIR)/sqlite3.o \
408	415	$(OBJDIR)/shell.o \
409	416	$(OBJDIR)/th.o \
		@@ -459,11 +466,11 @@
459	466
460	467	writeln "\$(OBJDIR)/sqlite3.o:\t\$(SRCDIR)/sqlite3.c"
461	468	set opt $SQLITE_OPTIONS
462	469	writeln "\t\$(XTCC) $opt -c \$(SRCDIR)/sqlite3.c -o \$(OBJDIR)/sqlite3.o\n"
463	470
464		-writeln "\$(OBJDIR)/shell.o:\t\$(SRCDIR)/shell.c"
	471	+writeln "\$(OBJDIR)/shell.o:\t\$(SRCDIR)/shell.c \$(SRCDIR)/sqlite3.h"
465	472	set opt {-Dmain=sqlite3_shell}
466	473	append opt " -DSQLITE_OMIT_LOAD_EXTENSION=1"
467	474	writeln "\t\$(XTCC) $opt -c \$(SRCDIR)/shell.c -o \$(OBJDIR)/shell.o\n"
468	475
469	476	writeln "\$(OBJDIR)/th.o:\t\$(SRCDIR)/th.c"
		@@ -556,11 +563,11 @@
556	563	$(BCC) -o$@ $**
557	564
558	565	mkindex$E: $(SRCDIR)\mkindex.c
559	566	$(BCC) -o$@ $**
560	567
561		-version$E: $B\win\version.c
	568	+version$E: $B\src\mkversion.c
562	569	$(BCC) -o$@ $**
563	570
564	571	$(OBJDIR)\shell$O : $(SRCDIR)\shell.c
565	572	$(TCC) -o$@ -c -Dmain=sqlite3_shell $(SQLITE_OPTIONS) $**
566	573
		@@ -571,11 +578,11 @@
571	578	$(TCC) -o$@ -c $**
572	579
573	580	$(OBJDIR)\th_lang$O : $(SRCDIR)\th_lang.c
574	581	$(TCC) -o$@ -c $**
575	582
576		-VERSION.h : version$E $B\manifest.uuid $B\manifest
	583	+VERSION.h : version$E $B\manifest.uuid $B\manifest $B\VERSION
577	584	+$** > $@
578	585
579	586	page_index.h: mkindex$E $(SRC)
580	587	+$** > $@
581	588
		@@ -582,11 +589,11 @@
582	589	clean:
583	590	-del $(OBJDIR)\*.obj
584	591	-del .obj _.c .h .map
585	592
586	593	realclean:
587		- -del $(APPNAME) translate$E mkindex$E makeheaders$E version$E
	594	+ -del $(APPNAME) translate$E mkindex$E makeheaders$E mkversion$E
588	595
589	596	}
590	597	foreach s [lsort $src] {
591	598	writeln "\$(OBJDIR)\\$s\$O : ${s}_.c ${s}.h"
592	599	writeln "\t\$(TCC) -o\$@ -c ${s}_.c\n"
		@@ -646,11 +653,11 @@
646	653	INCL = -I. -I$(SRCDIR) -I$B\win\include -I$(MSCDIR)\extra\include -I$(ZINCDIR)
647	654
648	655	CFLAGS = -nologo -MT -O2
649	656	BCC = $(CC) $(CFLAGS)
650	657	TCC = $(CC) -c $(CFLAGS) $(MSCDEF) $(SSL) $(INCL)
651		-LIBS = $(ZLIB) ws2_32.lib $(SSLLIB)
	658	+LIBS = $(ZLIB) ws2_32.lib advapi32.lib $(SSLLIB)
652	659	LIBDIR = -LIBPATH:$(MSCDIR)\extra\lib -LIBPATH:$(ZLIBDIR)
653	660	}
654	661	regsub -all {[-]D} $SQLITE_OPTIONS {/D} MSC_SQLITE_OPTIONS
655	662	writeln "SQLITE_OPTIONS = $MSC_SQLITE_OPTIONS\n"
656	663	writeln -nonewline "SRC = "
		@@ -693,11 +700,11 @@
693	700	$(BCC) $**
694	701
695	702	mkindex$E: $(SRCDIR)\mkindex.c
696	703	$(BCC) $**
697	704
698		-version$E: $B\win\version.c
	705	+mkversion$E: $B\src\mkversion.c
699	706	$(BCC) $**
700	707
701	708	$(OX)\shell$O : $(SRCDIR)\shell.c
702	709	$(TCC) /Fo$@ /Dmain=sqlite3_shell $(SQLITE_OPTIONS) -c $(SRCDIR)\shell.c
703	710
		@@ -708,11 +715,11 @@
708	715	$(TCC) /Fo$@ -c $**
709	716
710	717	$(OX)\th_lang$O : $(SRCDIR)\th_lang.c
711	718	$(TCC) /Fo$@ -c $**
712	719
713		-VERSION.h : version$E $B\manifest.uuid $B\manifest
	720	+VERSION.h : mkversion$E $B\manifest.uuid $B\manifest $B\VERSION
714	721	$** > $@
715	722
716	723	page_index.h: mkindex$E $(SRC)
717	724	$** > $@
718	725
		@@ -720,11 +727,11 @@
720	727	-del $(OX)\*.obj
721	728	-del .obj _.c .h .map
722	729	-del headers linkopts
723	730
724	731	realclean:
725		- -del $(APPNAME) translate$E mkindex$E makeheaders$E version$E
	732	+ -del $(APPNAME) translate$E mkindex$E makeheaders$E mkversion$E
726	733
727	734	}
728	735	foreach s [lsort $src] {
729	736	writeln "\$(OX)\\$s\$O : ${s}_.c ${s}.h"
730	737	writeln "\t\$(TCC) /Fo\$@ -c ${s}_.c\n"
		@@ -880,11 +887,11 @@
880	887	# compiling standard fossil utils
881	888	$(UTILS_OBJ): %.obj: $(SRCDIR)%.c
882	889	$(CC) $(CCFLAGS) $(INCLUDE) "$<" -Fo"$@"
883	890
884	891	# compile special windows utils
885		-version.obj: $(WINDIR)version.c
	892	+version.obj: $(SRCDIR)mkversion.c
886	893	$(CC) $(CCFLAGS) $(INCLUDE) "$<" -Fo"$@"
887	894
888	895	# generate the translated c-source files
889	896	$(TRANSLATEDSRC): %_.c: $(SRCDIR)%.c translate.exe
890	897	translate.exe $< >$@
		@@ -892,12 +899,12 @@
892	899	# generate the index source, containing all web references,..
893	900	page_index.h: $(TRANSLATEDSRC) mkindex.exe
894	901	mkindex.exe $(TRANSLATEDSRC) >$@
895	902
896	903	# extracting version info from manifest
897		-VERSION.h: version.exe ..\manifest.uuid ..\manifest
898		- version.exe ..\manifest.uuid ..\manifest > $@
	904	+VERSION.h: version.exe ..\manifest.uuid ..\manifest ..\VERSION
	905	+ version.exe ..\manifest.uuid ..\manifest ..\VERSION > $@
899	906
900	907	# generate the simplified headers
901	908	headers: makeheaders.exe page_index.h VERSION.h ../src/sqlite3.h ../src/th.h VERSION.h
902	909	makeheaders.exe $(foreach ts,$(TRANSLATEDSRC),$(ts):$(ts:_.c=.h)) ../src/sqlite3.h ../src/th.h VERSION.h
903	910	echo Done >$@
904	911

	--- src/makemake.tcl
	+++ src/makemake.tcl
	@@ -176,27 +176,34 @@
176	$(BCC) -o $(OBJDIR)/makeheaders $(SRCDIR)/makeheaders.c
177
178	$(OBJDIR)/mkindex: $(SRCDIR)/mkindex.c
179	$(BCC) -o $(OBJDIR)/mkindex $(SRCDIR)/mkindex.c
180



181	# WARNING. DANGER. Running the testsuite modifies the repository the
182	# build is done from, i.e. the checkout belongs to. Do not sync/push
183	# the repository after running the tests.
184	test: $(APPNAME)
185	$(TCLSH) test/tester.tcl $(APPNAME)
186
187	$(OBJDIR)/VERSION.h: $(SRCDIR)/../manifest.uuid $(SRCDIR)/../manifest
188	awk '{ printf "#define MANIFEST_UUID \"%s\"\n", $$1}' \
189	$(SRCDIR)/../manifest.uuid >$(OBJDIR)/VERSION.h
190	awk '{ printf "#define MANIFEST_VERSION \"[%.10s]\"\n", $$1}' \
191	$(SRCDIR)/../manifest.uuid >>$(OBJDIR)/VERSION.h
192	awk '$$1=="D"{printf "#define MANIFEST_DATE \"%s %s\"\n",\
193	substr($$2,1,10),substr($$2,12,8)}' \
194	$(SRCDIR)/../manifest >>$(OBJDIR)/VERSION.h




195
196	EXTRAOBJ = \
197	$(OBJDIR)/sqlite3.o \
198	$(OBJDIR)/shell.o \
199	$(OBJDIR)/th.o \
200	$(OBJDIR)/th_lang.o
201
202	$(APPNAME): $(OBJDIR)/headers $(OBJ) $(EXTRAOBJ)
	@@ -241,17 +248,17 @@
241
242	writeln "\$(OBJDIR)/sqlite3.o:\t\$(SRCDIR)/sqlite3.c"
243	set opt {-DSQLITE_OMIT_LOAD_EXTENSION=1}
244	append opt " -DSQLITE_THREADSAFE=0 -DSQLITE_DEFAULT_FILE_FORMAT=4"
245	#append opt " -DSQLITE_ENABLE_FTS3=1"
246	append opt " -DSQLITE_ENABLE_STAT2"
247	append opt " -Dlocaltime=fossil_localtime"
248	append opt " -DSQLITE_ENABLE_LOCKING_STYLE=0"
249	set SQLITE_OPTIONS $opt
250	writeln "\t\$(XTCC) $opt -c \$(SRCDIR)/sqlite3.c -o \$(OBJDIR)/sqlite3.o\n"
251
252	writeln "\$(OBJDIR)/shell.o:\t\$(SRCDIR)/shell.c"
253	set opt {-Dmain=sqlite3_shell}
254	append opt " -DSQLITE_OMIT_LOAD_EXTENSION=1"
255	writeln "\t\$(XTCC) $opt -c \$(SRCDIR)/shell.c -o \$(OBJDIR)/shell.o\n"
256
257	writeln "\$(OBJDIR)/th.o:\t\$(SRCDIR)/th.c"
	@@ -389,21 +396,21 @@
389	$(BCC) -o $(OBJDIR)/makeheaders $(SRCDIR)/makeheaders.c
390
391	$(OBJDIR)/mkindex: $(SRCDIR)/mkindex.c
392	$(BCC) -o $(OBJDIR)/mkindex $(SRCDIR)/mkindex.c
393
394	$(VERSION): $(SRCDIR)/../win/version.c
395	$(BCC) -o $(OBJDIR)/version $(SRCDIR)/../win/version.c
396
397	# WARNING. DANGER. Running the testsuite modifies the repository the
398	# build is done from, i.e. the checkout belongs to. Do not sync/push
399	# the repository after running the tests.
400	test: $(APPNAME)
401	$(TCLSH) test/tester.tcl $(APPNAME)
402
403	$(OBJDIR)/VERSION.h: $(SRCDIR)/../manifest.uuid $(SRCDIR)/../manifest $(VERSION)
404	$(VERSION) $(SRCDIR)/../manifest.uuid $(SRCDIR)/../manifest >$(OBJDIR)/VERSION.h
405
406	EXTRAOBJ = \
407	$(OBJDIR)/sqlite3.o \
408	$(OBJDIR)/shell.o \
409	$(OBJDIR)/th.o \
	@@ -459,11 +466,11 @@
459
460	writeln "\$(OBJDIR)/sqlite3.o:\t\$(SRCDIR)/sqlite3.c"
461	set opt $SQLITE_OPTIONS
462	writeln "\t\$(XTCC) $opt -c \$(SRCDIR)/sqlite3.c -o \$(OBJDIR)/sqlite3.o\n"
463
464	writeln "\$(OBJDIR)/shell.o:\t\$(SRCDIR)/shell.c"
465	set opt {-Dmain=sqlite3_shell}
466	append opt " -DSQLITE_OMIT_LOAD_EXTENSION=1"
467	writeln "\t\$(XTCC) $opt -c \$(SRCDIR)/shell.c -o \$(OBJDIR)/shell.o\n"
468
469	writeln "\$(OBJDIR)/th.o:\t\$(SRCDIR)/th.c"
	@@ -556,11 +563,11 @@
556	$(BCC) -o$@ $**
557
558	mkindex$E: $(SRCDIR)\mkindex.c
559	$(BCC) -o$@ $**
560
561	version$E: $B\win\version.c
562	$(BCC) -o$@ $**
563
564	$(OBJDIR)\shell$O : $(SRCDIR)\shell.c
565	$(TCC) -o$@ -c -Dmain=sqlite3_shell $(SQLITE_OPTIONS) $**
566
	@@ -571,11 +578,11 @@
571	$(TCC) -o$@ -c $**
572
573	$(OBJDIR)\th_lang$O : $(SRCDIR)\th_lang.c
574	$(TCC) -o$@ -c $**
575
576	VERSION.h : version$E $B\manifest.uuid $B\manifest
577	+$** > $@
578
579	page_index.h: mkindex$E $(SRC)
580	+$** > $@
581
	@@ -582,11 +589,11 @@
582	clean:
583	-del $(OBJDIR)\*.obj
584	-del .obj _.c .h .map
585
586	realclean:
587	-del $(APPNAME) translate$E mkindex$E makeheaders$E version$E
588
589	}
590	foreach s [lsort $src] {
591	writeln "\$(OBJDIR)\\$s\$O : ${s}_.c ${s}.h"
592	writeln "\t\$(TCC) -o\$@ -c ${s}_.c\n"
	@@ -646,11 +653,11 @@
646	INCL = -I. -I$(SRCDIR) -I$B\win\include -I$(MSCDIR)\extra\include -I$(ZINCDIR)
647
648	CFLAGS = -nologo -MT -O2
649	BCC = $(CC) $(CFLAGS)
650	TCC = $(CC) -c $(CFLAGS) $(MSCDEF) $(SSL) $(INCL)
651	LIBS = $(ZLIB) ws2_32.lib $(SSLLIB)
652	LIBDIR = -LIBPATH:$(MSCDIR)\extra\lib -LIBPATH:$(ZLIBDIR)
653	}
654	regsub -all {[-]D} $SQLITE_OPTIONS {/D} MSC_SQLITE_OPTIONS
655	writeln "SQLITE_OPTIONS = $MSC_SQLITE_OPTIONS\n"
656	writeln -nonewline "SRC = "
	@@ -693,11 +700,11 @@
693	$(BCC) $**
694
695	mkindex$E: $(SRCDIR)\mkindex.c
696	$(BCC) $**
697
698	version$E: $B\win\version.c
699	$(BCC) $**
700
701	$(OX)\shell$O : $(SRCDIR)\shell.c
702	$(TCC) /Fo$@ /Dmain=sqlite3_shell $(SQLITE_OPTIONS) -c $(SRCDIR)\shell.c
703
	@@ -708,11 +715,11 @@
708	$(TCC) /Fo$@ -c $**
709
710	$(OX)\th_lang$O : $(SRCDIR)\th_lang.c
711	$(TCC) /Fo$@ -c $**
712
713	VERSION.h : version$E $B\manifest.uuid $B\manifest
714	$** > $@
715
716	page_index.h: mkindex$E $(SRC)
717	$** > $@
718
	@@ -720,11 +727,11 @@
720	-del $(OX)\*.obj
721	-del .obj _.c .h .map
722	-del headers linkopts
723
724	realclean:
725	-del $(APPNAME) translate$E mkindex$E makeheaders$E version$E
726
727	}
728	foreach s [lsort $src] {
729	writeln "\$(OX)\\$s\$O : ${s}_.c ${s}.h"
730	writeln "\t\$(TCC) /Fo\$@ -c ${s}_.c\n"
	@@ -880,11 +887,11 @@
880	# compiling standard fossil utils
881	$(UTILS_OBJ): %.obj: $(SRCDIR)%.c
882	$(CC) $(CCFLAGS) $(INCLUDE) "$<" -Fo"$@"
883
884	# compile special windows utils
885	version.obj: $(WINDIR)version.c
886	$(CC) $(CCFLAGS) $(INCLUDE) "$<" -Fo"$@"
887
888	# generate the translated c-source files
889	$(TRANSLATEDSRC): %_.c: $(SRCDIR)%.c translate.exe
890	translate.exe $< >$@
	@@ -892,12 +899,12 @@
892	# generate the index source, containing all web references,..
893	page_index.h: $(TRANSLATEDSRC) mkindex.exe
894	mkindex.exe $(TRANSLATEDSRC) >$@
895
896	# extracting version info from manifest
897	VERSION.h: version.exe ..\manifest.uuid ..\manifest
898	version.exe ..\manifest.uuid ..\manifest > $@
899
900	# generate the simplified headers
901	headers: makeheaders.exe page_index.h VERSION.h ../src/sqlite3.h ../src/th.h VERSION.h
902	makeheaders.exe $(foreach ts,$(TRANSLATEDSRC),$(ts):$(ts:_.c=.h)) ../src/sqlite3.h ../src/th.h VERSION.h
903	echo Done >$@
904

	--- src/makemake.tcl
	+++ src/makemake.tcl
	@@ -176,27 +176,34 @@
176	$(BCC) -o $(OBJDIR)/makeheaders $(SRCDIR)/makeheaders.c
177
178	$(OBJDIR)/mkindex: $(SRCDIR)/mkindex.c
179	$(BCC) -o $(OBJDIR)/mkindex $(SRCDIR)/mkindex.c
180
181	$(OBJDIR)/mkversion: $(SRCDIR)/mkversion.c
182	$(BCC) -o $(OBJDIR)/mkversion $(SRCDIR)/mkversion.c
183
184	# WARNING. DANGER. Running the testsuite modifies the repository the
185	# build is done from, i.e. the checkout belongs to. Do not sync/push
186	# the repository after running the tests.
187	test: $(APPNAME)
188	$(TCLSH) test/tester.tcl $(APPNAME)
189
190	$(OBJDIR)/VERSION.h: $(SRCDIR)/../manifest.uuid $(SRCDIR)/../manifest $(SRCDIR)/../VERSION $(OBJDIR)/mkversion
191	$(OBJDIR)/mkversion $(SRCDIR)/../manifest.uuid \
192	$(SRCDIR)/../manifest \
193	$(SRCDIR)/../VERSION >$(OBJDIR)/VERSION.h
194
195	# The USE_SYSTEM_SQLITE variable may be undefined, set to 0, or set
196	# to 1. If it is set to 1, then there is no need to build or link
197	# the sqlite3.o object. Instead, the system sqlite will be linked
198	# using -lsqlite3.
199	SQLITE3_OBJ.1 =
200	SQLITE3_OBJ.0 = $(OBJDIR)/sqlite3.o
201	SQLITE3_OBJ. = $(SQLITE3_OBJ.0)
202
203	EXTRAOBJ = \
204	$(SQLITE3_OBJ.$(USE_SYSTEM_SQLITE)) \
205	$(OBJDIR)/shell.o \
206	$(OBJDIR)/th.o \
207	$(OBJDIR)/th_lang.o
208
209	$(APPNAME): $(OBJDIR)/headers $(OBJ) $(EXTRAOBJ)
	@@ -241,17 +248,17 @@
248
249	writeln "\$(OBJDIR)/sqlite3.o:\t\$(SRCDIR)/sqlite3.c"
250	set opt {-DSQLITE_OMIT_LOAD_EXTENSION=1}
251	append opt " -DSQLITE_THREADSAFE=0 -DSQLITE_DEFAULT_FILE_FORMAT=4"
252	#append opt " -DSQLITE_ENABLE_FTS3=1"
253	append opt " -DSQLITE_ENABLE_STAT3"
254	append opt " -Dlocaltime=fossil_localtime"
255	append opt " -DSQLITE_ENABLE_LOCKING_STYLE=0"
256	set SQLITE_OPTIONS $opt
257	writeln "\t\$(XTCC) $opt -c \$(SRCDIR)/sqlite3.c -o \$(OBJDIR)/sqlite3.o\n"
258
259	writeln "\$(OBJDIR)/shell.o:\t\$(SRCDIR)/shell.c \$(SRCDIR)/sqlite3.h"
260	set opt {-Dmain=sqlite3_shell}
261	append opt " -DSQLITE_OMIT_LOAD_EXTENSION=1"
262	writeln "\t\$(XTCC) $opt -c \$(SRCDIR)/shell.c -o \$(OBJDIR)/shell.o\n"
263
264	writeln "\$(OBJDIR)/th.o:\t\$(SRCDIR)/th.c"
	@@ -389,21 +396,21 @@
396	$(BCC) -o $(OBJDIR)/makeheaders $(SRCDIR)/makeheaders.c
397
398	$(OBJDIR)/mkindex: $(SRCDIR)/mkindex.c
399	$(BCC) -o $(OBJDIR)/mkindex $(SRCDIR)/mkindex.c
400
401	$(VERSION): $(SRCDIR)/mkversion.c
402	$(BCC) -o $(OBJDIR)/version $(SRCDIR)/mkversion.c
403
404	# WARNING. DANGER. Running the testsuite modifies the repository the
405	# build is done from, i.e. the checkout belongs to. Do not sync/push
406	# the repository after running the tests.
407	test: $(APPNAME)
408	$(TCLSH) test/tester.tcl $(APPNAME)
409
410	$(OBJDIR)/VERSION.h: $(SRCDIR)/../manifest.uuid $(SRCDIR)/../manifest $(VERSION)
411	$(VERSION) $(SRCDIR)/../manifest.uuid $(SRCDIR)/../manifest $(SRCDIR)/../VERSION >$(OBJDIR)/VERSION.h
412
413	EXTRAOBJ = \
414	$(OBJDIR)/sqlite3.o \
415	$(OBJDIR)/shell.o \
416	$(OBJDIR)/th.o \
	@@ -459,11 +466,11 @@
466
467	writeln "\$(OBJDIR)/sqlite3.o:\t\$(SRCDIR)/sqlite3.c"
468	set opt $SQLITE_OPTIONS
469	writeln "\t\$(XTCC) $opt -c \$(SRCDIR)/sqlite3.c -o \$(OBJDIR)/sqlite3.o\n"
470
471	writeln "\$(OBJDIR)/shell.o:\t\$(SRCDIR)/shell.c \$(SRCDIR)/sqlite3.h"
472	set opt {-Dmain=sqlite3_shell}
473	append opt " -DSQLITE_OMIT_LOAD_EXTENSION=1"
474	writeln "\t\$(XTCC) $opt -c \$(SRCDIR)/shell.c -o \$(OBJDIR)/shell.o\n"
475
476	writeln "\$(OBJDIR)/th.o:\t\$(SRCDIR)/th.c"
	@@ -556,11 +563,11 @@
563	$(BCC) -o$@ $**
564
565	mkindex$E: $(SRCDIR)\mkindex.c
566	$(BCC) -o$@ $**
567
568	version$E: $B\src\mkversion.c
569	$(BCC) -o$@ $**
570
571	$(OBJDIR)\shell$O : $(SRCDIR)\shell.c
572	$(TCC) -o$@ -c -Dmain=sqlite3_shell $(SQLITE_OPTIONS) $**
573
	@@ -571,11 +578,11 @@
578	$(TCC) -o$@ -c $**
579
580	$(OBJDIR)\th_lang$O : $(SRCDIR)\th_lang.c
581	$(TCC) -o$@ -c $**
582
583	VERSION.h : version$E $B\manifest.uuid $B\manifest $B\VERSION
584	+$** > $@
585
586	page_index.h: mkindex$E $(SRC)
587	+$** > $@
588
	@@ -582,11 +589,11 @@
589	clean:
590	-del $(OBJDIR)\*.obj
591	-del .obj _.c .h .map
592
593	realclean:
594	-del $(APPNAME) translate$E mkindex$E makeheaders$E mkversion$E
595
596	}
597	foreach s [lsort $src] {
598	writeln "\$(OBJDIR)\\$s\$O : ${s}_.c ${s}.h"
599	writeln "\t\$(TCC) -o\$@ -c ${s}_.c\n"
	@@ -646,11 +653,11 @@
653	INCL = -I. -I$(SRCDIR) -I$B\win\include -I$(MSCDIR)\extra\include -I$(ZINCDIR)
654
655	CFLAGS = -nologo -MT -O2
656	BCC = $(CC) $(CFLAGS)
657	TCC = $(CC) -c $(CFLAGS) $(MSCDEF) $(SSL) $(INCL)
658	LIBS = $(ZLIB) ws2_32.lib advapi32.lib $(SSLLIB)
659	LIBDIR = -LIBPATH:$(MSCDIR)\extra\lib -LIBPATH:$(ZLIBDIR)
660	}
661	regsub -all {[-]D} $SQLITE_OPTIONS {/D} MSC_SQLITE_OPTIONS
662	writeln "SQLITE_OPTIONS = $MSC_SQLITE_OPTIONS\n"
663	writeln -nonewline "SRC = "
	@@ -693,11 +700,11 @@
700	$(BCC) $**
701
702	mkindex$E: $(SRCDIR)\mkindex.c
703	$(BCC) $**
704
705	mkversion$E: $B\src\mkversion.c
706	$(BCC) $**
707
708	$(OX)\shell$O : $(SRCDIR)\shell.c
709	$(TCC) /Fo$@ /Dmain=sqlite3_shell $(SQLITE_OPTIONS) -c $(SRCDIR)\shell.c
710
	@@ -708,11 +715,11 @@
715	$(TCC) /Fo$@ -c $**
716
717	$(OX)\th_lang$O : $(SRCDIR)\th_lang.c
718	$(TCC) /Fo$@ -c $**
719
720	VERSION.h : mkversion$E $B\manifest.uuid $B\manifest $B\VERSION
721	$** > $@
722
723	page_index.h: mkindex$E $(SRC)
724	$** > $@
725
	@@ -720,11 +727,11 @@
727	-del $(OX)\*.obj
728	-del .obj _.c .h .map
729	-del headers linkopts
730
731	realclean:
732	-del $(APPNAME) translate$E mkindex$E makeheaders$E mkversion$E
733
734	}
735	foreach s [lsort $src] {
736	writeln "\$(OX)\\$s\$O : ${s}_.c ${s}.h"
737	writeln "\t\$(TCC) /Fo\$@ -c ${s}_.c\n"
	@@ -880,11 +887,11 @@
887	# compiling standard fossil utils
888	$(UTILS_OBJ): %.obj: $(SRCDIR)%.c
889	$(CC) $(CCFLAGS) $(INCLUDE) "$<" -Fo"$@"
890
891	# compile special windows utils
892	version.obj: $(SRCDIR)mkversion.c
893	$(CC) $(CCFLAGS) $(INCLUDE) "$<" -Fo"$@"
894
895	# generate the translated c-source files
896	$(TRANSLATEDSRC): %_.c: $(SRCDIR)%.c translate.exe
897	translate.exe $< >$@
	@@ -892,12 +899,12 @@
899	# generate the index source, containing all web references,..
900	page_index.h: $(TRANSLATEDSRC) mkindex.exe
901	mkindex.exe $(TRANSLATEDSRC) >$@
902
903	# extracting version info from manifest
904	VERSION.h: version.exe ..\manifest.uuid ..\manifest ..\VERSION
905	version.exe ..\manifest.uuid ..\manifest ..\VERSION > $@
906
907	# generate the simplified headers
908	headers: makeheaders.exe page_index.h VERSION.h ../src/sqlite3.h ../src/th.h VERSION.h
909	makeheaders.exe $(foreach ts,$(TRANSLATEDSRC),$(ts):$(ts:_.c=.h)) ../src/sqlite3.h ../src/th.h VERSION.h
910	echo Done >$@
911

M src/manifest.c

+3 -4

		--- src/manifest.c
		+++ src/manifest.c
		@@ -1323,14 +1323,13 @@
1323	1323	if( pChildFile ){
1324	1324	add_one_mlink(cid, 0, pChildFile->zUuid, pChildFile->zName, 0,
1325	1325	isPublic, manifest_file_mperm(pChildFile));
1326	1326	}
1327	1327	}
1328		- }else if( pChild->zBaseline==0 && pParent->zBaseline!=0 ){
1329		- /* Parent is a delta but pChild is a baseline. Look for files that are
1330		- ** present in pParent but which are missing from pChild and mark them
1331		- ** has having been deleted. */
	1328	+ }else if( pChild->zBaseline==0 ){
	1329	+ /* pChild is a baseline. Look for files that are present in pParent
	1330	+ ** but are missing from pChild and mark them as having been deleted. */
1332	1331	manifest_file_rewind(pParent);
1333	1332	while( (pParentFile = manifest_file_next(pParent,0))!=0 ){
1334	1333	pChildFile = manifest_file_seek(pChild, pParentFile->zName);
1335	1334	if( pChildFile==0 && pParentFile->zUuid!=0 ){
1336	1335	add_one_mlink(cid, pParentFile->zUuid, 0, pParentFile->zName, 0,
1337	1336

	--- src/manifest.c
	+++ src/manifest.c
	@@ -1323,14 +1323,13 @@
1323	if( pChildFile ){
1324	add_one_mlink(cid, 0, pChildFile->zUuid, pChildFile->zName, 0,
1325	isPublic, manifest_file_mperm(pChildFile));
1326	}
1327	}
1328	}else if( pChild->zBaseline==0 && pParent->zBaseline!=0 ){
1329	/* Parent is a delta but pChild is a baseline. Look for files that are
1330	** present in pParent but which are missing from pChild and mark them
1331	** has having been deleted. */
1332	manifest_file_rewind(pParent);
1333	while( (pParentFile = manifest_file_next(pParent,0))!=0 ){
1334	pChildFile = manifest_file_seek(pChild, pParentFile->zName);
1335	if( pChildFile==0 && pParentFile->zUuid!=0 ){
1336	add_one_mlink(cid, pParentFile->zUuid, 0, pParentFile->zName, 0,
1337

	--- src/manifest.c
	+++ src/manifest.c
	@@ -1323,14 +1323,13 @@
1323	if( pChildFile ){
1324	add_one_mlink(cid, 0, pChildFile->zUuid, pChildFile->zName, 0,
1325	isPublic, manifest_file_mperm(pChildFile));
1326	}
1327	}
1328	}else if( pChild->zBaseline==0 ){
1329	/* pChild is a baseline. Look for files that are present in pParent
1330	** but are missing from pChild and mark them as having been deleted. */

1331	manifest_file_rewind(pParent);
1332	while( (pParentFile = manifest_file_next(pParent,0))!=0 ){
1333	pChildFile = manifest_file_seek(pChild, pParentFile->zName);
1334	if( pChildFile==0 && pParentFile->zUuid!=0 ){
1335	add_one_mlink(cid, pParentFile->zUuid, 0, pParentFile->zName, 0,
1336

M src/merge.c

+10 -2

		--- src/merge.c
		+++ src/merge.c
		@@ -53,10 +53,14 @@
53	53	** and do not try to merge parallel changes. This
54	54	** option overrides the "binary-glob" setting.
55	55	**
56	56	** --nochange \| -n Dryrun: do not actually make any changes; just
57	57	** show what would have happened.
	58	+**
	59	+** --case-sensitive BOOL Overwrite the case-sensitive setting. If false,
	60	+** files whose names differ only in case are taken
	61	+** to be the same file.
58	62	*/
59	63	void merge_cmd(void){
60	64	int vid; /* Current version "V" */
61	65	int mid; /* Version we are merging from "M" */
62	66	int pid; /* The pivot version - most recent common ancestor P */
		@@ -69,10 +73,11 @@
69	73	int debugFlag; /* True if --debug is present */
70	74	int nChng; /* Number of file name changes */
71	75	int aChng; / An array of file name changes */
72	76	int i; /* Loop counter */
73	77	int nConflict = 0; /* Number of conflicts seen */
	78	+ int caseSensitive; /* True for case-sensitive filenames */
74	79	Stmt q;
75	80
76	81
77	82	/* Notation:
78	83	**
		@@ -87,14 +92,16 @@
87	92	backoutFlag = find_option("backout",0,0)!=0;
88	93	debugFlag = find_option("debug",0,0)!=0;
89	94	zBinGlob = find_option("binary",0,1);
90	95	nochangeFlag = find_option("nochange","n",0)!=0;
91	96	zPivot = find_option("baseline",0,1);
	97	+ capture_case_sensitive_option();
92	98	if( g.argc!=3 ){
93	99	usage("VERSION");
94	100	}
95	101	db_must_be_within_tree();
	102	+ caseSensitive = filenames_are_case_sensitive();
96	103	if( zBinGlob==0 ) zBinGlob = db_get("binary-glob",0);
97	104	vid = db_lget_int("checkout", 0);
98	105	if( vid==0 ){
99	106	fossil_fatal("nothing is checked out");
100	107	}
		@@ -150,11 +157,11 @@
150	157	** in the current checkout, the pivot, and the version being merged.
151	158	*/
152	159	db_multi_exec(
153	160	"DROP TABLE IF EXISTS fv;"
154	161	"CREATE TEMP TABLE fv("
155		- " fn TEXT PRIMARY KEY," /* The filename */
	162	+ " fn TEXT PRIMARY KEY COLLATE %s," /* The filename */
156	163	" idv INTEGER," /* VFILE entry for current version */
157	164	" idp INTEGER," /* VFILE entry for the pivot */
158	165	" idm INTEGER," /* VFILE entry for version merging in */
159	166	" chnged BOOLEAN," /* True if current version has been edited */
160	167	" ridv INTEGER," /* Record ID for current version */
		@@ -161,11 +168,12 @@
161	168	" ridp INTEGER," /* Record ID for pivot */
162	169	" ridm INTEGER," /* Record ID for merge */
163	170	" isexe BOOLEAN," /* Execute permission enabled */
164	171	" fnp TEXT," /* The filename in the pivot */
165	172	" fnm TEXT" /* the filename in the merged version */
166		- ");"
	173	+ ");",
	174	+ caseSensitive ? "binary" : "nocase"
167	175	);
168	176
169	177	/* Add files found in V
170	178	*/
171	179	db_multi_exec(
172	180
173	181	ADDED src/mkversion.c

	--- src/merge.c
	+++ src/merge.c
	@@ -53,10 +53,14 @@
53	** and do not try to merge parallel changes. This
54	** option overrides the "binary-glob" setting.
55	**
56	** --nochange \| -n Dryrun: do not actually make any changes; just
57	** show what would have happened.




58	*/
59	void merge_cmd(void){
60	int vid; /* Current version "V" */
61	int mid; /* Version we are merging from "M" */
62	int pid; /* The pivot version - most recent common ancestor P */
	@@ -69,10 +73,11 @@
69	int debugFlag; /* True if --debug is present */
70	int nChng; /* Number of file name changes */
71	int aChng; / An array of file name changes */
72	int i; /* Loop counter */
73	int nConflict = 0; /* Number of conflicts seen */

74	Stmt q;
75
76
77	/* Notation:
78	**
	@@ -87,14 +92,16 @@
87	backoutFlag = find_option("backout",0,0)!=0;
88	debugFlag = find_option("debug",0,0)!=0;
89	zBinGlob = find_option("binary",0,1);
90	nochangeFlag = find_option("nochange","n",0)!=0;
91	zPivot = find_option("baseline",0,1);

92	if( g.argc!=3 ){
93	usage("VERSION");
94	}
95	db_must_be_within_tree();

96	if( zBinGlob==0 ) zBinGlob = db_get("binary-glob",0);
97	vid = db_lget_int("checkout", 0);
98	if( vid==0 ){
99	fossil_fatal("nothing is checked out");
100	}
	@@ -150,11 +157,11 @@
150	** in the current checkout, the pivot, and the version being merged.
151	*/
152	db_multi_exec(
153	"DROP TABLE IF EXISTS fv;"
154	"CREATE TEMP TABLE fv("
155	" fn TEXT PRIMARY KEY," /* The filename */
156	" idv INTEGER," /* VFILE entry for current version */
157	" idp INTEGER," /* VFILE entry for the pivot */
158	" idm INTEGER," /* VFILE entry for version merging in */
159	" chnged BOOLEAN," /* True if current version has been edited */
160	" ridv INTEGER," /* Record ID for current version */
	@@ -161,11 +168,12 @@
161	" ridp INTEGER," /* Record ID for pivot */
162	" ridm INTEGER," /* Record ID for merge */
163	" isexe BOOLEAN," /* Execute permission enabled */
164	" fnp TEXT," /* The filename in the pivot */
165	" fnm TEXT" /* the filename in the merged version */
166	");"

167	);
168
169	/* Add files found in V
170	*/
171	db_multi_exec(
172
173	DDED src/mkversion.c

	--- src/merge.c
	+++ src/merge.c
	@@ -53,10 +53,14 @@
53	** and do not try to merge parallel changes. This
54	** option overrides the "binary-glob" setting.
55	**
56	** --nochange \| -n Dryrun: do not actually make any changes; just
57	** show what would have happened.
58	**
59	** --case-sensitive BOOL Overwrite the case-sensitive setting. If false,
60	** files whose names differ only in case are taken
61	** to be the same file.
62	*/
63	void merge_cmd(void){
64	int vid; /* Current version "V" */
65	int mid; /* Version we are merging from "M" */
66	int pid; /* The pivot version - most recent common ancestor P */
	@@ -69,10 +73,11 @@
73	int debugFlag; /* True if --debug is present */
74	int nChng; /* Number of file name changes */
75	int aChng; / An array of file name changes */
76	int i; /* Loop counter */
77	int nConflict = 0; /* Number of conflicts seen */
78	int caseSensitive; /* True for case-sensitive filenames */
79	Stmt q;
80
81
82	/* Notation:
83	**
	@@ -87,14 +92,16 @@
92	backoutFlag = find_option("backout",0,0)!=0;
93	debugFlag = find_option("debug",0,0)!=0;
94	zBinGlob = find_option("binary",0,1);
95	nochangeFlag = find_option("nochange","n",0)!=0;
96	zPivot = find_option("baseline",0,1);
97	capture_case_sensitive_option();
98	if( g.argc!=3 ){
99	usage("VERSION");
100	}
101	db_must_be_within_tree();
102	caseSensitive = filenames_are_case_sensitive();
103	if( zBinGlob==0 ) zBinGlob = db_get("binary-glob",0);
104	vid = db_lget_int("checkout", 0);
105	if( vid==0 ){
106	fossil_fatal("nothing is checked out");
107	}
	@@ -150,11 +157,11 @@
157	** in the current checkout, the pivot, and the version being merged.
158	*/
159	db_multi_exec(
160	"DROP TABLE IF EXISTS fv;"
161	"CREATE TEMP TABLE fv("
162	" fn TEXT PRIMARY KEY COLLATE %s," /* The filename */
163	" idv INTEGER," /* VFILE entry for current version */
164	" idp INTEGER," /* VFILE entry for the pivot */
165	" idm INTEGER," /* VFILE entry for version merging in */
166	" chnged BOOLEAN," /* True if current version has been edited */
167	" ridv INTEGER," /* Record ID for current version */
	@@ -161,11 +168,12 @@
168	" ridp INTEGER," /* Record ID for pivot */
169	" ridm INTEGER," /* Record ID for merge */
170	" isexe BOOLEAN," /* Execute permission enabled */
171	" fnp TEXT," /* The filename in the pivot */
172	" fnm TEXT" /* the filename in the merged version */
173	");",
174	caseSensitive ? "binary" : "nocase"
175	);
176
177	/* Add files found in V
178	*/
179	db_multi_exec(
180
181	DDED src/mkversion.c

M src/mkversion.c

+1 -6

		--- a/src/mkversion.c
		+++ b/src/mkversion.c
		@@ -1,10 +1,5 @@
1		-fopen(argv[1],"gv[1],exists to do the job that AWK would do for the unix
2		-** makefile - to extract information from the "mainfest" and "manifest.uuid"
3		-** files for this p)
	1	+fopen(argv[1],"gv[1],"fMSC_VER)
4	2	d = _MSC_VER / 1 return 0;
5	3	}
6	4	b2b2,0,sizeof(b2)strcpy(b2b2b2vx[1000]return 0;
7	5	}
8		-;
9		- char b[1024b[strlen(b)-1] =0; return 0"gv[1],"fMSC_VER)
10		- fopen(argv[1

	--- a/src/mkversion.c
	+++ b/src/mkversion.c
	@@ -1,10 +1,5 @@
1	fopen(argv[1],"gv[1],exists to do the job that AWK would do for the unix
2	** makefile - to extract information from the "mainfest" and "manifest.uuid"
3	** files for this p)
4	d = _MSC_VER / 1 return 0;
5	}
6	b2b2,0,sizeof(b2)strcpy(b2b2b2vx[1000]return 0;
7	}
8	;
9	char b[1024b[strlen(b)-1] =0; return 0"gv[1],"fMSC_VER)
10	fopen(argv[1

	--- a/src/mkversion.c
	+++ b/src/mkversion.c
	@@ -1,10 +1,5 @@
1	fopen(argv[1],"gv[1],"fMSC_VER)


2	d = _MSC_VER / 1 return 0;
3	}
4	b2b2,0,sizeof(b2)strcpy(b2b2b2vx[1000]return 0;
5	}

A src/mkversion.c

		--- a/src/mkversion.c
		+++ b/src/mkversion.c
		@@ -0,0 +1,5 @@
	1	+fopen(argv[1],"gv[1],"fMSC_VER)
	2	+ d = _MSC_VER / 1 return 0;
	3	+}
	4	+b2b2,0,sizeof(b2)strcpy(b2b2b2vx[1000]return 0;
	5	+}

	--- a/src/mkversion.c
	+++ b/src/mkversion.c
	@@ -0,0 +1,5 @@

	--- a/src/mkversion.c
	+++ b/src/mkversion.c
	@@ -0,0 +1,5 @@
1	fopen(argv[1],"gv[1],"fMSC_VER)
2	d = _MSC_VER / 1 return 0;
3	}
4	b2b2,0,sizeof(b2)strcpy(b2b2b2vx[1000]return 0;
5	}

M src/popen.c

+6 -2

		--- src/popen.c
		+++ src/popen.c
		@@ -172,17 +172,21 @@
172	172	close(pout[1]);
173	173	*pChildPid = 0;
174	174	return 1;
175	175	}
176	176	if( *pChildPid==0 ){
	177	+ int fd;
	178	+ int nErr = 0;
177	179	/* This is the child process */
178	180	close(0);
179		- dup(pout[0]);
	181	+ fd = dup(pout[0]);
	182	+ if( fd!=0 ) nErr++;
180	183	close(pout[0]);
181	184	close(pout[1]);
182	185	close(1);
183		- dup(pin[1]);
	186	+ fd = dup(pin[1]);
	187	+ if( fd!=1 ) nErr++;
184	188	close(pin[0]);
185	189	close(pin[1]);
186	190	execl("/bin/sh", "/bin/sh", "-c", zCmd, (char*)0);
187	191	return 1;
188	192	}else{
189	193

	--- src/popen.c
	+++ src/popen.c
	@@ -172,17 +172,21 @@
172	close(pout[1]);
173	*pChildPid = 0;
174	return 1;
175	}
176	if( *pChildPid==0 ){


177	/* This is the child process */
178	close(0);
179	dup(pout[0]);

180	close(pout[0]);
181	close(pout[1]);
182	close(1);
183	dup(pin[1]);

184	close(pin[0]);
185	close(pin[1]);
186	execl("/bin/sh", "/bin/sh", "-c", zCmd, (char*)0);
187	return 1;
188	}else{
189

	--- src/popen.c
	+++ src/popen.c
	@@ -172,17 +172,21 @@
172	close(pout[1]);
173	*pChildPid = 0;
174	return 1;
175	}
176	if( *pChildPid==0 ){
177	int fd;
178	int nErr = 0;
179	/* This is the child process */
180	close(0);
181	fd = dup(pout[0]);
182	if( fd!=0 ) nErr++;
183	close(pout[0]);
184	close(pout[1]);
185	close(1);
186	fd = dup(pin[1]);
187	if( fd!=1 ) nErr++;
188	close(pin[0]);
189	close(pin[1]);
190	execl("/bin/sh", "/bin/sh", "-c", zCmd, (char*)0);
191	return 1;
192	}else{
193

M src/printf.c

+1 -1

		--- src/printf.c
		+++ src/printf.c
		@@ -859,11 +859,11 @@
859	859	int a, b;
860	860	do{
861	861	a = *zA++;
862	862	b = *zB++;
863	863	}while( a==b && a!=0 );
864		- return a - b;
	864	+ return ((unsigned char)a) - (unsigned char)b;
865	865	}
866	866	}
867	867
868	868	/*
869	869	** Case insensitive string comparison.
870	870

	--- src/printf.c
	+++ src/printf.c
	@@ -859,11 +859,11 @@
859	int a, b;
860	do{
861	a = *zA++;
862	b = *zB++;
863	}while( a==b && a!=0 );
864	return a - b;
865	}
866	}
867
868	/*
869	** Case insensitive string comparison.
870

	--- src/printf.c
	+++ src/printf.c
	@@ -859,11 +859,11 @@
859	int a, b;
860	do{
861	a = *zA++;
862	b = *zB++;
863	}while( a==b && a!=0 );
864	return ((unsigned char)a) - (unsigned char)b;
865	}
866	}
867
868	/*
869	** Case insensitive string comparison.
870

M src/rebuild.c

+1 -1

		--- src/rebuild.c
		+++ src/rebuild.c
		@@ -882,11 +882,11 @@
882	882	}
883	883
884	884	/*
885	885	** COMMAND: deconstruct
886	886	**
887		-** Usage %fossil deconstruct ?OPTIONS? DESTIONATION
	887	+** Usage %fossil deconstruct ?OPTIONS? DESTINATION
888	888	**
889	889	** Options:
890	890	** -R\|--repository REPOSITORY
891	891	** -L\|--prefixlength N
892	892	**
893	893

	--- src/rebuild.c
	+++ src/rebuild.c
	@@ -882,11 +882,11 @@
882	}
883
884	/*
885	** COMMAND: deconstruct
886	**
887	** Usage %fossil deconstruct ?OPTIONS? DESTIONATION
888	**
889	** Options:
890	** -R\|--repository REPOSITORY
891	** -L\|--prefixlength N
892	**
893

	--- src/rebuild.c
	+++ src/rebuild.c
	@@ -882,11 +882,11 @@
882	}
883
884	/*
885	** COMMAND: deconstruct
886	**
887	** Usage %fossil deconstruct ?OPTIONS? DESTINATION
888	**
889	** Options:
890	** -R\|--repository REPOSITORY
891	** -L\|--prefixlength N
892	**
893

M src/search.c

+1 -1

		--- src/search.c
		+++ src/search.c
		@@ -203,8 +203,8 @@
203	203	db_prepare(&q,
204	204	"SELECT rid, uuid, date, comment, 0, 0 FROM srch"
205	205	" WHERE x>%d ORDER BY x DESC, date DESC",
206	206	iBest/3
207	207	);
208		- print_timeline(&q, 1000);
	208	+ print_timeline(&q, 1000, 0);
209	209	db_finalize(&q);
210	210	}
211	211

	--- src/search.c
	+++ src/search.c
	@@ -203,8 +203,8 @@
203	db_prepare(&q,
204	"SELECT rid, uuid, date, comment, 0, 0 FROM srch"
205	" WHERE x>%d ORDER BY x DESC, date DESC",
206	iBest/3
207	);
208	print_timeline(&q, 1000);
209	db_finalize(&q);
210	}
211

	--- src/search.c
	+++ src/search.c
	@@ -203,8 +203,8 @@
203	db_prepare(&q,
204	"SELECT rid, uuid, date, comment, 0, 0 FROM srch"
205	" WHERE x>%d ORDER BY x DESC, date DESC",
206	iBest/3
207	);
208	print_timeline(&q, 1000, 0);
209	db_finalize(&q);
210	}
211

M src/setup.c

+12 -2

		--- src/setup.c
		+++ src/setup.c
		@@ -1050,10 +1050,11 @@
1050	1050	if( !g.okSetup ){
1051	1051	login_needed();
1052	1052	}
1053	1053
1054	1054	style_header("Settings");
	1055	+ db_open_local();
1055	1056	db_begin_transaction();
1056	1057	@ <p>This page provides a simple interface to the "fossil setting" command.
1057	1058	@ See the "fossil help setting" output below for further information on
1058	1059	@ the meaning of each setting.</p><hr />
1059	1060	@ <form action="%s(g.zTop)/setup_settings" method="post"><div>
		@@ -1062,25 +1063,34 @@
1062	1063	for(pSet=ctrlSettings; pSet->name!=0; pSet++){
1063	1064	if( pSet->width==0 ){
1064	1065	onoff_attribute(pSet->name, pSet->name,
1065	1066	pSet->var!=0 ? pSet->var : pSet->name,
1066	1067	is_truth(pSet->def));
1067		- @ <br />
	1068	+ if( pSet->versionable ){
	1069	+ @ (v)<br />
	1070	+ } else {
	1071	+ @ <br />
	1072	+ }
1068	1073	}
1069	1074	}
1070	1075	@ </td><td style="width: 30;"></td><td valign="top">
1071	1076	for(pSet=ctrlSettings; pSet->name!=0; pSet++){
1072	1077	if( pSet->width!=0 ){
1073	1078	entry_attribute(pSet->name, /pSet->width/ 40, pSet->name,
1074	1079	pSet->var!=0 ? pSet->var : pSet->name,
1075	1080	(char*)pSet->def);
1076		- @ <br />
	1081	+ if( pSet->versionable ){
	1082	+ @ (v)<br />
	1083	+ } else {
	1084	+ @ <br />
	1085	+ }
1077	1086	}
1078	1087	}
1079	1088	@ </td></tr></table>
1080	1089	@ <p><input type="submit" name="submit" value="Apply Changes" /></p>
1081	1090	@ </div></form>
	1091	+ @ <p>Settings marked with (v) are 'versionable' and will be overridden by the contents of files named <tt>.fossil-settings/PROPERTY</tt>.</p>
1082	1092	@ <hr /><p>
1083	1093	@ These settings work in the same way, as the <kbd>set</kbd> commandline:<br />
1084	1094	@ </p><pre>%s(zHelp_setting_cmd)</pre>
1085	1095	db_end_transaction(0);
1086	1096	style_footer();
1087	1097

	--- src/setup.c
	+++ src/setup.c
	@@ -1050,10 +1050,11 @@
1050	if( !g.okSetup ){
1051	login_needed();
1052	}
1053
1054	style_header("Settings");

1055	db_begin_transaction();
1056	@ <p>This page provides a simple interface to the "fossil setting" command.
1057	@ See the "fossil help setting" output below for further information on
1058	@ the meaning of each setting.</p><hr />
1059	@ <form action="%s(g.zTop)/setup_settings" method="post"><div>
	@@ -1062,25 +1063,34 @@
1062	for(pSet=ctrlSettings; pSet->name!=0; pSet++){
1063	if( pSet->width==0 ){
1064	onoff_attribute(pSet->name, pSet->name,
1065	pSet->var!=0 ? pSet->var : pSet->name,
1066	is_truth(pSet->def));
1067	@ <br />




1068	}
1069	}
1070	@ </td><td style="width: 30;"></td><td valign="top">
1071	for(pSet=ctrlSettings; pSet->name!=0; pSet++){
1072	if( pSet->width!=0 ){
1073	entry_attribute(pSet->name, /pSet->width/ 40, pSet->name,
1074	pSet->var!=0 ? pSet->var : pSet->name,
1075	(char*)pSet->def);
1076	@ <br />




1077	}
1078	}
1079	@ </td></tr></table>
1080	@ <p><input type="submit" name="submit" value="Apply Changes" /></p>
1081	@ </div></form>

1082	@ <hr /><p>
1083	@ These settings work in the same way, as the <kbd>set</kbd> commandline:<br />
1084	@ </p><pre>%s(zHelp_setting_cmd)</pre>
1085	db_end_transaction(0);
1086	style_footer();
1087

	--- src/setup.c
	+++ src/setup.c
	@@ -1050,10 +1050,11 @@
1050	if( !g.okSetup ){
1051	login_needed();
1052	}
1053
1054	style_header("Settings");
1055	db_open_local();
1056	db_begin_transaction();
1057	@ <p>This page provides a simple interface to the "fossil setting" command.
1058	@ See the "fossil help setting" output below for further information on
1059	@ the meaning of each setting.</p><hr />
1060	@ <form action="%s(g.zTop)/setup_settings" method="post"><div>
	@@ -1062,25 +1063,34 @@
1063	for(pSet=ctrlSettings; pSet->name!=0; pSet++){
1064	if( pSet->width==0 ){
1065	onoff_attribute(pSet->name, pSet->name,
1066	pSet->var!=0 ? pSet->var : pSet->name,
1067	is_truth(pSet->def));
1068	if( pSet->versionable ){
1069	@ (v)<br />
1070	} else {
1071	@ <br />
1072	}
1073	}
1074	}
1075	@ </td><td style="width: 30;"></td><td valign="top">
1076	for(pSet=ctrlSettings; pSet->name!=0; pSet++){
1077	if( pSet->width!=0 ){
1078	entry_attribute(pSet->name, /pSet->width/ 40, pSet->name,
1079	pSet->var!=0 ? pSet->var : pSet->name,
1080	(char*)pSet->def);
1081	if( pSet->versionable ){
1082	@ (v)<br />
1083	} else {
1084	@ <br />
1085	}
1086	}
1087	}
1088	@ </td></tr></table>
1089	@ <p><input type="submit" name="submit" value="Apply Changes" /></p>
1090	@ </div></form>
1091	@ <p>Settings marked with (v) are 'versionable' and will be overridden by the contents of files named <tt>.fossil-settings/PROPERTY</tt>.</p>
1092	@ <hr /><p>
1093	@ These settings work in the same way, as the <kbd>set</kbd> commandline:<br />
1094	@ </p><pre>%s(zHelp_setting_cmd)</pre>
1095	db_end_transaction(0);
1096	style_footer();
1097

M src/sqlite3.c

+3777 -1278

		--- src/sqlite3.c
		+++ src/sqlite3.c
		@@ -1,8 +1,8 @@
1	1	/******************************************************************************
2	2	** This file is an amalgamation of many separate C source files from SQLite
3		-** version 3.7.7. By combining all the individual C code files into this
	3	+** version 3.7.8. By combining all the individual C code files into this
4	4	** single large file, the entire code can be compiled as a single translation
5	5	** unit. This allows many compilers to do optimizations that would not be
6	6	** possible if the files were compiled separately. Performance improvements
7	7	** of 5% or more are commonly seen when SQLite is compiled as a single
8	8	** translation unit.
		@@ -385,23 +385,29 @@
385	385	/*
386	386	** Exactly one of the following macros must be defined in order to
387	387	** specify which memory allocation subsystem to use.
388	388	**
389	389	** SQLITE_SYSTEM_MALLOC // Use normal system malloc()
	390	+** SQLITE_WIN32_MALLOC // Use Win32 native heap API
390	391	** SQLITE_MEMDEBUG // Debugging version of system malloc()
	392	+**
	393	+** On Windows, if the SQLITE_WIN32_MALLOC_VALIDATE macro is defined and the
	394	+** assert() macro is enabled, each call into the Win32 native heap subsystem
	395	+** will cause HeapValidate to be called. If heap validation should fail, an
	396	+** assertion will be triggered.
391	397	**
392	398	** (Historical note: There used to be several other options, but we've
393	399	** pared it down to just these two.)
394	400	**
395	401	** If none of the above are defined, then set SQLITE_SYSTEM_MALLOC as
396	402	** the default.
397	403	*/
398		-#if defined(SQLITE_SYSTEM_MALLOC)+defined(SQLITE_MEMDEBUG)>1
	404	+#if defined(SQLITE_SYSTEM_MALLOC)+defined(SQLITE_WIN32_MALLOC)+defined(SQLITE_MEMDEBUG)>1
399	405	# error "At most one of the following compile-time configuration options\
400		- is allows: SQLITE_SYSTEM_MALLOC, SQLITE_MEMDEBUG"
	406	+ is allows: SQLITE_SYSTEM_MALLOC, SQLITE_WIN32_MALLOC, SQLITE_MEMDEBUG"
401	407	#endif
402		-#if defined(SQLITE_SYSTEM_MALLOC)+defined(SQLITE_MEMDEBUG)==0
	408	+#if defined(SQLITE_SYSTEM_MALLOC)+defined(SQLITE_WIN32_MALLOC)+defined(SQLITE_MEMDEBUG)==0
403	409	# define SQLITE_SYSTEM_MALLOC 1
404	410	#endif
405	411
406	412	/*
407	413	** If SQLITE_MALLOC_SOFT_LIMIT is not zero, then try to keep the
		@@ -648,13 +654,13 @@
648	654	**
649	655	** See also: [sqlite3_libversion()],
650	656	** [sqlite3_libversion_number()], [sqlite3_sourceid()],
651	657	** [sqlite_version()] and [sqlite_source_id()].
652	658	*/
653		-#define SQLITE_VERSION "3.7.7"
654		-#define SQLITE_VERSION_NUMBER 3007007
655		-#define SQLITE_SOURCE_ID "2011-06-15 13:11:06 f9750870ee04935f338e4d808900fee5a8b2b389"
	659	+#define SQLITE_VERSION "3.7.8"
	660	+#define SQLITE_VERSION_NUMBER 3007008
	661	+#define SQLITE_SOURCE_ID "2011-08-29 11:56:14 639cc85a911454bffdcccb33f2976c683953ae64"
656	662
657	663	/*
658	664	** CAPI3REF: Run-Time Library Version Numbers
659	665	** KEYWORDS: sqlite3_version, sqlite3_sourceid
660	666	**
		@@ -851,11 +857,11 @@
851	857	** semicolon-separate SQL statements passed into its 2nd argument,
852	858	** in the context of the [database connection] passed in as its 1st
853	859	** argument. ^If the callback function of the 3rd argument to
854	860	** sqlite3_exec() is not NULL, then it is invoked for each result row
855	861	** coming out of the evaluated SQL statements. ^The 4th argument to
856		-** to sqlite3_exec() is relayed through to the 1st argument of each
	862	+** sqlite3_exec() is relayed through to the 1st argument of each
857	863	** callback invocation. ^If the callback pointer to sqlite3_exec()
858	864	** is NULL, then no callback is ever invoked and result rows are
859	865	** ignored.
860	866	**
861	867	** ^If an error occurs while evaluating the SQL statements passed into
		@@ -1282,20 +1288,52 @@
1282	1288	** when [PRAGMA synchronous \| PRAGMA synchronous=OFF] is set, but most
1283	1289	** VFSes do not need this signal and should silently ignore this opcode.
1284	1290	** Applications should not call [sqlite3_file_control()] with this
1285	1291	** opcode as doing so may disrupt the operation of the specialized VFSes
1286	1292	** that do require it.
	1293	+**
	1294	+** ^The [SQLITE_FCNTL_WIN32_AV_RETRY] opcode is used to configure automatic
	1295	+** retry counts and intervals for certain disk I/O operations for the
	1296	+** windows [VFS] in order to work to provide robustness against
	1297	+** anti-virus programs. By default, the windows VFS will retry file read,
	1298	+** file write, and file delete opertions up to 10 times, with a delay
	1299	+** of 25 milliseconds before the first retry and with the delay increasing
	1300	+** by an additional 25 milliseconds with each subsequent retry. This
	1301	+** opcode allows those to values (10 retries and 25 milliseconds of delay)
	1302	+** to be adjusted. The values are changed for all database connections
	1303	+** within the same process. The argument is a pointer to an array of two
	1304	+** integers where the first integer i the new retry count and the second
	1305	+** integer is the delay. If either integer is negative, then the setting
	1306	+** is not changed but instead the prior value of that setting is written
	1307	+** into the array entry, allowing the current retry settings to be
	1308	+** interrogated. The zDbName parameter is ignored.
	1309	+**
	1310	+** ^The [SQLITE_FCNTL_PERSIST_WAL] opcode is used to set or query the
	1311	+** persistent [WAL \| Write AHead Log] setting. By default, the auxiliary
	1312	+** write ahead log and shared memory files used for transaction control
	1313	+** are automatically deleted when the latest connection to the database
	1314	+** closes. Setting persistent WAL mode causes those files to persist after
	1315	+** close. Persisting the files is useful when other processes that do not
	1316	+** have write permission on the directory containing the database file want
	1317	+** to read the database file, as the WAL and shared memory files must exist
	1318	+** in order for the database to be readable. The fourth parameter to
	1319	+** [sqlite3_file_control()] for this opcode should be a pointer to an integer.
	1320	+** That integer is 0 to disable persistent WAL mode or 1 to enable persistent
	1321	+** WAL mode. If the integer is -1, then it is overwritten with the current
	1322	+** WAL persistence setting.
	1323	+**
1287	1324	*/
1288	1325	#define SQLITE_FCNTL_LOCKSTATE 1
1289	1326	#define SQLITE_GET_LOCKPROXYFILE 2
1290	1327	#define SQLITE_SET_LOCKPROXYFILE 3
1291	1328	#define SQLITE_LAST_ERRNO 4
1292	1329	#define SQLITE_FCNTL_SIZE_HINT 5
1293	1330	#define SQLITE_FCNTL_CHUNK_SIZE 6
1294	1331	#define SQLITE_FCNTL_FILE_POINTER 7
1295	1332	#define SQLITE_FCNTL_SYNC_OMITTED 8
1296		-
	1333	+#define SQLITE_FCNTL_WIN32_AV_RETRY 9
	1334	+#define SQLITE_FCNTL_PERSIST_WAL 10
1297	1335
1298	1336	/*
1299	1337	** CAPI3REF: Mutex Handle
1300	1338	**
1301	1339	** The mutex module within SQLite defines [sqlite3_mutex] to be an
		@@ -1443,11 +1481,11 @@
1443	1481	** The xSleep() method causes the calling thread to sleep for at
1444	1482	** least the number of microseconds given. ^The xCurrentTime()
1445	1483	** method returns a Julian Day Number for the current date and time as
1446	1484	** a floating point value.
1447	1485	** ^The xCurrentTimeInt64() method returns, as an integer, the Julian
1448		-** Day Number multipled by 86400000 (the number of milliseconds in
	1486	+** Day Number multiplied by 86400000 (the number of milliseconds in
1449	1487	** a 24-hour day).
1450	1488	** ^SQLite will use the xCurrentTimeInt64() method to get the current
1451	1489	** date and time if that method is available (if iVersion is 2 or
1452	1490	** greater and the function pointer is not NULL) and will fall back
1453	1491	** to xCurrentTime() if xCurrentTimeInt64() is unavailable.
		@@ -1719,20 +1757,14 @@
1719	1757	** also used during testing of SQLite in order to specify an alternative
1720	1758	** memory allocator that simulates memory out-of-memory conditions in
1721	1759	** order to verify that SQLite recovers gracefully from such
1722	1760	** conditions.
1723	1761	**
1724		-** The xMalloc and xFree methods must work like the
1725		-** malloc() and free() functions from the standard C library.
1726		-** The xRealloc method must work like realloc() from the standard C library
1727		-** with the exception that if the second argument to xRealloc is zero,
1728		-** xRealloc must be a no-op - it must not perform any allocation or
1729		-** deallocation. ^SQLite guarantees that the second argument to
	1762	+** The xMalloc, xRealloc, and xFree methods must work like the
	1763	+** malloc(), realloc() and free() functions from the standard C library.
	1764	+** ^SQLite guarantees that the second argument to
1730	1765	** xRealloc is always a value returned by a prior call to xRoundup.
1731		-** And so in cases where xRoundup always returns a positive number,
1732		-** xRealloc can perform exactly as the standard library realloc() and
1733		-** still be in compliance with this specification.
1734	1766	**
1735	1767	** xSize should return the allocated size of a memory allocation
1736	1768	** previously obtained from xMalloc or xRealloc. The allocated size
1737	1769	** is always at least as big as the requested size but may be larger.
1738	1770	**
		@@ -1881,11 +1913,11 @@
1881	1913	** scratch memory beyond what is provided by this configuration option, then
1882	1914	** [sqlite3_malloc()] will be used to obtain the memory needed.</dd>
1883	1915	**
1884	1916	** [[SQLITE_CONFIG_PAGECACHE]] <dt>SQLITE_CONFIG_PAGECACHE</dt>
1885	1917	** <dd> ^This option specifies a static memory buffer that SQLite can use for
1886		-** the database page cache with the default page cache implemenation.
	1918	+** the database page cache with the default page cache implementation.
1887	1919	** This configuration should not be used if an application-define page
1888	1920	** cache implementation is loaded using the SQLITE_CONFIG_PCACHE option.
1889	1921	** There are three arguments to this option: A pointer to 8-byte aligned
1890	1922	** memory, the size of each page buffer (sz), and the number of pages (N).
1891	1923	** The sz argument should be the size of the largest database page
		@@ -2979,16 +3011,16 @@
2979	3011	** [[URI filenames in sqlite3_open()]] <h3>URI Filenames</h3>
2980	3012	**
2981	3013	** ^If [URI filename] interpretation is enabled, and the filename argument
2982	3014	** begins with "file:", then the filename is interpreted as a URI. ^URI
2983	3015	** filename interpretation is enabled if the [SQLITE_OPEN_URI] flag is
2984		-** is set in the fourth argument to sqlite3_open_v2(), or if it has
	3016	+** set in the fourth argument to sqlite3_open_v2(), or if it has
2985	3017	** been enabled globally using the [SQLITE_CONFIG_URI] option with the
2986	3018	** [sqlite3_config()] method or by the [SQLITE_USE_URI] compile-time option.
2987	3019	** As of SQLite version 3.7.7, URI filename interpretation is turned off
2988	3020	** by default, but future releases of SQLite might enable URI filename
2989		-** intepretation by default. See "[URI filenames]" for additional
	3021	+** interpretation by default. See "[URI filenames]" for additional
2990	3022	** information.
2991	3023	**
2992	3024	** URI filenames are parsed according to RFC 3986. ^If the URI contains an
2993	3025	** authority, then it must be either an empty string or the string
2994	3026	** "localhost". ^If the authority is not an empty string or "localhost", an
		@@ -3803,11 +3835,11 @@
3803	3835	** [extended result codes] might be returned as well.
3804	3836	**
3805	3837	** ^[SQLITE_BUSY] means that the database engine was unable to acquire the
3806	3838	** database locks it needs to do its job. ^If the statement is a [COMMIT]
3807	3839	** or occurs outside of an explicit transaction, then you can retry the
3808		-** statement. If the statement is not a [COMMIT] and occurs within a
	3840	+** statement. If the statement is not a [COMMIT] and occurs within an
3809	3841	** explicit transaction then you should rollback the transaction before
3810	3842	** continuing.
3811	3843	**
3812	3844	** ^[SQLITE_DONE] means that the statement has finished executing
3813	3845	** successfully. sqlite3_step() should not be called again on this virtual
		@@ -4082,11 +4114,11 @@
4082	4114
4083	4115	/*
4084	4116	** CAPI3REF: Destroy A Prepared Statement Object
4085	4117	**
4086	4118	** ^The sqlite3_finalize() function is called to delete a [prepared statement].
4087		-** ^If the most recent evaluation of the statement encountered no errors or
	4119	+** ^If the most recent evaluation of the statement encountered no errors
4088	4120	** or if the statement is never been evaluated, then sqlite3_finalize() returns
4089	4121	** SQLITE_OK. ^If the most recent evaluation of statement S failed, then
4090	4122	** sqlite3_finalize(S) returns the appropriate [error code] or
4091	4123	** [extended error code].
4092	4124	**
		@@ -5996,11 +6028,11 @@
5996	6028	** versions of these routines, it should at least provide stubs that always
5997	6029	** return true so that one does not get spurious assertion failures.
5998	6030	**
5999	6031	** ^If the argument to sqlite3_mutex_held() is a NULL pointer then
6000	6032	** the routine should return 1. This seems counter-intuitive since
6001		-** clearly the mutex cannot be held if it does not exist. But the
	6033	+** clearly the mutex cannot be held if it does not exist. But
6002	6034	** the reason the mutex does not exist is because the build is not
6003	6035	** using mutexes. And we do not want the assert() containing the
6004	6036	** call to sqlite3_mutex_held() to fail, so a non-zero return is
6005	6037	** the appropriate thing to do. ^The sqlite3_mutex_notheld()
6006	6038	** interface should also return 1 when given a NULL pointer.
		@@ -6119,11 +6151,12 @@
6119	6151	#define SQLITE_TESTCTRL_RESERVE 14
6120	6152	#define SQLITE_TESTCTRL_OPTIMIZATIONS 15
6121	6153	#define SQLITE_TESTCTRL_ISKEYWORD 16
6122	6154	#define SQLITE_TESTCTRL_PGHDRSZ 17
6123	6155	#define SQLITE_TESTCTRL_SCRATCHMALLOC 18
6124		-#define SQLITE_TESTCTRL_LAST 18
	6156	+#define SQLITE_TESTCTRL_LOCALTIME_FAULT 19
	6157	+#define SQLITE_TESTCTRL_LAST 19
6125	6158
6126	6159	/*
6127	6160	** CAPI3REF: SQLite Runtime Status
6128	6161	**
6129	6162	** ^This interface is used to retrieve runtime status information
		@@ -6505,11 +6538,11 @@
6505	6538	** [[the xFetch() page cache methods]]
6506	6539	** The xFetch() method locates a page in the cache and returns a pointer to
6507	6540	** the page, or a NULL pointer.
6508	6541	** A "page", in this context, means a buffer of szPage bytes aligned at an
6509	6542	** 8-byte boundary. The page to be fetched is determined by the key. ^The
6510		-** mimimum key value is 1. After it has been retrieved using xFetch, the page
	6543	+** minimum key value is 1. After it has been retrieved using xFetch, the page
6511	6544	** is considered to be "pinned".
6512	6545	**
6513	6546	** If the requested page is already in the page cache, then the page cache
6514	6547	** implementation must return a pointer to the page buffer with its content
6515	6548	** intact. If the requested page is not already in the cache, then the
		@@ -7599,10 +7632,18 @@
7599	7632	*/
7600	7633	#ifndef SQLITE_TEMP_STORE
7601	7634	# define SQLITE_TEMP_STORE 1
7602	7635	#endif
7603	7636
	7637	+/*
	7638	+** If all temporary storage is in-memory, then omit the external merge-sort
	7639	+** logic since it is superfluous.
	7640	+*/
	7641	+#if SQLITE_TEMP_STORE==3 && !defined(SQLITE_OMIT_MERGE_SORT)
	7642	+# define SQLITE_OMIT_MERGE_SORT
	7643	+#endif
	7644	+
7604	7645	/*
7605	7646	** GCC does not define the offsetof() macro so we'll have to do it
7606	7647	** ourselves.
7607	7648	*/
7608	7649	#ifndef offsetof
		@@ -7941,10 +7982,11 @@
7941	7982	#define BTREE_OMIT_JOURNAL 1 /* Do not create or use a rollback journal */
7942	7983	#define BTREE_NO_READLOCK 2 /* Omit readlocks on readonly files */
7943	7984	#define BTREE_MEMORY 4 /* This is an in-memory DB */
7944	7985	#define BTREE_SINGLE 8 /* The file contains at most 1 b-tree */
7945	7986	#define BTREE_UNORDERED 16 /* Use of a hash implementation is OK */
	7987	+#define BTREE_SORTER 32 /* Used as accumulator in external merge sort */
7946	7988
7947	7989	SQLITE_PRIVATE int sqlite3BtreeClose(Btree*);
7948	7990	SQLITE_PRIVATE int sqlite3BtreeSetCacheSize(Btree*,int);
7949	7991	SQLITE_PRIVATE int sqlite3BtreeSetSafetyLevel(Btree*,int,int,int);
7950	7992	SQLITE_PRIVATE int sqlite3BtreeSyncDisabled(Btree*);
		@@ -8143,10 +8185,11 @@
8143	8185	** or VDBE. The VDBE implements an abstract machine that runs a
8144	8186	** simple program to access and modify the underlying database.
8145	8187	*/
8146	8188	#ifndef _SQLITE_VDBE_H_
8147	8189	#define _SQLITE_VDBE_H_
	8190	+/* #include <stdio.h> */
8148	8191
8149	8192	/*
8150	8193	** A single VDBE is an opaque structure named "Vdbe". Only routines
8151	8194	** in the source file sqliteVdbe.c are allowed to see the insides
8152	8195	** of this structure.
		@@ -8186,10 +8229,11 @@
8186	8229	Mem pMem; / Used when p4type is P4_MEM */
8187	8230	VTable pVtab; / Used when p4type is P4_VTAB */
8188	8231	KeyInfo pKeyInfo; / Used when p4type is P4_KEYINFO */
8189	8232	int ai; / Used when p4type is P4_INTARRAY */
8190	8233	SubProgram pProgram; / Used when p4type is P4_SUBPROGRAM */
	8234	+ int (xAdvance)(BtCursor , int *);
8191	8235	} p4;
8192	8236	#ifdef SQLITE_DEBUG
8193	8237	char zComment; / Comment to improve readability */
8194	8238	#endif
8195	8239	#ifdef VDBE_PROFILE
		@@ -8241,10 +8285,11 @@
8241	8285	#define P4_REAL (-12) /* P4 is a 64-bit floating point value */
8242	8286	#define P4_INT64 (-13) /* P4 is a 64-bit signed integer */
8243	8287	#define P4_INT32 (-14) /* P4 is a 32-bit signed integer */
8244	8288	#define P4_INTARRAY (-15) /* P4 is a vector of 32-bit integers */
8245	8289	#define P4_SUBPROGRAM (-18) /* P4 is a pointer to a SubProgram structure */
	8290	+#define P4_ADVANCE (-19) /* P4 is a pointer to BtreeNext() or BtreePrev() */
8246	8291
8247	8292	/* When adding a P4 argument using P4_KEYINFO, a copy of the KeyInfo structure
8248	8293	** is made. That copy is freed when the Vdbe is finalized. But if the
8249	8294	** argument is P4_KEYINFO_HANDOFF, the passed in pointer is used. It still
8250	8295	** gets freed when the Vdbe is finalized so it still should be obtained
		@@ -8354,89 +8399,89 @@
8354	8399	#define OP_ReadCookie 35
8355	8400	#define OP_SetCookie 36
8356	8401	#define OP_VerifyCookie 37
8357	8402	#define OP_OpenRead 38
8358	8403	#define OP_OpenWrite 39
8359		-#define OP_OpenAutoindex 40
8360		-#define OP_OpenEphemeral 41
8361		-#define OP_OpenPseudo 42
8362		-#define OP_Close 43
8363		-#define OP_SeekLt 44
8364		-#define OP_SeekLe 45
8365		-#define OP_SeekGe 46
8366		-#define OP_SeekGt 47
8367		-#define OP_Seek 48
8368		-#define OP_NotFound 49
8369		-#define OP_Found 50
8370		-#define OP_IsUnique 51
8371		-#define OP_NotExists 52
8372		-#define OP_Sequence 53
8373		-#define OP_NewRowid 54
8374		-#define OP_Insert 55
8375		-#define OP_InsertInt 56
8376		-#define OP_Delete 57
8377		-#define OP_ResetCount 58
8378		-#define OP_RowKey 59
8379		-#define OP_RowData 60
8380		-#define OP_Rowid 61
8381		-#define OP_NullRow 62
8382		-#define OP_Last 63
8383		-#define OP_Sort 64
8384		-#define OP_Rewind 65
8385		-#define OP_Prev 66
8386		-#define OP_Next 67
8387		-#define OP_IdxInsert 70
8388		-#define OP_IdxDelete 71
8389		-#define OP_IdxRowid 72
8390		-#define OP_IdxLT 81
8391		-#define OP_IdxGE 92
8392		-#define OP_Destroy 95
8393		-#define OP_Clear 96
8394		-#define OP_CreateIndex 97
8395		-#define OP_CreateTable 98
8396		-#define OP_ParseSchema 99
8397		-#define OP_LoadAnalysis 100
8398		-#define OP_DropTable 101
8399		-#define OP_DropIndex 102
8400		-#define OP_DropTrigger 103
8401		-#define OP_IntegrityCk 104
8402		-#define OP_RowSetAdd 105
8403		-#define OP_RowSetRead 106
8404		-#define OP_RowSetTest 107
8405		-#define OP_Program 108
8406		-#define OP_Param 109
8407		-#define OP_FkCounter 110
8408		-#define OP_FkIfZero 111
8409		-#define OP_MemMax 112
8410		-#define OP_IfPos 113
8411		-#define OP_IfNeg 114
8412		-#define OP_IfZero 115
8413		-#define OP_AggStep 116
8414		-#define OP_AggFinal 117
8415		-#define OP_Checkpoint 118
8416		-#define OP_JournalMode 119
8417		-#define OP_Vacuum 120
8418		-#define OP_IncrVacuum 121
8419		-#define OP_Expire 122
8420		-#define OP_TableLock 123
8421		-#define OP_VBegin 124
8422		-#define OP_VCreate 125
8423		-#define OP_VDestroy 126
8424		-#define OP_VOpen 127
8425		-#define OP_VFilter 128
8426		-#define OP_VColumn 129
8427		-#define OP_VNext 131
8428		-#define OP_VRename 132
8429		-#define OP_VUpdate 133
8430		-#define OP_Pagecount 134
8431		-#define OP_MaxPgcnt 135
8432		-#define OP_Trace 136
8433		-#define OP_Noop 137
8434		-#define OP_Explain 138
8435		-
8436		-/* The following opcode values are never used */
8437		-#define OP_NotUsed_139 139
	8404	+#define OP_OpenSorter 40
	8405	+#define OP_OpenAutoindex 41
	8406	+#define OP_OpenEphemeral 42
	8407	+#define OP_OpenPseudo 43
	8408	+#define OP_Close 44
	8409	+#define OP_SeekLt 45
	8410	+#define OP_SeekLe 46
	8411	+#define OP_SeekGe 47
	8412	+#define OP_SeekGt 48
	8413	+#define OP_Seek 49
	8414	+#define OP_NotFound 50
	8415	+#define OP_Found 51
	8416	+#define OP_IsUnique 52
	8417	+#define OP_NotExists 53
	8418	+#define OP_Sequence 54
	8419	+#define OP_NewRowid 55
	8420	+#define OP_Insert 56
	8421	+#define OP_InsertInt 57
	8422	+#define OP_Delete 58
	8423	+#define OP_ResetCount 59
	8424	+#define OP_RowKey 60
	8425	+#define OP_RowData 61
	8426	+#define OP_Rowid 62
	8427	+#define OP_NullRow 63
	8428	+#define OP_Last 64
	8429	+#define OP_Sort 65
	8430	+#define OP_Rewind 66
	8431	+#define OP_Prev 67
	8432	+#define OP_Next 70
	8433	+#define OP_IdxInsert 71
	8434	+#define OP_IdxDelete 72
	8435	+#define OP_IdxRowid 81
	8436	+#define OP_IdxLT 92
	8437	+#define OP_IdxGE 95
	8438	+#define OP_Destroy 96
	8439	+#define OP_Clear 97
	8440	+#define OP_CreateIndex 98
	8441	+#define OP_CreateTable 99
	8442	+#define OP_ParseSchema 100
	8443	+#define OP_LoadAnalysis 101
	8444	+#define OP_DropTable 102
	8445	+#define OP_DropIndex 103
	8446	+#define OP_DropTrigger 104
	8447	+#define OP_IntegrityCk 105
	8448	+#define OP_RowSetAdd 106
	8449	+#define OP_RowSetRead 107
	8450	+#define OP_RowSetTest 108
	8451	+#define OP_Program 109
	8452	+#define OP_Param 110
	8453	+#define OP_FkCounter 111
	8454	+#define OP_FkIfZero 112
	8455	+#define OP_MemMax 113
	8456	+#define OP_IfPos 114
	8457	+#define OP_IfNeg 115
	8458	+#define OP_IfZero 116
	8459	+#define OP_AggStep 117
	8460	+#define OP_AggFinal 118
	8461	+#define OP_Checkpoint 119
	8462	+#define OP_JournalMode 120
	8463	+#define OP_Vacuum 121
	8464	+#define OP_IncrVacuum 122
	8465	+#define OP_Expire 123
	8466	+#define OP_TableLock 124
	8467	+#define OP_VBegin 125
	8468	+#define OP_VCreate 126
	8469	+#define OP_VDestroy 127
	8470	+#define OP_VOpen 128
	8471	+#define OP_VFilter 129
	8472	+#define OP_VColumn 131
	8473	+#define OP_VNext 132
	8474	+#define OP_VRename 133
	8475	+#define OP_VUpdate 134
	8476	+#define OP_Pagecount 135
	8477	+#define OP_MaxPgcnt 136
	8478	+#define OP_Trace 137
	8479	+#define OP_Noop 138
	8480	+#define OP_Explain 139
	8481	+
	8482	+/* The following opcode values are never used */
8438	8483	#define OP_NotUsed_140 140
8439	8484
8440	8485
8441	8486	/* Properties such as "out2" or "jump" that are specified in
8442	8487	** comments following the "case" for each opcode in the vdbe.c
		@@ -8453,23 +8498,23 @@
8453	8498	/* 0 */ 0x00, 0x01, 0x05, 0x04, 0x04, 0x10, 0x00, 0x02,\
8454	8499	/* 8 */ 0x02, 0x02, 0x02, 0x02, 0x02, 0x00, 0x24, 0x24,\
8455	8500	/* 16 */ 0x00, 0x00, 0x00, 0x24, 0x04, 0x05, 0x04, 0x00,\
8456	8501	/* 24 */ 0x00, 0x01, 0x05, 0x05, 0x00, 0x00, 0x00, 0x02,\
8457	8502	/* 32 */ 0x00, 0x00, 0x00, 0x02, 0x10, 0x00, 0x00, 0x00,\
8458		-/* 40 */ 0x00, 0x00, 0x00, 0x00, 0x11, 0x11, 0x11, 0x11,\
8459		-/* 48 */ 0x08, 0x11, 0x11, 0x11, 0x11, 0x02, 0x02, 0x00,\
8460		-/* 56 */ 0x00, 0x00, 0x00, 0x00, 0x00, 0x02, 0x00, 0x01,\
8461		-/* 64 */ 0x01, 0x01, 0x01, 0x01, 0x4c, 0x4c, 0x08, 0x00,\
8462		-/* 72 */ 0x02, 0x05, 0x05, 0x15, 0x15, 0x15, 0x15, 0x15,\
8463		-/* 80 */ 0x15, 0x01, 0x4c, 0x4c, 0x4c, 0x4c, 0x4c, 0x4c,\
8464		-/* 88 */ 0x4c, 0x4c, 0x4c, 0x4c, 0x01, 0x24, 0x02, 0x02,\
8465		-/* 96 */ 0x00, 0x02, 0x02, 0x00, 0x00, 0x00, 0x00, 0x00,\
8466		-/* 104 */ 0x00, 0x0c, 0x45, 0x15, 0x01, 0x02, 0x00, 0x01,\
8467		-/* 112 */ 0x08, 0x05, 0x05, 0x05, 0x00, 0x00, 0x00, 0x02,\
8468		-/* 120 */ 0x00, 0x01, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00,\
8469		-/* 128 */ 0x01, 0x00, 0x02, 0x01, 0x00, 0x00, 0x02, 0x02,\
8470		-/* 136 */ 0x00, 0x00, 0x00, 0x00, 0x00, 0x04, 0x04, 0x04,\
	8503	+/* 40 */ 0x00, 0x00, 0x00, 0x00, 0x00, 0x11, 0x11, 0x11,\
	8504	+/* 48 */ 0x11, 0x08, 0x11, 0x11, 0x11, 0x11, 0x02, 0x02,\
	8505	+/* 56 */ 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x02, 0x00,\
	8506	+/* 64 */ 0x01, 0x01, 0x01, 0x01, 0x4c, 0x4c, 0x01, 0x08,\
	8507	+/* 72 */ 0x00, 0x05, 0x05, 0x15, 0x15, 0x15, 0x15, 0x15,\
	8508	+/* 80 */ 0x15, 0x02, 0x4c, 0x4c, 0x4c, 0x4c, 0x4c, 0x4c,\
	8509	+/* 88 */ 0x4c, 0x4c, 0x4c, 0x4c, 0x01, 0x24, 0x02, 0x01,\
	8510	+/* 96 */ 0x02, 0x00, 0x02, 0x02, 0x00, 0x00, 0x00, 0x00,\
	8511	+/* 104 */ 0x00, 0x00, 0x0c, 0x45, 0x15, 0x01, 0x02, 0x00,\
	8512	+/* 112 */ 0x01, 0x08, 0x05, 0x05, 0x05, 0x00, 0x00, 0x00,\
	8513	+/* 120 */ 0x02, 0x00, 0x01, 0x00, 0x00, 0x00, 0x00, 0x00,\
	8514	+/* 128 */ 0x00, 0x01, 0x02, 0x00, 0x01, 0x00, 0x00, 0x02,\
	8515	+/* 136 */ 0x02, 0x00, 0x00, 0x00, 0x00, 0x04, 0x04, 0x04,\
8471	8516	/* 144 */ 0x04, 0x04,}
8472	8517
8473	8518	/************ End of opcodes.h *******************************************/
8474	8519	/************ Continuing where we left off in vdbe.h *********************/
8475	8520
		@@ -8484,13 +8529,13 @@
8484	8529	SQLITE_PRIVATE int sqlite3VdbeAddOp3(Vdbe*,int,int,int,int);
8485	8530	SQLITE_PRIVATE int sqlite3VdbeAddOp4(Vdbe,int,int,int,int,const char zP4,int);
8486	8531	SQLITE_PRIVATE int sqlite3VdbeAddOp4Int(Vdbe*,int,int,int,int,int);
8487	8532	SQLITE_PRIVATE int sqlite3VdbeAddOpList(Vdbe, int nOp, VdbeOpList const aOp);
8488	8533	SQLITE_PRIVATE void sqlite3VdbeAddParseSchemaOp(Vdbe,int,char);
8489		-SQLITE_PRIVATE void sqlite3VdbeChangeP1(Vdbe*, int addr, int P1);
8490		-SQLITE_PRIVATE void sqlite3VdbeChangeP2(Vdbe*, int addr, int P2);
8491		-SQLITE_PRIVATE void sqlite3VdbeChangeP3(Vdbe*, int addr, int P3);
	8534	+SQLITE_PRIVATE void sqlite3VdbeChangeP1(Vdbe*, u32 addr, int P1);
	8535	+SQLITE_PRIVATE void sqlite3VdbeChangeP2(Vdbe*, u32 addr, int P2);
	8536	+SQLITE_PRIVATE void sqlite3VdbeChangeP3(Vdbe*, u32 addr, int P3);
8492	8537	SQLITE_PRIVATE void sqlite3VdbeChangeP5(Vdbe*, u8 P5);
8493	8538	SQLITE_PRIVATE void sqlite3VdbeJumpHere(Vdbe*, int addr);
8494	8539	SQLITE_PRIVATE void sqlite3VdbeChangeToNoop(Vdbe*, int addr, int N);
8495	8540	SQLITE_PRIVATE void sqlite3VdbeChangeP4(Vdbe, int addr, const char zP4, int N);
8496	8541	SQLITE_PRIVATE void sqlite3VdbeUsesBtree(Vdbe*, int);
		@@ -8608,10 +8653,11 @@
8608	8653	** NOTE: These values must match the corresponding BTREE_ values in btree.h.
8609	8654	*/
8610	8655	#define PAGER_OMIT_JOURNAL 0x0001 /* Do not use a rollback journal */
8611	8656	#define PAGER_NO_READLOCK 0x0002 /* Omit readlocks on readonly files */
8612	8657	#define PAGER_MEMORY 0x0004 /* In-memory database */
	8658	+#define PAGER_SORTER 0x0020 /* Accumulator in external merge sort */
8613	8659
8614	8660	/*
8615	8661	** Valid values for the second argument to sqlite3PagerLockingMode().
8616	8662	*/
8617	8663	#define PAGER_LOCKINGMODE_QUERY -1
		@@ -8703,10 +8749,13 @@
8703	8749	SQLITE_PRIVATE sqlite3_file sqlite3PagerFile(Pager);
8704	8750	SQLITE_PRIVATE const char sqlite3PagerJournalname(Pager);
8705	8751	SQLITE_PRIVATE int sqlite3PagerNosync(Pager*);
8706	8752	SQLITE_PRIVATE void sqlite3PagerTempSpace(Pager);
8707	8753	SQLITE_PRIVATE int sqlite3PagerIsMemdb(Pager*);
	8754	+#ifndef SQLITE_OMIT_MERGE_SORT
	8755	+SQLITE_PRIVATE int sqlite3PagerUnderStress(Pager*);
	8756	+#endif
8708	8757
8709	8758	/* Functions used to truncate the database file. */
8710	8759	SQLITE_PRIVATE void sqlite3PagerTruncateImage(Pager*,Pgno);
8711	8760
8712	8761	#if defined(SQLITE_HAS_CODEC) && !defined(SQLITE_OMIT_WAL)
		@@ -9556,10 +9605,12 @@
9556	9605	#define SQLITE_IndexSort 0x04 /* Disable indexes for sorting */
9557	9606	#define SQLITE_IndexSearch 0x08 /* Disable indexes for searching */
9558	9607	#define SQLITE_IndexCover 0x10 /* Disable index covering table */
9559	9608	#define SQLITE_GroupByOrder 0x20 /* Disable GROUPBY cover of ORDERBY */
9560	9609	#define SQLITE_FactorOutConst 0x40 /* Disable factoring out constants */
	9610	+#define SQLITE_IdxRealAsInt 0x80 /* Store REAL as INT in indices */
	9611	+#define SQLITE_DistinctOpt 0x80 /* DISTINCT using indexes */
9561	9612	#define SQLITE_OptMask 0xff /* Mask of all disablable opts */
9562	9613
9563	9614	/*
9564	9615	** Possible values for the sqlite.magic field.
9565	9616	** The numbers are obtained at random and have no special meaning, other
		@@ -10447,10 +10498,11 @@
10447	10498	Table pTab; / An SQL table corresponding to zName */
10448	10499	Select pSelect; / A SELECT statement used in place of a table name */
10449	10500	u8 isPopulated; /* Temporary table associated with SELECT is populated */
10450	10501	u8 jointype; /* Type of join between this able and the previous */
10451	10502	u8 notIndexed; /* True if there is a NOT INDEXED clause */
	10503	+ u8 isCorrelated; /* True if sub-query is correlated */
10452	10504	#ifndef SQLITE_OMIT_EXPLAIN
10453	10505	u8 iSelectId; /* If pSelect!=0, the id of the sub-select in EQP */
10454	10506	#endif
10455	10507	int iCursor; /* The VDBE cursor number used to access this table */
10456	10508	Expr pOn; / The ON clause of a join */
		@@ -10566,10 +10618,11 @@
10566	10618	struct WhereInfo {
10567	10619	Parse pParse; / Parsing and code generating context */
10568	10620	u16 wctrlFlags; /* Flags originally passed to sqlite3WhereBegin() */
10569	10621	u8 okOnePass; /* Ok to use one-pass algorithm for UPDATE or DELETE */
10570	10622	u8 untestedTerms; /* Not all WHERE terms resolved by outer loop */
	10623	+ u8 eDistinct;
10571	10624	SrcList pTabList; / List of tables in the join */
10572	10625	int iTop; /* The very beginning of the WHERE loop */
10573	10626	int iContinue; /* Jump here to continue with next record */
10574	10627	int iBreak; /* Jump here to break out of the loop */
10575	10628	int nLevel; /* Number of nested loop */
		@@ -10577,10 +10630,13 @@
10577	10630	double savedNQueryLoop; /* pParse->nQueryLoop outside the WHERE loop */
10578	10631	double nRowOut; /* Estimated number of output rows */
10579	10632	WhereLevel a[1]; /* Information about each nest loop in WHERE */
10580	10633	};
10581	10634
	10635	+#define WHERE_DISTINCT_UNIQUE 1
	10636	+#define WHERE_DISTINCT_ORDERED 2
	10637	+
10582	10638	/*
10583	10639	** A NameContext defines a context in which to resolve table and column
10584	10640	** names. The context consists of a list of tables (the pSrcList) field and
10585	10641	** a list of named expression (pEList). The named expression list may
10586	10642	** be NULL. The pSrc corresponds to the FROM clause of a SELECT or
		@@ -11053,10 +11109,11 @@
11053	11109	int isPCacheInit; /* True after malloc is initialized */
11054	11110	sqlite3_mutex pInitMutex; / Mutex used by sqlite3_initialize() */
11055	11111	int nRefInitMutex; /* Number of users of pInitMutex */
11056	11112	void (xLog)(void,int,const char); / Function for logging */
11057	11113	void pLogArg; / First argument to xLog() */
	11114	+ int bLocaltimeFault; /* True to fail localtime() calls */
11058	11115	};
11059	11116
11060	11117	/*
11061	11118	** Context pointer passed down through the tree-walk.
11062	11119	*/
		@@ -11337,11 +11394,11 @@
11337	11394	#if defined(SQLITE_ENABLE_UPDATE_DELETE_LIMIT) && !defined(SQLITE_OMIT_SUBQUERY)
11338	11395	SQLITE_PRIVATE Expr sqlite3LimitWhere(Parse , SrcList , Expr , ExprList , Expr , Expr , char );
11339	11396	#endif
11340	11397	SQLITE_PRIVATE void sqlite3DeleteFrom(Parse, SrcList, Expr*);
11341	11398	SQLITE_PRIVATE void sqlite3Update(Parse, SrcList, ExprList, Expr, int);
11342		-SQLITE_PRIVATE WhereInfo sqlite3WhereBegin(Parse, SrcList, Expr, ExprList**, u16);
	11399	+SQLITE_PRIVATE WhereInfo sqlite3WhereBegin(Parse, SrcList, Expr, ExprList*,ExprList,u16);
11343	11400	SQLITE_PRIVATE void sqlite3WhereEnd(WhereInfo*);
11344	11401	SQLITE_PRIVATE int sqlite3ExprCodeGetColumn(Parse, Table, int, int, int);
11345	11402	SQLITE_PRIVATE void sqlite3ExprCodeGetColumnOfTable(Vdbe, Table, int, int, int);
11346	11403	SQLITE_PRIVATE void sqlite3ExprCodeMove(Parse*, int, int, int);
11347	11404	SQLITE_PRIVATE void sqlite3ExprCodeCopy(Parse*, int, int, int);
		@@ -12013,10 +12070,11 @@
12013	12070	0, /* isPCacheInit */
12014	12071	0, /* pInitMutex */
12015	12072	0, /* nRefInitMutex */
12016	12073	0, /* xLog */
12017	12074	0, /* pLogArg */
	12075	+ 0, /* bLocaltimeFault */
12018	12076	};
12019	12077
12020	12078
12021	12079	/*
12022	12080	** Hash table for global functions - functions common to all
		@@ -12321,10 +12379,13 @@
12321	12379	#ifdef SQLITE_OMIT_LOOKASIDE
12322	12380	"OMIT_LOOKASIDE",
12323	12381	#endif
12324	12382	#ifdef SQLITE_OMIT_MEMORYDB
12325	12383	"OMIT_MEMORYDB",
	12384	+#endif
	12385	+#ifdef SQLITE_OMIT_MERGE_SORT
	12386	+ "OMIT_MERGE_SORT",
12326	12387	#endif
12327	12388	#ifdef SQLITE_OMIT_OR_OPTIMIZATION
12328	12389	"OMIT_OR_OPTIMIZATION",
12329	12390	#endif
12330	12391	#ifdef SQLITE_OMIT_PAGER_PRAGMAS
		@@ -12387,10 +12448,13 @@
12387	12448	#ifdef SQLITE_OMIT_WSD
12388	12449	"OMIT_WSD",
12389	12450	#endif
12390	12451	#ifdef SQLITE_OMIT_XFER_OPT
12391	12452	"OMIT_XFER_OPT",
	12453	+#endif
	12454	+#ifdef SQLITE_PAGECACHE_BLOCKALLOC
	12455	+ "PAGECACHE_BLOCKALLOC",
12392	12456	#endif
12393	12457	#ifdef SQLITE_PERFORMANCE_TRACE
12394	12458	"PERFORMANCE_TRACE",
12395	12459	#endif
12396	12460	#ifdef SQLITE_PROXY_DEBUG
		@@ -12508,10 +12572,13 @@
12508	12572	/*
12509	12573	** Boolean values
12510	12574	*/
12511	12575	typedef unsigned char Bool;
12512	12576
	12577	+/* Opaque type used by code in vdbesort.c */
	12578	+typedef struct VdbeSorter VdbeSorter;
	12579	+
12513	12580	/*
12514	12581	** A cursor is a pointer into a single BTree within a database file.
12515	12582	** The cursor can seek to a BTree entry with a particular key, or
12516	12583	** loop over all entries of the Btree. You can also insert new BTree
12517	12584	** entries or retrieve the key or data from the entry that the cursor
		@@ -12539,10 +12606,11 @@
12539	12606	sqlite3_vtab_cursor pVtabCursor; / The cursor for a virtual table */
12540	12607	const sqlite3_module pModule; / Module for cursor pVtabCursor */
12541	12608	i64 seqCount; /* Sequence counter */
12542	12609	i64 movetoTarget; /* Argument to the deferred sqlite3BtreeMoveto() */
12543	12610	i64 lastRowid; /* Last rowid from a Next or NextIdx operation */
	12611	+ VdbeSorter pSorter; / Sorter object for OP_OpenSorter cursors */
12544	12612
12545	12613	/* Result of last sqlite3BtreeMoveto() done by an OP_NotExists or
12546	12614	** OP_IsUnique opcode on this cursor. */
12547	12615	int seekResult;
12548	12616
		@@ -12858,17 +12926,36 @@
12858	12926	SQLITE_PRIVATE int sqlite3VdbeMemRealify(Mem*);
12859	12927	SQLITE_PRIVATE int sqlite3VdbeMemNumerify(Mem*);
12860	12928	SQLITE_PRIVATE int sqlite3VdbeMemFromBtree(BtCursor,int,int,int,Mem);
12861	12929	SQLITE_PRIVATE void sqlite3VdbeMemRelease(Mem *p);
12862	12930	SQLITE_PRIVATE void sqlite3VdbeMemReleaseExternal(Mem *p);
	12931	+#define MemReleaseExt(X) \
	12932	+ if((X)->flags&(MEM_Agg\|MEM_Dyn\|MEM_RowSet\|MEM_Frame)) \
	12933	+ sqlite3VdbeMemReleaseExternal(X);
12863	12934	SQLITE_PRIVATE int sqlite3VdbeMemFinalize(Mem, FuncDef);
12864	12935	SQLITE_PRIVATE const char *sqlite3OpcodeName(int);
12865	12936	SQLITE_PRIVATE int sqlite3VdbeMemGrow(Mem *pMem, int n, int preserve);
12866	12937	SQLITE_PRIVATE int sqlite3VdbeCloseStatement(Vdbe *, int);
12867	12938	SQLITE_PRIVATE void sqlite3VdbeFrameDelete(VdbeFrame*);
12868	12939	SQLITE_PRIVATE int sqlite3VdbeFrameRestore(VdbeFrame *);
12869	12940	SQLITE_PRIVATE void sqlite3VdbeMemStoreType(Mem *pMem);
	12941	+
	12942	+#ifdef SQLITE_OMIT_MERGE_SORT
	12943	+# define sqlite3VdbeSorterInit(Y,Z) SQLITE_OK
	12944	+# define sqlite3VdbeSorterWrite(X,Y,Z) SQLITE_OK
	12945	+# define sqlite3VdbeSorterClose(Y,Z)
	12946	+# define sqlite3VdbeSorterRowkey(Y,Z) SQLITE_OK
	12947	+# define sqlite3VdbeSorterRewind(X,Y,Z) SQLITE_OK
	12948	+# define sqlite3VdbeSorterNext(X,Y,Z) SQLITE_OK
	12949	+#else
	12950	+SQLITE_PRIVATE int sqlite3VdbeSorterInit(sqlite3 , VdbeCursor );
	12951	+SQLITE_PRIVATE int sqlite3VdbeSorterWrite(sqlite3 , VdbeCursor , int);
	12952	+SQLITE_PRIVATE void sqlite3VdbeSorterClose(sqlite3 , VdbeCursor );
	12953	+SQLITE_PRIVATE int sqlite3VdbeSorterRowkey(VdbeCursor , Mem );
	12954	+SQLITE_PRIVATE int sqlite3VdbeSorterRewind(sqlite3 , VdbeCursor , int *);
	12955	+SQLITE_PRIVATE int sqlite3VdbeSorterNext(sqlite3 , VdbeCursor , int *);
	12956	+#endif
12870	12957
12871	12958	#if !defined(SQLITE_OMIT_SHARED_CACHE) && SQLITE_THREADSAFE>0
12872	12959	SQLITE_PRIVATE void sqlite3VdbeEnter(Vdbe*);
12873	12960	SQLITE_PRIVATE void sqlite3VdbeLeave(Vdbe*);
12874	12961	#else
		@@ -13159,30 +13246,16 @@
13159	13246	** Astronomical Algorithms, 2nd Edition, 1998
13160	13247	** ISBM 0-943396-61-1
13161	13248	** Willmann-Bell, Inc
13162	13249	** Richmond, Virginia (USA)
13163	13250	*/
	13251	+/* #include <stdlib.h> */
	13252	+/* #include <assert.h> */
13164	13253	#include <time.h>
13165	13254
13166	13255	#ifndef SQLITE_OMIT_DATETIME_FUNCS
13167	13256
13168		-/*
13169		-** On recent Windows platforms, the localtime_s() function is available
13170		-** as part of the "Secure CRT". It is essentially equivalent to
13171		-** localtime_r() available under most POSIX platforms, except that the
13172		-** order of the parameters is reversed.
13173		-**
13174		-** See http://msdn.microsoft.com/en-us/library/a442x3ye(VS.80).aspx.
13175		-**
13176		-** If the user has not indicated to use localtime_r() or localtime_s()
13177		-** already, check for an MSVC build environment that provides
13178		-** localtime_s().
13179		-*/
13180		-#if !defined(HAVE_LOCALTIME_R) && !defined(HAVE_LOCALTIME_S) && \
13181		- defined(_MSC_VER) && defined(_CRT_INSECURE_DEPRECATE)
13182		-#define HAVE_LOCALTIME_S 1
13183		-#endif
13184	13257
13185	13258	/*
13186	13259	** A structure for holding a single date and time.
13187	13260	*/
13188	13261	typedef struct DateTime DateTime;
		@@ -13523,20 +13596,90 @@
13523	13596	static void clearYMD_HMS_TZ(DateTime *p){
13524	13597	p->validYMD = 0;
13525	13598	p->validHMS = 0;
13526	13599	p->validTZ = 0;
13527	13600	}
	13601	+
	13602	+/*
	13603	+** On recent Windows platforms, the localtime_s() function is available
	13604	+** as part of the "Secure CRT". It is essentially equivalent to
	13605	+** localtime_r() available under most POSIX platforms, except that the
	13606	+** order of the parameters is reversed.
	13607	+**
	13608	+** See http://msdn.microsoft.com/en-us/library/a442x3ye(VS.80).aspx.
	13609	+**
	13610	+** If the user has not indicated to use localtime_r() or localtime_s()
	13611	+** already, check for an MSVC build environment that provides
	13612	+** localtime_s().
	13613	+*/
	13614	+#if !defined(HAVE_LOCALTIME_R) && !defined(HAVE_LOCALTIME_S) && \
	13615	+ defined(_MSC_VER) && defined(_CRT_INSECURE_DEPRECATE)
	13616	+#define HAVE_LOCALTIME_S 1
	13617	+#endif
	13618	+
	13619	+#ifndef SQLITE_OMIT_LOCALTIME
	13620	+/*
	13621	+** The following routine implements the rough equivalent of localtime_r()
	13622	+** using whatever operating-system specific localtime facility that
	13623	+** is available. This routine returns 0 on success and
	13624	+** non-zero on any kind of error.
	13625	+**
	13626	+** If the sqlite3GlobalConfig.bLocaltimeFault variable is true then this
	13627	+** routine will always fail.
	13628	+*/
	13629	+static int osLocaltime(time_t t, struct tm pTm){
	13630	+ int rc;
	13631	+#if (!defined(HAVE_LOCALTIME_R) \|\| !HAVE_LOCALTIME_R) \
	13632	+ && (!defined(HAVE_LOCALTIME_S) \|\| !HAVE_LOCALTIME_S)
	13633	+ struct tm *pX;
	13634	+#if SQLITE_THREADSAFE>0
	13635	+ sqlite3_mutex *mutex = sqlite3MutexAlloc(SQLITE_MUTEX_STATIC_MASTER);
	13636	+#endif
	13637	+ sqlite3_mutex_enter(mutex);
	13638	+ pX = localtime(t);
	13639	+#ifndef SQLITE_OMIT_BUILTIN_TEST
	13640	+ if( sqlite3GlobalConfig.bLocaltimeFault ) pX = 0;
	13641	+#endif
	13642	+ if( pX ) pTm = pX;
	13643	+ sqlite3_mutex_leave(mutex);
	13644	+ rc = pX==0;
	13645	+#else
	13646	+#ifndef SQLITE_OMIT_BUILTIN_TEST
	13647	+ if( sqlite3GlobalConfig.bLocaltimeFault ) return 1;
	13648	+#endif
	13649	+#if defined(HAVE_LOCALTIME_R) && HAVE_LOCALTIME_R
	13650	+ rc = localtime_r(t, pTm)==0;
	13651	+#else
	13652	+ rc = localtime_s(pTm, t);
	13653	+#endif /* HAVE_LOCALTIME_R */
	13654	+#endif /* HAVE_LOCALTIME_R \|\| HAVE_LOCALTIME_S */
	13655	+ return rc;
	13656	+}
	13657	+#endif /* SQLITE_OMIT_LOCALTIME */
	13658	+
13528	13659
13529	13660	#ifndef SQLITE_OMIT_LOCALTIME
13530	13661	/*
13531		-** Compute the difference (in milliseconds)
13532		-** between localtime and UTC (a.k.a. GMT)
13533		-** for the time value p where p is in UTC.
	13662	+** Compute the difference (in milliseconds) between localtime and UTC
	13663	+** (a.k.a. GMT) for the time value p where p is in UTC. If no error occurs,
	13664	+** return this value and set *pRc to SQLITE_OK.
	13665	+**
	13666	+** Or, if an error does occur, set *pRc to SQLITE_ERROR. The returned value
	13667	+** is undefined in this case.
13534	13668	*/
13535		-static sqlite3_int64 localtimeOffset(DateTime *p){
	13669	+static sqlite3_int64 localtimeOffset(
	13670	+ DateTime p, / Date at which to calculate offset */
	13671	+ sqlite3_context pCtx, / Write error here if one occurs */
	13672	+ int pRc / OUT: Error code. SQLITE_OK or ERROR */
	13673	+){
13536	13674	DateTime x, y;
13537	13675	time_t t;
	13676	+ struct tm sLocal;
	13677	+
	13678	+ /* Initialize the contents of sLocal to avoid a compiler warning. */
	13679	+ memset(&sLocal, 0, sizeof(sLocal));
	13680	+
13538	13681	x = *p;
13539	13682	computeYMD_HMS(&x);
13540	13683	if( x.Y<1971 \|\| x.Y>=2038 ){
13541	13684	x.Y = 2000;
13542	13685	x.M = 1;
		@@ -13550,51 +13693,27 @@
13550	13693	}
13551	13694	x.tz = 0;
13552	13695	x.validJD = 0;
13553	13696	computeJD(&x);
13554	13697	t = (time_t)(x.iJD/1000 - 21086676*(i64)10000);
13555		-#ifdef HAVE_LOCALTIME_R
13556		- {
13557		- struct tm sLocal;
13558		- localtime_r(&t, &sLocal);
13559		- y.Y = sLocal.tm_year + 1900;
13560		- y.M = sLocal.tm_mon + 1;
13561		- y.D = sLocal.tm_mday;
13562		- y.h = sLocal.tm_hour;
13563		- y.m = sLocal.tm_min;
13564		- y.s = sLocal.tm_sec;
13565		- }
13566		-#elif defined(HAVE_LOCALTIME_S) && HAVE_LOCALTIME_S
13567		- {
13568		- struct tm sLocal;
13569		- localtime_s(&sLocal, &t);
13570		- y.Y = sLocal.tm_year + 1900;
13571		- y.M = sLocal.tm_mon + 1;
13572		- y.D = sLocal.tm_mday;
13573		- y.h = sLocal.tm_hour;
13574		- y.m = sLocal.tm_min;
13575		- y.s = sLocal.tm_sec;
13576		- }
13577		-#else
13578		- {
13579		- struct tm *pTm;
13580		- sqlite3_mutex_enter(sqlite3MutexAlloc(SQLITE_MUTEX_STATIC_MASTER));
13581		- pTm = localtime(&t);
13582		- y.Y = pTm->tm_year + 1900;
13583		- y.M = pTm->tm_mon + 1;
13584		- y.D = pTm->tm_mday;
13585		- y.h = pTm->tm_hour;
13586		- y.m = pTm->tm_min;
13587		- y.s = pTm->tm_sec;
13588		- sqlite3_mutex_leave(sqlite3MutexAlloc(SQLITE_MUTEX_STATIC_MASTER));
13589		- }
13590		-#endif
	13698	+ if( osLocaltime(&t, &sLocal) ){
	13699	+ sqlite3_result_error(pCtx, "local time unavailable", -1);
	13700	+ *pRc = SQLITE_ERROR;
	13701	+ return 0;
	13702	+ }
	13703	+ y.Y = sLocal.tm_year + 1900;
	13704	+ y.M = sLocal.tm_mon + 1;
	13705	+ y.D = sLocal.tm_mday;
	13706	+ y.h = sLocal.tm_hour;
	13707	+ y.m = sLocal.tm_min;
	13708	+ y.s = sLocal.tm_sec;
13591	13709	y.validYMD = 1;
13592	13710	y.validHMS = 1;
13593	13711	y.validJD = 0;
13594	13712	y.validTZ = 0;
13595	13713	computeJD(&y);
	13714	+ *pRc = SQLITE_OK;
13596	13715	return y.iJD - x.iJD;
13597	13716	}
13598	13717	#endif /* SQLITE_OMIT_LOCALTIME */
13599	13718
13600	13719	/*
		@@ -13614,13 +13733,16 @@
13614	13733	** weekday N
13615	13734	** unixepoch
13616	13735	** localtime
13617	13736	** utc
13618	13737	**
13619		-** Return 0 on success and 1 if there is any kind of error.
	13738	+** Return 0 on success and 1 if there is any kind of error. If the error
	13739	+** is in a system call (i.e. localtime()), then an error message is written
	13740	+** to context pCtx. If the error is an unrecognized modifier, no error is
	13741	+** written to pCtx.
13620	13742	*/
13621		-static int parseModifier(const char zMod, DateTime p){
	13743	+static int parseModifier(sqlite3_context pCtx, const char zMod, DateTime *p){
13622	13744	int rc = 1;
13623	13745	int n;
13624	13746	double r;
13625	13747	char *z, zBuf[30];
13626	13748	z = zBuf;
		@@ -13636,13 +13758,12 @@
13636	13758	** Assuming the current time value is UTC (a.k.a. GMT), shift it to
13637	13759	** show local time.
13638	13760	*/
13639	13761	if( strcmp(z, "localtime")==0 ){
13640	13762	computeJD(p);
13641		- p->iJD += localtimeOffset(p);
	13763	+ p->iJD += localtimeOffset(p, pCtx, &rc);
13642	13764	clearYMD_HMS_TZ(p);
13643		- rc = 0;
13644	13765	}
13645	13766	break;
13646	13767	}
13647	13768	#endif
13648	13769	case 'u': {
		@@ -13659,15 +13780,16 @@
13659	13780	}
13660	13781	#ifndef SQLITE_OMIT_LOCALTIME
13661	13782	else if( strcmp(z, "utc")==0 ){
13662	13783	sqlite3_int64 c1;
13663	13784	computeJD(p);
13664		- c1 = localtimeOffset(p);
13665		- p->iJD -= c1;
13666		- clearYMD_HMS_TZ(p);
13667		- p->iJD += c1 - localtimeOffset(p);
13668		- rc = 0;
	13785	+ c1 = localtimeOffset(p, pCtx, &rc);
	13786	+ if( rc==SQLITE_OK ){
	13787	+ p->iJD -= c1;
	13788	+ clearYMD_HMS_TZ(p);
	13789	+ p->iJD += c1 - localtimeOffset(p, pCtx, &rc);
	13790	+ }
13669	13791	}
13670	13792	#endif
13671	13793	break;
13672	13794	}
13673	13795	case 'w': {
		@@ -13844,13 +13966,12 @@
13844	13966	if( !z \|\| parseDateOrTime(context, (char*)z, p) ){
13845	13967	return 1;
13846	13968	}
13847	13969	}
13848	13970	for(i=1; i<argc; i++){
13849		- if( (z = sqlite3_value_text(argv[i]))==0 \|\| parseModifier((char*)z, p) ){
13850		- return 1;
13851		- }
	13971	+ z = sqlite3_value_text(argv[i]);
	13972	+ if( z==0 \|\| parseModifier(context, (char*)z, p) ) return 1;
13852	13973	}
13853	13974	return 0;
13854	13975	}
13855	13976
13856	13977
		@@ -14344,11 +14465,11 @@
14344	14465	DO_OS_MALLOC_TEST(0);
14345	14466	/* 0x87f3f is a mask of SQLITE_OPEN_ flags that are valid to be passed
14346	14467	** down into the VFS layer. Some SQLITE_OPEN_ flags (for example,
14347	14468	** SQLITE_OPEN_FULLMUTEX or SQLITE_OPEN_SHAREDCACHE) are blocked before
14348	14469	** reaching the VFS. */
14349		- rc = pVfs->xOpen(pVfs, zPath, pFile, flags & 0x87f3f, pFlagsOut);
	14470	+ rc = pVfs->xOpen(pVfs, zPath, pFile, flags & 0x87f7f, pFlagsOut);
14350	14471	assert( rc==SQLITE_OK \|\| pFile->pMethods==0 );
14351	14472	return rc;
14352	14473	}
14353	14474	SQLITE_PRIVATE int sqlite3OsDelete(sqlite3_vfs pVfs, const char zPath, int dirSync){
14354	14475	return pVfs->xDelete(pVfs, zPath, dirSync);
		@@ -14880,10 +15001,11 @@
14880	15001	extern void backtrace_symbols_fd(voidconst,int,int);
14881	15002	#else
14882	15003	# define backtrace(A,B) 1
14883	15004	# define backtrace_symbols_fd(A,B,C)
14884	15005	#endif
	15006	+/* #include <stdio.h> */
14885	15007
14886	15008	/*
14887	15009	** Each memory allocation looks like this:
14888	15010	**
14889	15011	** ------------------------------------------------------------------------
		@@ -15805,11 +15927,11 @@
15805	15927	** Free an outstanding memory allocation.
15806	15928	**
15807	15929	** This function assumes that the necessary mutexes, if any, are
15808	15930	** already held by the caller. Hence "Unsafe".
15809	15931	*/
15810		-void memsys3FreeUnsafe(void *pOld){
	15932	+static void memsys3FreeUnsafe(void *pOld){
15811	15933	Mem3Block p = (Mem3Block)pOld;
15812	15934	int i;
15813	15935	u32 size, x;
15814	15936	assert( sqlite3_mutex_held(mem3.mutex) );
15815	15937	assert( p>mem3.aPool && p<&mem3.aPool[mem3.nPool] );
		@@ -15880,21 +16002,21 @@
15880	16002	}
15881	16003
15882	16004	/*
15883	16005	** Free memory.
15884	16006	*/
15885		-void memsys3Free(void *pPrior){
	16007	+static void memsys3Free(void *pPrior){
15886	16008	assert( pPrior );
15887	16009	memsys3Enter();
15888	16010	memsys3FreeUnsafe(pPrior);
15889	16011	memsys3Leave();
15890	16012	}
15891	16013
15892	16014	/*
15893	16015	** Change the size of an existing memory allocation
15894	16016	*/
15895		-void memsys3Realloc(void pPrior, int nBytes){
	16017	+static void memsys3Realloc(void pPrior, int nBytes){
15896	16018	int nOld;
15897	16019	void *p;
15898	16020	if( pPrior==0 ){
15899	16021	return sqlite3_malloc(nBytes);
15900	16022	}
		@@ -17983,10 +18105,11 @@
17983	18105	**
17984	18106	*************************************************************************
17985	18107	**
17986	18108	** Memory allocation functions used throughout sqlite.
17987	18109	*/
	18110	+/* #include <stdarg.h> */
17988	18111
17989	18112	/*
17990	18113	** Attempt to release up to n bytes of non-essential memory currently
17991	18114	** held by SQLite. An example of non-essential memory is memory used to
17992	18115	** cache database pages that are not currently in use.
		@@ -19960,10 +20083,11 @@
19960	20083	** BOM or Byte Order Mark:
19961	20084	** 0xff 0xfe little-endian utf-16 follows
19962	20085	** 0xfe 0xff big-endian utf-16 follows
19963	20086	**
19964	20087	*/
	20088	+/* #include <assert.h> */
19965	20089
19966	20090	#ifndef SQLITE_AMALGAMATION
19967	20091	/*
19968	20092	** The following constant value is used by the SQLITE_BIGENDIAN and
19969	20093	** SQLITE_LITTLEENDIAN macros.
		@@ -20502,10 +20626,11 @@
20502	20626	**
20503	20627	** This file contains functions for allocating memory, comparing
20504	20628	** strings, and stuff like that.
20505	20629	**
20506	20630	*/
	20631	+/* #include <stdarg.h> */
20507	20632	#ifdef SQLITE_HAVE_ISNAN
20508	20633	# include <math.h>
20509	20634	#endif
20510	20635
20511	20636	/*
		@@ -21634,26 +21759,32 @@
21634	21759	return -x;
21635	21760	}
21636	21761
21637	21762	#ifdef SQLITE_ENABLE_8_3_NAMES
21638	21763	/*
21639		-** If SQLITE_ENABLE_8_3_NAME is set at compile-time and if the database
	21764	+** If SQLITE_ENABLE_8_3_NAMES is set at compile-time and if the database
21640	21765	** filename in zBaseFilename is a URI with the "8_3_names=1" parameter and
21641	21766	** if filename in z[] has a suffix (a.k.a. "extension") that is longer than
21642	21767	** three characters, then shorten the suffix on z[] to be the last three
21643	21768	** characters of the original suffix.
	21769	+**
	21770	+** If SQLITE_ENABLE_8_3_NAMES is set to 2 at compile-time, then always
	21771	+** do the suffix shortening regardless of URI parameter.
21644	21772	**
21645	21773	** Examples:
21646	21774	**
21647	21775	** test.db-journal => test.nal
21648	21776	** test.db-wal => test.wal
21649	21777	** test.db-shm => test.shm
21650	21778	*/
21651	21779	SQLITE_PRIVATE void sqlite3FileSuffix3(const char zBaseFilename, char z){
	21780	+#if SQLITE_ENABLE_8_3_NAMES<2
21652	21781	const char *zOk;
21653	21782	zOk = sqlite3_uri_parameter(zBaseFilename, "8_3_names");
21654		- if( zOk && sqlite3GetBoolean(zOk) ){
	21783	+ if( zOk && sqlite3GetBoolean(zOk) )
	21784	+#endif
	21785	+ {
21655	21786	int i, sz;
21656	21787	sz = sqlite3Strlen30(z);
21657	21788	for(i=sz-1; i>0 && z[i]!='/' && z[i]!='.'; i--){}
21658	21789	if( z[i]=='.' && ALWAYS(sz>i+4) ) memcpy(&z[i+1], &z[sz-3], 4);
21659	21790	}
		@@ -21674,10 +21805,11 @@
21674	21805	**
21675	21806	*************************************************************************
21676	21807	** This is the implementation of generic hash-tables
21677	21808	** used in SQLite.
21678	21809	*/
	21810	+/* #include <assert.h> */
21679	21811
21680	21812	/* Turn bulk memory into a hash table object by initializing the
21681	21813	** fields of the Hash structure.
21682	21814	**
21683	21815	** "pNew" is a pointer to the hash table that is to be initialized.
		@@ -21982,52 +22114,52 @@
21982	22114	/* 35 */ "ReadCookie",
21983	22115	/* 36 */ "SetCookie",
21984	22116	/* 37 */ "VerifyCookie",
21985	22117	/* 38 */ "OpenRead",
21986	22118	/* 39 */ "OpenWrite",
21987		- /* 40 */ "OpenAutoindex",
21988		- /* 41 */ "OpenEphemeral",
21989		- /* 42 */ "OpenPseudo",
21990		- /* 43 */ "Close",
21991		- /* 44 */ "SeekLt",
21992		- /* 45 */ "SeekLe",
21993		- /* 46 */ "SeekGe",
21994		- /* 47 */ "SeekGt",
21995		- /* 48 */ "Seek",
21996		- /* 49 */ "NotFound",
21997		- /* 50 */ "Found",
21998		- /* 51 */ "IsUnique",
21999		- /* 52 */ "NotExists",
22000		- /* 53 */ "Sequence",
22001		- /* 54 */ "NewRowid",
22002		- /* 55 */ "Insert",
22003		- /* 56 */ "InsertInt",
22004		- /* 57 */ "Delete",
22005		- /* 58 */ "ResetCount",
22006		- /* 59 */ "RowKey",
22007		- /* 60 */ "RowData",
22008		- /* 61 */ "Rowid",
22009		- /* 62 */ "NullRow",
22010		- /* 63 */ "Last",
22011		- /* 64 */ "Sort",
22012		- /* 65 */ "Rewind",
22013		- /* 66 */ "Prev",
22014		- /* 67 */ "Next",
	22119	+ /* 40 */ "OpenSorter",
	22120	+ /* 41 */ "OpenAutoindex",
	22121	+ /* 42 */ "OpenEphemeral",
	22122	+ /* 43 */ "OpenPseudo",
	22123	+ /* 44 */ "Close",
	22124	+ /* 45 */ "SeekLt",
	22125	+ /* 46 */ "SeekLe",
	22126	+ /* 47 */ "SeekGe",
	22127	+ /* 48 */ "SeekGt",
	22128	+ /* 49 */ "Seek",
	22129	+ /* 50 */ "NotFound",
	22130	+ /* 51 */ "Found",
	22131	+ /* 52 */ "IsUnique",
	22132	+ /* 53 */ "NotExists",
	22133	+ /* 54 */ "Sequence",
	22134	+ /* 55 */ "NewRowid",
	22135	+ /* 56 */ "Insert",
	22136	+ /* 57 */ "InsertInt",
	22137	+ /* 58 */ "Delete",
	22138	+ /* 59 */ "ResetCount",
	22139	+ /* 60 */ "RowKey",
	22140	+ /* 61 */ "RowData",
	22141	+ /* 62 */ "Rowid",
	22142	+ /* 63 */ "NullRow",
	22143	+ /* 64 */ "Last",
	22144	+ /* 65 */ "Sort",
	22145	+ /* 66 */ "Rewind",
	22146	+ /* 67 */ "Prev",
22015	22147	/* 68 */ "Or",
22016	22148	/* 69 */ "And",
22017		- /* 70 */ "IdxInsert",
22018		- /* 71 */ "IdxDelete",
22019		- /* 72 */ "IdxRowid",
	22149	+ /* 70 */ "Next",
	22150	+ /* 71 */ "IdxInsert",
	22151	+ /* 72 */ "IdxDelete",
22020	22152	/* 73 */ "IsNull",
22021	22153	/* 74 */ "NotNull",
22022	22154	/* 75 */ "Ne",
22023	22155	/* 76 */ "Eq",
22024	22156	/* 77 */ "Gt",
22025	22157	/* 78 */ "Le",
22026	22158	/* 79 */ "Lt",
22027	22159	/* 80 */ "Ge",
22028		- /* 81 */ "IdxLT",
	22160	+ /* 81 */ "IdxRowid",
22029	22161	/* 82 */ "BitAnd",
22030	22162	/* 83 */ "BitOr",
22031	22163	/* 84 */ "ShiftLeft",
22032	22164	/* 85 */ "ShiftRight",
22033	22165	/* 86 */ "Add",
		@@ -22034,58 +22166,58 @@
22034	22166	/* 87 */ "Subtract",
22035	22167	/* 88 */ "Multiply",
22036	22168	/* 89 */ "Divide",
22037	22169	/* 90 */ "Remainder",
22038	22170	/* 91 */ "Concat",
22039		- /* 92 */ "IdxGE",
	22171	+ /* 92 */ "IdxLT",
22040	22172	/* 93 */ "BitNot",
22041	22173	/* 94 */ "String8",
22042		- /* 95 */ "Destroy",
22043		- /* 96 */ "Clear",
22044		- /* 97 */ "CreateIndex",
22045		- /* 98 */ "CreateTable",
22046		- /* 99 */ "ParseSchema",
22047		- /* 100 */ "LoadAnalysis",
22048		- /* 101 */ "DropTable",
22049		- /* 102 */ "DropIndex",
22050		- /* 103 */ "DropTrigger",
22051		- /* 104 */ "IntegrityCk",
22052		- /* 105 */ "RowSetAdd",
22053		- /* 106 */ "RowSetRead",
22054		- /* 107 */ "RowSetTest",
22055		- /* 108 */ "Program",
22056		- /* 109 */ "Param",
22057		- /* 110 */ "FkCounter",
22058		- /* 111 */ "FkIfZero",
22059		- /* 112 */ "MemMax",
22060		- /* 113 */ "IfPos",
22061		- /* 114 */ "IfNeg",
22062		- /* 115 */ "IfZero",
22063		- /* 116 */ "AggStep",
22064		- /* 117 */ "AggFinal",
22065		- /* 118 */ "Checkpoint",
22066		- /* 119 */ "JournalMode",
22067		- /* 120 */ "Vacuum",
22068		- /* 121 */ "IncrVacuum",
22069		- /* 122 */ "Expire",
22070		- /* 123 */ "TableLock",
22071		- /* 124 */ "VBegin",
22072		- /* 125 */ "VCreate",
22073		- /* 126 */ "VDestroy",
22074		- /* 127 */ "VOpen",
22075		- /* 128 */ "VFilter",
22076		- /* 129 */ "VColumn",
	22174	+ /* 95 */ "IdxGE",
	22175	+ /* 96 */ "Destroy",
	22176	+ /* 97 */ "Clear",
	22177	+ /* 98 */ "CreateIndex",
	22178	+ /* 99 */ "CreateTable",
	22179	+ /* 100 */ "ParseSchema",
	22180	+ /* 101 */ "LoadAnalysis",
	22181	+ /* 102 */ "DropTable",
	22182	+ /* 103 */ "DropIndex",
	22183	+ /* 104 */ "DropTrigger",
	22184	+ /* 105 */ "IntegrityCk",
	22185	+ /* 106 */ "RowSetAdd",
	22186	+ /* 107 */ "RowSetRead",
	22187	+ /* 108 */ "RowSetTest",
	22188	+ /* 109 */ "Program",
	22189	+ /* 110 */ "Param",
	22190	+ /* 111 */ "FkCounter",
	22191	+ /* 112 */ "FkIfZero",
	22192	+ /* 113 */ "MemMax",
	22193	+ /* 114 */ "IfPos",
	22194	+ /* 115 */ "IfNeg",
	22195	+ /* 116 */ "IfZero",
	22196	+ /* 117 */ "AggStep",
	22197	+ /* 118 */ "AggFinal",
	22198	+ /* 119 */ "Checkpoint",
	22199	+ /* 120 */ "JournalMode",
	22200	+ /* 121 */ "Vacuum",
	22201	+ /* 122 */ "IncrVacuum",
	22202	+ /* 123 */ "Expire",
	22203	+ /* 124 */ "TableLock",
	22204	+ /* 125 */ "VBegin",
	22205	+ /* 126 */ "VCreate",
	22206	+ /* 127 */ "VDestroy",
	22207	+ /* 128 */ "VOpen",
	22208	+ /* 129 */ "VFilter",
22077	22209	/* 130 */ "Real",
22078		- /* 131 */ "VNext",
22079		- /* 132 */ "VRename",
22080		- /* 133 */ "VUpdate",
22081		- /* 134 */ "Pagecount",
22082		- /* 135 */ "MaxPgcnt",
22083		- /* 136 */ "Trace",
22084		- /* 137 */ "Noop",
22085		- /* 138 */ "Explain",
22086		- /* 139 */ "NotUsed_139",
	22210	+ /* 131 */ "VColumn",
	22211	+ /* 132 */ "VNext",
	22212	+ /* 133 */ "VRename",
	22213	+ /* 134 */ "VUpdate",
	22214	+ /* 135 */ "Pagecount",
	22215	+ /* 136 */ "MaxPgcnt",
	22216	+ /* 137 */ "Trace",
	22217	+ /* 138 */ "Noop",
	22218	+ /* 139 */ "Explain",
22087	22219	/* 140 */ "NotUsed_140",
22088	22220	/* 141 */ "ToText",
22089	22221	/* 142 */ "ToBlob",
22090	22222	/* 143 */ "ToNumeric",
22091	22223	/* 144 */ "ToInt",
		@@ -22183,14 +22315,17 @@
22183	22315	#ifdef MEMORY_DEBUG
22184	22316	# error "The MEMORY_DEBUG macro is obsolete. Use SQLITE_DEBUG instead."
22185	22317	#endif
22186	22318
22187	22319	#ifdef SQLITE_DEBUG
22188		-SQLITE_PRIVATE int sqlite3OSTrace = 0;
22189		-#define OSTRACE(X) if( sqlite3OSTrace ) sqlite3DebugPrintf X
	22320	+# ifndef SQLITE_DEBUG_OS_TRACE
	22321	+# define SQLITE_DEBUG_OS_TRACE 0
	22322	+# endif
	22323	+ int sqlite3OSTrace = SQLITE_DEBUG_OS_TRACE;
	22324	+# define OSTRACE(X) if( sqlite3OSTrace ) sqlite3DebugPrintf X
22190	22325	#else
22191		-#define OSTRACE(X)
	22326	+# define OSTRACE(X)
22192	22327	#endif
22193	22328
22194	22329	/*
22195	22330	** Macros for performance tracing. Normally turned off. Only works
22196	22331	** on i486 hardware.
		@@ -24343,10 +24478,11 @@
24343	24478	*/
24344	24479	#include <sys/types.h>
24345	24480	#include <sys/stat.h>
24346	24481	#include <fcntl.h>
24347	24482	#include <unistd.h>
	24483	+/* #include <time.h> */
24348	24484	#include <sys/time.h>
24349	24485	#include <errno.h>
24350	24486	#ifndef SQLITE_OMIT_WAL
24351	24487	#include <sys/mman.h>
24352	24488	#endif
		@@ -24364,10 +24500,14 @@
24364	24500
24365	24501	#if defined(__APPLE__) \|\| (SQLITE_ENABLE_LOCKING_STYLE && !OS_VXWORKS)
24366	24502	# include <sys/mount.h>
24367	24503	#endif
24368	24504
	24505	+#ifdef HAVE_UTIME
	24506	+# include <utime.h>
	24507	+#endif
	24508	+
24369	24509	/*
24370	24510	** Allowed values of unixFile.fsFlags
24371	24511	*/
24372	24512	#define SQLITE_FSFLAGS_IS_MSDOS 0x1
24373	24513
		@@ -24374,10 +24514,11 @@
24374	24514	/*
24375	24515	** If we are to be thread-safe, include the pthreads header and define
24376	24516	** the SQLITE_UNIX_THREADS macro.
24377	24517	*/
24378	24518	#if SQLITE_THREADSAFE
	24519	+/* # include <pthread.h> */
24379	24520	# define SQLITE_UNIX_THREADS 1
24380	24521	#endif
24381	24522
24382	24523	/*
24383	24524	** Default permissions when creating a new file
		@@ -24429,11 +24570,10 @@
24429	24570	typedef struct unixFile unixFile;
24430	24571	struct unixFile {
24431	24572	sqlite3_io_methods const pMethod; / Always the first entry */
24432	24573	unixInodeInfo pInode; / Info about locks on this inode */
24433	24574	int h; /* The file descriptor */
24434		- int dirfd; /* File descriptor for the directory */
24435	24575	unsigned char eFileLock; /* The type of lock held on this fd */
24436	24576	unsigned char ctrlFlags; /* Behavioral bits. UNIXFILE_* flags */
24437	24577	int lastErrno; /* The unix errno from last I/O error */
24438	24578	void lockingContext; / Locking style specific state */
24439	24579	UnixUnusedFd pUnused; / Pre-allocated UnixUnusedFd */
		@@ -24471,12 +24611,18 @@
24471	24611	};
24472	24612
24473	24613	/*
24474	24614	** Allowed values for the unixFile.ctrlFlags bitmask:
24475	24615	*/
24476		-#define UNIXFILE_EXCL 0x01 /* Connections from one process only */
24477		-#define UNIXFILE_RDONLY 0x02 /* Connection is read only */
	24616	+#define UNIXFILE_EXCL 0x01 /* Connections from one process only */
	24617	+#define UNIXFILE_RDONLY 0x02 /* Connection is read only */
	24618	+#define UNIXFILE_PERSIST_WAL 0x04 /* Persistent WAL mode */
	24619	+#ifndef SQLITE_DISABLE_DIRSYNC
	24620	+# define UNIXFILE_DIRSYNC 0x08 /* Directory sync needed */
	24621	+#else
	24622	+# define UNIXFILE_DIRSYNC 0x00
	24623	+#endif
24478	24624
24479	24625	/*
24480	24626	** Include code that is common to all os_*.c files
24481	24627	*/
24482	24628	/************ Include os_common.h in the middle of os_unix.c *************/
		@@ -24511,14 +24657,17 @@
24511	24657	#ifdef MEMORY_DEBUG
24512	24658	# error "The MEMORY_DEBUG macro is obsolete. Use SQLITE_DEBUG instead."
24513	24659	#endif
24514	24660
24515	24661	#ifdef SQLITE_DEBUG
24516		-SQLITE_PRIVATE int sqlite3OSTrace = 0;
24517		-#define OSTRACE(X) if( sqlite3OSTrace ) sqlite3DebugPrintf X
	24662	+# ifndef SQLITE_DEBUG_OS_TRACE
	24663	+# define SQLITE_DEBUG_OS_TRACE 0
	24664	+# endif
	24665	+ int sqlite3OSTrace = SQLITE_DEBUG_OS_TRACE;
	24666	+# define OSTRACE(X) if( sqlite3OSTrace ) sqlite3DebugPrintf X
24518	24667	#else
24519		-#define OSTRACE(X)
	24668	+# define OSTRACE(X)
24520	24669	#endif
24521	24670
24522	24671	/*
24523	24672	** Macros for performance tracing. Normally turned off. Only works
24524	24673	** on i486 hardware.
		@@ -24723,10 +24872,13 @@
24723	24872	*/
24724	24873	static int posixOpen(const char *zFile, int flags, int mode){
24725	24874	return open(zFile, flags, mode);
24726	24875	}
24727	24876
	24877	+/* Forward reference */
	24878	+static int openDirectory(const char, int);
	24879	+
24728	24880	/*
24729	24881	** Many system calls are accessed through pointer-to-functions so that
24730	24882	** they may be overridden at runtime to facilitate fault injection during
24731	24883	** testing and sandboxing. The following array holds the names and pointers
24732	24884	** to all overrideable system calls.
		@@ -24819,10 +24971,16 @@
24819	24971	#else
24820	24972	{ "fallocate", (sqlite3_syscall_ptr)0, 0 },
24821	24973	#endif
24822	24974	#define osFallocate ((int(*)(int,off_t,off_t))aSyscall[15].pCurrent)
24823	24975
	24976	+ { "unlink", (sqlite3_syscall_ptr)unlink, 0 },
	24977	+#define osUnlink ((int()(const char))aSyscall[16].pCurrent)
	24978	+
	24979	+ { "openDirectory", (sqlite3_syscall_ptr)openDirectory, 0 },
	24980	+#define osOpenDirectory ((int()(const char,int*))aSyscall[17].pCurrent)
	24981	+
24824	24982	}; /* End of the overrideable system calls */
24825	24983
24826	24984	/*
24827	24985	** This is the xSetSystemCall() method of sqlite3_vfs for all of the
24828	24986	** "unix" VFSes. Return SQLITE_OK opon successfully updating the
		@@ -25103,11 +25261,13 @@
25103	25261	case EINVAL:
25104	25262	case ENOTCONN:
25105	25263	case ENODEV:
25106	25264	case ENXIO:
25107	25265	case ENOENT:
	25266	+#ifdef ESTALE /* ESTALE is not defined on Interix systems */
25108	25267	case ESTALE:
	25268	+#endif
25109	25269	case ENOSYS:
25110	25270	/* these should force the client to close the file and reconnect */
25111	25271
25112	25272	default:
25113	25273	return sqliteIOErr;
		@@ -26173,22 +26333,18 @@
26173	26333	** even on VxWorks. A mutex will be acquired on VxWorks by the
26174	26334	** vxworksReleaseFileId() routine.
26175	26335	*/
26176	26336	static int closeUnixFile(sqlite3_file *id){
26177	26337	unixFile pFile = (unixFile)id;
26178		- if( pFile->dirfd>=0 ){
26179		- robust_close(pFile, pFile->dirfd, __LINE__);
26180		- pFile->dirfd=-1;
26181		- }
26182	26338	if( pFile->h>=0 ){
26183	26339	robust_close(pFile, pFile->h, __LINE__);
26184	26340	pFile->h = -1;
26185	26341	}
26186	26342	#if OS_VXWORKS
26187	26343	if( pFile->pId ){
26188	26344	if( pFile->isDelete ){
26189		- unlink(pFile->pId->zCanonicalName);
	26345	+ osUnlink(pFile->pId->zCanonicalName);
26190	26346	}
26191	26347	vxworksReleaseFileId(pFile->pId);
26192	26348	pFile->pId = 0;
26193	26349	}
26194	26350	#endif
		@@ -26369,12 +26525,14 @@
26369	26525	/* If we have any lock, then the lock file already exists. All we have
26370	26526	** to do is adjust our internal record of the lock level.
26371	26527	*/
26372	26528	if( pFile->eFileLock > NO_LOCK ){
26373	26529	pFile->eFileLock = eFileLock;
26374		-#if !OS_VXWORKS
26375	26530	/* Always update the timestamp on the old file */
	26531	+#ifdef HAVE_UTIME
	26532	+ utime(zLockFile, NULL);
	26533	+#else
26376	26534	utimes(zLockFile, NULL);
26377	26535	#endif
26378	26536	return SQLITE_OK;
26379	26537	}
26380	26538
		@@ -26431,11 +26589,11 @@
26431	26589	return SQLITE_OK;
26432	26590	}
26433	26591
26434	26592	/* To fully unlock the database, delete the lock file */
26435	26593	assert( eFileLock==NO_LOCK );
26436		- if( unlink(zLockFile) ){
	26594	+ if( osUnlink(zLockFile) ){
26437	26595	int rc = 0;
26438	26596	int tErrno = errno;
26439	26597	if( ENOENT != tErrno ){
26440	26598	rc = SQLITE_IOERR_UNLOCK;
26441	26599	}
		@@ -26937,15 +27095,16 @@
26937	27095	*/
26938	27096	static int afpCheckReservedLock(sqlite3_file id, int pResOut){
26939	27097	int rc = SQLITE_OK;
26940	27098	int reserved = 0;
26941	27099	unixFile pFile = (unixFile)id;
	27100	+ afpLockingContext *context;
26942	27101
26943	27102	SimulateIOError( return SQLITE_IOERR_CHECKRESERVEDLOCK; );
26944	27103
26945	27104	assert( pFile );
26946		- afpLockingContext context = (afpLockingContext ) pFile->lockingContext;
	27105	+ context = (afpLockingContext *) pFile->lockingContext;
26947	27106	if( context->reserved ){
26948	27107	*pResOut = 1;
26949	27108	return SQLITE_OK;
26950	27109	}
26951	27110	unixEnterMutex(); /* Because pFile->pInode is shared across threads */
		@@ -27081,11 +27240,11 @@
27081	27240
27082	27241	/* If control gets to this point, then actually go ahead and make
27083	27242	** operating system calls for the specified lock.
27084	27243	*/
27085	27244	if( eFileLock==SHARED_LOCK ){
27086		- int lrc1, lrc2, lrc1Errno;
	27245	+ int lrc1, lrc2, lrc1Errno = 0;
27087	27246	long lk, mask;
27088	27247
27089	27248	assert( pInode->nShared==0 );
27090	27249	assert( pInode->eFileLock==0 );
27091	27250
		@@ -27455,21 +27614,23 @@
27455	27614	#if defined(USE_PREAD)
27456	27615	do{ got = osPwrite(id->h, pBuf, cnt, offset); }while( got<0 && errno==EINTR );
27457	27616	#elif defined(USE_PREAD64)
27458	27617	do{ got = osPwrite64(id->h, pBuf, cnt, offset);}while( got<0 && errno==EINTR);
27459	27618	#else
27460		- newOffset = lseek(id->h, offset, SEEK_SET);
27461		- SimulateIOError( newOffset-- );
27462		- if( newOffset!=offset ){
27463		- if( newOffset == -1 ){
27464		- ((unixFile*)id)->lastErrno = errno;
27465		- }else{
27466		- ((unixFile*)id)->lastErrno = 0;
27467		- }
27468		- return -1;
27469		- }
27470		- do{ got = osWrite(id->h, pBuf, cnt); }while( got<0 && errno==EINTR );
	27619	+ do{
	27620	+ newOffset = lseek(id->h, offset, SEEK_SET);
	27621	+ SimulateIOError( newOffset-- );
	27622	+ if( newOffset!=offset ){
	27623	+ if( newOffset == -1 ){
	27624	+ ((unixFile*)id)->lastErrno = errno;
	27625	+ }else{
	27626	+ ((unixFile*)id)->lastErrno = 0;
	27627	+ }
	27628	+ return -1;
	27629	+ }
	27630	+ got = osWrite(id->h, pBuf, cnt);
	27631	+ }while( got<0 && errno==EINTR );
27471	27632	#endif
27472	27633	TIMER_END;
27473	27634	if( got<0 ){
27474	27635	((unixFile*)id)->lastErrno = errno;
27475	27636	}
		@@ -27666,10 +27827,54 @@
27666	27827	if( OS_VXWORKS && rc!= -1 ){
27667	27828	rc = 0;
27668	27829	}
27669	27830	return rc;
27670	27831	}
	27832	+
	27833	+/*
	27834	+** Open a file descriptor to the directory containing file zFilename.
	27835	+** If successful, *pFd is set to the opened file descriptor and
	27836	+** SQLITE_OK is returned. If an error occurs, either SQLITE_NOMEM
	27837	+** or SQLITE_CANTOPEN is returned and *pFd is set to an undefined
	27838	+** value.
	27839	+**
	27840	+** The directory file descriptor is used for only one thing - to
	27841	+** fsync() a directory to make sure file creation and deletion events
	27842	+** are flushed to disk. Such fsyncs are not needed on newer
	27843	+** journaling filesystems, but are required on older filesystems.
	27844	+**
	27845	+** This routine can be overridden using the xSetSysCall interface.
	27846	+** The ability to override this routine was added in support of the
	27847	+** chromium sandbox. Opening a directory is a security risk (we are
	27848	+** told) so making it overrideable allows the chromium sandbox to
	27849	+** replace this routine with a harmless no-op. To make this routine
	27850	+** a no-op, replace it with a stub that returns SQLITE_OK but leaves
	27851	+** *pFd set to a negative number.
	27852	+**
	27853	+** If SQLITE_OK is returned, the caller is responsible for closing
	27854	+** the file descriptor *pFd using close().
	27855	+*/
	27856	+static int openDirectory(const char zFilename, int pFd){
	27857	+ int ii;
	27858	+ int fd = -1;
	27859	+ char zDirname[MAX_PATHNAME+1];
	27860	+
	27861	+ sqlite3_snprintf(MAX_PATHNAME, zDirname, "%s", zFilename);
	27862	+ for(ii=(int)strlen(zDirname); ii>1 && zDirname[ii]!='/'; ii--);
	27863	+ if( ii>0 ){
	27864	+ zDirname[ii] = '\0';
	27865	+ fd = robust_open(zDirname, O_RDONLY\|O_BINARY, 0);
	27866	+ if( fd>=0 ){
	27867	+#ifdef FD_CLOEXEC
	27868	+ osFcntl(fd, F_SETFD, osFcntl(fd, F_GETFD, 0) \| FD_CLOEXEC);
	27869	+#endif
	27870	+ OSTRACE(("OPENDIR %-3d %s\n", fd, zDirname));
	27871	+ }
	27872	+ }
	27873	+ *pFd = fd;
	27874	+ return (fd>=0?SQLITE_OK:unixLogError(SQLITE_CANTOPEN_BKPT, "open", zDirname));
	27875	+}
27671	27876
27672	27877	/*
27673	27878	** Make sure all writes to a particular file are committed to disk.
27674	27879	**
27675	27880	** If dataOnly==0 then both the file itself and its metadata (file
		@@ -27707,32 +27912,27 @@
27707	27912	SimulateIOError( rc=1 );
27708	27913	if( rc ){
27709	27914	pFile->lastErrno = errno;
27710	27915	return unixLogError(SQLITE_IOERR_FSYNC, "full_fsync", pFile->zPath);
27711	27916	}
27712		- if( pFile->dirfd>=0 ){
27713		- OSTRACE(("DIRSYNC %-3d (have_fullfsync=%d fullsync=%d)\n", pFile->dirfd,
	27917	+
	27918	+ /* Also fsync the directory containing the file if the DIRSYNC flag
	27919	+ ** is set. This is a one-time occurrance. Many systems (examples: AIX)
	27920	+ ** are unable to fsync a directory, so ignore errors on the fsync.
	27921	+ */
	27922	+ if( pFile->ctrlFlags & UNIXFILE_DIRSYNC ){
	27923	+ int dirfd;
	27924	+ OSTRACE(("DIRSYNC %s (have_fullfsync=%d fullsync=%d)\n", pFile->zPath,
27714	27925	HAVE_FULLFSYNC, isFullsync));
27715		-#ifndef SQLITE_DISABLE_DIRSYNC
27716		- /* The directory sync is only attempted if full_fsync is
27717		- ** turned off or unavailable. If a full_fsync occurred above,
27718		- ** then the directory sync is superfluous.
27719		- */
27720		- if( (!HAVE_FULLFSYNC \|\| !isFullsync) && full_fsync(pFile->dirfd,0,0) ){
27721		- /*
27722		- ** We have received multiple reports of fsync() returning
27723		- ** errors when applied to directories on certain file systems.
27724		- ** A failed directory sync is not a big deal. So it seems
27725		- ** better to ignore the error. Ticket #1657
27726		- */
27727		- /* pFile->lastErrno = errno; */
27728		- /* return SQLITE_IOERR; */
27729		- }
27730		-#endif
27731		- /* Only need to sync once, so close the directory when we are done */
27732		- robust_close(pFile, pFile->dirfd, __LINE__);
27733		- pFile->dirfd = -1;
	27926	+ rc = osOpenDirectory(pFile->zPath, &dirfd);
	27927	+ if( rc==SQLITE_OK && dirfd>=0 ){
	27928	+ full_fsync(dirfd, 0, 0);
	27929	+ robust_close(pFile, dirfd, __LINE__);
	27930	+ }else if( rc==SQLITE_CANTOPEN ){
	27931	+ rc = SQLITE_OK;
	27932	+ }
	27933	+ pFile->ctrlFlags &= ~UNIXFILE_DIRSYNC;
27734	27934	}
27735	27935	return rc;
27736	27936	}
27737	27937
27738	27938	/*
		@@ -27810,15 +28010,13 @@
27810	28010	static int proxyFileControl(sqlite3_file,int,void);
27811	28011	#endif
27812	28012
27813	28013	/*
27814	28014	** This function is called to handle the SQLITE_FCNTL_SIZE_HINT
27815		-** file-control operation.
27816		-**
27817		-** If the user has configured a chunk-size for this file, it could be
27818		-** that the file needs to be extended at this point. Otherwise, the
27819		-** SQLITE_FCNTL_SIZE_HINT operation is a no-op for Unix.
	28015	+** file-control operation. Enlarge the database to nBytes in size
	28016	+** (rounded up to the next chunk-size). If the database is already
	28017	+** nBytes or larger, this routine is a no-op.
27820	28018	*/
27821	28019	static int fcntlSizeHint(unixFile *pFile, i64 nByte){
27822	28020	if( pFile->szChunk ){
27823	28021	i64 nSize; /* Required file size */
27824	28022	struct stat buf; /* Used to hold return values of fstat() */
		@@ -27866,25 +28064,41 @@
27866	28064
27867	28065	/*
27868	28066	** Information and control of an open file handle.
27869	28067	*/
27870	28068	static int unixFileControl(sqlite3_file id, int op, void pArg){
	28069	+ unixFile pFile = (unixFile)id;
27871	28070	switch( op ){
27872	28071	case SQLITE_FCNTL_LOCKSTATE: {
27873		- (int)pArg = ((unixFile*)id)->eFileLock;
	28072	+ (int)pArg = pFile->eFileLock;
27874	28073	return SQLITE_OK;
27875	28074	}
27876	28075	case SQLITE_LAST_ERRNO: {
27877		- (int)pArg = ((unixFile*)id)->lastErrno;
	28076	+ (int)pArg = pFile->lastErrno;
27878	28077	return SQLITE_OK;
27879	28078	}
27880	28079	case SQLITE_FCNTL_CHUNK_SIZE: {
27881		- ((unixFile)id)->szChunk = (int *)pArg;
	28080	+ pFile->szChunk = (int )pArg;
27882	28081	return SQLITE_OK;
27883	28082	}
27884	28083	case SQLITE_FCNTL_SIZE_HINT: {
27885		- return fcntlSizeHint((unixFile )id, (i64 *)pArg);
	28084	+ int rc;
	28085	+ SimulateIOErrorBenign(1);
	28086	+ rc = fcntlSizeHint(pFile, (i64 )pArg);
	28087	+ SimulateIOErrorBenign(0);
	28088	+ return rc;
	28089	+ }
	28090	+ case SQLITE_FCNTL_PERSIST_WAL: {
	28091	+ int bPersist = (int)pArg;
	28092	+ if( bPersist<0 ){
	28093	+ (int)pArg = (pFile->ctrlFlags & UNIXFILE_PERSIST_WAL)!=0;
	28094	+ }else if( bPersist==0 ){
	28095	+ pFile->ctrlFlags &= ~UNIXFILE_PERSIST_WAL;
	28096	+ }else{
	28097	+ pFile->ctrlFlags \|= UNIXFILE_PERSIST_WAL;
	28098	+ }
	28099	+ return SQLITE_OK;
27886	28100	}
27887	28101	#ifndef NDEBUG
27888	28102	/* The pager calls this method to signal that it has done
27889	28103	** a rollback and that the database is therefore unchanged and
27890	28104	** it hence it is OK for the transaction change counter to be
		@@ -28096,11 +28310,11 @@
28096	28310	unixShmNode *p = pFd->pInode->pShmNode;
28097	28311	assert( unixMutexHeld() );
28098	28312	if( p && p->nRef==0 ){
28099	28313	int i;
28100	28314	assert( p->pInode==pFd->pInode );
28101		- if( p->mutex ) sqlite3_mutex_free(p->mutex);
	28315	+ sqlite3_mutex_free(p->mutex);
28102	28316	for(i=0; i<p->nRegion; i++){
28103	28317	if( p->h>=0 ){
28104	28318	munmap(p->apRegion[i], p->szRegion);
28105	28319	}else{
28106	28320	sqlite3_free(p->apRegion[i]);
		@@ -28565,11 +28779,11 @@
28565	28779	** shared-memory file, too */
28566	28780	unixEnterMutex();
28567	28781	assert( pShmNode->nRef>0 );
28568	28782	pShmNode->nRef--;
28569	28783	if( pShmNode->nRef==0 ){
28570		- if( deleteFlag && pShmNode->h>=0 ) unlink(pShmNode->zFilename);
	28784	+ if( deleteFlag && pShmNode->h>=0 ) osUnlink(pShmNode->zFilename);
28571	28785	unixShmPurge(pDbFd);
28572	28786	}
28573	28787	unixLeaveMutex();
28574	28788
28575	28789	return SQLITE_OK;
		@@ -28878,11 +29092,11 @@
28878	29092	** Initialize the contents of the unixFile structure pointed to by pId.
28879	29093	*/
28880	29094	static int fillInUnixFile(
28881	29095	sqlite3_vfs pVfs, / Pointer to vfs object */
28882	29096	int h, /* Open file descriptor of file being opened */
28883		- int dirfd, /* Directory file descriptor */
	29097	+ int syncDir, /* True to sync directory on first sync */
28884	29098	sqlite3_file pId, / Write to the unixFile structure here */
28885	29099	const char zFilename, / Name of the file being opened */
28886	29100	int noLock, /* Omit locking if true */
28887	29101	int isDelete, /* Delete on close if true */
28888	29102	int isReadOnly /* True if the file is opened read-only */
		@@ -28909,20 +29123,22 @@
28909	29123	assert( zFilename==0 \|\| zFilename[0]=='/' );
28910	29124	#endif
28911	29125
28912	29126	OSTRACE(("OPEN %-3d %s\n", h, zFilename));
28913	29127	pNew->h = h;
28914		- pNew->dirfd = dirfd;
28915	29128	pNew->zPath = zFilename;
28916	29129	if( memcmp(pVfs->zName,"unix-excl",10)==0 ){
28917	29130	pNew->ctrlFlags = UNIXFILE_EXCL;
28918	29131	}else{
28919	29132	pNew->ctrlFlags = 0;
28920	29133	}
28921	29134	if( isReadOnly ){
28922	29135	pNew->ctrlFlags \|= UNIXFILE_RDONLY;
28923	29136	}
	29137	+ if( syncDir ){
	29138	+ pNew->ctrlFlags \|= UNIXFILE_DIRSYNC;
	29139	+ }
28924	29140
28925	29141	#if OS_VXWORKS
28926	29142	pNew->pId = vxworksFindFileId(zFilename);
28927	29143	if( pNew->pId==0 ){
28928	29144	noLock = 1;
		@@ -29045,56 +29261,24 @@
29045	29261	pNew->lastErrno = 0;
29046	29262	#if OS_VXWORKS
29047	29263	if( rc!=SQLITE_OK ){
29048	29264	if( h>=0 ) robust_close(pNew, h, __LINE__);
29049	29265	h = -1;
29050		- unlink(zFilename);
	29266	+ osUnlink(zFilename);
29051	29267	isDelete = 0;
29052	29268	}
29053	29269	pNew->isDelete = isDelete;
29054	29270	#endif
29055	29271	if( rc!=SQLITE_OK ){
29056		- if( dirfd>=0 ) robust_close(pNew, dirfd, __LINE__);
29057	29272	if( h>=0 ) robust_close(pNew, h, __LINE__);
29058	29273	}else{
29059	29274	pNew->pMethod = pLockingStyle;
29060	29275	OpenCounter(+1);
29061	29276	}
29062	29277	return rc;
29063	29278	}
29064	29279
29065		-/*
29066		-** Open a file descriptor to the directory containing file zFilename.
29067		-** If successful, *pFd is set to the opened file descriptor and
29068		-** SQLITE_OK is returned. If an error occurs, either SQLITE_NOMEM
29069		-** or SQLITE_CANTOPEN is returned and *pFd is set to an undefined
29070		-** value.
29071		-**
29072		-** If SQLITE_OK is returned, the caller is responsible for closing
29073		-** the file descriptor *pFd using close().
29074		-*/
29075		-static int openDirectory(const char zFilename, int pFd){
29076		- int ii;
29077		- int fd = -1;
29078		- char zDirname[MAX_PATHNAME+1];
29079		-
29080		- sqlite3_snprintf(MAX_PATHNAME, zDirname, "%s", zFilename);
29081		- for(ii=(int)strlen(zDirname); ii>1 && zDirname[ii]!='/'; ii--);
29082		- if( ii>0 ){
29083		- zDirname[ii] = '\0';
29084		- fd = robust_open(zDirname, O_RDONLY\|O_BINARY, 0);
29085		- if( fd>=0 ){
29086		-#ifdef FD_CLOEXEC
29087		- osFcntl(fd, F_SETFD, osFcntl(fd, F_GETFD, 0) \| FD_CLOEXEC);
29088		-#endif
29089		- OSTRACE(("OPENDIR %-3d %s\n", fd, zDirname));
29090		- }
29091		- }
29092		- *pFd = fd;
29093		- return (fd>=0?SQLITE_OK:unixLogError(SQLITE_CANTOPEN_BKPT, "open", zDirname));
29094		-}
29095		-
29096	29280	/*
29097	29281	** Return the name of a directory in which to put temporary files.
29098	29282	** If no suitable temporary file directory can be found, return NULL.
29099	29283	*/
29100	29284	static const char *unixTempFileDir(void){
		@@ -29205,11 +29389,11 @@
29205	29389	** ignored and -1 is returned. The caller will try to open a new file
29206	29390	** descriptor on the same path, fail, and return an error to SQLite.
29207	29391	**
29208	29392	** Even if a subsequent open() call does succeed, the consequences of
29209	29393	** not searching for a resusable file descriptor are not dire. */
29210		- if( 0==stat(zPath, &sStat) ){
	29394	+ if( 0==osStat(zPath, &sStat) ){
29211	29395	unixInodeInfo *pInode;
29212	29396
29213	29397	unixEnterMutex();
29214	29398	pInode = inodeList;
29215	29399	while( pInode && (pInode->fileId.dev!=sStat.st_dev
		@@ -29281,11 +29465,11 @@
29281	29465	while( nDb>0 && zPath[nDb]!='-' ) nDb--;
29282	29466	if( nDb==0 ) return SQLITE_OK;
29283	29467	memcpy(zDb, zPath, nDb);
29284	29468	zDb[nDb] = '\0';
29285	29469
29286		- if( 0==stat(zDb, &sStat) ){
	29470	+ if( 0==osStat(zDb, &sStat) ){
29287	29471	*pMode = sStat.st_mode & 0777;
29288	29472	}else{
29289	29473	rc = SQLITE_IOERR_FSTAT;
29290	29474	}
29291	29475	}else if( flags & SQLITE_OPEN_DELETEONCLOSE ){
		@@ -29323,11 +29507,10 @@
29323	29507	int flags, /* Input flags to control the opening */
29324	29508	int pOutFlags / Output flags returned to SQLite core */
29325	29509	){
29326	29510	unixFile p = (unixFile )pFile;
29327	29511	int fd = -1; /* File descriptor returned by open() */
29328		- int dirfd = -1; /* Directory file descriptor */
29329	29512	int openFlags = 0; /* Flags to pass to open() */
29330	29513	int eType = flags&0xFFFFFF00; /* Type of file to open */
29331	29514	int noLock; /* True to omit locking primitives */
29332	29515	int rc = SQLITE_OK; /* Function Return Code */
29333	29516
		@@ -29337,16 +29520,19 @@
29337	29520	int isReadonly = (flags & SQLITE_OPEN_READONLY);
29338	29521	int isReadWrite = (flags & SQLITE_OPEN_READWRITE);
29339	29522	#if SQLITE_ENABLE_LOCKING_STYLE
29340	29523	int isAutoProxy = (flags & SQLITE_OPEN_AUTOPROXY);
29341	29524	#endif
	29525	+#if defined(__APPLE__) \|\| SQLITE_ENABLE_LOCKING_STYLE
	29526	+ struct statfs fsInfo;
	29527	+#endif
29342	29528
29343	29529	/* If creating a master or main-file journal, this function will open
29344	29530	** a file-descriptor on the directory too. The first time unixSync()
29345	29531	** is called the directory file descriptor will be fsync()ed and close()d.
29346	29532	*/
29347		- int isOpenDirectory = (isCreate && (
	29533	+ int syncDir = (isCreate && (
29348	29534	eType==SQLITE_OPEN_MASTER_JOURNAL
29349	29535	\|\| eType==SQLITE_OPEN_MAIN_JOURNAL
29350	29536	\|\| eType==SQLITE_OPEN_WAL
29351	29537	));
29352	29538
		@@ -29396,11 +29582,11 @@
29396	29582	}
29397	29583	}
29398	29584	p->pUnused = pUnused;
29399	29585	}else if( !zName ){
29400	29586	/* If zName is NULL, the upper layer is requesting a temp file. */
29401		- assert(isDelete && !isOpenDirectory);
	29587	+ assert(isDelete && !syncDir);
29402	29588	rc = unixGetTempname(MAX_PATHNAME+1, zTmpname);
29403	29589	if( rc!=SQLITE_OK ){
29404	29590	return rc;
29405	29591	}
29406	29592	zName = zTmpname;
		@@ -29452,44 +29638,29 @@
29452	29638
29453	29639	if( isDelete ){
29454	29640	#if OS_VXWORKS
29455	29641	zPath = zName;
29456	29642	#else
29457		- unlink(zName);
	29643	+ osUnlink(zName);
29458	29644	#endif
29459	29645	}
29460	29646	#if SQLITE_ENABLE_LOCKING_STYLE
29461	29647	else{
29462	29648	p->openFlags = openFlags;
29463	29649	}
29464	29650	#endif
29465	29651
29466		- if( isOpenDirectory ){
29467		- rc = openDirectory(zPath, &dirfd);
29468		- if( rc!=SQLITE_OK ){
29469		- /* It is safe to close fd at this point, because it is guaranteed not
29470		- ** to be open on a database file. If it were open on a database file,
29471		- ** it would not be safe to close as this would release any locks held
29472		- ** on the file by this process. */
29473		- assert( eType!=SQLITE_OPEN_MAIN_DB );
29474		- robust_close(p, fd, __LINE__);
29475		- goto open_finished;
29476		- }
29477		- }
29478		-
29479	29652	#ifdef FD_CLOEXEC
29480	29653	osFcntl(fd, F_SETFD, osFcntl(fd, F_GETFD, 0) \| FD_CLOEXEC);
29481	29654	#endif
29482	29655
29483	29656	noLock = eType!=SQLITE_OPEN_MAIN_DB;
29484	29657
29485	29658
29486	29659	#if defined(__APPLE__) \|\| SQLITE_ENABLE_LOCKING_STYLE
29487		- struct statfs fsInfo;
29488	29660	if( fstatfs(fd, &fsInfo) == -1 ){
29489	29661	((unixFile*)pFile)->lastErrno = errno;
29490		- if( dirfd>=0 ) robust_close(p, dirfd, __LINE__);
29491	29662	robust_close(p, fd, __LINE__);
29492	29663	return SQLITE_IOERR_ACCESS;
29493	29664	}
29494	29665	if (0 == strncmp("msdos", fsInfo.f_fstypename, 5)) {
29495	29666	((unixFile*)pFile)->fsFlags \|= SQLITE_FSFLAGS_IS_MSDOS;
		@@ -29507,31 +29678,27 @@
29507	29678	/* SQLITE_FORCE_PROXY_LOCKING==1 means force always use proxy, 0 means
29508	29679	** never use proxy, NULL means use proxy for non-local files only. */
29509	29680	if( envforce!=NULL ){
29510	29681	useProxy = atoi(envforce)>0;
29511	29682	}else{
29512		- struct statfs fsInfo;
29513	29683	if( statfs(zPath, &fsInfo) == -1 ){
29514	29684	/* In theory, the close(fd) call is sub-optimal. If the file opened
29515	29685	** with fd is a database file, and there are other connections open
29516	29686	** on that file that are currently holding advisory locks on it,
29517	29687	** then the call to close() will cancel those locks. In practice,
29518	29688	** we're assuming that statfs() doesn't fail very often. At least
29519	29689	** not while other file descriptors opened by the same process on
29520	29690	** the same file are working. */
29521	29691	p->lastErrno = errno;
29522		- if( dirfd>=0 ){
29523		- robust_close(p, dirfd, __LINE__);
29524		- }
29525	29692	robust_close(p, fd, __LINE__);
29526	29693	rc = SQLITE_IOERR_ACCESS;
29527	29694	goto open_finished;
29528	29695	}
29529	29696	useProxy = !(fsInfo.f_flags&MNT_LOCAL);
29530	29697	}
29531	29698	if( useProxy ){
29532		- rc = fillInUnixFile(pVfs, fd, dirfd, pFile, zPath, noLock,
	29699	+ rc = fillInUnixFile(pVfs, fd, syncDir, pFile, zPath, noLock,
29533	29700	isDelete, isReadonly);
29534	29701	if( rc==SQLITE_OK ){
29535	29702	rc = proxyTransformUnixFile((unixFile*)pFile, ":auto:");
29536	29703	if( rc!=SQLITE_OK ){
29537	29704	/* Use unixClose to clean up the resources added in fillInUnixFile
		@@ -29545,11 +29712,11 @@
29545	29712	goto open_finished;
29546	29713	}
29547	29714	}
29548	29715	#endif
29549	29716
29550		- rc = fillInUnixFile(pVfs, fd, dirfd, pFile, zPath, noLock,
	29717	+ rc = fillInUnixFile(pVfs, fd, syncDir, pFile, zPath, noLock,
29551	29718	isDelete, isReadonly);
29552	29719	open_finished:
29553	29720	if( rc!=SQLITE_OK ){
29554	29721	sqlite3_free(p->pUnused);
29555	29722	}
		@@ -29567,17 +29734,17 @@
29567	29734	int dirSync /* If true, fsync() directory after deleting file */
29568	29735	){
29569	29736	int rc = SQLITE_OK;
29570	29737	UNUSED_PARAMETER(NotUsed);
29571	29738	SimulateIOError(return SQLITE_IOERR_DELETE);
29572		- if( unlink(zPath)==(-1) && errno!=ENOENT ){
	29739	+ if( osUnlink(zPath)==(-1) && errno!=ENOENT ){
29573	29740	return unixLogError(SQLITE_IOERR_DELETE, "unlink", zPath);
29574	29741	}
29575	29742	#ifndef SQLITE_DISABLE_DIRSYNC
29576	29743	if( dirSync ){
29577	29744	int fd;
29578		- rc = openDirectory(zPath, &fd);
	29745	+ rc = osOpenDirectory(zPath, &fd);
29579	29746	if( rc==SQLITE_OK ){
29580	29747	#if OS_VXWORKS
29581	29748	if( fsync(fd)==-1 )
29582	29749	#else
29583	29750	if( fsync(fd) )
		@@ -29584,10 +29751,12 @@
29584	29751	#endif
29585	29752	{
29586	29753	rc = unixLogError(SQLITE_IOERR_DIR_FSYNC, "fsync", zPath);
29587	29754	}
29588	29755	robust_close(0, fd, __LINE__);
	29756	+ }else if( rc==SQLITE_CANTOPEN ){
	29757	+ rc = SQLITE_OK;
29589	29758	}
29590	29759	}
29591	29760	#endif
29592	29761	return rc;
29593	29762	}
		@@ -29626,11 +29795,11 @@
29626	29795	assert(!"Invalid flags argument");
29627	29796	}
29628	29797	*pResOut = (osAccess(zPath, amode)==0);
29629	29798	if( flags==SQLITE_ACCESS_EXISTS && *pResOut ){
29630	29799	struct stat buf;
29631		- if( 0==stat(zPath, &buf) && buf.st_size==0 ){
	29800	+ if( 0==osStat(zPath, &buf) && buf.st_size==0 ){
29632	29801	*pResOut = 0;
29633	29802	}
29634	29803	}
29635	29804	return SQLITE_OK;
29636	29805	}
		@@ -30145,11 +30314,10 @@
30145	30314	const char path, / path for the new unixFile */
30146	30315	unixFile *ppFile, / unixFile created and returned by ref */
30147	30316	int islockfile /* if non zero missing dirs will be created */
30148	30317	) {
30149	30318	int fd = -1;
30150		- int dirfd = -1;
30151	30319	unixFile *pNew;
30152	30320	int rc = SQLITE_OK;
30153	30321	int openFlags = O_RDWR \| O_CREAT;
30154	30322	sqlite3_vfs dummyVfs;
30155	30323	int terrno = 0;
		@@ -30210,11 +30378,11 @@
30210	30378	dummyVfs.zName = "dummy";
30211	30379	pUnused->fd = fd;
30212	30380	pUnused->flags = openFlags;
30213	30381	pNew->pUnused = pUnused;
30214	30382
30215		- rc = fillInUnixFile(&dummyVfs, fd, dirfd, (sqlite3_file*)pNew, path, 0, 0, 0);
	30383	+ rc = fillInUnixFile(&dummyVfs, fd, 0, (sqlite3_file*)pNew, path, 0, 0, 0);
30216	30384	if( rc==SQLITE_OK ){
30217	30385	*ppFile = pNew;
30218	30386	return SQLITE_OK;
30219	30387	}
30220	30388	end_create_proxy:
		@@ -30250,10 +30418,12 @@
30250	30418	*pError = err;
30251	30419	}
30252	30420	return SQLITE_IOERR;
30253	30421	}
30254	30422	}
	30423	+#else
	30424	+ UNUSED_PARAMETER(pError);
30255	30425	#endif
30256	30426	#ifdef SQLITE_TEST
30257	30427	/* simulate multiple hosts by creating unique hostid file paths */
30258	30428	if( sqlite3_hostid_num != 0){
30259	30429	pHostID[0] = (char)(pHostID[0] + (char)(sqlite3_hostid_num & 0xFF));
		@@ -30324,11 +30494,11 @@
30324	30494	conchFile->openFlags = O_RDWR \| O_CREAT;
30325	30495
30326	30496	end_breaklock:
30327	30497	if( rc ){
30328	30498	if( fd>=0 ){
30329		- unlink(tPath);
	30499	+ osUnlink(tPath);
30330	30500	robust_close(pFile, fd, __LINE__);
30331	30501	}
30332	30502	fprintf(stderr, "failed to break stale lock on %s, %s\n", cPath, errmsg);
30333	30503	}
30334	30504	return rc;
		@@ -30342,10 +30512,11 @@
30342	30512	unixFile *conchFile = pCtx->conchFile;
30343	30513	int rc = SQLITE_OK;
30344	30514	int nTries = 0;
30345	30515	struct timespec conchModTime;
30346	30516
	30517	+ memset(&conchModTime, 0, sizeof(conchModTime));
30347	30518	do {
30348	30519	rc = conchFile->pMethod->xLock((sqlite3_file*)conchFile, lockType);
30349	30520	nTries ++;
30350	30521	if( rc==SQLITE_BUSY ){
30351	30522	/* If the lock failed (busy):
		@@ -30573,15 +30744,16 @@
30573	30744	conchFile->pMethod->xUnlock((sqlite3_file*)conchFile, SHARED_LOCK);
30574	30745
30575	30746	end_takeconch:
30576	30747	OSTRACE(("TRANSPROXY: CLOSE %d\n", pFile->h));
30577	30748	if( rc==SQLITE_OK && pFile->openFlags ){
	30749	+ int fd;
30578	30750	if( pFile->h>=0 ){
30579	30751	robust_close(pFile, pFile->h, __LINE__);
30580	30752	}
30581	30753	pFile->h = -1;
30582		- int fd = robust_open(pCtx->dbPath, pFile->openFlags,
	30754	+ fd = robust_open(pCtx->dbPath, pFile->openFlags,
30583	30755	SQLITE_DEFAULT_FILE_PERMISSIONS);
30584	30756	OSTRACE(("TRANSPROXY: OPEN %d\n", fd));
30585	30757	if( fd>=0 ){
30586	30758	pFile->h = fd;
30587	30759	}else{
		@@ -31147,11 +31319,11 @@
31147	31319	};
31148	31320	unsigned int i; /* Loop counter */
31149	31321
31150	31322	/* Double-check that the aSyscall[] array has been constructed
31151	31323	** correctly. See ticket [bb3a86e890c8e96ab] */
31152		- assert( ArraySize(aSyscall)==16 );
	31324	+ assert( ArraySize(aSyscall)==18 );
31153	31325
31154	31326	/* Register all VFSes defined in the aVfs[] array */
31155	31327	for(i=0; i<(sizeof(aVfs)/sizeof(sqlite3_vfs)); i++){
31156	31328	sqlite3_vfs_register(&aVfs[i], i==0);
31157	31329	}
		@@ -31264,14 +31436,17 @@
31264	31436	#ifdef MEMORY_DEBUG
31265	31437	# error "The MEMORY_DEBUG macro is obsolete. Use SQLITE_DEBUG instead."
31266	31438	#endif
31267	31439
31268	31440	#ifdef SQLITE_DEBUG
31269		-SQLITE_PRIVATE int sqlite3OSTrace = 0;
31270		-#define OSTRACE(X) if( sqlite3OSTrace ) sqlite3DebugPrintf X
	31441	+# ifndef SQLITE_DEBUG_OS_TRACE
	31442	+# define SQLITE_DEBUG_OS_TRACE 0
	31443	+# endif
	31444	+ int sqlite3OSTrace = SQLITE_DEBUG_OS_TRACE;
	31445	+# define OSTRACE(X) if( sqlite3OSTrace ) sqlite3DebugPrintf X
31271	31446	#else
31272		-#define OSTRACE(X)
	31447	+# define OSTRACE(X)
31273	31448	#endif
31274	31449
31275	31450	/*
31276	31451	** Macros for performance tracing. Normally turned off. Only works
31277	31452	** on i486 hardware.
		@@ -31479,12 +31654,13 @@
31479	31654	typedef struct winFile winFile;
31480	31655	struct winFile {
31481	31656	const sqlite3_io_methods pMethod; / Must be first */
31482	31657	sqlite3_vfs pVfs; / The VFS used to open this file */
31483	31658	HANDLE h; /* Handle for accessing the file */
31484		- unsigned char locktype; /* Type of lock currently held on this file */
	31659	+ u8 locktype; /* Type of lock currently held on this file */
31485	31660	short sharedLockByte; /* Randomly chosen byte used as a shared lock */
	31661	+ u8 bPersistWal; /* True to persist WAL files */
31486	31662	DWORD lastErrno; /* The Windows errno from the last I/O error */
31487	31663	DWORD sectorSize; /* Sector size of the device file is on */
31488	31664	winShm pShm; / Instance of shared memory on this file */
31489	31665	const char zPath; / Full pathname of this file */
31490	31666	int szChunk; /* Chunk size configured by FCNTL_CHUNK_SIZE */
		@@ -31495,10 +31671,80 @@
31495	31671	winceLock local; /* Locks obtained by this instance of winFile */
31496	31672	winceLock shared; / Global shared lock memory for the file */
31497	31673	#endif
31498	31674	};
31499	31675
	31676	+/*
	31677	+ * If compiled with SQLITE_WIN32_MALLOC on Windows, we will use the
	31678	+ * various Win32 API heap functions instead of our own.
	31679	+ */
	31680	+#ifdef SQLITE_WIN32_MALLOC
	31681	+/*
	31682	+ * The initial size of the Win32-specific heap. This value may be zero.
	31683	+ */
	31684	+#ifndef SQLITE_WIN32_HEAP_INIT_SIZE
	31685	+# define SQLITE_WIN32_HEAP_INIT_SIZE ((SQLITE_DEFAULT_CACHE_SIZE) * \
	31686	+ (SQLITE_DEFAULT_PAGE_SIZE) + 4194304)
	31687	+#endif
	31688	+
	31689	+/*
	31690	+ * The maximum size of the Win32-specific heap. This value may be zero.
	31691	+ */
	31692	+#ifndef SQLITE_WIN32_HEAP_MAX_SIZE
	31693	+# define SQLITE_WIN32_HEAP_MAX_SIZE (0)
	31694	+#endif
	31695	+
	31696	+/*
	31697	+ * The extra flags to use in calls to the Win32 heap APIs. This value may be
	31698	+ * zero for the default behavior.
	31699	+ */
	31700	+#ifndef SQLITE_WIN32_HEAP_FLAGS
	31701	+# define SQLITE_WIN32_HEAP_FLAGS (0)
	31702	+#endif
	31703	+
	31704	+/*
	31705	+** The winMemData structure stores information required by the Win32-specific
	31706	+** sqlite3_mem_methods implementation.
	31707	+*/
	31708	+typedef struct winMemData winMemData;
	31709	+struct winMemData {
	31710	+#ifndef NDEBUG
	31711	+ u32 magic; /* Magic number to detect structure corruption. */
	31712	+#endif
	31713	+ HANDLE hHeap; /* The handle to our heap. */
	31714	+ BOOL bOwned; /* Do we own the heap (i.e. destroy it on shutdown)? */
	31715	+};
	31716	+
	31717	+#ifndef NDEBUG
	31718	+#define WINMEM_MAGIC 0x42b2830b
	31719	+#endif
	31720	+
	31721	+static struct winMemData win_mem_data = {
	31722	+#ifndef NDEBUG
	31723	+ WINMEM_MAGIC,
	31724	+#endif
	31725	+ NULL, FALSE
	31726	+};
	31727	+
	31728	+#ifndef NDEBUG
	31729	+#define winMemAssertMagic() assert( win_mem_data.magic==WINMEM_MAGIC )
	31730	+#else
	31731	+#define winMemAssertMagic()
	31732	+#endif
	31733	+
	31734	+#define winMemGetHeap() win_mem_data.hHeap
	31735	+
	31736	+static void *winMemMalloc(int nBytes);
	31737	+static void winMemFree(void *pPrior);
	31738	+static void winMemRealloc(void pPrior, int nBytes);
	31739	+static int winMemSize(void *p);
	31740	+static int winMemRoundup(int n);
	31741	+static int winMemInit(void *pAppData);
	31742	+static void winMemShutdown(void *pAppData);
	31743	+
	31744	+SQLITE_PRIVATE const sqlite3_mem_methods *sqlite3MemGetWin32(void);
	31745	+#endif /* SQLITE_WIN32_MALLOC */
31500	31746
31501	31747	/*
31502	31748	** Forward prototypes.
31503	31749	*/
31504	31750	static int getSectorSize(
		@@ -31547,10 +31793,192 @@
31547	31793	}
31548	31794	return sqlite3_os_type==2;
31549	31795	}
31550	31796	#endif /* SQLITE_OS_WINCE */
31551	31797
	31798	+#ifdef SQLITE_WIN32_MALLOC
	31799	+/*
	31800	+** Allocate nBytes of memory.
	31801	+*/
	31802	+static void *winMemMalloc(int nBytes){
	31803	+ HANDLE hHeap;
	31804	+ void *p;
	31805	+
	31806	+ winMemAssertMagic();
	31807	+ hHeap = winMemGetHeap();
	31808	+ assert( hHeap!=0 );
	31809	+ assert( hHeap!=INVALID_HANDLE_VALUE );
	31810	+#ifdef SQLITE_WIN32_MALLOC_VALIDATE
	31811	+ assert ( HeapValidate(hHeap, SQLITE_WIN32_HEAP_FLAGS, NULL) );
	31812	+#endif
	31813	+ assert( nBytes>=0 );
	31814	+ p = HeapAlloc(hHeap, SQLITE_WIN32_HEAP_FLAGS, (SIZE_T)nBytes);
	31815	+ if( !p ){
	31816	+ sqlite3_log(SQLITE_NOMEM, "failed to HeapAlloc %u bytes (%d), heap=%p",
	31817	+ nBytes, GetLastError(), (void*)hHeap);
	31818	+ }
	31819	+ return p;
	31820	+}
	31821	+
	31822	+/*
	31823	+** Free memory.
	31824	+*/
	31825	+static void winMemFree(void *pPrior){
	31826	+ HANDLE hHeap;
	31827	+
	31828	+ winMemAssertMagic();
	31829	+ hHeap = winMemGetHeap();
	31830	+ assert( hHeap!=0 );
	31831	+ assert( hHeap!=INVALID_HANDLE_VALUE );
	31832	+#ifdef SQLITE_WIN32_MALLOC_VALIDATE
	31833	+ assert ( HeapValidate(hHeap, SQLITE_WIN32_HEAP_FLAGS, pPrior) );
	31834	+#endif
	31835	+ if( !pPrior ) return; /* Passing NULL to HeapFree is undefined. */
	31836	+ if( !HeapFree(hHeap, SQLITE_WIN32_HEAP_FLAGS, pPrior) ){
	31837	+ sqlite3_log(SQLITE_NOMEM, "failed to HeapFree block %p (%d), heap=%p",
	31838	+ pPrior, GetLastError(), (void*)hHeap);
	31839	+ }
	31840	+}
	31841	+
	31842	+/*
	31843	+** Change the size of an existing memory allocation
	31844	+*/
	31845	+static void winMemRealloc(void pPrior, int nBytes){
	31846	+ HANDLE hHeap;
	31847	+ void *p;
	31848	+
	31849	+ winMemAssertMagic();
	31850	+ hHeap = winMemGetHeap();
	31851	+ assert( hHeap!=0 );
	31852	+ assert( hHeap!=INVALID_HANDLE_VALUE );
	31853	+#ifdef SQLITE_WIN32_MALLOC_VALIDATE
	31854	+ assert ( HeapValidate(hHeap, SQLITE_WIN32_HEAP_FLAGS, pPrior) );
	31855	+#endif
	31856	+ assert( nBytes>=0 );
	31857	+ if( !pPrior ){
	31858	+ p = HeapAlloc(hHeap, SQLITE_WIN32_HEAP_FLAGS, (SIZE_T)nBytes);
	31859	+ }else{
	31860	+ p = HeapReAlloc(hHeap, SQLITE_WIN32_HEAP_FLAGS, pPrior, (SIZE_T)nBytes);
	31861	+ }
	31862	+ if( !p ){
	31863	+ sqlite3_log(SQLITE_NOMEM, "failed to %s %u bytes (%d), heap=%p",
	31864	+ pPrior ? "HeapReAlloc" : "HeapAlloc", nBytes, GetLastError(),
	31865	+ (void*)hHeap);
	31866	+ }
	31867	+ return p;
	31868	+}
	31869	+
	31870	+/*
	31871	+** Return the size of an outstanding allocation, in bytes.
	31872	+*/
	31873	+static int winMemSize(void *p){
	31874	+ HANDLE hHeap;
	31875	+ SIZE_T n;
	31876	+
	31877	+ winMemAssertMagic();
	31878	+ hHeap = winMemGetHeap();
	31879	+ assert( hHeap!=0 );
	31880	+ assert( hHeap!=INVALID_HANDLE_VALUE );
	31881	+#ifdef SQLITE_WIN32_MALLOC_VALIDATE
	31882	+ assert ( HeapValidate(hHeap, SQLITE_WIN32_HEAP_FLAGS, NULL) );
	31883	+#endif
	31884	+ if( !p ) return 0;
	31885	+ n = HeapSize(hHeap, SQLITE_WIN32_HEAP_FLAGS, p);
	31886	+ if( n==(SIZE_T)-1 ){
	31887	+ sqlite3_log(SQLITE_NOMEM, "failed to HeapSize block %p (%d), heap=%p",
	31888	+ p, GetLastError(), (void*)hHeap);
	31889	+ return 0;
	31890	+ }
	31891	+ return (int)n;
	31892	+}
	31893	+
	31894	+/*
	31895	+** Round up a request size to the next valid allocation size.
	31896	+*/
	31897	+static int winMemRoundup(int n){
	31898	+ return n;
	31899	+}
	31900	+
	31901	+/*
	31902	+** Initialize this module.
	31903	+*/
	31904	+static int winMemInit(void *pAppData){
	31905	+ winMemData pWinMemData = (winMemData )pAppData;
	31906	+
	31907	+ if( !pWinMemData ) return SQLITE_ERROR;
	31908	+ assert( pWinMemData->magic==WINMEM_MAGIC );
	31909	+ if( !pWinMemData->hHeap ){
	31910	+ pWinMemData->hHeap = HeapCreate(SQLITE_WIN32_HEAP_FLAGS,
	31911	+ SQLITE_WIN32_HEAP_INIT_SIZE,
	31912	+ SQLITE_WIN32_HEAP_MAX_SIZE);
	31913	+ if( !pWinMemData->hHeap ){
	31914	+ sqlite3_log(SQLITE_NOMEM,
	31915	+ "failed to HeapCreate (%d), flags=%u, initSize=%u, maxSize=%u",
	31916	+ GetLastError(), SQLITE_WIN32_HEAP_FLAGS, SQLITE_WIN32_HEAP_INIT_SIZE,
	31917	+ SQLITE_WIN32_HEAP_MAX_SIZE);
	31918	+ return SQLITE_NOMEM;
	31919	+ }
	31920	+ pWinMemData->bOwned = TRUE;
	31921	+ }
	31922	+ assert( pWinMemData->hHeap!=0 );
	31923	+ assert( pWinMemData->hHeap!=INVALID_HANDLE_VALUE );
	31924	+#ifdef SQLITE_WIN32_MALLOC_VALIDATE
	31925	+ assert( HeapValidate(pWinMemData->hHeap, SQLITE_WIN32_HEAP_FLAGS, NULL) );
	31926	+#endif
	31927	+ return SQLITE_OK;
	31928	+}
	31929	+
	31930	+/*
	31931	+** Deinitialize this module.
	31932	+*/
	31933	+static void winMemShutdown(void *pAppData){
	31934	+ winMemData pWinMemData = (winMemData )pAppData;
	31935	+
	31936	+ if( !pWinMemData ) return;
	31937	+ if( pWinMemData->hHeap ){
	31938	+ assert( pWinMemData->hHeap!=INVALID_HANDLE_VALUE );
	31939	+#ifdef SQLITE_WIN32_MALLOC_VALIDATE
	31940	+ assert( HeapValidate(pWinMemData->hHeap, SQLITE_WIN32_HEAP_FLAGS, NULL) );
	31941	+#endif
	31942	+ if( pWinMemData->bOwned ){
	31943	+ if( !HeapDestroy(pWinMemData->hHeap) ){
	31944	+ sqlite3_log(SQLITE_NOMEM, "failed to HeapDestroy (%d), heap=%p",
	31945	+ GetLastError(), (void*)pWinMemData->hHeap);
	31946	+ }
	31947	+ pWinMemData->bOwned = FALSE;
	31948	+ }
	31949	+ pWinMemData->hHeap = NULL;
	31950	+ }
	31951	+}
	31952	+
	31953	+/*
	31954	+** Populate the low-level memory allocation function pointers in
	31955	+** sqlite3GlobalConfig.m with pointers to the routines in this file. The
	31956	+** arguments specify the block of memory to manage.
	31957	+**
	31958	+** This routine is only called by sqlite3_config(), and therefore
	31959	+** is not required to be threadsafe (it is not).
	31960	+*/
	31961	+SQLITE_PRIVATE const sqlite3_mem_methods *sqlite3MemGetWin32(void){
	31962	+ static const sqlite3_mem_methods winMemMethods = {
	31963	+ winMemMalloc,
	31964	+ winMemFree,
	31965	+ winMemRealloc,
	31966	+ winMemSize,
	31967	+ winMemRoundup,
	31968	+ winMemInit,
	31969	+ winMemShutdown,
	31970	+ &win_mem_data
	31971	+ };
	31972	+ return &winMemMethods;
	31973	+}
	31974	+
	31975	+SQLITE_PRIVATE void sqlite3MemSetDefault(void){
	31976	+ sqlite3_config(SQLITE_CONFIG_MALLOC, sqlite3MemGetWin32());
	31977	+}
	31978	+#endif /* SQLITE_WIN32_MALLOC */
	31979	+
31552	31980	/*
31553	31981	** Convert a UTF-8 string to microsoft unicode (UTF-16?).
31554	31982	**
31555	31983	** Space to hold the returned string is obtained from malloc.
31556	31984	*/
		@@ -31778,19 +32206,68 @@
31778	32206	iLine, iErrno, zFunc, zPath, zMsg
31779	32207	);
31780	32208
31781	32209	return errcode;
31782	32210	}
	32211	+
	32212	+/*
	32213	+** The number of times that a ReadFile(), WriteFile(), and DeleteFile()
	32214	+** will be retried following a locking error - probably caused by
	32215	+** antivirus software. Also the initial delay before the first retry.
	32216	+** The delay increases linearly with each retry.
	32217	+*/
	32218	+#ifndef SQLITE_WIN32_IOERR_RETRY
	32219	+# define SQLITE_WIN32_IOERR_RETRY 10
	32220	+#endif
	32221	+#ifndef SQLITE_WIN32_IOERR_RETRY_DELAY
	32222	+# define SQLITE_WIN32_IOERR_RETRY_DELAY 25
	32223	+#endif
	32224	+static int win32IoerrRetry = SQLITE_WIN32_IOERR_RETRY;
	32225	+static int win32IoerrRetryDelay = SQLITE_WIN32_IOERR_RETRY_DELAY;
	32226	+
	32227	+/*
	32228	+** If a ReadFile() or WriteFile() error occurs, invoke this routine
	32229	+** to see if it should be retried. Return TRUE to retry. Return FALSE
	32230	+** to give up with an error.
	32231	+*/
	32232	+static int retryIoerr(int *pnRetry){
	32233	+ DWORD e;
	32234	+ if( *pnRetry>=win32IoerrRetry ){
	32235	+ return 0;
	32236	+ }
	32237	+ e = GetLastError();
	32238	+ if( e==ERROR_ACCESS_DENIED \|\|
	32239	+ e==ERROR_LOCK_VIOLATION \|\|
	32240	+ e==ERROR_SHARING_VIOLATION ){
	32241	+ Sleep(win32IoerrRetryDelay(1+pnRetry));
	32242	+ ++*pnRetry;
	32243	+ return 1;
	32244	+ }
	32245	+ return 0;
	32246	+}
	32247	+
	32248	+/*
	32249	+** Log a I/O error retry episode.
	32250	+*/
	32251	+static void logIoerr(int nRetry){
	32252	+ if( nRetry ){
	32253	+ sqlite3_log(SQLITE_IOERR,
	32254	+ "delayed %dms for lock/sharing conflict",
	32255	+ win32IoerrRetryDelaynRetry(nRetry+1)/2
	32256	+ );
	32257	+ }
	32258	+}
31783	32259
31784	32260	#if SQLITE_OS_WINCE
31785	32261	/*************************************************************************
31786	32262	** This section contains code for WinCE only.
31787	32263	*/
31788	32264	/*
31789	32265	** WindowsCE does not have a localtime() function. So create a
31790	32266	** substitute.
31791	32267	*/
	32268	+/* #include <time.h> */
31792	32269	struct tm __cdecl localtime(const time_t t)
31793	32270	{
31794	32271	static struct tm y;
31795	32272	FILETIME uTm, lTm;
31796	32273	SYSTEMTIME pTm;
		@@ -32196,22 +32673,25 @@
32196	32673	int amt, /* Number of bytes to read */
32197	32674	sqlite3_int64 offset /* Begin reading at this offset */
32198	32675	){
32199	32676	winFile pFile = (winFile)id; /* file handle */
32200	32677	DWORD nRead; /* Number of bytes actually read from file */
	32678	+ int nRetry = 0; /* Number of retrys */
32201	32679
32202	32680	assert( id!=0 );
32203	32681	SimulateIOError(return SQLITE_IOERR_READ);
32204	32682	OSTRACE(("READ %d lock=%d\n", pFile->h, pFile->locktype));
32205	32683
32206	32684	if( seekWinFile(pFile, offset) ){
32207	32685	return SQLITE_FULL;
32208	32686	}
32209		- if( !ReadFile(pFile->h, pBuf, amt, &nRead, 0) ){
	32687	+ while( !ReadFile(pFile->h, pBuf, amt, &nRead, 0) ){
	32688	+ if( retryIoerr(&nRetry) ) continue;
32210	32689	pFile->lastErrno = GetLastError();
32211	32690	return winLogError(SQLITE_IOERR_READ, "winRead", pFile->zPath);
32212	32691	}
	32692	+ logIoerr(nRetry);
32213	32693	if( nRead<(DWORD)amt ){
32214	32694	/* Unread parts of the buffer must be zero-filled */
32215	32695	memset(&((char*)pBuf)[nRead], 0, amt-nRead);
32216	32696	return SQLITE_IOERR_SHORT_READ;
32217	32697	}
		@@ -32229,10 +32709,11 @@
32229	32709	int amt, /* Number of bytes to write */
32230	32710	sqlite3_int64 offset /* Offset into the file to begin writing at */
32231	32711	){
32232	32712	int rc; /* True if error has occured, else false */
32233	32713	winFile pFile = (winFile)id; /* File handle */
	32714	+ int nRetry = 0; /* Number of retries */
32234	32715
32235	32716	assert( amt>0 );
32236	32717	assert( pFile );
32237	32718	SimulateIOError(return SQLITE_IOERR_WRITE);
32238	32719	SimulateDiskfullError(return SQLITE_FULL);
		@@ -32243,11 +32724,16 @@
32243	32724	if( rc==0 ){
32244	32725	u8 aRem = (u8 )pBuf; /* Data yet to be written */
32245	32726	int nRem = amt; /* Number of bytes yet to be written */
32246	32727	DWORD nWrite; /* Bytes written by each WriteFile() call */
32247	32728
32248		- while( nRem>0 && WriteFile(pFile->h, aRem, nRem, &nWrite, 0) && nWrite>0 ){
	32729	+ while( nRem>0 ){
	32730	+ if( !WriteFile(pFile->h, aRem, nRem, &nWrite, 0) ){
	32731	+ if( retryIoerr(&nRetry) ) continue;
	32732	+ break;
	32733	+ }
	32734	+ if( nWrite<=0 ) break;
32249	32735	aRem += nWrite;
32250	32736	nRem -= nWrite;
32251	32737	}
32252	32738	if( nRem>0 ){
32253	32739	pFile->lastErrno = GetLastError();
		@@ -32254,14 +32740,17 @@
32254	32740	rc = 1;
32255	32741	}
32256	32742	}
32257	32743
32258	32744	if( rc ){
32259		- if( pFile->lastErrno==ERROR_HANDLE_DISK_FULL ){
	32745	+ if( ( pFile->lastErrno==ERROR_HANDLE_DISK_FULL )
	32746	+ \|\| ( pFile->lastErrno==ERROR_DISK_FULL )){
32260	32747	return SQLITE_FULL;
32261	32748	}
32262	32749	return winLogError(SQLITE_IOERR_WRITE, "winWrite", pFile->zPath);
	32750	+ }else{
	32751	+ logIoerr(nRetry);
32263	32752	}
32264	32753	return SQLITE_OK;
32265	32754	}
32266	32755
32267	32756	/*
		@@ -32651,32 +33140,63 @@
32651	33140
32652	33141	/*
32653	33142	** Control and query of the open file handle.
32654	33143	*/
32655	33144	static int winFileControl(sqlite3_file id, int op, void pArg){
	33145	+ winFile pFile = (winFile)id;
32656	33146	switch( op ){
32657	33147	case SQLITE_FCNTL_LOCKSTATE: {
32658		- (int)pArg = ((winFile*)id)->locktype;
	33148	+ (int)pArg = pFile->locktype;
32659	33149	return SQLITE_OK;
32660	33150	}
32661	33151	case SQLITE_LAST_ERRNO: {
32662		- (int)pArg = (int)((winFile*)id)->lastErrno;
	33152	+ (int)pArg = (int)pFile->lastErrno;
32663	33153	return SQLITE_OK;
32664	33154	}
32665	33155	case SQLITE_FCNTL_CHUNK_SIZE: {
32666		- ((winFile)id)->szChunk = (int *)pArg;
	33156	+ pFile->szChunk = (int )pArg;
32667	33157	return SQLITE_OK;
32668	33158	}
32669	33159	case SQLITE_FCNTL_SIZE_HINT: {
32670		- sqlite3_int64 sz = (sqlite3_int64)pArg;
32671		- SimulateIOErrorBenign(1);
32672		- winTruncate(id, sz);
32673		- SimulateIOErrorBenign(0);
	33160	+ winFile pFile = (winFile)id;
	33161	+ sqlite3_int64 oldSz;
	33162	+ int rc = winFileSize(id, &oldSz);
	33163	+ if( rc==SQLITE_OK ){
	33164	+ sqlite3_int64 newSz = (sqlite3_int64)pArg;
	33165	+ if( newSz>oldSz ){
	33166	+ SimulateIOErrorBenign(1);
	33167	+ rc = winTruncate(id, newSz);
	33168	+ SimulateIOErrorBenign(0);
	33169	+ }
	33170	+ }
	33171	+ return rc;
	33172	+ }
	33173	+ case SQLITE_FCNTL_PERSIST_WAL: {
	33174	+ int bPersist = (int)pArg;
	33175	+ if( bPersist<0 ){
	33176	+ (int)pArg = pFile->bPersistWal;
	33177	+ }else{
	33178	+ pFile->bPersistWal = bPersist!=0;
	33179	+ }
32674	33180	return SQLITE_OK;
32675	33181	}
32676	33182	case SQLITE_FCNTL_SYNC_OMITTED: {
32677	33183	return SQLITE_OK;
	33184	+ }
	33185	+ case SQLITE_FCNTL_WIN32_AV_RETRY: {
	33186	+ int a = (int)pArg;
	33187	+ if( a[0]>0 ){
	33188	+ win32IoerrRetry = a[0];
	33189	+ }else{
	33190	+ a[0] = win32IoerrRetry;
	33191	+ }
	33192	+ if( a[1]>0 ){
	33193	+ win32IoerrRetryDelay = a[1];
	33194	+ }else{
	33195	+ a[1] = win32IoerrRetryDelay;
	33196	+ }
	33197	+ return SQLITE_OK;
32678	33198	}
32679	33199	}
32680	33200	return SQLITE_NOTFOUND;
32681	33201	}
32682	33202
		@@ -33482,10 +34002,11 @@
33482	34002	int isTemp = 0;
33483	34003	#endif
33484	34004	winFile pFile = (winFile)id;
33485	34005	void zConverted; / Filename in OS encoding */
33486	34006	const char zUtf8Name = zName; / Filename in UTF-8 encoding */
	34007	+ int cnt = 0;
33487	34008
33488	34009	/* If argument zPath is a NULL pointer, this function is required to open
33489	34010	** a temporary file. Use this buffer to store the file name in.
33490	34011	*/
33491	34012	char zTmpname[MAX_PATH+1]; /* Buffer used to create temp filename */
		@@ -33601,34 +34122,34 @@
33601	34122	#if SQLITE_OS_WINCE
33602	34123	dwFlagsAndAttributes \|= FILE_FLAG_RANDOM_ACCESS;
33603	34124	#endif
33604	34125
33605	34126	if( isNT() ){
33606		- h = CreateFileW((WCHAR*)zConverted,
33607		- dwDesiredAccess,
33608		- dwShareMode,
33609		- NULL,
33610		- dwCreationDisposition,
33611		- dwFlagsAndAttributes,
33612		- NULL
33613		- );
	34127	+ while( (h = CreateFileW((WCHAR*)zConverted,
	34128	+ dwDesiredAccess,
	34129	+ dwShareMode, NULL,
	34130	+ dwCreationDisposition,
	34131	+ dwFlagsAndAttributes,
	34132	+ NULL))==INVALID_HANDLE_VALUE &&
	34133	+ retryIoerr(&cnt) ){}
33614	34134	/* isNT() is 1 if SQLITE_OS_WINCE==1, so this else is never executed.
33615	34135	** Since the ASCII version of these Windows API do not exist for WINCE,
33616	34136	** it's important to not reference them for WINCE builds.
33617	34137	*/
33618	34138	#if SQLITE_OS_WINCE==0
33619	34139	}else{
33620		- h = CreateFileA((char*)zConverted,
33621		- dwDesiredAccess,
33622		- dwShareMode,
33623		- NULL,
33624		- dwCreationDisposition,
33625		- dwFlagsAndAttributes,
33626		- NULL
33627		- );
	34140	+ while( (h = CreateFileA((char*)zConverted,
	34141	+ dwDesiredAccess,
	34142	+ dwShareMode, NULL,
	34143	+ dwCreationDisposition,
	34144	+ dwFlagsAndAttributes,
	34145	+ NULL))==INVALID_HANDLE_VALUE &&
	34146	+ retryIoerr(&cnt) ){}
33628	34147	#endif
33629	34148	}
	34149	+
	34150	+ logIoerr(cnt);
33630	34151
33631	34152	OSTRACE(("OPEN %d %s 0x%lx %s\n",
33632	34153	h, zName, dwDesiredAccess,
33633	34154	h==INVALID_HANDLE_VALUE ? "failed" : "ok"));
33634	34155
		@@ -33691,19 +34212,17 @@
33691	34212	** file open, we will be unable to delete it. To work around this
33692	34213	** problem, we delay 100 milliseconds and try to delete again. Up
33693	34214	** to MX_DELETION_ATTEMPTs deletion attempts are run before giving
33694	34215	** up and returning an error.
33695	34216	*/
33696		-#define MX_DELETION_ATTEMPTS 5
33697	34217	static int winDelete(
33698	34218	sqlite3_vfs pVfs, / Not used on win32 */
33699	34219	const char zFilename, / Name of file to delete */
33700	34220	int syncDir /* Not used on win32 */
33701	34221	){
33702	34222	int cnt = 0;
33703		- DWORD rc;
33704		- DWORD error = 0;
	34223	+ int rc;
33705	34224	void *zConverted;
33706	34225	UNUSED_PARAMETER(pVfs);
33707	34226	UNUSED_PARAMETER(syncDir);
33708	34227
33709	34228	SimulateIOError(return SQLITE_IOERR_DELETE);
		@@ -33710,38 +34229,34 @@
33710	34229	zConverted = convertUtf8Filename(zFilename);
33711	34230	if( zConverted==0 ){
33712	34231	return SQLITE_NOMEM;
33713	34232	}
33714	34233	if( isNT() ){
33715		- do{
33716		- DeleteFileW(zConverted);
33717		- }while( ( ((rc = GetFileAttributesW(zConverted)) != INVALID_FILE_ATTRIBUTES)
33718		- \|\| ((error = GetLastError()) == ERROR_ACCESS_DENIED))
33719		- && (++cnt < MX_DELETION_ATTEMPTS)
33720		- && (Sleep(100), 1) );
	34234	+ rc = 1;
	34235	+ while( GetFileAttributesW(zConverted)!=INVALID_FILE_ATTRIBUTES &&
	34236	+ (rc = DeleteFileW(zConverted))==0 && retryIoerr(&cnt) ){}
	34237	+ rc = rc ? SQLITE_OK : SQLITE_ERROR;
33721	34238	/* isNT() is 1 if SQLITE_OS_WINCE==1, so this else is never executed.
33722	34239	** Since the ASCII version of these Windows API do not exist for WINCE,
33723	34240	** it's important to not reference them for WINCE builds.
33724	34241	*/
33725	34242	#if SQLITE_OS_WINCE==0
33726	34243	}else{
33727		- do{
33728		- DeleteFileA(zConverted);
33729		- }while( ( ((rc = GetFileAttributesA(zConverted)) != INVALID_FILE_ATTRIBUTES)
33730		- \|\| ((error = GetLastError()) == ERROR_ACCESS_DENIED))
33731		- && (++cnt < MX_DELETION_ATTEMPTS)
33732		- && (Sleep(100), 1) );
	34244	+ rc = 1;
	34245	+ while( GetFileAttributesA(zConverted)!=INVALID_FILE_ATTRIBUTES &&
	34246	+ (rc = DeleteFileA(zConverted))==0 && retryIoerr(&cnt) ){}
	34247	+ rc = rc ? SQLITE_OK : SQLITE_ERROR;
33733	34248	#endif
33734	34249	}
	34250	+ if( rc ){
	34251	+ rc = winLogError(SQLITE_IOERR_DELETE, "winDelete", zFilename);
	34252	+ }else{
	34253	+ logIoerr(cnt);
	34254	+ }
33735	34255	free(zConverted);
33736		- OSTRACE(("DELETE \"%s\" %s\n", zFilename,
33737		- ( (rc==INVALID_FILE_ATTRIBUTES) && (error==ERROR_FILE_NOT_FOUND)) ?
33738		- "ok" : "failed" ));
33739		-
33740		- return ( (rc == INVALID_FILE_ATTRIBUTES)
33741		- && (error == ERROR_FILE_NOT_FOUND)) ? SQLITE_OK :
33742		- winLogError(SQLITE_IOERR_DELETE, "winDelete", zFilename);
	34256	+ OSTRACE(("DELETE \"%s\" %s\n", zFilename, (rc ? "failed" : "ok" )));
	34257	+ return rc;
33743	34258	}
33744	34259
33745	34260	/*
33746	34261	** Check the existance and status of a file.
33747	34262	*/
		@@ -33760,15 +34275,17 @@
33760	34275	zConverted = convertUtf8Filename(zFilename);
33761	34276	if( zConverted==0 ){
33762	34277	return SQLITE_NOMEM;
33763	34278	}
33764	34279	if( isNT() ){
	34280	+ int cnt = 0;
33765	34281	WIN32_FILE_ATTRIBUTE_DATA sAttrData;
33766	34282	memset(&sAttrData, 0, sizeof(sAttrData));
33767		- if( GetFileAttributesExW((WCHAR*)zConverted,
	34283	+ while( !(rc = GetFileAttributesExW((WCHAR*)zConverted,
33768	34284	GetFileExInfoStandard,
33769		- &sAttrData) ){
	34285	+ &sAttrData)) && retryIoerr(&cnt) ){}
	34286	+ if( rc ){
33770	34287	/* For an SQLITE_ACCESS_EXISTS query, treat a zero-length file
33771	34288	** as if it does not exist.
33772	34289	*/
33773	34290	if( flags==SQLITE_ACCESS_EXISTS
33774	34291	&& sAttrData.nFileSizeHigh==0
		@@ -33776,10 +34293,11 @@
33776	34293	attr = INVALID_FILE_ATTRIBUTES;
33777	34294	}else{
33778	34295	attr = sAttrData.dwFileAttributes;
33779	34296	}
33780	34297	}else{
	34298	+ logIoerr(cnt);
33781	34299	if( GetLastError()!=ERROR_FILE_NOT_FOUND ){
33782	34300	winLogError(SQLITE_IOERR_ACCESS, "winAccess", zFilename);
33783	34301	free(zConverted);
33784	34302	return SQLITE_IOERR_ACCESS;
33785	34303	}else{
		@@ -33800,11 +34318,12 @@
33800	34318	case SQLITE_ACCESS_READ:
33801	34319	case SQLITE_ACCESS_EXISTS:
33802	34320	rc = attr!=INVALID_FILE_ATTRIBUTES;
33803	34321	break;
33804	34322	case SQLITE_ACCESS_READWRITE:
33805		- rc = (attr & FILE_ATTRIBUTE_READONLY)==0;
	34323	+ rc = attr!=INVALID_FILE_ATTRIBUTES &&
	34324	+ (attr & FILE_ATTRIBUTE_READONLY)==0;
33806	34325	break;
33807	34326	default:
33808	34327	assert(!"Invalid flags argument");
33809	34328	}
33810	34329	*pResOut = rc;
		@@ -35256,10 +35775,13 @@
35256	35775	typedef struct PCache1 PCache1;
35257	35776	typedef struct PgHdr1 PgHdr1;
35258	35777	typedef struct PgFreeslot PgFreeslot;
35259	35778	typedef struct PGroup PGroup;
35260	35779
	35780	+typedef struct PGroupBlock PGroupBlock;
	35781	+typedef struct PGroupBlockList PGroupBlockList;
	35782	+
35261	35783	/* Each page cache (or PCache) belongs to a PGroup. A PGroup is a set
35262	35784	** of one or more PCaches that are able to recycle each others unpinned
35263	35785	** pages when they are under memory pressure. A PGroup is an instance of
35264	35786	** the following object.
35265	35787	**
		@@ -35285,12 +35807,70 @@
35285	35807	int nMaxPage; /* Sum of nMax for purgeable caches */
35286	35808	int nMinPage; /* Sum of nMin for purgeable caches */
35287	35809	int mxPinned; /* nMaxpage + 10 - nMinPage */
35288	35810	int nCurrentPage; /* Number of purgeable pages allocated */
35289	35811	PgHdr1 pLruHead, pLruTail; /* LRU list of unpinned pages */
	35812	+#ifdef SQLITE_PAGECACHE_BLOCKALLOC
	35813	+ int isBusy; /* Do not run ReleaseMemory() if true */
	35814	+ PGroupBlockList pBlockList; / List of block-lists for this group */
	35815	+#endif
35290	35816	};
35291	35817
	35818	+/*
	35819	+** If SQLITE_PAGECACHE_BLOCKALLOC is defined when the library is built,
	35820	+** each PGroup structure has a linked list of the the following starting
	35821	+** at PGroup.pBlockList. There is one entry for each distinct page-size
	35822	+** currently used by members of the PGroup (i.e. 1024 bytes, 4096 bytes
	35823	+** etc.). Variable PGroupBlockList.nByte is set to the actual allocation
	35824	+** size requested by each pcache, which is the database page-size plus
	35825	+** the various header structures used by the pcache, pager and btree layers.
	35826	+** Usually around (pgsz+200) bytes.
	35827	+**
	35828	+** This size (pgsz+200) bytes is not allocated efficiently by some
	35829	+** implementations of malloc. In particular, some implementations are only
	35830	+** able to allocate blocks of memory chunks of 2^N bytes, where N is some
	35831	+** integer value. Since the page-size is a power of 2, this means we
	35832	+** end up wasting (pgsz-200) bytes in each allocation.
	35833	+**
	35834	+** If SQLITE_PAGECACHE_BLOCKALLOC is defined, the (pgsz+200) byte blocks
	35835	+** are not allocated directly. Instead, blocks of roughly M*(pgsz+200) bytes
	35836	+** are requested from malloc allocator. After a block is returned,
	35837	+** sqlite3MallocSize() is used to determine how many (pgsz+200) byte
	35838	+** allocations can fit in the space returned by malloc(). This value may
	35839	+** be more than M.
	35840	+**
	35841	+** The blocks are stored in a doubly-linked list. Variable PGroupBlock.nEntry
	35842	+** contains the number of allocations that will fit in the aData[] space.
	35843	+** nEntry is limited to the number of bits in bitmask mUsed. If a slot
	35844	+** within aData is in use, the corresponding bit in mUsed is set. Thus
	35845	+** when (mUsed+1==(1 << nEntry)) the block is completely full.
	35846	+**
	35847	+** Each time a slot within a block is freed, the block is moved to the start
	35848	+** of the linked-list. And if a block becomes completely full, then it is
	35849	+** moved to the end of the list. As a result, when searching for a free
	35850	+** slot, only the first block in the list need be examined. If it is full,
	35851	+** then it is guaranteed that all blocks are full.
	35852	+*/
	35853	+struct PGroupBlockList {
	35854	+ int nByte; /* Size of each allocation in bytes */
	35855	+ PGroupBlock pFirst; / First PGroupBlock in list */
	35856	+ PGroupBlock pLast; / Last PGroupBlock in list */
	35857	+ PGroupBlockList pNext; / Next block-list attached to group */
	35858	+};
	35859	+
	35860	+struct PGroupBlock {
	35861	+ Bitmask mUsed; /* Mask of used slots */
	35862	+ int nEntry; /* Maximum number of allocations in aData[] */
	35863	+ u8 aData; / Pointer to data block */
	35864	+ PGroupBlock pNext; / Next PGroupBlock in list */
	35865	+ PGroupBlock pPrev; / Previous PGroupBlock in list */
	35866	+ PGroupBlockList pList; / Owner list */
	35867	+};
	35868	+
	35869	+/* Minimum value for PGroupBlock.nEntry */
	35870	+#define PAGECACHE_BLOCKALLOC_MINENTRY 15
	35871	+
35292	35872	/* Each page cache is an instance of the following object. Every
35293	35873	** open database file (including each in-memory database and each
35294	35874	** temporary or transient database) has a single page cache which
35295	35875	** is an instance of this object.
35296	35876	**
		@@ -35389,10 +35969,21 @@
35389	35969	**
35390	35970	** assert( PGHDR1_TO_PAGE(PAGE_TO_PGHDR1(pCache, X))==X );
35391	35971	*/
35392	35972	#define PGHDR1_TO_PAGE(p) (void)(((char)p) - p->pCache->szPage)
35393	35973	#define PAGE_TO_PGHDR1(c, p) (PgHdr1)(((char)p) + c->szPage)
	35974	+
	35975	+/*
	35976	+** Blocks used by the SQLITE_PAGECACHE_BLOCKALLOC blocks to store/retrieve
	35977	+** a PGroupBlock pointer based on a pointer to a page buffer.
	35978	+*/
	35979	+#define PAGE_SET_BLOCKPTR(pCache, pPg, pBlock) \
	35980	+ ( (PGroupBlock )&(((u8)pPg)[sizeof(PgHdr1) + pCache->szPage]) = pBlock )
	35981	+
	35982	+#define PAGE_GET_BLOCKPTR(pCache, pPg) \
	35983	+ ( (PGroupBlock )&(((u8)pPg)[sizeof(PgHdr1) + pCache->szPage]) )
	35984	+
35394	35985
35395	35986	/*
35396	35987	** Macros to enter and leave the PCache LRU mutex.
35397	35988	*/
35398	35989	#define pcache1EnterMutex(X) sqlite3_mutex_enter((X)->mutex)
		@@ -35515,25 +36106,159 @@
35515	36106	return iSize;
35516	36107	}
35517	36108	}
35518	36109	#endif /* SQLITE_ENABLE_MEMORY_MANAGEMENT */
35519	36110
	36111	+#ifdef SQLITE_PAGECACHE_BLOCKALLOC
	36112	+/*
	36113	+** The block pBlock belongs to list pList but is not currently linked in.
	36114	+** Insert it into the start of the list.
	36115	+*/
	36116	+static void addBlockToList(PGroupBlockList pList, PGroupBlock pBlock){
	36117	+ pBlock->pPrev = 0;
	36118	+ pBlock->pNext = pList->pFirst;
	36119	+ pList->pFirst = pBlock;
	36120	+ if( pBlock->pNext ){
	36121	+ pBlock->pNext->pPrev = pBlock;
	36122	+ }else{
	36123	+ assert( pList->pLast==0 );
	36124	+ pList->pLast = pBlock;
	36125	+ }
	36126	+}
	36127	+
	36128	+/*
	36129	+** If there are no blocks in the list headed by pList, remove pList
	36130	+** from the pGroup->pBlockList list and free it with sqlite3_free().
	36131	+*/
	36132	+static void freeListIfEmpty(PGroup pGroup, PGroupBlockList pList){
	36133	+ assert( sqlite3_mutex_held(pGroup->mutex) );
	36134	+ if( pList->pFirst==0 ){
	36135	+ PGroupBlockList **pp;
	36136	+ for(pp=&pGroup->pBlockList; pp!=pList; pp=&(pp)->pNext);
	36137	+ pp = (pp)->pNext;
	36138	+ sqlite3_free(pList);
	36139	+ }
	36140	+}
	36141	+#endif /* SQLITE_PAGECACHE_BLOCKALLOC */
	36142	+
35520	36143	/*
35521	36144	** Allocate a new page object initially associated with cache pCache.
35522	36145	*/
35523	36146	static PgHdr1 pcache1AllocPage(PCache1 pCache){
35524	36147	int nByte = sizeof(PgHdr1) + pCache->szPage;
35525		- void *pPg = pcache1Alloc(nByte);
	36148	+ void *pPg = 0;
35526	36149	PgHdr1 *p;
	36150	+
	36151	+#ifdef SQLITE_PAGECACHE_BLOCKALLOC
	36152	+ PGroup *pGroup = pCache->pGroup;
	36153	+ PGroupBlockList *pList;
	36154	+ PGroupBlock *pBlock;
	36155	+ int i;
	36156	+
	36157	+ nByte += sizeof(PGroupBlockList *);
	36158	+ nByte = ROUND8(nByte);
	36159	+
	36160	+ for(pList=pGroup->pBlockList; pList; pList=pList->pNext){
	36161	+ if( pList->nByte==nByte ) break;
	36162	+ }
	36163	+ if( pList==0 ){
	36164	+ PGroupBlockList *pNew;
	36165	+ assert( pGroup->isBusy==0 );
	36166	+ assert( sqlite3_mutex_held(pGroup->mutex) );
	36167	+ pGroup->isBusy = 1; /* Disable sqlite3PcacheReleaseMemory() */
	36168	+ pNew = (PGroupBlockList *)sqlite3MallocZero(sizeof(PGroupBlockList));
	36169	+ pGroup->isBusy = 0; /* Reenable sqlite3PcacheReleaseMemory() */
	36170	+ if( pNew==0 ){
	36171	+ /* malloc() failure. Return early. */
	36172	+ return 0;
	36173	+ }
	36174	+#ifdef SQLITE_DEBUG
	36175	+ for(pList=pGroup->pBlockList; pList; pList=pList->pNext){
	36176	+ assert( pList->nByte!=nByte );
	36177	+ }
	36178	+#endif
	36179	+ pNew->nByte = nByte;
	36180	+ pNew->pNext = pGroup->pBlockList;
	36181	+ pGroup->pBlockList = pNew;
	36182	+ pList = pNew;
	36183	+ }
	36184	+
	36185	+ pBlock = pList->pFirst;
	36186	+ if( pBlock==0 \|\| pBlock->mUsed==(((Bitmask)1<<pBlock->nEntry)-1) ){
	36187	+ int sz;
	36188	+
	36189	+ /* Allocate a new block. Try to allocate enough space for the PGroupBlock
	36190	+ ** structure and MINENTRY allocations of nByte bytes each. If the
	36191	+ ** allocator returns more memory than requested, then more than MINENTRY
	36192	+ ** allocations may fit in it. */
	36193	+ assert( sqlite3_mutex_held(pGroup->mutex) );
	36194	+ pcache1LeaveMutex(pCache->pGroup);
	36195	+ sz = sizeof(PGroupBlock) + PAGECACHE_BLOCKALLOC_MINENTRY * nByte;
	36196	+ pBlock = (PGroupBlock *)sqlite3Malloc(sz);
	36197	+ pcache1EnterMutex(pCache->pGroup);
	36198	+
	36199	+ if( !pBlock ){
	36200	+ freeListIfEmpty(pGroup, pList);
	36201	+ return 0;
	36202	+ }
	36203	+ pBlock->nEntry = (sqlite3MallocSize(pBlock) - sizeof(PGroupBlock)) / nByte;
	36204	+ if( pBlock->nEntry>=BMS ){
	36205	+ pBlock->nEntry = BMS-1;
	36206	+ }
	36207	+ pBlock->pList = pList;
	36208	+ pBlock->mUsed = 0;
	36209	+ pBlock->aData = (u8 *)&pBlock[1];
	36210	+ addBlockToList(pList, pBlock);
	36211	+
	36212	+ sz = sqlite3MallocSize(pBlock);
	36213	+ sqlite3_mutex_enter(pcache1.mutex);
	36214	+ sqlite3StatusAdd(SQLITE_STATUS_PAGECACHE_OVERFLOW, sz);
	36215	+ sqlite3_mutex_leave(pcache1.mutex);
	36216	+ }
	36217	+
	36218	+ for(i=0; pPg==0 && ALWAYS(i<pBlock->nEntry); i++){
	36219	+ if( 0==(pBlock->mUsed & ((Bitmask)1<<i)) ){
	36220	+ pBlock->mUsed \|= ((Bitmask)1<<i);
	36221	+ pPg = (void )&pBlock->aData[pList->nByte i];
	36222	+ }
	36223	+ }
	36224	+ assert( pPg );
	36225	+ PAGE_SET_BLOCKPTR(pCache, pPg, pBlock);
	36226	+
	36227	+ /* If the block is now full, shift it to the end of the list */
	36228	+ if( pBlock->mUsed==(((Bitmask)1<<pBlock->nEntry)-1) && pList->pLast!=pBlock ){
	36229	+ assert( pList->pFirst==pBlock );
	36230	+ assert( pBlock->pPrev==0 );
	36231	+ assert( pList->pLast->pNext==0 );
	36232	+ pList->pFirst = pBlock->pNext;
	36233	+ pList->pFirst->pPrev = 0;
	36234	+ pBlock->pPrev = pList->pLast;
	36235	+ pBlock->pNext = 0;
	36236	+ pList->pLast->pNext = pBlock;
	36237	+ pList->pLast = pBlock;
	36238	+ }
	36239	+ p = PAGE_TO_PGHDR1(pCache, pPg);
	36240	+ if( pCache->bPurgeable ){
	36241	+ pCache->pGroup->nCurrentPage++;
	36242	+ }
	36243	+#else
	36244	+ /* The group mutex must be released before pcache1Alloc() is called. This
	36245	+ ** is because it may call sqlite3_release_memory(), which assumes that
	36246	+ ** this mutex is not held. */
	36247	+ assert( sqlite3_mutex_held(pCache->pGroup->mutex) );
	36248	+ pcache1LeaveMutex(pCache->pGroup);
	36249	+ pPg = pcache1Alloc(nByte);
	36250	+ pcache1EnterMutex(pCache->pGroup);
35527	36251	if( pPg ){
35528	36252	p = PAGE_TO_PGHDR1(pCache, pPg);
35529	36253	if( pCache->bPurgeable ){
35530	36254	pCache->pGroup->nCurrentPage++;
35531	36255	}
35532	36256	}else{
35533	36257	p = 0;
35534	36258	}
	36259	+#endif
35535	36260	return p;
35536	36261	}
35537	36262
35538	36263	/*
35539	36264	** Free a page object allocated by pcache1AllocPage().
		@@ -35543,14 +36268,56 @@
35543	36268	** with a NULL pointer, so we mark the NULL test with ALWAYS().
35544	36269	*/
35545	36270	static void pcache1FreePage(PgHdr1 *p){
35546	36271	if( ALWAYS(p) ){
35547	36272	PCache1 *pCache = p->pCache;
	36273	+ void *pPg = PGHDR1_TO_PAGE(p);
	36274	+
	36275	+#ifdef SQLITE_PAGECACHE_BLOCKALLOC
	36276	+ PGroupBlock *pBlock = PAGE_GET_BLOCKPTR(pCache, pPg);
	36277	+ PGroupBlockList *pList = pBlock->pList;
	36278	+ int i = ((u8 *)pPg - pBlock->aData) / pList->nByte;
	36279	+
	36280	+ assert( pPg==(void )&pBlock->aData[ipList->nByte] );
	36281	+ assert( pBlock->mUsed & ((Bitmask)1<<i) );
	36282	+ pBlock->mUsed &= ~((Bitmask)1<<i);
	36283	+
	36284	+ /* Remove the block from the list. If it is completely empty, free it.
	36285	+ ** Or if it is not completely empty, re-insert it at the start of the
	36286	+ ** list. */
	36287	+ if( pList->pFirst==pBlock ){
	36288	+ pList->pFirst = pBlock->pNext;
	36289	+ if( pList->pFirst ) pList->pFirst->pPrev = 0;
	36290	+ }else{
	36291	+ pBlock->pPrev->pNext = pBlock->pNext;
	36292	+ }
	36293	+ if( pList->pLast==pBlock ){
	36294	+ pList->pLast = pBlock->pPrev;
	36295	+ if( pList->pLast ) pList->pLast->pNext = 0;
	36296	+ }else{
	36297	+ pBlock->pNext->pPrev = pBlock->pPrev;
	36298	+ }
	36299	+
	36300	+ if( pBlock->mUsed==0 ){
	36301	+ PGroup *pGroup = p->pCache->pGroup;
	36302	+
	36303	+ int sz = sqlite3MallocSize(pBlock);
	36304	+ sqlite3_mutex_enter(pcache1.mutex);
	36305	+ sqlite3StatusAdd(SQLITE_STATUS_PAGECACHE_OVERFLOW, -sz);
	36306	+ sqlite3_mutex_leave(pcache1.mutex);
	36307	+ freeListIfEmpty(pGroup, pList);
	36308	+ sqlite3_free(pBlock);
	36309	+ }else{
	36310	+ addBlockToList(pList, pBlock);
	36311	+ }
	36312	+#else
	36313	+ assert( sqlite3_mutex_held(p->pCache->pGroup->mutex) );
	36314	+ pcache1Free(pPg);
	36315	+#endif
35548	36316	if( pCache->bPurgeable ){
35549	36317	pCache->pGroup->nCurrentPage--;
35550	36318	}
35551		- pcache1Free(PGHDR1_TO_PAGE(p));
35552	36319	}
35553	36320	}
35554	36321
35555	36322	/*
35556	36323	** Malloc function used by SQLite to obtain space from the buffer configured
		@@ -35984,13 +36751,11 @@
35984	36751	/* Step 5. If a usable page buffer has still not been found,
35985	36752	** attempt to allocate a new one.
35986	36753	*/
35987	36754	if( !pPage ){
35988	36755	if( createFlag==1 ) sqlite3BeginBenignMalloc();
35989		- pcache1LeaveMutex(pGroup);
35990	36756	pPage = pcache1AllocPage(pCache);
35991		- pcache1EnterMutex(pGroup);
35992	36757	if( createFlag==1 ) sqlite3EndBenignMalloc();
35993	36758	}
35994	36759
35995	36760	if( pPage ){
35996	36761	unsigned int h = iKey % pCache->nHash;
		@@ -36156,10 +36921,13 @@
36156	36921	** been released, the function returns. The return value is the total number
36157	36922	** of bytes of memory released.
36158	36923	*/
36159	36924	SQLITE_PRIVATE int sqlite3PcacheReleaseMemory(int nReq){
36160	36925	int nFree = 0;
	36926	+#ifdef SQLITE_PAGECACHE_BLOCKALLOC
	36927	+ if( pcache1.grp.isBusy ) return 0;
	36928	+#endif
36161	36929	assert( sqlite3_mutex_notheld(pcache1.grp.mutex) );
36162	36930	assert( sqlite3_mutex_notheld(pcache1.mutex) );
36163	36931	if( pcache1.pStart==0 ){
36164	36932	PgHdr1 *p;
36165	36933	pcache1EnterMutex(&pcache1.grp);
		@@ -37368,10 +38136,12 @@
37368	38136	u8 ckptSyncFlags; /* SYNC_NORMAL or SYNC_FULL for checkpoint */
37369	38137	u8 syncFlags; /* SYNC_NORMAL or SYNC_FULL otherwise */
37370	38138	u8 tempFile; /* zFilename is a temporary file */
37371	38139	u8 readOnly; /* True for a read-only database */
37372	38140	u8 memDb; /* True to inhibit all file I/O */
	38141	+ u8 hasSeenStress; /* pagerStress() called one or more times */
	38142	+ u8 isSorter; /* True for a PAGER_SORTER */
37373	38143
37374	38144	/**************************************************************************
37375	38145	** The following block contains those class members that change during
37376	38146	** routine opertion. Class members not in this block are either fixed
37377	38147	** when the pager is first created or else only change when there is a
		@@ -37590,10 +38360,19 @@
37590	38360	\|\| p->journalMode==PAGER_JOURNALMODE_MEMORY
37591	38361	);
37592	38362	assert( p->eState!=PAGER_ERROR && p->eState!=PAGER_OPEN );
37593	38363	assert( pagerUseWal(p)==0 );
37594	38364	}
	38365	+
	38366	+ /* A sorter is a temp file that never spills to disk and always has
	38367	+ ** the doNotSpill flag set
	38368	+ */
	38369	+ if( p->isSorter ){
	38370	+ assert( p->tempFile );
	38371	+ assert( p->doNotSpill );
	38372	+ assert( p->fd->pMethods==0 );
	38373	+ }
37595	38374
37596	38375	/* If changeCountDone is set, a RESERVED lock or greater must be held
37597	38376	** on the file.
37598	38377	*/
37599	38378	assert( pPager->changeCountDone==0 \|\| pPager->eLock>=RESERVED_LOCK );
		@@ -40487,10 +41266,11 @@
40487	41266	** to the caller.
40488	41267	*/
40489	41268	SQLITE_PRIVATE int sqlite3PagerClose(Pager *pPager){
40490	41269	u8 pTmp = (u8 )pPager->pTmpSpace;
40491	41270
	41271	+ assert( assert_pager_state(pPager) );
40492	41272	disable_simulated_io_errors();
40493	41273	sqlite3BeginBenignMalloc();
40494	41274	/* pPager->errCode = 0; */
40495	41275	pPager->exclusiveMode = 0;
40496	41276	#ifndef SQLITE_OMIT_WAL
		@@ -40921,10 +41701,11 @@
40921	41701	** is impossible for sqlite3PCacheFetch() to be called with createFlag==1
40922	41702	** while in the error state, hence it is impossible for this routine to
40923	41703	** be called in the error state. Nevertheless, we include a NEVER()
40924	41704	** test for the error state as a safeguard against future changes.
40925	41705	*/
	41706	+ pPager->hasSeenStress = 1;
40926	41707	if( NEVER(pPager->errCode) ) return SQLITE_OK;
40927	41708	if( pPager->doNotSpill ) return SQLITE_OK;
40928	41709	if( pPager->doNotSyncSpill && (pPg->flags & PGHDR_NEED_SYNC)!=0 ){
40929	41710	return SQLITE_OK;
40930	41711	}
		@@ -41292,10 +42073,16 @@
41292	42073	}
41293	42074	/* pPager->xBusyHandler = 0; */
41294	42075	/* pPager->pBusyHandlerArg = 0; */
41295	42076	pPager->xReiniter = xReinit;
41296	42077	/* memset(pPager->aHash, 0, sizeof(pPager->aHash)); */
	42078	+#ifndef SQLITE_OMIT_MERGE_SORT
	42079	+ if( flags & PAGER_SORTER ){
	42080	+ pPager->doNotSpill = 1;
	42081	+ pPager->isSorter = 1;
	42082	+ }
	42083	+#endif
41297	42084
41298	42085	*ppPager = pPager;
41299	42086	return SQLITE_OK;
41300	42087	}
41301	42088
		@@ -42835,10 +43622,21 @@
42835	43622	** Return true if this is an in-memory pager.
42836	43623	*/
42837	43624	SQLITE_PRIVATE int sqlite3PagerIsMemdb(Pager *pPager){
42838	43625	return MEMDB;
42839	43626	}
	43627	+
	43628	+#ifndef SQLITE_OMIT_MERGE_SORT
	43629	+/*
	43630	+** Return true if the pager has seen a pagerStress callback.
	43631	+*/
	43632	+SQLITE_PRIVATE int sqlite3PagerUnderStress(Pager *pPager){
	43633	+ assert( pPager->isSorter );
	43634	+ assert( pPager->doNotSpill );
	43635	+ return pPager->hasSeenStress;
	43636	+}
	43637	+#endif
42840	43638
42841	43639	/*
42842	43640	** Check that there are at least nSavepoint savepoints open. If there are
42843	43641	** currently less than nSavepoints open, then open one or more savepoints
42844	43642	** to make up the difference. If the number of savepoints is already
		@@ -45384,17 +46182,19 @@
45384	46182	**
45385	46183	** The EXCLUSIVE lock is not released before returning.
45386	46184	*/
45387	46185	rc = sqlite3OsLock(pWal->pDbFd, SQLITE_LOCK_EXCLUSIVE);
45388	46186	if( rc==SQLITE_OK ){
	46187	+ int bPersistWal = -1;
45389	46188	if( pWal->exclusiveMode==WAL_NORMAL_MODE ){
45390	46189	pWal->exclusiveMode = WAL_EXCLUSIVE_MODE;
45391	46190	}
45392	46191	rc = sqlite3WalCheckpoint(
45393	46192	pWal, SQLITE_CHECKPOINT_PASSIVE, 0, 0, sync_flags, nBuf, zBuf, 0, 0
45394	46193	);
45395		- if( rc==SQLITE_OK ){
	46194	+ sqlite3OsFileControl(pWal->pDbFd, SQLITE_FCNTL_PERSIST_WAL, &bPersistWal);
	46195	+ if( rc==SQLITE_OK && bPersistWal!=1 ){
45396	46196	isDelete = 1;
45397	46197	}
45398	46198	}
45399	46199
45400	46200	walIndexClose(pWal, isDelete);
		@@ -49202,15 +50002,26 @@
49202	50002	assert( (flags & BTREE_UNORDERED)==0 \|\| (flags & BTREE_SINGLE)!=0 );
49203	50003
49204	50004	/* A BTREE_SINGLE database is always a temporary and/or ephemeral */
49205	50005	assert( (flags & BTREE_SINGLE)==0 \|\| isTempDb );
49206	50006
	50007	+ /* The BTREE_SORTER flag is only used if SQLITE_OMIT_MERGE_SORT is undef */
	50008	+#ifdef SQLITE_OMIT_MERGE_SORT
	50009	+ assert( (flags & BTREE_SORTER)==0 );
	50010	+#endif
	50011	+
	50012	+ /* BTREE_SORTER is always on a BTREE_SINGLE, BTREE_OMIT_JOURNAL */
	50013	+ assert( (flags & BTREE_SORTER)==0 \|\|
	50014	+ (flags & (BTREE_SINGLE\|BTREE_OMIT_JOURNAL))
	50015	+ ==(BTREE_SINGLE\|BTREE_OMIT_JOURNAL) );
	50016	+
49207	50017	if( db->flags & SQLITE_NoReadlock ){
49208	50018	flags \|= BTREE_NO_READLOCK;
49209	50019	}
49210	50020	if( isMemdb ){
49211	50021	flags \|= BTREE_MEMORY;
	50022	+ flags &= ~BTREE_SORTER;
49212	50023	}
49213	50024	if( (vfsFlags & SQLITE_OPEN_MAIN_DB)!=0 && (isMemdb \|\| isTempDb) ){
49214	50025	vfsFlags = (vfsFlags & ~SQLITE_OPEN_MAIN_DB) \| SQLITE_OPEN_TEMP_DB;
49215	50026	}
49216	50027	p = sqlite3MallocZero(sizeof(Btree));
		@@ -50936,11 +51747,12 @@
50936	51747
50937	51748	if( NEVER(wrFlag && pBt->readOnly) ){
50938	51749	return SQLITE_READONLY;
50939	51750	}
50940	51751	if( iTable==1 && btreePagecount(pBt)==0 ){
50941		- return SQLITE_EMPTY;
	51752	+ assert( wrFlag==0 );
	51753	+ iTable = 0;
50942	51754	}
50943	51755
50944	51756	/* Now that no other errors can occur, finish filling in the BtCursor
50945	51757	** variables and link the cursor into the BtShared list. */
50946	51758	pCur->pgnoRoot = (Pgno)iTable;
		@@ -51690,10 +52502,13 @@
51690	52502	int i;
51691	52503	for(i=1; i<=pCur->iPage; i++){
51692	52504	releasePage(pCur->apPage[i]);
51693	52505	}
51694	52506	pCur->iPage = 0;
	52507	+ }else if( pCur->pgnoRoot==0 ){
	52508	+ pCur->eState = CURSOR_INVALID;
	52509	+ return SQLITE_OK;
51695	52510	}else{
51696	52511	rc = getAndInitPage(pBt, pCur->pgnoRoot, &pCur->apPage[0]);
51697	52512	if( rc!=SQLITE_OK ){
51698	52513	pCur->eState = CURSOR_INVALID;
51699	52514	return rc;
		@@ -51799,11 +52614,11 @@
51799	52614	assert( cursorHoldsMutex(pCur) );
51800	52615	assert( sqlite3_mutex_held(pCur->pBtree->db->mutex) );
51801	52616	rc = moveToRoot(pCur);
51802	52617	if( rc==SQLITE_OK ){
51803	52618	if( pCur->eState==CURSOR_INVALID ){
51804		- assert( pCur->apPage[pCur->iPage]->nCell==0 );
	52619	+ assert( pCur->pgnoRoot==0 \|\| pCur->apPage[pCur->iPage]->nCell==0 );
51805	52620	*pRes = 1;
51806	52621	}else{
51807	52622	assert( pCur->apPage[pCur->iPage]->nCell>0 );
51808	52623	*pRes = 0;
51809	52624	rc = moveToLeftmost(pCur);
		@@ -51838,11 +52653,11 @@
51838	52653	}
51839	52654
51840	52655	rc = moveToRoot(pCur);
51841	52656	if( rc==SQLITE_OK ){
51842	52657	if( CURSOR_INVALID==pCur->eState ){
51843		- assert( pCur->apPage[pCur->iPage]->nCell==0 );
	52658	+ assert( pCur->pgnoRoot==0 \|\| pCur->apPage[pCur->iPage]->nCell==0 );
51844	52659	*pRes = 1;
51845	52660	}else{
51846	52661	assert( pCur->eState==CURSOR_VALID );
51847	52662	*pRes = 0;
51848	52663	rc = moveToRightmost(pCur);
		@@ -51911,16 +52726,16 @@
51911	52726
51912	52727	rc = moveToRoot(pCur);
51913	52728	if( rc ){
51914	52729	return rc;
51915	52730	}
51916		- assert( pCur->apPage[pCur->iPage] );
51917		- assert( pCur->apPage[pCur->iPage]->isInit );
51918		- assert( pCur->apPage[pCur->iPage]->nCell>0 \|\| pCur->eState==CURSOR_INVALID );
	52731	+ assert( pCur->pgnoRoot==0 \|\| pCur->apPage[pCur->iPage] );
	52732	+ assert( pCur->pgnoRoot==0 \|\| pCur->apPage[pCur->iPage]->isInit );
	52733	+ assert( pCur->eState==CURSOR_INVALID \|\| pCur->apPage[pCur->iPage]->nCell>0 );
51919	52734	if( pCur->eState==CURSOR_INVALID ){
51920	52735	*pRes = -1;
51921		- assert( pCur->apPage[pCur->iPage]->nCell==0 );
	52736	+ assert( pCur->pgnoRoot==0 \|\| pCur->apPage[pCur->iPage]->nCell==0 );
51922	52737	return SQLITE_OK;
51923	52738	}
51924	52739	assert( pCur->apPage[0]->intKey \|\| pIdxKey );
51925	52740	for(;;){
51926	52741	int lwr, upr, idx;
		@@ -54745,13 +55560,20 @@
54745	55560	releasePage(pPage);
54746	55561	}
54747	55562	return rc;
54748	55563	}
54749	55564	SQLITE_PRIVATE int sqlite3BtreeDropTable(Btree p, int iTable, int piMoved){
	55565	+ BtShared *pBt = p->pBt;
54750	55566	int rc;
54751	55567	sqlite3BtreeEnter(p);
54752		- rc = btreeDropTable(p, iTable, piMoved);
	55568	+ if( (pBt->openFlags&BTREE_SINGLE) ){
	55569	+ pBt->nPage = 0;
	55570	+ sqlite3PagerTruncateImage(pBt->pPager, 1);
	55571	+ rc = newDatabase(pBt);
	55572	+ }else{
	55573	+ rc = btreeDropTable(p, iTable, piMoved);
	55574	+ }
54753	55575	sqlite3BtreeLeave(p);
54754	55576	return rc;
54755	55577	}
54756	55578
54757	55579
		@@ -54826,10 +55648,15 @@
54826	55648	** corruption) an SQLite error code is returned.
54827	55649	*/
54828	55650	SQLITE_PRIVATE int sqlite3BtreeCount(BtCursor pCur, i64 pnEntry){
54829	55651	i64 nEntry = 0; /* Value to return in pnEntry /
54830	55652	int rc; /* Return code */
	55653	+
	55654	+ if( pCur->pgnoRoot==0 ){
	55655	+ *pnEntry = 0;
	55656	+ return SQLITE_OK;
	55657	+ }
54831	55658	rc = moveToRoot(pCur);
54832	55659
54833	55660	/* Unless an error occurs, the following loop runs one iteration for each
54834	55661	** page in the B-Tree structure (not including overflow pages).
54835	55662	*/
		@@ -55610,11 +56437,10 @@
55610	56437	*/
55611	56438	SQLITE_PRIVATE int sqlite3BtreeSetVersion(Btree *pBtree, int iVersion){
55612	56439	BtShared *pBt = pBtree->pBt;
55613	56440	int rc; /* Return code */
55614	56441
55615		- assert( pBtree->inTrans==TRANS_NONE );
55616	56442	assert( iVersion==1 \|\| iVersion==2 );
55617	56443
55618	56444	/* If setting the version fields to 1, do not automatically open the
55619	56445	** WAL connection, even if the version fields are currently set to 2.
55620	56446	*/
		@@ -56049,106 +56875,110 @@
56049	56875	/* Update the schema version field in the destination database. This
56050	56876	** is to make sure that the schema-version really does change in
56051	56877	** the case where the source and destination databases have the
56052	56878	** same schema version.
56053	56879	*/
56054		- if( rc==SQLITE_DONE
56055		- && (rc = sqlite3BtreeUpdateMeta(p->pDest,1,p->iDestSchema+1))==SQLITE_OK
56056		- ){
56057		- int nDestTruncate;
56058		-
56059		- if( p->pDestDb ){
56060		- sqlite3ResetInternalSchema(p->pDestDb, -1);
56061		- }
56062		-
56063		- /* Set nDestTruncate to the final number of pages in the destination
56064		- ** database. The complication here is that the destination page
56065		- ** size may be different to the source page size.
56066		- **
56067		- ** If the source page size is smaller than the destination page size,
56068		- ** round up. In this case the call to sqlite3OsTruncate() below will
56069		- ** fix the size of the file. However it is important to call
56070		- ** sqlite3PagerTruncateImage() here so that any pages in the
56071		- ** destination file that lie beyond the nDestTruncate page mark are
56072		- ** journalled by PagerCommitPhaseOne() before they are destroyed
56073		- ** by the file truncation.
56074		- */
56075		- assert( pgszSrc==sqlite3BtreeGetPageSize(p->pSrc) );
56076		- assert( pgszDest==sqlite3BtreeGetPageSize(p->pDest) );
56077		- if( pgszSrc<pgszDest ){
56078		- int ratio = pgszDest/pgszSrc;
56079		- nDestTruncate = (nSrcPage+ratio-1)/ratio;
56080		- if( nDestTruncate==(int)PENDING_BYTE_PAGE(p->pDest->pBt) ){
56081		- nDestTruncate--;
56082		- }
56083		- }else{
56084		- nDestTruncate = nSrcPage * (pgszSrc/pgszDest);
56085		- }
56086		- sqlite3PagerTruncateImage(pDestPager, nDestTruncate);
56087		-
56088		- if( pgszSrc<pgszDest ){
56089		- /* If the source page-size is smaller than the destination page-size,
56090		- ** two extra things may need to happen:
56091		- **
56092		- ** * The destination may need to be truncated, and
56093		- **
56094		- ** * Data stored on the pages immediately following the
56095		- ** pending-byte page in the source database may need to be
56096		- ** copied into the destination database.
56097		- */
56098		- const i64 iSize = (i64)pgszSrc * (i64)nSrcPage;
56099		- sqlite3_file * const pFile = sqlite3PagerFile(pDestPager);
56100		- i64 iOff;
56101		- i64 iEnd;
56102		-
56103		- assert( pFile );
56104		- assert( (i64)nDestTruncate*(i64)pgszDest >= iSize \|\| (
56105		- nDestTruncate==(int)(PENDING_BYTE_PAGE(p->pDest->pBt)-1)
56106		- && iSize>=PENDING_BYTE && iSize<=PENDING_BYTE+pgszDest
56107		- ));
56108		-
56109		- /* This call ensures that all data required to recreate the original
56110		- ** database has been stored in the journal for pDestPager and the
56111		- ** journal synced to disk. So at this point we may safely modify
56112		- ** the database file in any way, knowing that if a power failure
56113		- ** occurs, the original database will be reconstructed from the
56114		- ** journal file. */
56115		- rc = sqlite3PagerCommitPhaseOne(pDestPager, 0, 1);
56116		-
56117		- /* Write the extra pages and truncate the database file as required. */
56118		- iEnd = MIN(PENDING_BYTE + pgszDest, iSize);
56119		- for(
56120		- iOff=PENDING_BYTE+pgszSrc;
56121		- rc==SQLITE_OK && iOff<iEnd;
56122		- iOff+=pgszSrc
56123		- ){
56124		- PgHdr *pSrcPg = 0;
56125		- const Pgno iSrcPg = (Pgno)((iOff/pgszSrc)+1);
56126		- rc = sqlite3PagerGet(pSrcPager, iSrcPg, &pSrcPg);
56127		- if( rc==SQLITE_OK ){
56128		- u8 *zData = sqlite3PagerGetData(pSrcPg);
56129		- rc = sqlite3OsWrite(pFile, zData, pgszSrc, iOff);
56130		- }
56131		- sqlite3PagerUnref(pSrcPg);
56132		- }
56133		- if( rc==SQLITE_OK ){
56134		- rc = backupTruncateFile(pFile, iSize);
56135		- }
56136		-
56137		- /* Sync the database file to disk. */
56138		- if( rc==SQLITE_OK ){
56139		- rc = sqlite3PagerSync(pDestPager);
56140		- }
56141		- }else{
56142		- rc = sqlite3PagerCommitPhaseOne(pDestPager, 0, 0);
56143		- }
56144		-
56145		- /* Finish committing the transaction to the destination database. */
56146		- if( SQLITE_OK==rc
56147		- && SQLITE_OK==(rc = sqlite3BtreeCommitPhaseTwo(p->pDest, 0))
56148		- ){
56149		- rc = SQLITE_DONE;
	56880	+ if( rc==SQLITE_DONE ){
	56881	+ rc = sqlite3BtreeUpdateMeta(p->pDest,1,p->iDestSchema+1);
	56882	+ if( rc==SQLITE_OK ){
	56883	+ if( p->pDestDb ){
	56884	+ sqlite3ResetInternalSchema(p->pDestDb, -1);
	56885	+ }
	56886	+ if( destMode==PAGER_JOURNALMODE_WAL ){
	56887	+ rc = sqlite3BtreeSetVersion(p->pDest, 2);
	56888	+ }
	56889	+ }
	56890	+ if( rc==SQLITE_OK ){
	56891	+ int nDestTruncate;
	56892	+ /* Set nDestTruncate to the final number of pages in the destination
	56893	+ ** database. The complication here is that the destination page
	56894	+ ** size may be different to the source page size.
	56895	+ **
	56896	+ ** If the source page size is smaller than the destination page size,
	56897	+ ** round up. In this case the call to sqlite3OsTruncate() below will
	56898	+ ** fix the size of the file. However it is important to call
	56899	+ ** sqlite3PagerTruncateImage() here so that any pages in the
	56900	+ ** destination file that lie beyond the nDestTruncate page mark are
	56901	+ ** journalled by PagerCommitPhaseOne() before they are destroyed
	56902	+ ** by the file truncation.
	56903	+ */
	56904	+ assert( pgszSrc==sqlite3BtreeGetPageSize(p->pSrc) );
	56905	+ assert( pgszDest==sqlite3BtreeGetPageSize(p->pDest) );
	56906	+ if( pgszSrc<pgszDest ){
	56907	+ int ratio = pgszDest/pgszSrc;
	56908	+ nDestTruncate = (nSrcPage+ratio-1)/ratio;
	56909	+ if( nDestTruncate==(int)PENDING_BYTE_PAGE(p->pDest->pBt) ){
	56910	+ nDestTruncate--;
	56911	+ }
	56912	+ }else{
	56913	+ nDestTruncate = nSrcPage * (pgszSrc/pgszDest);
	56914	+ }
	56915	+ sqlite3PagerTruncateImage(pDestPager, nDestTruncate);
	56916	+
	56917	+ if( pgszSrc<pgszDest ){
	56918	+ /* If the source page-size is smaller than the destination page-size,
	56919	+ ** two extra things may need to happen:
	56920	+ **
	56921	+ ** * The destination may need to be truncated, and
	56922	+ **
	56923	+ ** * Data stored on the pages immediately following the
	56924	+ ** pending-byte page in the source database may need to be
	56925	+ ** copied into the destination database.
	56926	+ */
	56927	+ const i64 iSize = (i64)pgszSrc * (i64)nSrcPage;
	56928	+ sqlite3_file * const pFile = sqlite3PagerFile(pDestPager);
	56929	+ i64 iOff;
	56930	+ i64 iEnd;
	56931	+
	56932	+ assert( pFile );
	56933	+ assert( (i64)nDestTruncate*(i64)pgszDest >= iSize \|\| (
	56934	+ nDestTruncate==(int)(PENDING_BYTE_PAGE(p->pDest->pBt)-1)
	56935	+ && iSize>=PENDING_BYTE && iSize<=PENDING_BYTE+pgszDest
	56936	+ ));
	56937	+
	56938	+ /* This call ensures that all data required to recreate the original
	56939	+ ** database has been stored in the journal for pDestPager and the
	56940	+ ** journal synced to disk. So at this point we may safely modify
	56941	+ ** the database file in any way, knowing that if a power failure
	56942	+ ** occurs, the original database will be reconstructed from the
	56943	+ ** journal file. */
	56944	+ rc = sqlite3PagerCommitPhaseOne(pDestPager, 0, 1);
	56945	+
	56946	+ /* Write the extra pages and truncate the database file as required */
	56947	+ iEnd = MIN(PENDING_BYTE + pgszDest, iSize);
	56948	+ for(
	56949	+ iOff=PENDING_BYTE+pgszSrc;
	56950	+ rc==SQLITE_OK && iOff<iEnd;
	56951	+ iOff+=pgszSrc
	56952	+ ){
	56953	+ PgHdr *pSrcPg = 0;
	56954	+ const Pgno iSrcPg = (Pgno)((iOff/pgszSrc)+1);
	56955	+ rc = sqlite3PagerGet(pSrcPager, iSrcPg, &pSrcPg);
	56956	+ if( rc==SQLITE_OK ){
	56957	+ u8 *zData = sqlite3PagerGetData(pSrcPg);
	56958	+ rc = sqlite3OsWrite(pFile, zData, pgszSrc, iOff);
	56959	+ }
	56960	+ sqlite3PagerUnref(pSrcPg);
	56961	+ }
	56962	+ if( rc==SQLITE_OK ){
	56963	+ rc = backupTruncateFile(pFile, iSize);
	56964	+ }
	56965	+
	56966	+ /* Sync the database file to disk. */
	56967	+ if( rc==SQLITE_OK ){
	56968	+ rc = sqlite3PagerSync(pDestPager);
	56969	+ }
	56970	+ }else{
	56971	+ rc = sqlite3PagerCommitPhaseOne(pDestPager, 0, 0);
	56972	+ }
	56973	+
	56974	+ /* Finish committing the transaction to the destination database. */
	56975	+ if( SQLITE_OK==rc
	56976	+ && SQLITE_OK==(rc = sqlite3BtreeCommitPhaseTwo(p->pDest, 0))
	56977	+ ){
	56978	+ rc = SQLITE_DONE;
	56979	+ }
56150	56980	}
56151	56981	}
56152	56982
56153	56983	/* If bCloseTrans is true, then this function opened a read transaction
56154	56984	** on the source database. Close the read transaction here. There is
		@@ -56612,38 +57442,32 @@
56612	57442	** invoking an external callback, free it now. Calling this function
56613	57443	** does not free any Mem.zMalloc buffer.
56614	57444	*/
56615	57445	SQLITE_PRIVATE void sqlite3VdbeMemReleaseExternal(Mem *p){
56616	57446	assert( p->db==0 \|\| sqlite3_mutex_held(p->db->mutex) );
56617		- testcase( p->flags & MEM_Agg );
56618		- testcase( p->flags & MEM_Dyn );
56619		- testcase( p->flags & MEM_RowSet );
56620		- testcase( p->flags & MEM_Frame );
56621		- if( p->flags&(MEM_Agg\|MEM_Dyn\|MEM_RowSet\|MEM_Frame) ){
56622		- if( p->flags&MEM_Agg ){
56623		- sqlite3VdbeMemFinalize(p, p->u.pDef);
56624		- assert( (p->flags & MEM_Agg)==0 );
56625		- sqlite3VdbeMemRelease(p);
56626		- }else if( p->flags&MEM_Dyn && p->xDel ){
56627		- assert( (p->flags&MEM_RowSet)==0 );
56628		- p->xDel((void *)p->z);
56629		- p->xDel = 0;
56630		- }else if( p->flags&MEM_RowSet ){
56631		- sqlite3RowSetClear(p->u.pRowSet);
56632		- }else if( p->flags&MEM_Frame ){
56633		- sqlite3VdbeMemSetNull(p);
56634		- }
	57447	+ if( p->flags&MEM_Agg ){
	57448	+ sqlite3VdbeMemFinalize(p, p->u.pDef);
	57449	+ assert( (p->flags & MEM_Agg)==0 );
	57450	+ sqlite3VdbeMemRelease(p);
	57451	+ }else if( p->flags&MEM_Dyn && p->xDel ){
	57452	+ assert( (p->flags&MEM_RowSet)==0 );
	57453	+ p->xDel((void *)p->z);
	57454	+ p->xDel = 0;
	57455	+ }else if( p->flags&MEM_RowSet ){
	57456	+ sqlite3RowSetClear(p->u.pRowSet);
	57457	+ }else if( p->flags&MEM_Frame ){
	57458	+ sqlite3VdbeMemSetNull(p);
56635	57459	}
56636	57460	}
56637	57461
56638	57462	/*
56639	57463	** Release any memory held by the Mem. This may leave the Mem in an
56640	57464	** inconsistent state, for example with (Mem.z==0) and
56641	57465	** (Mem.type==SQLITE_TEXT).
56642	57466	*/
56643	57467	SQLITE_PRIVATE void sqlite3VdbeMemRelease(Mem *p){
56644		- sqlite3VdbeMemReleaseExternal(p);
	57468	+ MemReleaseExt(p);
56645	57469	sqlite3DbFree(p->db, p->zMalloc);
56646	57470	p->z = 0;
56647	57471	p->zMalloc = 0;
56648	57472	p->xDel = 0;
56649	57473	}
		@@ -56961,11 +57785,11 @@
56961	57785	** pFrom->z is used, then pTo->z points to the same thing as pFrom->z
56962	57786	** and flags gets srcType (either MEM_Ephem or MEM_Static).
56963	57787	*/
56964	57788	SQLITE_PRIVATE void sqlite3VdbeMemShallowCopy(Mem pTo, const Mem pFrom, int srcType){
56965	57789	assert( (pFrom->flags & MEM_RowSet)==0 );
56966		- sqlite3VdbeMemReleaseExternal(pTo);
	57790	+ MemReleaseExt(pTo);
56967	57791	memcpy(pTo, pFrom, MEMCELLSIZE);
56968	57792	pTo->xDel = 0;
56969	57793	if( (pFrom->flags&MEM_Static)==0 ){
56970	57794	pTo->flags &= ~(MEM_Dyn\|MEM_Static\|MEM_Ephem);
56971	57795	assert( srcType==MEM_Ephem \|\| srcType==MEM_Static );
		@@ -56979,11 +57803,11 @@
56979	57803	*/
56980	57804	SQLITE_PRIVATE int sqlite3VdbeMemCopy(Mem pTo, const Mem pFrom){
56981	57805	int rc = SQLITE_OK;
56982	57806
56983	57807	assert( (pFrom->flags & MEM_RowSet)==0 );
56984		- sqlite3VdbeMemReleaseExternal(pTo);
	57808	+ MemReleaseExt(pTo);
56985	57809	memcpy(pTo, pFrom, MEMCELLSIZE);
56986	57810	pTo->flags &= ~MEM_Dyn;
56987	57811
56988	57812	if( pTo->flags&(MEM_Str\|MEM_Blob) ){
56989	57813	if( 0==(pFrom->flags&MEM_Static) ){
		@@ -57934,10 +58758,16 @@
57934	58758	assert( p->nOp - i >= 3 );
57935	58759	assert( pOp[-1].opcode==OP_Integer );
57936	58760	n = pOp[-1].p1;
57937	58761	if( n>nMaxArgs ) nMaxArgs = n;
57938	58762	#endif
	58763	+ }else if( opcode==OP_Next ){
	58764	+ pOp->p4.xAdvance = sqlite3BtreeNext;
	58765	+ pOp->p4type = P4_ADVANCE;
	58766	+ }else if( opcode==OP_Prev ){
	58767	+ pOp->p4.xAdvance = sqlite3BtreePrevious;
	58768	+ pOp->p4type = P4_ADVANCE;
57939	58769	}
57940	58770
57941	58771	if( (pOp->opflags & OPFLG_JUMP)!=0 && pOp->p2<0 ){
57942	58772	assert( -1-pOp->p2<p->nLabel );
57943	58773	pOp->p2 = aLabel[-1-pOp->p2];
		@@ -58025,37 +58855,34 @@
58025	58855	** Change the value of the P1 operand for a specific instruction.
58026	58856	** This routine is useful when a large program is loaded from a
58027	58857	** static array using sqlite3VdbeAddOpList but we want to make a
58028	58858	** few minor changes to the program.
58029	58859	*/
58030		-SQLITE_PRIVATE void sqlite3VdbeChangeP1(Vdbe *p, int addr, int val){
	58860	+SQLITE_PRIVATE void sqlite3VdbeChangeP1(Vdbe *p, u32 addr, int val){
58031	58861	assert( p!=0 );
58032		- assert( addr>=0 );
58033		- if( p->nOp>addr ){
	58862	+ if( ((u32)p->nOp)>addr ){
58034	58863	p->aOp[addr].p1 = val;
58035	58864	}
58036	58865	}
58037	58866
58038	58867	/*
58039	58868	** Change the value of the P2 operand for a specific instruction.
58040	58869	** This routine is useful for setting a jump destination.
58041	58870	*/
58042		-SQLITE_PRIVATE void sqlite3VdbeChangeP2(Vdbe *p, int addr, int val){
	58871	+SQLITE_PRIVATE void sqlite3VdbeChangeP2(Vdbe *p, u32 addr, int val){
58043	58872	assert( p!=0 );
58044		- assert( addr>=0 );
58045		- if( p->nOp>addr ){
	58873	+ if( ((u32)p->nOp)>addr ){
58046	58874	p->aOp[addr].p2 = val;
58047	58875	}
58048	58876	}
58049	58877
58050	58878	/*
58051	58879	** Change the value of the P3 operand for a specific instruction.
58052	58880	*/
58053		-SQLITE_PRIVATE void sqlite3VdbeChangeP3(Vdbe *p, int addr, int val){
	58881	+SQLITE_PRIVATE void sqlite3VdbeChangeP3(Vdbe *p, u32 addr, int val){
58054	58882	assert( p!=0 );
58055		- assert( addr>=0 );
58056		- if( p->nOp>addr ){
	58883	+ if( ((u32)p->nOp)>addr ){
58057	58884	p->aOp[addr].p3 = val;
58058	58885	}
58059	58886	}
58060	58887
58061	58888	/*
		@@ -58442,10 +59269,14 @@
58442	59269	break;
58443	59270	}
58444	59271	case P4_SUBPROGRAM: {
58445	59272	sqlite3_snprintf(nTemp, zTemp, "program");
58446	59273	break;
	59274	+ }
	59275	+ case P4_ADVANCE: {
	59276	+ zTemp[0] = 0;
	59277	+ break;
58447	59278	}
58448	59279	default: {
58449	59280	zP4 = pOp->p4.z;
58450	59281	if( zP4==0 ){
58451	59282	zP4 = zTemp;
		@@ -59006,10 +59837,11 @@
59006	59837	nMem = 10;
59007	59838	}
59008	59839	memset(zCsr, 0, zEnd-zCsr);
59009	59840	zCsr += (zCsr - (u8*)0)&7;
59010	59841	assert( EIGHT_BYTE_ALIGNMENT(zCsr) );
	59842	+ p->expired = 0;
59011	59843
59012	59844	/* Memory for registers, parameters, cursor, etc, is allocated in two
59013	59845	** passes. On the first pass, we try to reuse unused space at the
59014	59846	** end of the opcode array. If we are unable to satisfy all memory
59015	59847	** requirements by reusing the opcode array tail, then the second
		@@ -59065,10 +59897,11 @@
59065	59897	*/
59066	59898	SQLITE_PRIVATE void sqlite3VdbeFreeCursor(Vdbe p, VdbeCursor pCx){
59067	59899	if( pCx==0 ){
59068	59900	return;
59069	59901	}
	59902	+ sqlite3VdbeSorterClose(p->db, pCx);
59070	59903	if( pCx->pBt ){
59071	59904	sqlite3BtreeClose(pCx->pBt);
59072	59905	/* The pCx->pCursor will be close automatically, if it exists, by
59073	59906	** the call above. */
59074	59907	}else if( pCx->pCursor ){
		@@ -61205,10 +62038,18 @@
61205	62038	rc = db->errCode = p->rc;
61206	62039	}
61207	62040	return (rc&db->errMask);
61208	62041	}
61209	62042
	62043	+/*
	62044	+** The maximum number of times that a statement will try to reparse
	62045	+** itself before giving up and returning SQLITE_SCHEMA.
	62046	+*/
	62047	+#ifndef SQLITE_MAX_SCHEMA_RETRY
	62048	+# define SQLITE_MAX_SCHEMA_RETRY 5
	62049	+#endif
	62050	+
61210	62051	/*
61211	62052	** This is the top-level implementation of sqlite3_step(). Call
61212	62053	** sqlite3Step() to do most of the work. If a schema error occurs,
61213	62054	** call sqlite3Reprepare() and try again.
61214	62055	*/
		@@ -61223,14 +62064,14 @@
61223	62064	return SQLITE_MISUSE_BKPT;
61224	62065	}
61225	62066	db = v->db;
61226	62067	sqlite3_mutex_enter(db->mutex);
61227	62068	while( (rc = sqlite3Step(v))==SQLITE_SCHEMA
61228		- && cnt++ < 5
	62069	+ && cnt++ < SQLITE_MAX_SCHEMA_RETRY
61229	62070	&& (rc2 = rc = sqlite3Reprepare(v))==SQLITE_OK ){
61230	62071	sqlite3_reset(pStmt);
61231		- v->expired = 0;
	62072	+ assert( v->expired==0 );
61232	62073	}
61233	62074	if( rc2!=SQLITE_OK && ALWAYS(v->isPrepareV2) && ALWAYS(db->pErr) ){
61234	62075	/* This case occurs after failing to recompile an sql statement.
61235	62076	** The error message from the SQL compiler has already been loaded
61236	62077	** into the database handle. This block copies the error message
		@@ -62357,10 +63198,17 @@
62357	63198	** Call sqlite3VdbeMemExpandBlob() on the supplied value (type Mem*)
62358	63199	** P if required.
62359	63200	*/
62360	63201	#define ExpandBlob(P) (((P)->flags&MEM_Zero)?sqlite3VdbeMemExpandBlob(P):0)
62361	63202
	63203	+/* Return true if the cursor was opened using the OP_OpenSorter opcode. */
	63204	+#ifdef SQLITE_OMIT_MERGE_SORT
	63205	+# define isSorter(x) 0
	63206	+#else
	63207	+# define isSorter(x) ((x)->pSorter!=0)
	63208	+#endif
	63209	+
62362	63210	/*
62363	63211	** Argument pMem points at a register that will be passed to a
62364	63212	** user-defined function or returned to the user as the result of a query.
62365	63213	** This routine sets the pMem->type variable used by the sqlite3_value_*()
62366	63214	** routines.
		@@ -62951,10 +63799,11 @@
62951	63799	u8 zEndHdr; / Pointer to first byte after the header */
62952	63800	u32 offset; /* Offset into the data */
62953	63801	u32 szField; /* Number of bytes in the content of a field */
62954	63802	int szHdr; /* Size of the header size field at start of record */
62955	63803	int avail; /* Number of bytes of available data */
	63804	+ u32 t; /* A type code from the record header */
62956	63805	Mem pReg; / PseudoTable input register */
62957	63806	} am;
62958	63807	struct OP_Affinity_stack_vars {
62959	63808	const char zAffinity; / The affinity to be applied */
62960	63809	char cAff; /* A single character of affinity */
		@@ -63109,11 +63958,10 @@
63109	63958	BtCursor *pCrsr;
63110	63959	int res;
63111	63960	} bl;
63112	63961	struct OP_Next_stack_vars {
63113	63962	VdbeCursor *pC;
63114		- BtCursor *pCrsr;
63115	63963	int res;
63116	63964	} bm;
63117	63965	struct OP_IdxInsert_stack_vars {
63118	63966	VdbeCursor *pC;
63119	63967	BtCursor *pCrsr;
		@@ -63363,11 +64211,11 @@
63363	64211	if( pOp->opflags & OPFLG_OUT2_PRERELEASE ){
63364	64212	assert( pOp->p2>0 );
63365	64213	assert( pOp->p2<=p->nMem );
63366	64214	pOut = &aMem[pOp->p2];
63367	64215	memAboutToChange(p, pOut);
63368		- sqlite3VdbeMemReleaseExternal(pOut);
	64216	+ MemReleaseExt(pOut);
63369	64217	pOut->flags = MEM_Int;
63370	64218	}
63371	64219
63372	64220	/* Sanity checking on other operands */
63373	64221	#ifdef SQLITE_DEBUG
		@@ -63496,11 +64344,11 @@
63496	64344	break;
63497	64345	}
63498	64346
63499	64347	/* Opcode: HaltIfNull P1 P2 P3 P4 *
63500	64348	**
63501		-** Check the value in register P3. If is is NULL then Halt using
	64349	+** Check the value in register P3. If it is NULL then Halt using
63502	64350	** parameter P1, P2, and P4 as if this were a Halt instruction. If the
63503	64351	** value in register P3 is not NULL, then this routine is a no-op.
63504	64352	*/
63505	64353	case OP_HaltIfNull: { /* in3 */
63506	64354	pIn3 = &aMem[pOp->p3];
		@@ -63730,10 +64578,15 @@
63730	64578	assert( memIsValid(pIn1) );
63731	64579	memAboutToChange(p, pOut);
63732	64580	u.ac.zMalloc = pOut->zMalloc;
63733	64581	pOut->zMalloc = 0;
63734	64582	sqlite3VdbeMemMove(pOut, pIn1);
	64583	+#ifdef SQLITE_DEBUG
	64584	+ if( pOut->pScopyFrom>=&aMem[u.ac.p1] && pOut->pScopyFrom<&aMem[u.ac.p1+pOp->p3] ){
	64585	+ pOut->pScopyFrom += u.ac.p1 - pOp->p2;
	64586	+ }
	64587	+#endif
63735	64588	pIn1->zMalloc = u.ac.zMalloc;
63736	64589	REGISTER_TRACE(u.ac.p2++, pOut);
63737	64590	pIn1++;
63738	64591	pOut++;
63739	64592	}
		@@ -64433,11 +65286,11 @@
64433	65286	** additional information.
64434	65287	**
64435	65288	** If SQLITE_NULLEQ is set in P5 then the result of comparison is always either
64436	65289	** true or false and is never NULL. If both operands are NULL then the result
64437	65290	** of comparison is false. If either operand is NULL then the result is true.
64438		-** If neither operand is NULL the the result is the same as it would be if
	65291	+** If neither operand is NULL the result is the same as it would be if
64439	65292	** the SQLITE_NULLEQ flag were omitted from P5.
64440	65293	*/
64441	65294	/* Opcode: Eq P1 P2 P3 P4 P5
64442	65295	**
64443	65296	** This works just like the Lt opcode except that the jump is taken if
		@@ -64445,11 +65298,11 @@
64445	65298	** See the Lt opcode for additional information.
64446	65299	**
64447	65300	** If SQLITE_NULLEQ is set in P5 then the result of comparison is always either
64448	65301	** true or false and is never NULL. If both operands are NULL then the result
64449	65302	** of comparison is true. If either operand is NULL then the result is false.
64450		-** If neither operand is NULL the the result is the same as it would be if
	65303	+** If neither operand is NULL the result is the same as it would be if
64451	65304	** the SQLITE_NULLEQ flag were omitted from P5.
64452	65305	*/
64453	65306	/* Opcode: Le P1 P2 P3 P4 P5
64454	65307	**
64455	65308	** This works just like the Lt opcode except that the jump is taken if
		@@ -64730,17 +65583,17 @@
64730	65583	break;
64731	65584	}
64732	65585
64733	65586	/* Opcode: If P1 P2 P3 * *
64734	65587	**
64735		-** Jump to P2 if the value in register P1 is true. The value is
	65588	+** Jump to P2 if the value in register P1 is true. The value
64736	65589	** is considered true if it is numeric and non-zero. If the value
64737	65590	** in P1 is NULL then take the jump if P3 is true.
64738	65591	*/
64739	65592	/* Opcode: IfNot P1 P2 P3 * *
64740	65593	**
64741		-** Jump to P2 if the value in register P1 is False. The value is
	65594	+** Jump to P2 if the value in register P1 is False. The value
64742	65595	** is considered true if it has a numeric value of zero. If the value
64743	65596	** in P1 is NULL then take the jump if P3 is true.
64744	65597	*/
64745	65598	case OP_If: /* jump, in1 */
64746	65599	case OP_IfNot: { /* jump, in1 */
		@@ -64828,10 +65681,11 @@
64828	65681	u8 zEndHdr; / Pointer to first byte after the header */
64829	65682	u32 offset; /* Offset into the data */
64830	65683	u32 szField; /* Number of bytes in the content of a field */
64831	65684	int szHdr; /* Size of the header size field at start of record */
64832	65685	int avail; /* Number of bytes of available data */
	65686	+ u32 t; /* A type code from the record header */
64833	65687	Mem pReg; / PseudoTable input register */
64834	65688	#endif /* local variables moved into u.am */
64835	65689
64836	65690
64837	65691	u.am.p1 = pOp->p1;
		@@ -64840,11 +65694,10 @@
64840	65694	memset(&u.am.sMem, 0, sizeof(u.am.sMem));
64841	65695	assert( u.am.p1<p->nCursor );
64842	65696	assert( pOp->p3>0 && pOp->p3<=p->nMem );
64843	65697	u.am.pDest = &aMem[pOp->p3];
64844	65698	memAboutToChange(p, u.am.pDest);
64845		- MemSetTypeFlag(u.am.pDest, MEM_Null);
64846	65699	u.am.zRec = 0;
64847	65700
64848	65701	/* This block sets the variable u.am.payloadSize to be the total number of
64849	65702	** bytes in the record.
64850	65703	**
		@@ -64884,11 +65737,11 @@
64884	65737	}else{
64885	65738	assert( sqlite3BtreeCursorIsValid(u.am.pCrsr) );
64886	65739	rc = sqlite3BtreeDataSize(u.am.pCrsr, &u.am.payloadSize);
64887	65740	assert( rc==SQLITE_OK ); /* DataSize() cannot fail */
64888	65741	}
64889		- }else if( u.am.pC->pseudoTableReg>0 ){
	65742	+ }else if( ALWAYS(u.am.pC->pseudoTableReg>0) ){
64890	65743	u.am.pReg = &aMem[u.am.pC->pseudoTableReg];
64891	65744	assert( u.am.pReg->flags & MEM_Blob );
64892	65745	assert( memIsValid(u.am.pReg) );
64893	65746	u.am.payloadSize = u.am.pReg->n;
64894	65747	u.am.zRec = u.am.pReg->z;
		@@ -64897,13 +65750,14 @@
64897	65750	}else{
64898	65751	/* Consider the row to be NULL */
64899	65752	u.am.payloadSize = 0;
64900	65753	}
64901	65754
64902		- /* If u.am.payloadSize is 0, then just store a NULL */
	65755	+ /* If u.am.payloadSize is 0, then just store a NULL. This can happen because of
	65756	+ ** nullRow or because of a corrupt database. */
64903	65757	if( u.am.payloadSize==0 ){
64904		- assert( u.am.pDest->flags&MEM_Null );
	65758	+ MemSetTypeFlag(u.am.pDest, MEM_Null);
64905	65759	goto op_column_out;
64906	65760	}
64907	65761	assert( db->aLimit[SQLITE_LIMIT_LENGTH]>=0 );
64908	65762	if( u.am.payloadSize > (u32)db->aLimit[SQLITE_LIMIT_LENGTH] ){
64909	65763	goto too_big;
		@@ -65006,12 +65860,18 @@
65006	65860	** of the record to the start of the data for the u.am.i-th column
65007	65861	*/
65008	65862	for(u.am.i=0; u.am.i<u.am.nField; u.am.i++){
65009	65863	if( u.am.zIdx<u.am.zEndHdr ){
65010	65864	u.am.aOffset[u.am.i] = u.am.offset;
65011		- u.am.zIdx += getVarint32(u.am.zIdx, u.am.aType[u.am.i]);
65012		- u.am.szField = sqlite3VdbeSerialTypeLen(u.am.aType[u.am.i]);
	65865	+ if( u.am.zIdx[0]<0x80 ){
	65866	+ u.am.t = u.am.zIdx[0];
	65867	+ u.am.zIdx++;
	65868	+ }else{
	65869	+ u.am.zIdx += sqlite3GetVarint32(u.am.zIdx, &u.am.t);
	65870	+ }
	65871	+ u.am.aType[u.am.i] = u.am.t;
	65872	+ u.am.szField = sqlite3VdbeSerialTypeLen(u.am.t);
65013	65873	u.am.offset += u.am.szField;
65014	65874	if( u.am.offset<u.am.szField ){ /* True if u.am.offset overflows */
65015	65875	u.am.zIdx = &u.am.zEndHdr[1]; /* Forces SQLITE_CORRUPT return below */
65016	65876	break;
65017	65877	}
		@@ -65048,11 +65908,11 @@
65048	65908	** a pointer to a Mem object.
65049	65909	*/
65050	65910	if( u.am.aOffset[u.am.p2] ){
65051	65911	assert( rc==SQLITE_OK );
65052	65912	if( u.am.zRec ){
65053		- sqlite3VdbeMemReleaseExternal(u.am.pDest);
	65913	+ MemReleaseExt(u.am.pDest);
65054	65914	sqlite3VdbeSerialGet((u8 *)&u.am.zRec[u.am.aOffset[u.am.p2]], u.am.aType[u.am.p2], u.am.pDest);
65055	65915	}else{
65056	65916	u.am.len = sqlite3VdbeSerialTypeLen(u.am.aType[u.am.p2]);
65057	65917	sqlite3VdbeMemMove(&u.am.sMem, u.am.pDest);
65058	65918	rc = sqlite3VdbeMemFromBtree(u.am.pCrsr, u.am.aOffset[u.am.p2], u.am.len, u.am.pC->isIndex, &u.am.sMem);
		@@ -65065,11 +65925,11 @@
65065	65925	u.am.pDest->enc = encoding;
65066	65926	}else{
65067	65927	if( pOp->p4type==P4_MEM ){
65068	65928	sqlite3VdbeMemShallowCopy(u.am.pDest, pOp->p4.pMem, MEM_Static);
65069	65929	}else{
65070		- assert( u.am.pDest->flags&MEM_Null );
	65930	+ MemSetTypeFlag(u.am.pDest, MEM_Null);
65071	65931	}
65072	65932	}
65073	65933
65074	65934	/* If we dynamically allocated space to hold the data (in the
65075	65935	** sqlite3VdbeMemFromBtree() call above) then transfer control of that
		@@ -65267,11 +66127,11 @@
65267	66127	i64 nEntry;
65268	66128	BtCursor *pCrsr;
65269	66129	#endif /* local variables moved into u.ap */
65270	66130
65271	66131	u.ap.pCrsr = p->apCsr[pOp->p1]->pCursor;
65272		- if( u.ap.pCrsr ){
	66132	+ if( ALWAYS(u.ap.pCrsr) ){
65273	66133	rc = sqlite3BtreeCount(u.ap.pCrsr, &u.ap.nEntry);
65274	66134	}else{
65275	66135	u.ap.nEntry = 0;
65276	66136	}
65277	66137	pOut->u.i = u.ap.nEntry;
		@@ -65843,19 +66703,13 @@
65843	66703	u.aw.pCur->nullRow = 1;
65844	66704	u.aw.pCur->isOrdered = 1;
65845	66705	rc = sqlite3BtreeCursor(u.aw.pX, u.aw.p2, u.aw.wrFlag, u.aw.pKeyInfo, u.aw.pCur->pCursor);
65846	66706	u.aw.pCur->pKeyInfo = u.aw.pKeyInfo;
65847	66707
65848		- /* Since it performs no memory allocation or IO, the only values that
65849		- ** sqlite3BtreeCursor() may return are SQLITE_EMPTY and SQLITE_OK.
65850		- ** SQLITE_EMPTY is only returned when attempting to open the table
65851		- ** rooted at page 1 of a zero-byte database. */
65852		- assert( rc==SQLITE_EMPTY \|\| rc==SQLITE_OK );
65853		- if( rc==SQLITE_EMPTY ){
65854		- u.aw.pCur->pCursor = 0;
65855		- rc = SQLITE_OK;
65856		- }
	66708	+ /* Since it performs no memory allocation or IO, the only value that
	66709	+ ** sqlite3BtreeCursor() may return is SQLITE_OK. */
	66710	+ assert( rc==SQLITE_OK );
65857	66711
65858	66712	/* Set the VdbeCursor.isTable and isIndex variables. Previous versions of
65859	66713	** SQLite used to check if the root-page flags were sane at this point
65860	66714	** and report database corruption if they were not, but this check has
65861	66715	** since moved into the btree layer. */
		@@ -65862,11 +66716,11 @@
65862	66716	u.aw.pCur->isTable = pOp->p4type!=P4_KEYINFO;
65863	66717	u.aw.pCur->isIndex = !u.aw.pCur->isTable;
65864	66718	break;
65865	66719	}
65866	66720
65867		-/* Opcode: OpenEphemeral P1 P2 * P4 *
	66721	+/* Opcode: OpenEphemeral P1 P2 * P4 P5
65868	66722	**
65869	66723	** Open a new cursor P1 to a transient table.
65870	66724	** The cursor is always opened read/write even if
65871	66725	** the main database is read-only. The ephemeral
65872	66726	** table is deleted automatically when the cursor is closed.
		@@ -65879,18 +66733,30 @@
65879	66733	** This opcode was once called OpenTemp. But that created
65880	66734	** confusion because the term "temp table", might refer either
65881	66735	** to a TEMP table at the SQL level, or to a table opened by
65882	66736	** this opcode. Then this opcode was call OpenVirtual. But
65883	66737	** that created confusion with the whole virtual-table idea.
	66738	+**
	66739	+** The P5 parameter can be a mask of the BTREE_* flags defined
	66740	+** in btree.h. These flags control aspects of the operation of
	66741	+** the btree. The BTREE_OMIT_JOURNAL and BTREE_SINGLE flags are
	66742	+** added automatically.
65884	66743	*/
65885	66744	/* Opcode: OpenAutoindex P1 P2 * P4 *
65886	66745	**
65887	66746	** This opcode works the same as OP_OpenEphemeral. It has a
65888	66747	** different name to distinguish its use. Tables created using
65889	66748	** by this opcode will be used for automatically created transient
65890	66749	** indices in joins.
65891	66750	*/
	66751	+/* Opcode: OpenSorter P1 P2 * P4 *
	66752	+**
	66753	+** This opcode works like OP_OpenEphemeral except that it opens
	66754	+** a transient index that is specifically designed to sort large
	66755	+** tables using an external merge-sort algorithm.
	66756	+*/
	66757	+case OP_OpenSorter:
65892	66758	case OP_OpenAutoindex:
65893	66759	case OP_OpenEphemeral: {
65894	66760	#if 0 /* local variables moved into u.ax */
65895	66761	VdbeCursor *pCx;
65896	66762	#endif /* local variables moved into u.ax */
		@@ -65900,10 +66766,11 @@
65900	66766	SQLITE_OPEN_EXCLUSIVE \|
65901	66767	SQLITE_OPEN_DELETEONCLOSE \|
65902	66768	SQLITE_OPEN_TRANSIENT_DB;
65903	66769
65904	66770	assert( pOp->p1>=0 );
	66771	+ assert( (pOp->opcode==OP_OpenSorter)==((pOp->p5 & BTREE_SORTER)!=0) );
65905	66772	u.ax.pCx = allocateCursor(p, pOp->p1, pOp->p2, -1, 1);
65906	66773	if( u.ax.pCx==0 ) goto no_mem;
65907	66774	u.ax.pCx->nullRow = 1;
65908	66775	rc = sqlite3BtreeOpen(db->pVfs, 0, db, &u.ax.pCx->pBt,
65909	66776	BTREE_OMIT_JOURNAL \| BTREE_SINGLE \| pOp->p5, vfsFlags);
		@@ -65917,11 +66784,11 @@
65917	66784	** automatically created table with root-page 1 (an BLOB_INTKEY table).
65918	66785	*/
65919	66786	if( pOp->p4.pKeyInfo ){
65920	66787	int pgno;
65921	66788	assert( pOp->p4type==P4_KEYINFO );
65922		- rc = sqlite3BtreeCreateTable(u.ax.pCx->pBt, &pgno, BTREE_BLOBKEY);
	66789	+ rc = sqlite3BtreeCreateTable(u.ax.pCx->pBt, &pgno, BTREE_BLOBKEY \| pOp->p5);
65923	66790	if( rc==SQLITE_OK ){
65924	66791	assert( pgno==MASTER_ROOT+1 );
65925	66792	rc = sqlite3BtreeCursor(u.ax.pCx->pBt, pgno, 1,
65926	66793	(KeyInfo*)pOp->p4.z, u.ax.pCx->pCursor);
65927	66794	u.ax.pCx->pKeyInfo = pOp->p4.pKeyInfo;
		@@ -65933,10 +66800,15 @@
65933	66800	u.ax.pCx->isTable = 1;
65934	66801	}
65935	66802	}
65936	66803	u.ax.pCx->isOrdered = (pOp->p5!=BTREE_UNORDERED);
65937	66804	u.ax.pCx->isIndex = !u.ax.pCx->isTable;
	66805	+#ifndef SQLITE_OMIT_MERGE_SORT
	66806	+ if( rc==SQLITE_OK && pOp->opcode==OP_OpenSorter ){
	66807	+ rc = sqlite3VdbeSorterInit(db, u.ax.pCx);
	66808	+ }
	66809	+#endif
65938	66810	break;
65939	66811	}
65940	66812
65941	66813	/* Opcode: OpenPseudo P1 P2 P3 * *
65942	66814	**
		@@ -66052,11 +66924,11 @@
66052	66924	assert( u.az.pC->pseudoTableReg==0 );
66053	66925	assert( OP_SeekLe == OP_SeekLt+1 );
66054	66926	assert( OP_SeekGe == OP_SeekLt+2 );
66055	66927	assert( OP_SeekGt == OP_SeekLt+3 );
66056	66928	assert( u.az.pC->isOrdered );
66057		- if( u.az.pC->pCursor!=0 ){
	66929	+ if( ALWAYS(u.az.pC->pCursor!=0) ){
66058	66930	u.az.oc = pOp->opcode;
66059	66931	u.az.pC->nullRow = 0;
66060	66932	if( u.az.pC->isTable ){
66061	66933	/* The input value in P3 might be of any type: integer, real, string,
66062	66934	** blob, or NULL. But it needs to be an integer before we can do
		@@ -66390,11 +67262,11 @@
66390	67262	break;
66391	67263	}
66392	67264
66393	67265	/* Opcode: NotExists P1 P2 P3 * *
66394	67266	**
66395		-** Use the content of register P3 as a integer key. If a record
	67267	+** Use the content of register P3 as an integer key. If a record
66396	67268	** with that key does not exist in table of P1, then jump to P2.
66397	67269	** If the record does exist, then fall through. The cursor is left
66398	67270	** pointing to the record if it exists.
66399	67271	**
66400	67272	** The difference between this operation and NotFound is that this
		@@ -66418,11 +67290,11 @@
66418	67290	u.bd.pC = p->apCsr[pOp->p1];
66419	67291	assert( u.bd.pC!=0 );
66420	67292	assert( u.bd.pC->isTable );
66421	67293	assert( u.bd.pC->pseudoTableReg==0 );
66422	67294	u.bd.pCrsr = u.bd.pC->pCursor;
66423		- if( u.bd.pCrsr!=0 ){
	67295	+ if( ALWAYS(u.bd.pCrsr!=0) ){
66424	67296	u.bd.res = 0;
66425	67297	u.bd.iKey = pIn3->u.i;
66426	67298	rc = sqlite3BtreeMovetoUnpacked(u.bd.pCrsr, 0, u.bd.iKey, 0, &u.bd.res);
66427	67299	u.bd.pC->lastRowid = pIn3->u.i;
66428	67300	u.bd.pC->rowidIsValid = u.bd.res==0 ?1:0;
		@@ -66468,11 +67340,11 @@
66468	67340	** written to register P2.
66469	67341	**
66470	67342	** If P3>0 then P3 is a register in the root frame of this VDBE that holds
66471	67343	** the largest previously generated record number. No new record numbers are
66472	67344	** allowed to be less than this value. When this value reaches its maximum,
66473		-** a SQLITE_FULL error is generated. The P3 register is updated with the '
	67345	+** an SQLITE_FULL error is generated. The P3 register is updated with the '
66474	67346	** generated record number. This P3 mechanism is used to help implement the
66475	67347	** AUTOINCREMENT feature.
66476	67348	*/
66477	67349	case OP_NewRowid: { /* out2-prerelease */
66478	67350	#if 0 /* local variables moved into u.be */
		@@ -66842,10 +67714,17 @@
66842	67714	assert( u.bh.pC->isTable \|\| pOp->opcode==OP_RowKey );
66843	67715	assert( u.bh.pC->isIndex \|\| pOp->opcode==OP_RowData );
66844	67716	assert( u.bh.pC!=0 );
66845	67717	assert( u.bh.pC->nullRow==0 );
66846	67718	assert( u.bh.pC->pseudoTableReg==0 );
	67719	+
	67720	+ if( isSorter(u.bh.pC) ){
	67721	+ assert( pOp->opcode==OP_RowKey );
	67722	+ rc = sqlite3VdbeSorterRowkey(u.bh.pC, pOut);
	67723	+ break;
	67724	+ }
	67725	+
66847	67726	assert( u.bh.pC->pCursor!=0 );
66848	67727	u.bh.pCrsr = u.bh.pC->pCursor;
66849	67728	assert( sqlite3BtreeCursorIsValid(u.bh.pCrsr) );
66850	67729
66851	67730	/* The OP_RowKey and OP_RowData opcodes always follow OP_NotExists or
		@@ -66950,10 +67829,11 @@
66950	67829	assert( pOp->p1>=0 && pOp->p1<p->nCursor );
66951	67830	u.bj.pC = p->apCsr[pOp->p1];
66952	67831	assert( u.bj.pC!=0 );
66953	67832	u.bj.pC->nullRow = 1;
66954	67833	u.bj.pC->rowidIsValid = 0;
	67834	+ assert( u.bj.pC->pCursor \|\| u.bj.pC->pVtabCursor );
66955	67835	if( u.bj.pC->pCursor ){
66956	67836	sqlite3BtreeClearCursor(u.bj.pC->pCursor);
66957	67837	}
66958	67838	break;
66959	67839	}
		@@ -66975,11 +67855,11 @@
66975	67855
66976	67856	assert( pOp->p1>=0 && pOp->p1<p->nCursor );
66977	67857	u.bk.pC = p->apCsr[pOp->p1];
66978	67858	assert( u.bk.pC!=0 );
66979	67859	u.bk.pCrsr = u.bk.pC->pCursor;
66980		- if( u.bk.pCrsr==0 ){
	67860	+ if( NEVER(u.bk.pCrsr==0) ){
66981	67861	u.bk.res = 1;
66982	67862	}else{
66983	67863	rc = sqlite3BtreeLast(u.bk.pCrsr, &u.bk.res);
66984	67864	}
66985	67865	u.bk.pC->nullRow = (u8)u.bk.res;
		@@ -67030,11 +67910,15 @@
67030	67910
67031	67911	assert( pOp->p1>=0 && pOp->p1<p->nCursor );
67032	67912	u.bl.pC = p->apCsr[pOp->p1];
67033	67913	assert( u.bl.pC!=0 );
67034	67914	u.bl.res = 1;
67035		- if( (u.bl.pCrsr = u.bl.pC->pCursor)!=0 ){
	67915	+ if( isSorter(u.bl.pC) ){
	67916	+ rc = sqlite3VdbeSorterRewind(db, u.bl.pC, &u.bl.res);
	67917	+ }else{
	67918	+ u.bl.pCrsr = u.bl.pC->pCursor;
	67919	+ assert( u.bl.pCrsr );
67036	67920	rc = sqlite3BtreeFirst(u.bl.pCrsr, &u.bl.res);
67037	67921	u.bl.pC->atFirst = u.bl.res==0 ?1:0;
67038	67922	u.bl.pC->deferredMoveto = 0;
67039	67923	u.bl.pC->cacheStatus = CACHE_STALE;
67040	67924	u.bl.pC->rowidIsValid = 0;
		@@ -67045,18 +67929,21 @@
67045	67929	pc = pOp->p2 - 1;
67046	67930	}
67047	67931	break;
67048	67932	}
67049	67933
67050		-/* Opcode: Next P1 P2 * * P5
	67934	+/* Opcode: Next P1 P2 * P4 P5
67051	67935	**
67052	67936	** Advance cursor P1 so that it points to the next key/data pair in its
67053	67937	** table or index. If there are no more key/value pairs then fall through
67054	67938	** to the following instruction. But if the cursor advance was successful,
67055	67939	** jump immediately to P2.
67056	67940	**
67057	67941	** The P1 cursor must be for a real table, not a pseudo-table.
	67942	+**
	67943	+** P4 is always of type P4_ADVANCE. The function pointer points to
	67944	+** sqlite3BtreeNext().
67058	67945	**
67059	67946	** If P5 is positive and the jump is taken, then event counter
67060	67947	** number P5-1 in the prepared statement is incremented.
67061	67948	**
67062	67949	** See also: Prev
		@@ -67067,19 +67954,21 @@
67067	67954	** table or index. If there is no previous key/value pairs then fall through
67068	67955	** to the following instruction. But if the cursor backup was successful,
67069	67956	** jump immediately to P2.
67070	67957	**
67071	67958	** The P1 cursor must be for a real table, not a pseudo-table.
	67959	+**
	67960	+** P4 is always of type P4_ADVANCE. The function pointer points to
	67961	+** sqlite3BtreePrevious().
67072	67962	**
67073	67963	** If P5 is positive and the jump is taken, then event counter
67074	67964	** number P5-1 in the prepared statement is incremented.
67075	67965	*/
67076	67966	case OP_Prev: /* jump */
67077	67967	case OP_Next: { /* jump */
67078	67968	#if 0 /* local variables moved into u.bm */
67079	67969	VdbeCursor *pC;
67080		- BtCursor *pCrsr;
67081	67970	int res;
67082	67971	#endif /* local variables moved into u.bm */
67083	67972
67084	67973	CHECK_FOR_INTERRUPT;
67085	67974	assert( pOp->p1>=0 && pOp->p1<p->nCursor );
		@@ -67086,19 +67975,21 @@
67086	67975	assert( pOp->p5<=ArraySize(p->aCounter) );
67087	67976	u.bm.pC = p->apCsr[pOp->p1];
67088	67977	if( u.bm.pC==0 ){
67089	67978	break; /* See ticket #2273 */
67090	67979	}
67091		- u.bm.pCrsr = u.bm.pC->pCursor;
67092		- if( u.bm.pCrsr==0 ){
67093		- u.bm.pC->nullRow = 1;
67094		- break;
67095		- }
67096		- u.bm.res = 1;
67097		- assert( u.bm.pC->deferredMoveto==0 );
67098		- rc = pOp->opcode==OP_Next ? sqlite3BtreeNext(u.bm.pCrsr, &u.bm.res) :
67099		- sqlite3BtreePrevious(u.bm.pCrsr, &u.bm.res);
	67980	+ if( isSorter(u.bm.pC) ){
	67981	+ assert( pOp->opcode==OP_Next );
	67982	+ rc = sqlite3VdbeSorterNext(db, u.bm.pC, &u.bm.res);
	67983	+ }else{
	67984	+ u.bm.res = 1;
	67985	+ assert( u.bm.pC->deferredMoveto==0 );
	67986	+ assert( u.bm.pC->pCursor );
	67987	+ assert( pOp->opcode!=OP_Next \|\| pOp->p4.xAdvance==sqlite3BtreeNext );
	67988	+ assert( pOp->opcode!=OP_Prev \|\| pOp->p4.xAdvance==sqlite3BtreePrevious );
	67989	+ rc = pOp->p4.xAdvance(u.bm.pC->pCursor, &u.bm.res);
	67990	+ }
67100	67991	u.bm.pC->nullRow = (u8)u.bm.res;
67101	67992	u.bm.pC->cacheStatus = CACHE_STALE;
67102	67993	if( u.bm.res==0 ){
67103	67994	pc = pOp->p2 - 1;
67104	67995	if( pOp->p5 ) p->aCounter[pOp->p5-1]++;
		@@ -67110,11 +68001,11 @@
67110	68001	break;
67111	68002	}
67112	68003
67113	68004	/* Opcode: IdxInsert P1 P2 P3 * P5
67114	68005	**
67115		-** Register P2 holds a SQL index key made using the
	68006	+** Register P2 holds an SQL index key made using the
67116	68007	** MakeRecord instructions. This opcode writes that key
67117	68008	** into the index P1. Data for the entry is nil.
67118	68009	**
67119	68010	** P3 is a flag that provides a hint to the b-tree layer that this
67120	68011	** insert is likely to be an append.
		@@ -67140,14 +68031,17 @@
67140	68031	assert( u.bn.pC->isTable==0 );
67141	68032	rc = ExpandBlob(pIn2);
67142	68033	if( rc==SQLITE_OK ){
67143	68034	u.bn.nKey = pIn2->n;
67144	68035	u.bn.zKey = pIn2->z;
67145		- rc = sqlite3BtreeInsert(u.bn.pCrsr, u.bn.zKey, u.bn.nKey, "", 0, 0, pOp->p3,
67146		- ((pOp->p5 & OPFLAG_USESEEKRESULT) ? u.bn.pC->seekResult : 0)
67147		- );
67148		- assert( u.bn.pC->deferredMoveto==0 );
	68036	+ rc = sqlite3VdbeSorterWrite(db, u.bn.pC, u.bn.nKey);
	68037	+ if( rc==SQLITE_OK ){
	68038	+ rc = sqlite3BtreeInsert(u.bn.pCrsr, u.bn.zKey, u.bn.nKey, "", 0, 0, pOp->p3,
	68039	+ ((pOp->p5 & OPFLAG_USESEEKRESULT) ? u.bn.pC->seekResult : 0)
	68040	+ );
	68041	+ assert( u.bn.pC->deferredMoveto==0 );
	68042	+ }
67149	68043	u.bn.pC->cacheStatus = CACHE_STALE;
67150	68044	}
67151	68045	}
67152	68046	break;
67153	68047	}
		@@ -69323,10 +70217,722 @@
69323	70217	}
69324	70218
69325	70219	#endif /* #ifndef SQLITE_OMIT_INCRBLOB */
69326	70220
69327	70221	/************ End of vdbeblob.c ******************************************/
	70222	+/************ Begin file vdbesort.c **************************************/
	70223	+/*
	70224	+** 2011 July 9
	70225	+**
	70226	+** The author disclaims copyright to this source code. In place of
	70227	+** a legal notice, here is a blessing:
	70228	+**
	70229	+** May you do good and not evil.
	70230	+** May you find forgiveness for yourself and forgive others.
	70231	+** May you share freely, never taking more than you give.
	70232	+**
	70233	+*************************************************************************
	70234	+** This file contains code for the VdbeSorter object, used in concert with
	70235	+** a VdbeCursor to sort large numbers of keys (as may be required, for
	70236	+** example, by CREATE INDEX statements on tables too large to fit in main
	70237	+** memory).
	70238	+*/
	70239	+
	70240	+
	70241	+#ifndef SQLITE_OMIT_MERGE_SORT
	70242	+
	70243	+typedef struct VdbeSorterIter VdbeSorterIter;
	70244	+
	70245	+/*
	70246	+** NOTES ON DATA STRUCTURE USED FOR N-WAY MERGES:
	70247	+**
	70248	+** As keys are added to the sorter, they are written to disk in a series
	70249	+** of sorted packed-memory-arrays (PMAs). The size of each PMA is roughly
	70250	+** the same as the cache-size allowed for temporary databases. In order
	70251	+** to allow the caller to extract keys from the sorter in sorted order,
	70252	+** all PMAs currently stored on disk must be merged together. This comment
	70253	+** describes the data structure used to do so. The structure supports
	70254	+** merging any number of arrays in a single pass with no redundant comparison
	70255	+** operations.
	70256	+**
	70257	+** The aIter[] array contains an iterator for each of the PMAs being merged.
	70258	+** An aIter[] iterator either points to a valid key or else is at EOF. For
	70259	+** the purposes of the paragraphs below, we assume that the array is actually
	70260	+** N elements in size, where N is the smallest power of 2 greater to or equal
	70261	+** to the number of iterators being merged. The extra aIter[] elements are
	70262	+** treated as if they are empty (always at EOF).
	70263	+**
	70264	+** The aTree[] array is also N elements in size. The value of N is stored in
	70265	+** the VdbeSorter.nTree variable.
	70266	+**
	70267	+** The final (N/2) elements of aTree[] contain the results of comparing
	70268	+** pairs of iterator keys together. Element i contains the result of
	70269	+** comparing aIter[2i-N] and aIter[2i-N+1]. Whichever key is smaller, the
	70270	+** aTree element is set to the index of it.
	70271	+**
	70272	+** For the purposes of this comparison, EOF is considered greater than any
	70273	+** other key value. If the keys are equal (only possible with two EOF
	70274	+** values), it doesn't matter which index is stored.
	70275	+**
	70276	+** The (N/4) elements of aTree[] that preceed the final (N/2) described
	70277	+** above contains the index of the smallest of each block of 4 iterators.
	70278	+** And so on. So that aTree[1] contains the index of the iterator that
	70279	+** currently points to the smallest key value. aTree[0] is unused.
	70280	+**
	70281	+** Example:
	70282	+**
	70283	+** aIter[0] -> Banana
	70284	+** aIter[1] -> Feijoa
	70285	+** aIter[2] -> Elderberry
	70286	+** aIter[3] -> Currant
	70287	+** aIter[4] -> Grapefruit
	70288	+** aIter[5] -> Apple
	70289	+** aIter[6] -> Durian
	70290	+** aIter[7] -> EOF
	70291	+**
	70292	+** aTree[] = { X, 5 0, 5 0, 3, 5, 6 }
	70293	+**
	70294	+** The current element is "Apple" (the value of the key indicated by
	70295	+** iterator 5). When the Next() operation is invoked, iterator 5 will
	70296	+** be advanced to the next key in its segment. Say the next key is
	70297	+** "Eggplant":
	70298	+**
	70299	+** aIter[5] -> Eggplant
	70300	+**
	70301	+** The contents of aTree[] are updated first by comparing the new iterator
	70302	+** 5 key to the current key of iterator 4 (still "Grapefruit"). The iterator
	70303	+** 5 value is still smaller, so aTree[6] is set to 5. And so on up the tree.
	70304	+** The value of iterator 6 - "Durian" - is now smaller than that of iterator
	70305	+** 5, so aTree[3] is set to 6. Key 0 is smaller than key 6 (Banana<Durian),
	70306	+** so the value written into element 1 of the array is 0. As follows:
	70307	+**
	70308	+** aTree[] = { X, 0 0, 6 0, 3, 5, 6 }
	70309	+**
	70310	+** In other words, each time we advance to the next sorter element, log2(N)
	70311	+** key comparison operations are required, where N is the number of segments
	70312	+** being merged (rounded up to the next power of 2).
	70313	+*/
	70314	+struct VdbeSorter {
	70315	+ int nWorking; /* Start a new b-tree after this many pages */
	70316	+ int nBtree; /* Current size of b-tree contents as PMA */
	70317	+ int nTree; /* Used size of aTree/aIter (power of 2) */
	70318	+ VdbeSorterIter aIter; / Array of iterators to merge */
	70319	+ int aTree; / Current state of incremental merge */
	70320	+ i64 iWriteOff; /* Current write offset within file pTemp1 */
	70321	+ i64 iReadOff; /* Current read offset within file pTemp1 */
	70322	+ sqlite3_file pTemp1; / PMA file 1 */
	70323	+ int nPMA; /* Number of PMAs stored in pTemp1 */
	70324	+};
	70325	+
	70326	+/*
	70327	+** The following type is an iterator for a PMA. It caches the current key in
	70328	+** variables nKey/aKey. If the iterator is at EOF, pFile==0.
	70329	+*/
	70330	+struct VdbeSorterIter {
	70331	+ i64 iReadOff; /* Current read offset */
	70332	+ i64 iEof; /* 1 byte past EOF for this iterator */
	70333	+ sqlite3_file pFile; / File iterator is reading from */
	70334	+ int nAlloc; /* Bytes of space at aAlloc */
	70335	+ u8 aAlloc; / Allocated space */
	70336	+ int nKey; /* Number of bytes in key */
	70337	+ u8 aKey; / Pointer to current key */
	70338	+};
	70339	+
	70340	+/* Minimum allowable value for the VdbeSorter.nWorking variable */
	70341	+#define SORTER_MIN_WORKING 10
	70342	+
	70343	+/* Maximum number of segments to merge in a single pass. */
	70344	+#define SORTER_MAX_MERGE_COUNT 16
	70345	+
	70346	+/*
	70347	+** Free all memory belonging to the VdbeSorterIter object passed as the second
	70348	+** argument. All structure fields are set to zero before returning.
	70349	+*/
	70350	+static void vdbeSorterIterZero(sqlite3 db, VdbeSorterIter pIter){
	70351	+ sqlite3DbFree(db, pIter->aAlloc);
	70352	+ memset(pIter, 0, sizeof(VdbeSorterIter));
	70353	+}
	70354	+
	70355	+/*
	70356	+** Advance iterator pIter to the next key in its PMA. Return SQLITE_OK if
	70357	+** no error occurs, or an SQLite error code if one does.
	70358	+*/
	70359	+static int vdbeSorterIterNext(
	70360	+ sqlite3 db, / Database handle (for sqlite3DbMalloc() ) */
	70361	+ VdbeSorterIter pIter / Iterator to advance */
	70362	+){
	70363	+ int rc; /* Return Code */
	70364	+ int nRead; /* Number of bytes read */
	70365	+ int nRec; /* Size of record in bytes */
	70366	+ int iOff; /* Size of serialized size varint in bytes */
	70367	+
	70368	+ nRead = pIter->iEof - pIter->iReadOff;
	70369	+ if( nRead>5 ) nRead = 5;
	70370	+ if( nRead<=0 ){
	70371	+ /* This is an EOF condition */
	70372	+ vdbeSorterIterZero(db, pIter);
	70373	+ return SQLITE_OK;
	70374	+ }
	70375	+
	70376	+ rc = sqlite3OsRead(pIter->pFile, pIter->aAlloc, nRead, pIter->iReadOff);
	70377	+ iOff = getVarint32(pIter->aAlloc, nRec);
	70378	+
	70379	+ if( rc==SQLITE_OK && (iOff+nRec)>nRead ){
	70380	+ int nRead2; /* Number of extra bytes to read */
	70381	+ if( (iOff+nRec)>pIter->nAlloc ){
	70382	+ int nNew = pIter->nAlloc*2;
	70383	+ while( (iOff+nRec)>nNew ) nNew = nNew*2;
	70384	+ pIter->aAlloc = sqlite3DbReallocOrFree(db, pIter->aAlloc, nNew);
	70385	+ if( !pIter->aAlloc ) return SQLITE_NOMEM;
	70386	+ pIter->nAlloc = nNew;
	70387	+ }
	70388	+
	70389	+ nRead2 = iOff + nRec - nRead;
	70390	+ rc = sqlite3OsRead(
	70391	+ pIter->pFile, &pIter->aAlloc[nRead], nRead2, pIter->iReadOff+nRead
	70392	+ );
	70393	+ }
	70394	+
	70395	+ assert( nRec>0 \|\| rc!=SQLITE_OK );
	70396	+ pIter->iReadOff += iOff+nRec;
	70397	+ pIter->nKey = nRec;
	70398	+ pIter->aKey = &pIter->aAlloc[iOff];
	70399	+ return rc;
	70400	+}
	70401	+
	70402	+/*
	70403	+** Write a single varint, value iVal, to file-descriptor pFile. Return
	70404	+** SQLITE_OK if successful, or an SQLite error code if some error occurs.
	70405	+**
	70406	+** The value of *piOffset when this function is called is used as the byte
	70407	+** offset in file pFile to write to. Before returning, *piOffset is
	70408	+** incremented by the number of bytes written.
	70409	+*/
	70410	+static int vdbeSorterWriteVarint(
	70411	+ sqlite3_file pFile, / File to write to */
	70412	+ i64 iVal, /* Value to write as a varint */
	70413	+ i64 piOffset / IN/OUT: Write offset in file pFile */
	70414	+){
	70415	+ u8 aVarint[9]; /* Buffer large enough for a varint */
	70416	+ int nVarint; /* Number of used bytes in varint */
	70417	+ int rc; /* Result of write() call */
	70418	+
	70419	+ nVarint = sqlite3PutVarint(aVarint, iVal);
	70420	+ rc = sqlite3OsWrite(pFile, aVarint, nVarint, *piOffset);
	70421	+ *piOffset += nVarint;
	70422	+
	70423	+ return rc;
	70424	+}
	70425	+
	70426	+/*
	70427	+** Read a single varint from file-descriptor pFile. Return SQLITE_OK if
	70428	+** successful, or an SQLite error code if some error occurs.
	70429	+**
	70430	+** The value of *piOffset when this function is called is used as the
	70431	+** byte offset in file pFile from whence to read the varint. If successful
	70432	+** (i.e. if no IO error occurs), then *piOffset is set to the offset of
	70433	+** the first byte past the end of the varint before returning. *piVal is
	70434	+** set to the integer value read. If an error occurs, the final values of
	70435	+** both piOffset and piVal are undefined.
	70436	+*/
	70437	+static int vdbeSorterReadVarint(
	70438	+ sqlite3_file pFile, / File to read from */
	70439	+ i64 iEof, /* Total number of bytes in file */
	70440	+ i64 piOffset, / IN/OUT: Read offset in pFile */
	70441	+ i64 piVal / OUT: Value read from file */
	70442	+){
	70443	+ u8 aVarint[9]; /* Buffer large enough for a varint */
	70444	+ i64 iOff = piOffset; / Offset in file to read from */
	70445	+ int nRead = 9; /* Number of bytes to read from file */
	70446	+ int rc; /* Return code */
	70447	+
	70448	+ assert( iEof>iOff );
	70449	+ if( (iEof-iOff)<nRead ){
	70450	+ nRead = iEof-iOff;
	70451	+ }
	70452	+
	70453	+ rc = sqlite3OsRead(pFile, aVarint, nRead, iOff);
	70454	+ if( rc==SQLITE_OK ){
	70455	+ piOffset += getVarint(aVarint, (u64 )piVal);
	70456	+ }
	70457	+
	70458	+ return rc;
	70459	+}
	70460	+
	70461	+/*
	70462	+** Initialize iterator pIter to scan through the PMA stored in file pFile
	70463	+** starting at offset iStart and ending at offset iEof-1. This function
	70464	+** leaves the iterator pointing to the first key in the PMA (or EOF if the
	70465	+** PMA is empty).
	70466	+*/
	70467	+static int vdbeSorterIterInit(
	70468	+ sqlite3 db, / Database handle */
	70469	+ VdbeSorter pSorter, / Sorter object */
	70470	+ i64 iStart, /* Start offset in pFile */
	70471	+ VdbeSorterIter pIter, / Iterator to populate */
	70472	+ i64 pnByte / IN/OUT: Increment this value by PMA size */
	70473	+){
	70474	+ int rc;
	70475	+
	70476	+ assert( pSorter->iWriteOff>iStart );
	70477	+ assert( pIter->aAlloc==0 );
	70478	+ pIter->pFile = pSorter->pTemp1;
	70479	+ pIter->iReadOff = iStart;
	70480	+ pIter->nAlloc = 128;
	70481	+ pIter->aAlloc = (u8 *)sqlite3DbMallocRaw(db, pIter->nAlloc);
	70482	+ if( !pIter->aAlloc ){
	70483	+ rc = SQLITE_NOMEM;
	70484	+ }else{
	70485	+ i64 iEof = pSorter->iWriteOff; /* EOF of file pSorter->pTemp1 */
	70486	+ i64 nByte; /* Total size of PMA in bytes */
	70487	+ rc = vdbeSorterReadVarint(pSorter->pTemp1, iEof, &pIter->iReadOff, &nByte);
	70488	+ *pnByte += nByte;
	70489	+ pIter->iEof = pIter->iReadOff + nByte;
	70490	+ }
	70491	+ if( rc==SQLITE_OK ){
	70492	+ rc = vdbeSorterIterNext(db, pIter);
	70493	+ }
	70494	+ return rc;
	70495	+}
	70496	+
	70497	+/*
	70498	+** This function is called to compare two iterator keys when merging
	70499	+** multiple b-tree segments. Parameter iOut is the index of the aTree[]
	70500	+** value to recalculate.
	70501	+*/
	70502	+static int vdbeSorterDoCompare(VdbeCursor *pCsr, int iOut){
	70503	+ VdbeSorter *pSorter = pCsr->pSorter;
	70504	+ int i1;
	70505	+ int i2;
	70506	+ int iRes;
	70507	+ VdbeSorterIter *p1;
	70508	+ VdbeSorterIter *p2;
	70509	+
	70510	+ assert( iOut<pSorter->nTree && iOut>0 );
	70511	+
	70512	+ if( iOut>=(pSorter->nTree/2) ){
	70513	+ i1 = (iOut - pSorter->nTree/2) * 2;
	70514	+ i2 = i1 + 1;
	70515	+ }else{
	70516	+ i1 = pSorter->aTree[iOut*2];
	70517	+ i2 = pSorter->aTree[iOut*2+1];
	70518	+ }
	70519	+
	70520	+ p1 = &pSorter->aIter[i1];
	70521	+ p2 = &pSorter->aIter[i2];
	70522	+
	70523	+ if( p1->pFile==0 ){
	70524	+ iRes = i2;
	70525	+ }else if( p2->pFile==0 ){
	70526	+ iRes = i1;
	70527	+ }else{
	70528	+ char aSpace[150];
	70529	+ UnpackedRecord *r1;
	70530	+
	70531	+ r1 = sqlite3VdbeRecordUnpack(
	70532	+ pCsr->pKeyInfo, p1->nKey, p1->aKey, aSpace, sizeof(aSpace)
	70533	+ );
	70534	+ if( r1==0 ) return SQLITE_NOMEM;
	70535	+
	70536	+ if( sqlite3VdbeRecordCompare(p2->nKey, p2->aKey, r1)>=0 ){
	70537	+ iRes = i1;
	70538	+ }else{
	70539	+ iRes = i2;
	70540	+ }
	70541	+ sqlite3VdbeDeleteUnpackedRecord(r1);
	70542	+ }
	70543	+
	70544	+ pSorter->aTree[iOut] = iRes;
	70545	+ return SQLITE_OK;
	70546	+}
	70547	+
	70548	+/*
	70549	+** Initialize the temporary index cursor just opened as a sorter cursor.
	70550	+*/
	70551	+SQLITE_PRIVATE int sqlite3VdbeSorterInit(sqlite3 db, VdbeCursor pCsr){
	70552	+ assert( pCsr->pKeyInfo && pCsr->pBt );
	70553	+ pCsr->pSorter = sqlite3DbMallocZero(db, sizeof(VdbeSorter));
	70554	+ return (pCsr->pSorter ? SQLITE_OK : SQLITE_NOMEM);
	70555	+}
	70556	+
	70557	+/*
	70558	+** Free any cursor components allocated by sqlite3VdbeSorterXXX routines.
	70559	+*/
	70560	+SQLITE_PRIVATE void sqlite3VdbeSorterClose(sqlite3 db, VdbeCursor pCsr){
	70561	+ VdbeSorter *pSorter = pCsr->pSorter;
	70562	+ if( pSorter ){
	70563	+ if( pSorter->aIter ){
	70564	+ int i;
	70565	+ for(i=0; i<pSorter->nTree; i++){
	70566	+ vdbeSorterIterZero(db, &pSorter->aIter[i]);
	70567	+ }
	70568	+ sqlite3DbFree(db, pSorter->aIter);
	70569	+ }
	70570	+ if( pSorter->pTemp1 ){
	70571	+ sqlite3OsCloseFree(pSorter->pTemp1);
	70572	+ }
	70573	+ sqlite3DbFree(db, pSorter);
	70574	+ pCsr->pSorter = 0;
	70575	+ }
	70576	+}
	70577	+
	70578	+/*
	70579	+** Allocate space for a file-handle and open a temporary file. If successful,
	70580	+** set *ppFile to point to the malloc'd file-handle and return SQLITE_OK.
	70581	+** Otherwise, set *ppFile to 0 and return an SQLite error code.
	70582	+*/
	70583	+static int vdbeSorterOpenTempFile(sqlite3 db, sqlite3_file *ppFile){
	70584	+ int dummy;
	70585	+ return sqlite3OsOpenMalloc(db->pVfs, 0, ppFile,
	70586	+ SQLITE_OPEN_TEMP_JOURNAL \|
	70587	+ SQLITE_OPEN_READWRITE \| SQLITE_OPEN_CREATE \|
	70588	+ SQLITE_OPEN_EXCLUSIVE \| SQLITE_OPEN_DELETEONCLOSE, &dummy
	70589	+ );
	70590	+}
	70591	+
	70592	+
	70593	+/*
	70594	+** Write the current contents of the b-tree to a PMA. Return SQLITE_OK
	70595	+** if successful, or an SQLite error code otherwise.
	70596	+**
	70597	+** The format of a PMA is:
	70598	+**
	70599	+** * A varint. This varint contains the total number of bytes of content
	70600	+** in the PMA (not including the varint itself).
	70601	+**
	70602	+** * One or more records packed end-to-end in order of ascending keys.
	70603	+** Each record consists of a varint followed by a blob of data (the
	70604	+** key). The varint is the number of bytes in the blob of data.
	70605	+*/
	70606	+static int vdbeSorterBtreeToPMA(sqlite3 db, VdbeCursor pCsr){
	70607	+ int rc = SQLITE_OK; /* Return code */
	70608	+ VdbeSorter *pSorter = pCsr->pSorter;
	70609	+ int res = 0;
	70610	+
	70611	+ /* sqlite3BtreeFirst() cannot fail because sorter btrees are always held
	70612	+ ** in memory and so an I/O error is not possible. */
	70613	+ rc = sqlite3BtreeFirst(pCsr->pCursor, &res);
	70614	+ if( NEVER(rc!=SQLITE_OK) \|\| res ) return rc;
	70615	+ assert( pSorter->nBtree>0 );
	70616	+
	70617	+ /* If the first temporary PMA file has not been opened, open it now. */
	70618	+ if( pSorter->pTemp1==0 ){
	70619	+ rc = vdbeSorterOpenTempFile(db, &pSorter->pTemp1);
	70620	+ assert( rc!=SQLITE_OK \|\| pSorter->pTemp1 );
	70621	+ assert( pSorter->iWriteOff==0 );
	70622	+ assert( pSorter->nPMA==0 );
	70623	+ }
	70624	+
	70625	+ if( rc==SQLITE_OK ){
	70626	+ i64 iWriteOff = pSorter->iWriteOff;
	70627	+ void aMalloc = 0; / Array used to hold a single record */
	70628	+ int nMalloc = 0; /* Allocated size of aMalloc[] in bytes */
	70629	+
	70630	+ pSorter->nPMA++;
	70631	+ for(
	70632	+ rc = vdbeSorterWriteVarint(pSorter->pTemp1, pSorter->nBtree, &iWriteOff);
	70633	+ rc==SQLITE_OK && res==0;
	70634	+ rc = sqlite3BtreeNext(pCsr->pCursor, &res)
	70635	+ ){
	70636	+ i64 nKey; /* Size of this key in bytes */
	70637	+
	70638	+ /* Write the size of the record in bytes to the output file */
	70639	+ (void)sqlite3BtreeKeySize(pCsr->pCursor, &nKey);
	70640	+ rc = vdbeSorterWriteVarint(pSorter->pTemp1, nKey, &iWriteOff);
	70641	+
	70642	+ /* Make sure the aMalloc[] buffer is large enough for the record */
	70643	+ if( rc==SQLITE_OK && nKey>nMalloc ){
	70644	+ aMalloc = sqlite3DbReallocOrFree(db, aMalloc, nKey);
	70645	+ if( !aMalloc ){
	70646	+ rc = SQLITE_NOMEM;
	70647	+ }else{
	70648	+ nMalloc = nKey;
	70649	+ }
	70650	+ }
	70651	+
	70652	+ /* Write the record itself to the output file */
	70653	+ if( rc==SQLITE_OK ){
	70654	+ /* sqlite3BtreeKey() cannot fail because sorter btrees held in memory */
	70655	+ rc = sqlite3BtreeKey(pCsr->pCursor, 0, nKey, aMalloc);
	70656	+ if( ALWAYS(rc==SQLITE_OK) ){
	70657	+ rc = sqlite3OsWrite(pSorter->pTemp1, aMalloc, nKey, iWriteOff);
	70658	+ iWriteOff += nKey;
	70659	+ }
	70660	+ }
	70661	+
	70662	+ if( rc!=SQLITE_OK ) break;
	70663	+ }
	70664	+
	70665	+ /* This assert verifies that unless an error has occurred, the size of
	70666	+ ** the PMA on disk is the same as the expected size stored in
	70667	+ ** pSorter->nBtree. */
	70668	+ assert( rc!=SQLITE_OK \|\| pSorter->nBtree==(
	70669	+ iWriteOff-pSorter->iWriteOff-sqlite3VarintLen(pSorter->nBtree)
	70670	+ ));
	70671	+
	70672	+ pSorter->iWriteOff = iWriteOff;
	70673	+ sqlite3DbFree(db, aMalloc);
	70674	+ }
	70675	+
	70676	+ pSorter->nBtree = 0;
	70677	+ return rc;
	70678	+}
	70679	+
	70680	+/*
	70681	+** This function is called on a sorter cursor by the VDBE before each row
	70682	+** is inserted into VdbeCursor.pCsr. Argument nKey is the size of the key, in
	70683	+** bytes, about to be inserted.
	70684	+**
	70685	+** If it is determined that the temporary b-tree accessed via VdbeCursor.pCsr
	70686	+** is large enough, its contents are written to a sorted PMA on disk and the
	70687	+** tree emptied. This prevents the b-tree (which must be small enough to
	70688	+** fit entirely in the cache in order to support efficient inserts) from
	70689	+** growing too large.
	70690	+**
	70691	+** An SQLite error code is returned if an error occurs. Otherwise, SQLITE_OK.
	70692	+*/
	70693	+SQLITE_PRIVATE int sqlite3VdbeSorterWrite(sqlite3 db, VdbeCursor pCsr, int nKey){
	70694	+ int rc = SQLITE_OK; /* Return code */
	70695	+ VdbeSorter *pSorter = pCsr->pSorter;
	70696	+ if( pSorter ){
	70697	+ Pager *pPager = sqlite3BtreePager(pCsr->pBt);
	70698	+ int nPage; /* Current size of temporary file in pages */
	70699	+
	70700	+ /* Sorters never spill to disk */
	70701	+ assert( sqlite3PagerFile(pPager)->pMethods==0 );
	70702	+
	70703	+ /* Determine how many pages the temporary b-tree has grown to */
	70704	+ sqlite3PagerPagecount(pPager, &nPage);
	70705	+
	70706	+ /* If pSorter->nWorking is still zero, but the temporary file has been
	70707	+ ** created in the file-system, then the most recent insert into the
	70708	+ ** current b-tree segment probably caused the cache to overflow (it is
	70709	+ ** also possible that sqlite3_release_memory() was called). So set the
	70710	+ ** size of the working set to a little less than the current size of the
	70711	+ ** file in pages. */
	70712	+ if( pSorter->nWorking==0 && sqlite3PagerUnderStress(pPager) ){
	70713	+ pSorter->nWorking = nPage-5;
	70714	+ if( pSorter->nWorking<SORTER_MIN_WORKING ){
	70715	+ pSorter->nWorking = SORTER_MIN_WORKING;
	70716	+ }
	70717	+ }
	70718	+
	70719	+ /* If the number of pages used by the current b-tree segment is greater
	70720	+ ** than the size of the working set (VdbeSorter.nWorking), start a new
	70721	+ ** segment b-tree. */
	70722	+ if( pSorter->nWorking && nPage>=pSorter->nWorking ){
	70723	+ BtCursor p = pCsr->pCursor;/ Cursor structure to close and reopen */
	70724	+ int iRoot; /* Root page of new tree */
	70725	+
	70726	+ /* Copy the current contents of the b-tree into a PMA in sorted order.
	70727	+ ** Close the currently open b-tree cursor. */
	70728	+ rc = vdbeSorterBtreeToPMA(db, pCsr);
	70729	+ sqlite3BtreeCloseCursor(p);
	70730	+
	70731	+ if( rc==SQLITE_OK ){
	70732	+ rc = sqlite3BtreeDropTable(pCsr->pBt, 2, 0);
	70733	+#ifdef SQLITE_DEBUG
	70734	+ sqlite3PagerPagecount(pPager, &nPage);
	70735	+ assert( rc!=SQLITE_OK \|\| nPage==1 );
	70736	+#endif
	70737	+ }
	70738	+ if( rc==SQLITE_OK ){
	70739	+ rc = sqlite3BtreeCreateTable(pCsr->pBt, &iRoot, BTREE_BLOBKEY);
	70740	+ }
	70741	+ if( rc==SQLITE_OK ){
	70742	+ assert( iRoot==2 );
	70743	+ rc = sqlite3BtreeCursor(pCsr->pBt, iRoot, 1, pCsr->pKeyInfo, p);
	70744	+ }
	70745	+ }
	70746	+
	70747	+ pSorter->nBtree += sqlite3VarintLen(nKey) + nKey;
	70748	+ }
	70749	+ return rc;
	70750	+}
	70751	+
	70752	+/*
	70753	+** Helper function for sqlite3VdbeSorterRewind().
	70754	+*/
	70755	+static int vdbeSorterInitMerge(
	70756	+ sqlite3 db, / Database handle */
	70757	+ VdbeCursor pCsr, / Cursor handle for this sorter */
	70758	+ i64 pnByte / Sum of bytes in all opened PMAs */
	70759	+){
	70760	+ VdbeSorter *pSorter = pCsr->pSorter;
	70761	+ int rc = SQLITE_OK; /* Return code */
	70762	+ int i; /* Used to iterator through aIter[] */
	70763	+ i64 nByte = 0; /* Total bytes in all opened PMAs */
	70764	+
	70765	+ /* Initialize the iterators. */
	70766	+ for(i=0; rc==SQLITE_OK && i<SORTER_MAX_MERGE_COUNT; i++){
	70767	+ VdbeSorterIter *pIter = &pSorter->aIter[i];
	70768	+ rc = vdbeSorterIterInit(db, pSorter, pSorter->iReadOff, pIter, &nByte);
	70769	+ pSorter->iReadOff = pIter->iEof;
	70770	+ assert( pSorter->iReadOff<=pSorter->iWriteOff \|\| rc!=SQLITE_OK );
	70771	+ if( pSorter->iReadOff>=pSorter->iWriteOff ) break;
	70772	+ }
	70773	+
	70774	+ /* Initialize the aTree[] array. */
	70775	+ for(i=pSorter->nTree-1; rc==SQLITE_OK && i>0; i--){
	70776	+ rc = vdbeSorterDoCompare(pCsr, i);
	70777	+ }
	70778	+
	70779	+ *pnByte = nByte;
	70780	+ return rc;
	70781	+}
	70782	+
	70783	+/*
	70784	+** Once the sorter has been populated, this function is called to prepare
	70785	+** for iterating through its contents in sorted order.
	70786	+*/
	70787	+SQLITE_PRIVATE int sqlite3VdbeSorterRewind(sqlite3 db, VdbeCursor pCsr, int *pbEof){
	70788	+ VdbeSorter *pSorter = pCsr->pSorter;
	70789	+ int rc; /* Return code */
	70790	+ sqlite3_file pTemp2 = 0; / Second temp file to use */
	70791	+ i64 iWrite2 = 0; /* Write offset for pTemp2 */
	70792	+ int nIter; /* Number of iterators used */
	70793	+ int nByte; /* Bytes of space required for aIter/aTree */
	70794	+ int N = 2; /* Power of 2 >= nIter */
	70795	+
	70796	+ assert( pSorter );
	70797	+
	70798	+ /* Write the current b-tree to a PMA. Close the b-tree cursor. */
	70799	+ rc = vdbeSorterBtreeToPMA(db, pCsr);
	70800	+ sqlite3BtreeCloseCursor(pCsr->pCursor);
	70801	+ if( rc!=SQLITE_OK ) return rc;
	70802	+ if( pSorter->nPMA==0 ){
	70803	+ *pbEof = 1;
	70804	+ return SQLITE_OK;
	70805	+ }
	70806	+
	70807	+ /* Allocate space for aIter[] and aTree[]. */
	70808	+ nIter = pSorter->nPMA;
	70809	+ if( nIter>SORTER_MAX_MERGE_COUNT ) nIter = SORTER_MAX_MERGE_COUNT;
	70810	+ assert( nIter>0 );
	70811	+ while( N<nIter ) N += N;
	70812	+ nByte = N * (sizeof(int) + sizeof(VdbeSorterIter));
	70813	+ pSorter->aIter = (VdbeSorterIter *)sqlite3DbMallocZero(db, nByte);
	70814	+ if( !pSorter->aIter ) return SQLITE_NOMEM;
	70815	+ pSorter->aTree = (int *)&pSorter->aIter[N];
	70816	+ pSorter->nTree = N;
	70817	+
	70818	+ do {
	70819	+ int iNew; /* Index of new, merged, PMA */
	70820	+
	70821	+ for(iNew=0;
	70822	+ rc==SQLITE_OK && iNew*SORTER_MAX_MERGE_COUNT<pSorter->nPMA;
	70823	+ iNew++
	70824	+ ){
	70825	+ i64 nWrite; /* Number of bytes in new PMA */
	70826	+
	70827	+ /* If there are SORTER_MAX_MERGE_COUNT or less PMAs in file pTemp1,
	70828	+ ** initialize an iterator for each of them and break out of the loop.
	70829	+ ** These iterators will be incrementally merged as the VDBE layer calls
	70830	+ ** sqlite3VdbeSorterNext().
	70831	+ **
	70832	+ ** Otherwise, if pTemp1 contains more than SORTER_MAX_MERGE_COUNT PMAs,
	70833	+ ** initialize interators for SORTER_MAX_MERGE_COUNT of them. These PMAs
	70834	+ ** are merged into a single PMA that is written to file pTemp2.
	70835	+ */
	70836	+ rc = vdbeSorterInitMerge(db, pCsr, &nWrite);
	70837	+ assert( rc!=SQLITE_OK \|\| pSorter->aIter[ pSorter->aTree[1] ].pFile );
	70838	+ if( rc!=SQLITE_OK \|\| pSorter->nPMA<=SORTER_MAX_MERGE_COUNT ){
	70839	+ break;
	70840	+ }
	70841	+
	70842	+ /* Open the second temp file, if it is not already open. */
	70843	+ if( pTemp2==0 ){
	70844	+ assert( iWrite2==0 );
	70845	+ rc = vdbeSorterOpenTempFile(db, &pTemp2);
	70846	+ }
	70847	+
	70848	+ if( rc==SQLITE_OK ){
	70849	+ rc = vdbeSorterWriteVarint(pTemp2, nWrite, &iWrite2);
	70850	+ }
	70851	+
	70852	+ if( rc==SQLITE_OK ){
	70853	+ int bEof = 0;
	70854	+ while( rc==SQLITE_OK && bEof==0 ){
	70855	+ int nToWrite;
	70856	+ VdbeSorterIter *pIter = &pSorter->aIter[ pSorter->aTree[1] ];
	70857	+ assert( pIter->pFile );
	70858	+ nToWrite = pIter->nKey + sqlite3VarintLen(pIter->nKey);
	70859	+ rc = sqlite3OsWrite(pTemp2, pIter->aAlloc, nToWrite, iWrite2);
	70860	+ iWrite2 += nToWrite;
	70861	+ if( rc==SQLITE_OK ){
	70862	+ rc = sqlite3VdbeSorterNext(db, pCsr, &bEof);
	70863	+ }
	70864	+ }
	70865	+ }
	70866	+ }
	70867	+
	70868	+ if( pSorter->nPMA<=SORTER_MAX_MERGE_COUNT ){
	70869	+ break;
	70870	+ }else{
	70871	+ sqlite3_file *pTmp = pSorter->pTemp1;
	70872	+ pSorter->nPMA = iNew;
	70873	+ pSorter->pTemp1 = pTemp2;
	70874	+ pTemp2 = pTmp;
	70875	+ pSorter->iWriteOff = iWrite2;
	70876	+ pSorter->iReadOff = 0;
	70877	+ iWrite2 = 0;
	70878	+ }
	70879	+ }while( rc==SQLITE_OK );
	70880	+
	70881	+ if( pTemp2 ){
	70882	+ sqlite3OsCloseFree(pTemp2);
	70883	+ }
	70884	+ *pbEof = (pSorter->aIter[pSorter->aTree[1]].pFile==0);
	70885	+ return rc;
	70886	+}
	70887	+
	70888	+/*
	70889	+** Advance to the next element in the sorter.
	70890	+*/
	70891	+SQLITE_PRIVATE int sqlite3VdbeSorterNext(sqlite3 db, VdbeCursor pCsr, int *pbEof){
	70892	+ VdbeSorter *pSorter = pCsr->pSorter;
	70893	+ int iPrev = pSorter->aTree[1]; /* Index of iterator to advance */
	70894	+ int i; /* Index of aTree[] to recalculate */
	70895	+ int rc; /* Return code */
	70896	+
	70897	+ rc = vdbeSorterIterNext(db, &pSorter->aIter[iPrev]);
	70898	+ for(i=(pSorter->nTree+iPrev)/2; rc==SQLITE_OK && i>0; i=i/2){
	70899	+ rc = vdbeSorterDoCompare(pCsr, i);
	70900	+ }
	70901	+
	70902	+ *pbEof = (pSorter->aIter[pSorter->aTree[1]].pFile==0);
	70903	+ return rc;
	70904	+}
	70905	+
	70906	+/*
	70907	+** Copy the current sorter key into the memory cell pOut.
	70908	+*/
	70909	+SQLITE_PRIVATE int sqlite3VdbeSorterRowkey(VdbeCursor pCsr, Mem pOut){
	70910	+ VdbeSorter *pSorter = pCsr->pSorter;
	70911	+ VdbeSorterIter *pIter;
	70912	+
	70913	+ pIter = &pSorter->aIter[ pSorter->aTree[1] ];
	70914	+
	70915	+ /* Coverage testing note: As things are currently, this call will always
	70916	+ ** succeed. This is because the memory cell passed by the VDBE layer
	70917	+ ** happens to be the same one as was used to assemble the keys before they
	70918	+ ** were passed to the sorter - meaning it is always large enough for the
	70919	+ ** largest key. But this could change very easily, so we leave the call
	70920	+ ** to sqlite3VdbeMemGrow() in. */
	70921	+ if( NEVER(sqlite3VdbeMemGrow(pOut, pIter->nKey, 0)) ){
	70922	+ return SQLITE_NOMEM;
	70923	+ }
	70924	+ pOut->n = pIter->nKey;
	70925	+ MemSetTypeFlag(pOut, MEM_Blob);
	70926	+ memcpy(pOut->z, pIter->aKey, pIter->nKey);
	70927	+
	70928	+ return SQLITE_OK;
	70929	+}
	70930	+
	70931	+#endif /* #ifndef SQLITE_OMIT_MERGE_SORT */
	70932	+
	70933	+/************ End of vdbesort.c ******************************************/
69328	70934	/************ Begin file journal.c ***************************************/
69329	70935	/*
69330	70936	** 2007 August 22
69331	70937	**
69332	70938	** The author disclaims copyright to this source code. In place of
		@@ -69839,10 +71445,12 @@
69839	71445	**
69840	71446	*************************************************************************
69841	71447	** This file contains routines used for walking the parser tree for
69842	71448	** an SQL statement.
69843	71449	*/
	71450	+/* #include <stdlib.h> */
	71451	+/* #include <string.h> */
69844	71452
69845	71453
69846	71454	/*
69847	71455	** Walk an expression tree. Invoke the callback once for each node
69848	71456	** of the expression, while decending. (In other words, the callback
		@@ -69977,10 +71585,12 @@
69977	71585	**
69978	71586	** This file contains routines used for walking the parser tree and
69979	71587	** resolve all identifiers by associating them with a particular
69980	71588	** table and column.
69981	71589	*/
	71590	+/* #include <stdlib.h> */
	71591	+/* #include <string.h> */
69982	71592
69983	71593	/*
69984	71594	** Turn the pExpr expression into an alias for the iCol-th column of the
69985	71595	** result set in pEList.
69986	71596	**
		@@ -70956,15 +72566,29 @@
70956	72566	/* Recursively resolve names in all subqueries
70957	72567	*/
70958	72568	for(i=0; i<p->pSrc->nSrc; i++){
70959	72569	struct SrcList_item *pItem = &p->pSrc->a[i];
70960	72570	if( pItem->pSelect ){
	72571	+ NameContext pNC; / Used to iterate name contexts */
	72572	+ int nRef = 0; /* Refcount for pOuterNC and outer contexts */
70961	72573	const char *zSavedContext = pParse->zAuthContext;
	72574	+
	72575	+ /* Count the total number of references to pOuterNC and all of its
	72576	+ ** parent contexts. After resolving references to expressions in
	72577	+ ** pItem->pSelect, check if this value has changed. If so, then
	72578	+ ** SELECT statement pItem->pSelect must be correlated. Set the
	72579	+ ** pItem->isCorrelated flag if this is the case. */
	72580	+ for(pNC=pOuterNC; pNC; pNC=pNC->pNext) nRef += pNC->nRef;
	72581	+
70962	72582	if( pItem->zName ) pParse->zAuthContext = pItem->zName;
70963	72583	sqlite3ResolveSelectNames(pParse, pItem->pSelect, pOuterNC);
70964	72584	pParse->zAuthContext = zSavedContext;
70965	72585	if( pParse->nErr \|\| db->mallocFailed ) return WRC_Abort;
	72586	+
	72587	+ for(pNC=pOuterNC; pNC; pNC=pNC->pNext) nRef -= pNC->nRef;
	72588	+ assert( pItem->isCorrelated==0 && nRef<=0 );
	72589	+ pItem->isCorrelated = (nRef!=0);
70966	72590	}
70967	72591	}
70968	72592
70969	72593	/* If there are no aggregate functions in the result-set, and no GROUP BY
70970	72594	** expression, do not allow aggregates in any of the other expressions.
		@@ -72068,10 +73692,11 @@
72068	73692	pNewItem->zName = sqlite3DbStrDup(db, pOldItem->zName);
72069	73693	pNewItem->zAlias = sqlite3DbStrDup(db, pOldItem->zAlias);
72070	73694	pNewItem->jointype = pOldItem->jointype;
72071	73695	pNewItem->iCursor = pOldItem->iCursor;
72072	73696	pNewItem->isPopulated = pOldItem->isPopulated;
	73697	+ pNewItem->isCorrelated = pOldItem->isCorrelated;
72073	73698	pNewItem->zIndex = sqlite3DbStrDup(db, pOldItem->zIndex);
72074	73699	pNewItem->notIndexed = pOldItem->notIndexed;
72075	73700	pNewItem->pIndex = pOldItem->pIndex;
72076	73701	pTab = pNewItem->pTab = pOldItem->pTab;
72077	73702	if( pTab ){
		@@ -78958,11 +80583,11 @@
78958	80583	Table *p;
78959	80584	int n;
78960	80585	const char *z;
78961	80586	Token sEnd;
78962	80587	DbFixer sFix;
78963		- Token *pName;
	80588	+ Token *pName = 0;
78964	80589	int iDb;
78965	80590	sqlite3 *db = pParse->db;
78966	80591
78967	80592	if( pParse->nVar>0 ){
78968	80593	sqlite3ErrorMsg(pParse, "parameters are not allowed in views");
		@@ -79263,10 +80888,33 @@
79263	80888	iDestroyed = iLargest;
79264	80889	}
79265	80890	}
79266	80891	#endif
79267	80892	}
	80893	+
	80894	+/*
	80895	+** Remove entries from the sqlite_stat1 and sqlite_stat2 tables
	80896	+** after a DROP INDEX or DROP TABLE command.
	80897	+*/
	80898	+static void sqlite3ClearStatTables(
	80899	+ Parse pParse, / The parsing context */
	80900	+ int iDb, /* The database number */
	80901	+ const char zType, / "idx" or "tbl" */
	80902	+ const char zName / Name of index or table */
	80903	+){
	80904	+ static const char *azStatTab[] = { "sqlite_stat1", "sqlite_stat2" };
	80905	+ int i;
	80906	+ const char *zDbName = pParse->db->aDb[iDb].zName;
	80907	+ for(i=0; i<ArraySize(azStatTab); i++){
	80908	+ if( sqlite3FindTable(pParse->db, azStatTab[i], zDbName) ){
	80909	+ sqlite3NestedParse(pParse,
	80910	+ "DELETE FROM %Q.%s WHERE %s=%Q",
	80911	+ zDbName, azStatTab[i], zType, zName
	80912	+ );
	80913	+ }
	80914	+ }
	80915	+}
79268	80916
79269	80917	/*
79270	80918	** This routine is called to do the work of a DROP TABLE statement.
79271	80919	** pName is the name of the table to be dropped.
79272	80920	*/
		@@ -79403,18 +81051,11 @@
79403	81051	** database.
79404	81052	*/
79405	81053	sqlite3NestedParse(pParse,
79406	81054	"DELETE FROM %Q.%s WHERE tbl_name=%Q and type!='trigger'",
79407	81055	pDb->zName, SCHEMA_TABLE(iDb), pTab->zName);
79408		-
79409		- /* Drop any statistics from the sqlite_stat1 table, if it exists */
79410		- if( sqlite3FindTable(db, "sqlite_stat1", db->aDb[iDb].zName) ){
79411		- sqlite3NestedParse(pParse,
79412		- "DELETE FROM %Q.sqlite_stat1 WHERE tbl=%Q", pDb->zName, pTab->zName
79413		- );
79414		- }
79415		-
	81056	+ sqlite3ClearStatTables(pParse, iDb, "tbl", pTab->zName);
79416	81057	if( !isView && !IsVirtual(pTab) ){
79417	81058	destroyTable(pParse, pTab);
79418	81059	}
79419	81060
79420	81061	/* Remove the table entry from SQLite's internal schema and modify
		@@ -79592,18 +81233,28 @@
79592	81233	*/
79593	81234	static void sqlite3RefillIndex(Parse pParse, Index pIndex, int memRootPage){
79594	81235	Table pTab = pIndex->pTable; / The table that is indexed */
79595	81236	int iTab = pParse->nTab++; /* Btree cursor used for pTab */
79596	81237	int iIdx = pParse->nTab++; /* Btree cursor used for pIndex */
	81238	+ int iSorter = iTab; /* Cursor opened by OpenSorter (if in use) */
79597	81239	int addr1; /* Address of top of loop */
79598	81240	int tnum; /* Root page of index */
79599	81241	Vdbe v; / Generate code into this virtual machine */
79600	81242	KeyInfo pKey; / KeyInfo for index */
79601	81243	int regIdxKey; /* Registers containing the index key */
79602	81244	int regRecord; /* Register holding assemblied index record */
79603	81245	sqlite3 db = pParse->db; / The database connection */
79604	81246	int iDb = sqlite3SchemaToIndex(db, pIndex->pSchema);
	81247	+
	81248	+ /* Set bUseSorter to use OP_OpenSorter, or clear it to insert directly
	81249	+ ** into the index. The sorter is used unless either OMIT_MERGE_SORT is
	81250	+ ** defined or the system is configured to store temp files in-memory. */
	81251	+#ifdef SQLITE_OMIT_MERGE_SORT
	81252	+ static const int bUseSorter = 0;
	81253	+#else
	81254	+ const int bUseSorter = !sqlite3TempInMemory(pParse->db);
	81255	+#endif
79605	81256
79606	81257	#ifndef SQLITE_OMIT_AUTHORIZATION
79607	81258	if( sqlite3AuthCheck(pParse, SQLITE_REINDEX, pIndex->zName, 0,
79608	81259	db->aDb[iDb].zName ) ){
79609	81260	return;
		@@ -79625,14 +81276,33 @@
79625	81276	sqlite3VdbeAddOp4(v, OP_OpenWrite, iIdx, tnum, iDb,
79626	81277	(char *)pKey, P4_KEYINFO_HANDOFF);
79627	81278	if( memRootPage>=0 ){
79628	81279	sqlite3VdbeChangeP5(v, 1);
79629	81280	}
	81281	+
	81282	+ /* Open the sorter cursor if we are to use one. */
	81283	+ if( bUseSorter ){
	81284	+ iSorter = pParse->nTab++;
	81285	+ sqlite3VdbeAddOp4(v, OP_OpenSorter, iSorter, 0, 0, (char*)pKey, P4_KEYINFO);
	81286	+ sqlite3VdbeChangeP5(v, BTREE_SORTER);
	81287	+ }
	81288	+
	81289	+ /* Open the table. Loop through all rows of the table, inserting index
	81290	+ ** records into the sorter. */
79630	81291	sqlite3OpenTable(pParse, iTab, iDb, pTab, OP_OpenRead);
79631	81292	addr1 = sqlite3VdbeAddOp2(v, OP_Rewind, iTab, 0);
79632	81293	regRecord = sqlite3GetTempReg(pParse);
79633	81294	regIdxKey = sqlite3GenerateIndexKey(pParse, pIndex, iTab, regRecord, 1);
	81295	+
	81296	+ if( bUseSorter ){
	81297	+ sqlite3VdbeAddOp2(v, OP_IdxInsert, iSorter, regRecord);
	81298	+ sqlite3VdbeAddOp2(v, OP_Next, iTab, addr1+1);
	81299	+ sqlite3VdbeJumpHere(v, addr1);
	81300	+ addr1 = sqlite3VdbeAddOp2(v, OP_Sort, iSorter, 0);
	81301	+ sqlite3VdbeAddOp2(v, OP_RowKey, iSorter, regRecord);
	81302	+ }
	81303	+
79634	81304	if( pIndex->onError!=OE_None ){
79635	81305	const int regRowid = regIdxKey + pIndex->nColumn;
79636	81306	const int j2 = sqlite3VdbeCurrentAddr(v) + 2;
79637	81307	void * const pRegKey = SQLITE_INT_TO_PTR(regIdxKey);
79638	81308
		@@ -79647,17 +81317,19 @@
79647	81317	*/
79648	81318	sqlite3VdbeAddOp4(v, OP_IsUnique, iIdx, j2, regRowid, pRegKey, P4_INT32);
79649	81319	sqlite3HaltConstraint(
79650	81320	pParse, OE_Abort, "indexed columns are not unique", P4_STATIC);
79651	81321	}
79652		- sqlite3VdbeAddOp2(v, OP_IdxInsert, iIdx, regRecord);
	81322	+ sqlite3VdbeAddOp3(v, OP_IdxInsert, iIdx, regRecord, bUseSorter);
79653	81323	sqlite3VdbeChangeP5(v, OPFLAG_USESEEKRESULT);
79654	81324	sqlite3ReleaseTempReg(pParse, regRecord);
79655		- sqlite3VdbeAddOp2(v, OP_Next, iTab, addr1+1);
	81325	+ sqlite3VdbeAddOp2(v, OP_Next, iSorter, addr1+1);
79656	81326	sqlite3VdbeJumpHere(v, addr1);
	81327	+
79657	81328	sqlite3VdbeAddOp1(v, OP_Close, iTab);
79658	81329	sqlite3VdbeAddOp1(v, OP_Close, iIdx);
	81330	+ sqlite3VdbeAddOp1(v, OP_Close, iSorter);
79659	81331	}
79660	81332
79661	81333	/*
79662	81334	** Create a new index for an SQL table. pName1.pName2 is the name of the index
79663	81335	** and pTblList is the name of the table that is to be indexed. Both will
		@@ -80075,11 +81747,11 @@
80075	81747	if( pStart ){
80076	81748	assert( pEnd!=0 );
80077	81749	/* A named index with an explicit CREATE INDEX statement */
80078	81750	zStmt = sqlite3MPrintf(db, "CREATE%s INDEX %.*s",
80079	81751	onError==OE_None ? "" : " UNIQUE",
80080		- pEnd->z - pName->z + 1,
	81752	+ (int)(pEnd->z - pName->z) + 1,
80081	81753	pName->z);
80082	81754	}else{
80083	81755	/* An automatic index created by a PRIMARY KEY or UNIQUE constraint */
80084	81756	/* zStmt = sqlite3MPrintf(""); */
80085	81757	zStmt = 0;
		@@ -80233,19 +81905,13 @@
80233	81905	v = sqlite3GetVdbe(pParse);
80234	81906	if( v ){
80235	81907	sqlite3BeginWriteOperation(pParse, 1, iDb);
80236	81908	sqlite3NestedParse(pParse,
80237	81909	"DELETE FROM %Q.%s WHERE name=%Q AND type='index'",
80238		- db->aDb[iDb].zName, SCHEMA_TABLE(iDb),
80239		- pIndex->zName
80240		- );
80241		- if( sqlite3FindTable(db, "sqlite_stat1", db->aDb[iDb].zName) ){
80242		- sqlite3NestedParse(pParse,
80243		- "DELETE FROM %Q.sqlite_stat1 WHERE idx=%Q",
80244		- db->aDb[iDb].zName, pIndex->zName
80245		- );
80246		- }
	81910	+ db->aDb[iDb].zName, SCHEMA_TABLE(iDb), pIndex->zName
	81911	+ );
	81912	+ sqlite3ClearStatTables(pParse, iDb, "idx", pIndex->zName);
80247	81913	sqlite3ChangeCookie(pParse, iDb);
80248	81914	destroyRootPage(pParse, pIndex->tnum, iDb);
80249	81915	sqlite3VdbeAddOp4(v, OP_DropIndex, iDb, 0, 0, pIndex->zName, 0);
80250	81916	}
80251	81917
		@@ -80613,12 +82279,13 @@
80613	82279	** The operator is "natural cross join". The A and B operands are stored
80614	82280	** in p->a[0] and p->a[1], respectively. The parser initially stores the
80615	82281	** operator with A. This routine shifts that operator over to B.
80616	82282	*/
80617	82283	SQLITE_PRIVATE void sqlite3SrcListShiftJoinType(SrcList *p){
80618		- if( p && p->a ){
	82284	+ if( p ){
80619	82285	int i;
	82286	+ assert( p->a \|\| p->nSrc==0 );
80620	82287	for(i=p->nSrc-1; i>0; i--){
80621	82288	p->a[i].jointype = p->a[i-1].jointype;
80622	82289	}
80623	82290	p->a[0].jointype = 0;
80624	82291	}
		@@ -81870,11 +83537,13 @@
81870	83537	int regRowid; /* Actual register containing rowids */
81871	83538
81872	83539	/* Collect rowids of every row to be deleted.
81873	83540	*/
81874	83541	sqlite3VdbeAddOp2(v, OP_Null, 0, iRowSet);
81875		- pWInfo = sqlite3WhereBegin(pParse, pTabList, pWhere,0,WHERE_DUPLICATES_OK);
	83542	+ pWInfo = sqlite3WhereBegin(
	83543	+ pParse, pTabList, pWhere, 0, 0, WHERE_DUPLICATES_OK
	83544	+ );
81876	83545	if( pWInfo==0 ) goto delete_from_cleanup;
81877	83546	regRowid = sqlite3ExprCodeGetColumn(pParse, pTab, -1, iCur, iRowid);
81878	83547	sqlite3VdbeAddOp2(v, OP_RowSetAdd, iRowSet, regRowid);
81879	83548	if( db->flags & SQLITE_CountRows ){
81880	83549	sqlite3VdbeAddOp2(v, OP_AddImm, memCnt, 1);
		@@ -82135,12 +83804,18 @@
82135	83804	sqlite3VdbeAddOp3(v, OP_Column, iCur, idx, regBase+j);
82136	83805	sqlite3ColumnDefault(v, pTab, idx, -1);
82137	83806	}
82138	83807	}
82139	83808	if( doMakeRec ){
	83809	+ const char *zAff;
	83810	+ if( pTab->pSelect \|\| (pParse->db->flags & SQLITE_IdxRealAsInt)!=0 ){
	83811	+ zAff = 0;
	83812	+ }else{
	83813	+ zAff = sqlite3IndexAffinityStr(v, pIdx);
	83814	+ }
82140	83815	sqlite3VdbeAddOp3(v, OP_MakeRecord, regBase, nCol+1, regOut);
82141		- sqlite3VdbeChangeP4(v, -1, sqlite3IndexAffinityStr(v, pIdx), P4_TRANSIENT);
	83816	+ sqlite3VdbeChangeP4(v, -1, zAff, P4_TRANSIENT);
82142	83817	}
82143	83818	sqlite3ReleaseTempRange(pParse, regBase, nCol+1);
82144	83819	return regBase;
82145	83820	}
82146	83821
		@@ -82162,10 +83837,12 @@
82162	83837	**
82163	83838	** There is only one exported symbol in this file - the function
82164	83839	** sqliteRegisterBuildinFunctions() found at the bottom of the file.
82165	83840	** All other code has file scope.
82166	83841	*/
	83842	+/* #include <stdlib.h> */
	83843	+/* #include <assert.h> */
82167	83844
82168	83845	/*
82169	83846	** Return the collating function associated with a function.
82170	83847	*/
82171	83848	static CollSeq sqlite3GetFuncCollSeq(sqlite3_context context){
		@@ -84311,11 +85988,11 @@
84311	85988
84312	85989	/* Create VDBE to loop through the entries in pSrc that match the WHERE
84313	85990	** clause. If the constraint is not deferred, throw an exception for
84314	85991	** each row found. Otherwise, for deferred constraints, increment the
84315	85992	** deferred constraint counter by nIncr for each row selected. */
84316		- pWInfo = sqlite3WhereBegin(pParse, pSrc, pWhere, 0, 0);
	85993	+ pWInfo = sqlite3WhereBegin(pParse, pSrc, pWhere, 0, 0, 0);
84317	85994	if( nIncr>0 && pFKey->isDeferred==0 ){
84318	85995	sqlite3ParseToplevel(pParse)->mayAbort = 1;
84319	85996	}
84320	85997	sqlite3VdbeAddOp2(v, OP_FkCounter, pFKey->isDeferred, nIncr);
84321	85998	if( pWInfo ){
		@@ -84485,11 +86162,28 @@
84485	86162	pTo = sqlite3FindTable(db, pFKey->zTo, zDb);
84486	86163	}else{
84487	86164	pTo = sqlite3LocateTable(pParse, 0, pFKey->zTo, zDb);
84488	86165	}
84489	86166	if( !pTo \|\| locateFkeyIndex(pParse, pTo, pFKey, &pIdx, &aiFree) ){
	86167	+ assert( isIgnoreErrors==0 \|\| (regOld!=0 && regNew==0) );
84490	86168	if( !isIgnoreErrors \|\| db->mallocFailed ) return;
	86169	+ if( pTo==0 ){
	86170	+ /* If isIgnoreErrors is true, then a table is being dropped. In this
	86171	+ ** case SQLite runs a "DELETE FROM xxx" on the table being dropped
	86172	+ ** before actually dropping it in order to check FK constraints.
	86173	+ ** If the parent table of an FK constraint on the current table is
	86174	+ ** missing, behave as if it is empty. i.e. decrement the relevant
	86175	+ ** FK counter for each row of the current table with non-NULL keys.
	86176	+ */
	86177	+ Vdbe *v = sqlite3GetVdbe(pParse);
	86178	+ int iJump = sqlite3VdbeCurrentAddr(v) + pFKey->nCol + 1;
	86179	+ for(i=0; i<pFKey->nCol; i++){
	86180	+ int iReg = pFKey->aCol[i].iFrom + regOld + 1;
	86181	+ sqlite3VdbeAddOp2(v, OP_IsNull, iReg, iJump);
	86182	+ }
	86183	+ sqlite3VdbeAddOp2(v, OP_FkCounter, pFKey->isDeferred, -1);
	86184	+ }
84491	86185	continue;
84492	86186	}
84493	86187	assert( pFKey->nCol==1 \|\| (aiFree && pIdx) );
84494	86188
84495	86189	if( aiFree ){
		@@ -87178,10 +88872,13 @@
87178	88872	int (strnicmp)(const char,const char*,int);
87179	88873	int (unlock_notify)(sqlite3,void()(void,int),void);
87180	88874	int (wal_autocheckpoint)(sqlite3,int);
87181	88875	int (wal_checkpoint)(sqlite3,const char*);
87182	88876	void (wal_hook)(sqlite3,int()(void,sqlite3,const char,int),void);
	88877	+ int (blob_reopen)(sqlite3_blob,sqlite3_int64);
	88878	+ int (vtab_config)(sqlite3,int op,...);
	88879	+ int (vtab_on_conflict)(sqlite3);
87183	88880	};
87184	88881
87185	88882	/*
87186	88883	** The following macros redefine the API routines so that they are
87187	88884	** redirected throught the global sqlite3_api structure.
		@@ -87378,19 +89075,23 @@
87378	89075	#define sqlite3_strnicmp sqlite3_api->strnicmp
87379	89076	#define sqlite3_unlock_notify sqlite3_api->unlock_notify
87380	89077	#define sqlite3_wal_autocheckpoint sqlite3_api->wal_autocheckpoint
87381	89078	#define sqlite3_wal_checkpoint sqlite3_api->wal_checkpoint
87382	89079	#define sqlite3_wal_hook sqlite3_api->wal_hook
	89080	+#define sqlite3_blob_reopen sqlite3_api->blob_reopen
	89081	+#define sqlite3_vtab_config sqlite3_api->vtab_config
	89082	+#define sqlite3_vtab_on_conflict sqlite3_api->vtab_on_conflict
87383	89083	#endif /* SQLITE_CORE */
87384	89084
87385	89085	#define SQLITE_EXTENSION_INIT1 const sqlite3_api_routines *sqlite3_api = 0;
87386	89086	#define SQLITE_EXTENSION_INIT2(v) sqlite3_api = v;
87387	89087
87388	89088	#endif /* _SQLITE3EXT_H_ */
87389	89089
87390	89090	/************ End of sqlite3ext.h ****************************************/
87391	89091	/************ Continuing where we left off in loadext.c ******************/
	89092	+/* #include <string.h> */
87392	89093
87393	89094	#ifndef SQLITE_OMIT_LOAD_EXTENSION
87394	89095
87395	89096	/*
87396	89097	** Some API routines are omitted when various features are
		@@ -87452,10 +89153,12 @@
87452	89153
87453	89154	#ifdef SQLITE_OMIT_VIRTUALTABLE
87454	89155	# define sqlite3_create_module 0
87455	89156	# define sqlite3_create_module_v2 0
87456	89157	# define sqlite3_declare_vtab 0
	89158	+# define sqlite3_vtab_config 0
	89159	+# define sqlite3_vtab_on_conflict 0
87457	89160	#endif
87458	89161
87459	89162	#ifdef SQLITE_OMIT_SHARED_CACHE
87460	89163	# define sqlite3_enable_shared_cache 0
87461	89164	#endif
		@@ -87475,10 +89178,11 @@
87475	89178	#define sqlite3_blob_bytes 0
87476	89179	#define sqlite3_blob_close 0
87477	89180	#define sqlite3_blob_open 0
87478	89181	#define sqlite3_blob_read 0
87479	89182	#define sqlite3_blob_write 0
	89183	+#define sqlite3_blob_reopen 0
87480	89184	#endif
87481	89185
87482	89186	/*
87483	89187	** The following structure contains pointers to all SQLite API routines.
87484	89188	** A pointer to this structure is passed into extensions when they are
		@@ -87740,10 +89444,13 @@
87740	89444	#else
87741	89445	0,
87742	89446	0,
87743	89447	0,
87744	89448	#endif
	89449	+ sqlite3_blob_reopen,
	89450	+ sqlite3_vtab_config,
	89451	+ sqlite3_vtab_on_conflict,
87745	89452	};
87746	89453
87747	89454	/*
87748	89455	** Attempt to load an SQLite extension library contained in the file
87749	89456	** zFile. The entry point is zProc. zProc may be 0 in which case a
		@@ -94129,10 +95836,11 @@
94129	95836	Expr pHaving; / The HAVING clause. May be NULL */
94130	95837	int isDistinct; /* True if the DISTINCT keyword is present */
94131	95838	int distinct; /* Table to use for the distinct set */
94132	95839	int rc = 1; /* Value to return from this function */
94133	95840	int addrSortIndex; /* Address of an OP_OpenEphemeral instruction */
	95841	+ int addrDistinctIndex; /* Address of an OP_OpenEphemeral instruction */
94134	95842	AggInfo sAggInfo; /* Information used by aggregate queries */
94135	95843	int iEnd; /* Address of the end of the query */
94136	95844	sqlite3 db; / The database connection */
94137	95845
94138	95846	#ifndef SQLITE_OMIT_EXPLAIN
		@@ -94255,20 +95963,10 @@
94255	95963	explainSetInteger(pParse->iSelectId, iRestoreSelectId);
94256	95964	return rc;
94257	95965	}
94258	95966	#endif
94259	95967
94260		- /* If possible, rewrite the query to use GROUP BY instead of DISTINCT.
94261		- ** GROUP BY might use an index, DISTINCT never does.
94262		- */
94263		- assert( p->pGroupBy==0 \|\| (p->selFlags & SF_Aggregate)!=0 );
94264		- if( (p->selFlags & (SF_Distinct\|SF_Aggregate))==SF_Distinct ){
94265		- p->pGroupBy = sqlite3ExprListDup(db, p->pEList, 0);
94266		- pGroupBy = p->pGroupBy;
94267		- p->selFlags &= ~SF_Distinct;
94268		- }
94269		-
94270	95968	/* If there is both a GROUP BY and an ORDER BY clause and they are
94271	95969	** identical, then disable the ORDER BY clause since the GROUP BY
94272	95970	** will cause elements to come out in the correct order. This is
94273	95971	** an optimization - the correct answer should result regardless.
94274	95972	** Use the SQLITE_GroupByOrder flag with SQLITE_TESTCTRL_OPTIMIZER
		@@ -94276,10 +95974,34 @@
94276	95974	*/
94277	95975	if( sqlite3ExprListCompare(p->pGroupBy, pOrderBy)==0
94278	95976	&& (db->flags & SQLITE_GroupByOrder)==0 ){
94279	95977	pOrderBy = 0;
94280	95978	}
	95979	+
	95980	+ /* If the query is DISTINCT with an ORDER BY but is not an aggregate, and
	95981	+ ** if the select-list is the same as the ORDER BY list, then this query
	95982	+ ** can be rewritten as a GROUP BY. In other words, this:
	95983	+ **
	95984	+ ** SELECT DISTINCT xyz FROM ... ORDER BY xyz
	95985	+ **
	95986	+ ** is transformed to:
	95987	+ **
	95988	+ ** SELECT xyz FROM ... GROUP BY xyz
	95989	+ **
	95990	+ ** The second form is preferred as a single index (or temp-table) may be
	95991	+ ** used for both the ORDER BY and DISTINCT processing. As originally
	95992	+ ** written the query must use a temp-table for at least one of the ORDER
	95993	+ ** BY and DISTINCT, and an index or separate temp-table for the other.
	95994	+ */
	95995	+ if( (p->selFlags & (SF_Distinct\|SF_Aggregate))==SF_Distinct
	95996	+ && sqlite3ExprListCompare(pOrderBy, p->pEList)==0
	95997	+ ){
	95998	+ p->selFlags &= ~SF_Distinct;
	95999	+ p->pGroupBy = sqlite3ExprListDup(db, p->pEList, 0);
	96000	+ pGroupBy = p->pGroupBy;
	96001	+ pOrderBy = 0;
	96002	+ }
94281	96003
94282	96004	/* If there is an ORDER BY clause, then this sorting
94283	96005	** index might end up being unused if the data can be
94284	96006	** extracted in pre-sorted order. If that is the case, then the
94285	96007	** OP_OpenEphemeral instruction will be changed to an OP_Noop once
		@@ -94312,26 +96034,25 @@
94312	96034
94313	96035	/* Open a virtual index to use for the distinct set.
94314	96036	*/
94315	96037	if( p->selFlags & SF_Distinct ){
94316	96038	KeyInfo *pKeyInfo;
94317		- assert( isAgg \|\| pGroupBy );
94318	96039	distinct = pParse->nTab++;
94319	96040	pKeyInfo = keyInfoFromExprList(pParse, p->pEList);
94320		- sqlite3VdbeAddOp4(v, OP_OpenEphemeral, distinct, 0, 0,
94321		- (char*)pKeyInfo, P4_KEYINFO_HANDOFF);
	96041	+ addrDistinctIndex = sqlite3VdbeAddOp4(v, OP_OpenEphemeral, distinct, 0, 0,
	96042	+ (char*)pKeyInfo, P4_KEYINFO_HANDOFF);
94322	96043	sqlite3VdbeChangeP5(v, BTREE_UNORDERED);
94323	96044	}else{
94324		- distinct = -1;
	96045	+ distinct = addrDistinctIndex = -1;
94325	96046	}
94326	96047
94327	96048	/* Aggregate and non-aggregate queries are handled differently */
94328	96049	if( !isAgg && pGroupBy==0 ){
94329		- /* This case is for non-aggregate queries
94330		- ** Begin the database scan
94331		- */
94332		- pWInfo = sqlite3WhereBegin(pParse, pTabList, pWhere, &pOrderBy, 0);
	96050	+ ExprList *pDist = (isDistinct ? p->pEList : 0);
	96051	+
	96052	+ /* Begin the database scan. */
	96053	+ pWInfo = sqlite3WhereBegin(pParse, pTabList, pWhere, &pOrderBy, pDist, 0);
94333	96054	if( pWInfo==0 ) goto select_end;
94334	96055	if( pWInfo->nRowOut < p->nSelectRow ) p->nSelectRow = pWInfo->nRowOut;
94335	96056
94336	96057	/* If sorting index that was created by a prior OP_OpenEphemeral
94337	96058	** instruction ended up not being needed, then change the OP_OpenEphemeral
		@@ -94340,14 +96061,56 @@
94340	96061	if( addrSortIndex>=0 && pOrderBy==0 ){
94341	96062	sqlite3VdbeChangeToNoop(v, addrSortIndex, 1);
94342	96063	p->addrOpenEphm[2] = -1;
94343	96064	}
94344	96065
94345		- /* Use the standard inner loop
94346		- */
94347		- assert(!isDistinct);
94348		- selectInnerLoop(pParse, p, pEList, 0, 0, pOrderBy, -1, pDest,
	96066	+ if( pWInfo->eDistinct ){
	96067	+ VdbeOp pOp; / No longer required OpenEphemeral instr. */
	96068	+
	96069	+ assert( addrDistinctIndex>0 );
	96070	+ pOp = sqlite3VdbeGetOp(v, addrDistinctIndex);
	96071	+
	96072	+ assert( isDistinct );
	96073	+ assert( pWInfo->eDistinct==WHERE_DISTINCT_ORDERED
	96074	+ \|\| pWInfo->eDistinct==WHERE_DISTINCT_UNIQUE
	96075	+ );
	96076	+ distinct = -1;
	96077	+ if( pWInfo->eDistinct==WHERE_DISTINCT_ORDERED ){
	96078	+ int iJump;
	96079	+ int iExpr;
	96080	+ int iFlag = ++pParse->nMem;
	96081	+ int iBase = pParse->nMem+1;
	96082	+ int iBase2 = iBase + pEList->nExpr;
	96083	+ pParse->nMem += (pEList->nExpr*2);
	96084	+
	96085	+ /* Change the OP_OpenEphemeral coded earlier to an OP_Integer. The
	96086	+ ** OP_Integer initializes the "first row" flag. */
	96087	+ pOp->opcode = OP_Integer;
	96088	+ pOp->p1 = 1;
	96089	+ pOp->p2 = iFlag;
	96090	+
	96091	+ sqlite3ExprCodeExprList(pParse, pEList, iBase, 1);
	96092	+ iJump = sqlite3VdbeCurrentAddr(v) + 1 + pEList->nExpr + 1 + 1;
	96093	+ sqlite3VdbeAddOp2(v, OP_If, iFlag, iJump-1);
	96094	+ for(iExpr=0; iExpr<pEList->nExpr; iExpr++){
	96095	+ CollSeq *pColl = sqlite3ExprCollSeq(pParse, pEList->a[iExpr].pExpr);
	96096	+ sqlite3VdbeAddOp3(v, OP_Ne, iBase+iExpr, iJump, iBase2+iExpr);
	96097	+ sqlite3VdbeChangeP4(v, -1, (const char *)pColl, P4_COLLSEQ);
	96098	+ sqlite3VdbeChangeP5(v, SQLITE_NULLEQ);
	96099	+ }
	96100	+ sqlite3VdbeAddOp2(v, OP_Goto, 0, pWInfo->iContinue);
	96101	+
	96102	+ sqlite3VdbeAddOp2(v, OP_Integer, 0, iFlag);
	96103	+ assert( sqlite3VdbeCurrentAddr(v)==iJump );
	96104	+ sqlite3VdbeAddOp3(v, OP_Move, iBase, iBase2, pEList->nExpr);
	96105	+ }else{
	96106	+ pOp->opcode = OP_Noop;
	96107	+ }
	96108	+ }
	96109	+
	96110	+ /* Use the standard inner loop. */
	96111	+ selectInnerLoop(pParse, p, pEList, 0, 0, pOrderBy, distinct, pDest,
94349	96112	pWInfo->iContinue, pWInfo->iBreak);
94350	96113
94351	96114	/* End the database scan loop.
94352	96115	*/
94353	96116	sqlite3WhereEnd(pWInfo);
		@@ -94453,11 +96216,11 @@
94453	96216	** This might involve two separate loops with an OP_Sort in between, or
94454	96217	** it might be a single loop that uses an index to extract information
94455	96218	** in the right order to begin with.
94456	96219	*/
94457	96220	sqlite3VdbeAddOp2(v, OP_Gosub, regReset, addrReset);
94458		- pWInfo = sqlite3WhereBegin(pParse, pTabList, pWhere, &pGroupBy, 0);
	96221	+ pWInfo = sqlite3WhereBegin(pParse, pTabList, pWhere, &pGroupBy, 0, 0);
94459	96222	if( pWInfo==0 ) goto select_end;
94460	96223	if( pGroupBy==0 ){
94461	96224	/* The optimizer is able to deliver rows in group by order so
94462	96225	** we do not have to sort. The OP_OpenEphemeral table will be
94463	96226	** cancelled later because we still need to use the pKeyInfo
		@@ -94715,11 +96478,11 @@
94715	96478	/* This case runs if the aggregate has no GROUP BY clause. The
94716	96479	** processing is much simpler since there is only a single row
94717	96480	** of output.
94718	96481	*/
94719	96482	resetAccumulator(pParse, &sAggInfo);
94720		- pWInfo = sqlite3WhereBegin(pParse, pTabList, pWhere, &pMinMax, flag);
	96483	+ pWInfo = sqlite3WhereBegin(pParse, pTabList, pWhere, &pMinMax, 0, flag);
94721	96484	if( pWInfo==0 ){
94722	96485	sqlite3ExprListDelete(db, pDel);
94723	96486	goto select_end;
94724	96487	}
94725	96488	updateAccumulator(pParse, &sAggInfo);
		@@ -94894,10 +96657,12 @@
94894	96657	** interface routine of sqlite3_exec().
94895	96658	**
94896	96659	** These routines are in a separate files so that they will not be linked
94897	96660	** if they are not used.
94898	96661	*/
	96662	+/* #include <stdlib.h> */
	96663	+/* #include <string.h> */
94899	96664
94900	96665	#ifndef SQLITE_OMIT_GET_TABLE
94901	96666
94902	96667	/*
94903	96668	** This structure is used to pass data from sqlite3_get_table() through
		@@ -95191,19 +96956,32 @@
95191	96956	iDb = sqlite3TwoPartName(pParse, pName1, pName2, &pName);
95192	96957	if( iDb<0 ){
95193	96958	goto trigger_cleanup;
95194	96959	}
95195	96960	}
	96961	+ if( !pTableName \|\| db->mallocFailed ){
	96962	+ goto trigger_cleanup;
	96963	+ }
	96964	+
	96965	+ /* A long-standing parser bug is that this syntax was allowed:
	96966	+ **
	96967	+ ** CREATE TRIGGER attached.demo AFTER INSERT ON attached.tab ....
	96968	+ ** ^^^^^^^^
	96969	+ **
	96970	+ ** To maintain backwards compatibility, ignore the database
	96971	+ ** name on pTableName if we are reparsing our of SQLITE_MASTER.
	96972	+ */
	96973	+ if( db->init.busy && iDb!=1 ){
	96974	+ sqlite3DbFree(db, pTableName->a[0].zDatabase);
	96975	+ pTableName->a[0].zDatabase = 0;
	96976	+ }
95196	96977
95197	96978	/* If the trigger name was unqualified, and the table is a temp table,
95198	96979	** then set iDb to 1 to create the trigger in the temporary database.
95199	96980	** If sqlite3SrcListLookup() returns 0, indicating the table does not
95200	96981	** exist, the error is caught by the block below.
95201	96982	*/
95202		- if( !pTableName \|\| db->mallocFailed ){
95203		- goto trigger_cleanup;
95204		- }
95205	96983	pTab = sqlite3SrcListLookup(pParse, pTableName);
95206	96984	if( db->init.busy==0 && pName2->n==0 && pTab
95207	96985	&& pTab->pSchema==db->aDb[1].pSchema ){
95208	96986	iDb = 1;
95209	96987	}
		@@ -96497,11 +98275,13 @@
96497	98275	}
96498	98276
96499	98277	/* Begin the database scan
96500	98278	*/
96501	98279	sqlite3VdbeAddOp2(v, OP_Null, 0, regOldRowid);
96502		- pWInfo = sqlite3WhereBegin(pParse, pTabList, pWhere,0, WHERE_ONEPASS_DESIRED);
	98280	+ pWInfo = sqlite3WhereBegin(
	98281	+ pParse, pTabList, pWhere, 0, 0, WHERE_ONEPASS_DESIRED
	98282	+ );
96503	98283	if( pWInfo==0 ) goto update_cleanup;
96504	98284	okOnePass = pWInfo->okOnePass;
96505	98285
96506	98286	/* Remember the rowid of every item to be updated.
96507	98287	*/
		@@ -98523,10 +100303,11 @@
98523	100303	#define WHERE_REVERSE 0x02000000 /* Scan in reverse order */
98524	100304	#define WHERE_UNIQUE 0x04000000 /* Selects no more than one row */
98525	100305	#define WHERE_VIRTUALTABLE 0x08000000 /* Use virtual-table processing */
98526	100306	#define WHERE_MULTI_OR 0x10000000 /* OR using multiple indices */
98527	100307	#define WHERE_TEMP_INDEX 0x20000000 /* Uses an ephemeral index */
	100308	+#define WHERE_DISTINCT 0x40000000 /* Correct order for DISTINCT */
98528	100309
98529	100310	/*
98530	100311	** Initialize a preallocated WhereClause structure.
98531	100312	*/
98532	100313	static void whereClauseInit(
		@@ -99667,10 +101448,166 @@
99667	101448	}
99668	101449	}
99669	101450	return 0;
99670	101451	}
99671	101452
	101453	+/*
	101454	+** This function searches the expression list passed as the second argument
	101455	+** for an expression of type TK_COLUMN that refers to the same column and
	101456	+** uses the same collation sequence as the iCol'th column of index pIdx.
	101457	+** Argument iBase is the cursor number used for the table that pIdx refers
	101458	+** to.
	101459	+**
	101460	+** If such an expression is found, its index in pList->a[] is returned. If
	101461	+** no expression is found, -1 is returned.
	101462	+*/
	101463	+static int findIndexCol(
	101464	+ Parse pParse, / Parse context */
	101465	+ ExprList pList, / Expression list to search */
	101466	+ int iBase, /* Cursor for table associated with pIdx */
	101467	+ Index pIdx, / Index to match column of */
	101468	+ int iCol /* Column of index to match */
	101469	+){
	101470	+ int i;
	101471	+ const char *zColl = pIdx->azColl[iCol];
	101472	+
	101473	+ for(i=0; i<pList->nExpr; i++){
	101474	+ Expr *p = pList->a[i].pExpr;
	101475	+ if( p->op==TK_COLUMN
	101476	+ && p->iColumn==pIdx->aiColumn[iCol]
	101477	+ && p->iTable==iBase
	101478	+ ){
	101479	+ CollSeq *pColl = sqlite3ExprCollSeq(pParse, p);
	101480	+ if( ALWAYS(pColl) && 0==sqlite3StrICmp(pColl->zName, zColl) ){
	101481	+ return i;
	101482	+ }
	101483	+ }
	101484	+ }
	101485	+
	101486	+ return -1;
	101487	+}
	101488	+
	101489	+/*
	101490	+** This routine determines if pIdx can be used to assist in processing a
	101491	+** DISTINCT qualifier. In other words, it tests whether or not using this
	101492	+** index for the outer loop guarantees that rows with equal values for
	101493	+** all expressions in the pDistinct list are delivered grouped together.
	101494	+**
	101495	+** For example, the query
	101496	+**
	101497	+** SELECT DISTINCT a, b, c FROM tbl WHERE a = ?
	101498	+**
	101499	+** can benefit from any index on columns "b" and "c".
	101500	+*/
	101501	+static int isDistinctIndex(
	101502	+ Parse pParse, / Parsing context */
	101503	+ WhereClause pWC, / The WHERE clause */
	101504	+ Index pIdx, / The index being considered */
	101505	+ int base, /* Cursor number for the table pIdx is on */
	101506	+ ExprList pDistinct, / The DISTINCT expressions */
	101507	+ int nEqCol /* Number of index columns with == */
	101508	+){
	101509	+ Bitmask mask = 0; /* Mask of unaccounted for pDistinct exprs */
	101510	+ int i; /* Iterator variable */
	101511	+
	101512	+ if( pIdx->zName==0 \|\| pDistinct==0 \|\| pDistinct->nExpr>=BMS ) return 0;
	101513	+ testcase( pDistinct->nExpr==BMS-1 );
	101514	+
	101515	+ /* Loop through all the expressions in the distinct list. If any of them
	101516	+ ** are not simple column references, return early. Otherwise, test if the
	101517	+ ** WHERE clause contains a "col=X" clause. If it does, the expression
	101518	+ ** can be ignored. If it does not, and the column does not belong to the
	101519	+ ** same table as index pIdx, return early. Finally, if there is no
	101520	+ ** matching "col=X" expression and the column is on the same table as pIdx,
	101521	+ ** set the corresponding bit in variable mask.
	101522	+ */
	101523	+ for(i=0; i<pDistinct->nExpr; i++){
	101524	+ WhereTerm *pTerm;
	101525	+ Expr *p = pDistinct->a[i].pExpr;
	101526	+ if( p->op!=TK_COLUMN ) return 0;
	101527	+ pTerm = findTerm(pWC, p->iTable, p->iColumn, ~(Bitmask)0, WO_EQ, 0);
	101528	+ if( pTerm ){
	101529	+ Expr *pX = pTerm->pExpr;
	101530	+ CollSeq *p1 = sqlite3BinaryCompareCollSeq(pParse, pX->pLeft, pX->pRight);
	101531	+ CollSeq *p2 = sqlite3ExprCollSeq(pParse, p);
	101532	+ if( p1==p2 ) continue;
	101533	+ }
	101534	+ if( p->iTable!=base ) return 0;
	101535	+ mask \|= (((Bitmask)1) << i);
	101536	+ }
	101537	+
	101538	+ for(i=nEqCol; mask && i<pIdx->nColumn; i++){
	101539	+ int iExpr = findIndexCol(pParse, pDistinct, base, pIdx, i);
	101540	+ if( iExpr<0 ) break;
	101541	+ mask &= ~(((Bitmask)1) << iExpr);
	101542	+ }
	101543	+
	101544	+ return (mask==0);
	101545	+}
	101546	+
	101547	+
	101548	+/*
	101549	+** Return true if the DISTINCT expression-list passed as the third argument
	101550	+** is redundant. A DISTINCT list is redundant if the database contains a
	101551	+** UNIQUE index that guarantees that the result of the query will be distinct
	101552	+** anyway.
	101553	+*/
	101554	+static int isDistinctRedundant(
	101555	+ Parse *pParse,
	101556	+ SrcList *pTabList,
	101557	+ WhereClause *pWC,
	101558	+ ExprList *pDistinct
	101559	+){
	101560	+ Table *pTab;
	101561	+ Index *pIdx;
	101562	+ int i;
	101563	+ int iBase;
	101564	+
	101565	+ /* If there is more than one table or sub-select in the FROM clause of
	101566	+ ** this query, then it will not be possible to show that the DISTINCT
	101567	+ ** clause is redundant. */
	101568	+ if( pTabList->nSrc!=1 ) return 0;
	101569	+ iBase = pTabList->a[0].iCursor;
	101570	+ pTab = pTabList->a[0].pTab;
	101571	+
	101572	+ /* If any of the expressions is an IPK column on table iBase, then return
	101573	+ ** true. Note: The (p->iTable==iBase) part of this test may be false if the
	101574	+ ** current SELECT is a correlated sub-query.
	101575	+ */
	101576	+ for(i=0; i<pDistinct->nExpr; i++){
	101577	+ Expr *p = pDistinct->a[i].pExpr;
	101578	+ if( p->op==TK_COLUMN && p->iTable==iBase && p->iColumn<0 ) return 1;
	101579	+ }
	101580	+
	101581	+ /* Loop through all indices on the table, checking each to see if it makes
	101582	+ ** the DISTINCT qualifier redundant. It does so if:
	101583	+ **
	101584	+ ** 1. The index is itself UNIQUE, and
	101585	+ **
	101586	+ ** 2. All of the columns in the index are either part of the pDistinct
	101587	+ ** list, or else the WHERE clause contains a term of the form "col=X",
	101588	+ ** where X is a constant value. The collation sequences of the
	101589	+ ** comparison and select-list expressions must match those of the index.
	101590	+ */
	101591	+ for(pIdx=pTab->pIndex; pIdx; pIdx=pIdx->pNext){
	101592	+ if( pIdx->onError==OE_None ) continue;
	101593	+ for(i=0; i<pIdx->nColumn; i++){
	101594	+ int iCol = pIdx->aiColumn[i];
	101595	+ if( 0==findTerm(pWC, iBase, iCol, ~(Bitmask)0, WO_EQ, pIdx)
	101596	+ && 0>findIndexCol(pParse, pDistinct, iBase, pIdx, i)
	101597	+ ){
	101598	+ break;
	101599	+ }
	101600	+ }
	101601	+ if( i==pIdx->nColumn ){
	101602	+ /* This index implies that the DISTINCT qualifier is redundant. */
	101603	+ return 1;
	101604	+ }
	101605	+ }
	101606	+
	101607	+ return 0;
	101608	+}
99672	101609
99673	101610	/*
99674	101611	** This routine decides if pIdx can be used to satisfy the ORDER BY
99675	101612	** clause. If it can, it returns 1. If pIdx cannot satisfy the
99676	101613	** ORDER BY clause, this routine returns 0.
		@@ -99703,11 +101640,14 @@
99703	101640	int sortOrder = 0; /* XOR of index and ORDER BY sort direction */
99704	101641	int nTerm; /* Number of ORDER BY terms */
99705	101642	struct ExprList_item pTerm; / A term of the ORDER BY clause */
99706	101643	sqlite3 *db = pParse->db;
99707	101644
99708		- assert( pOrderBy!=0 );
	101645	+ if( !pOrderBy ) return 0;
	101646	+ if( wsFlags & WHERE_COLUMN_IN ) return 0;
	101647	+ if( pIdx->bUnordered ) return 0;
	101648	+
99709	101649	nTerm = pOrderBy->nExpr;
99710	101650	assert( nTerm>0 );
99711	101651
99712	101652	/* Argument pIdx must either point to a 'real' named index structure,
99713	101653	** or an index structure allocated on the stack by bestBtreeIndex() to
		@@ -100016,10 +101956,14 @@
100016	101956	double costTempIdx; /* per-query cost of the transient index */
100017	101957	WhereTerm pTerm; / A single term of the WHERE clause */
100018	101958	WhereTerm pWCEnd; / End of pWC->a[] */
100019	101959	Table pTable; / Table tht might be indexed */
100020	101960
	101961	+ if( pParse->nQueryLoop<=(double)1 ){
	101962	+ /* There is no point in building an automatic index for a single scan */
	101963	+ return;
	101964	+ }
100021	101965	if( (pParse->db->flags & SQLITE_AutoIndex)==0 ){
100022	101966	/* Automatic indices are disabled at run-time */
100023	101967	return;
100024	101968	}
100025	101969	if( (pCost->plan.wsFlags & WHERE_NOT_FULLSCAN)!=0 ){
		@@ -100027,10 +101971,14 @@
100027	101971	return;
100028	101972	}
100029	101973	if( pSrc->notIndexed ){
100030	101974	/* The NOT INDEXED clause appears in the SQL. */
100031	101975	return;
	101976	+ }
	101977	+ if( pSrc->isCorrelated ){
	101978	+ /* The source is a correlated sub-query. No point in indexing it. */
	101979	+ return;
100032	101980	}
100033	101981
100034	101982	assert( pParse->nQueryLoop >= (double)1 );
100035	101983	pTable = pSrc->pTab;
100036	101984	nTableRow = pTable->nRowEst;
		@@ -100244,10 +102192,11 @@
100244	102192	if( pTerm->leftCursor != pSrc->iCursor ) continue;
100245	102193	assert( (pTerm->eOperator&(pTerm->eOperator-1))==0 );
100246	102194	testcase( pTerm->eOperator==WO_IN );
100247	102195	testcase( pTerm->eOperator==WO_ISNULL );
100248	102196	if( pTerm->eOperator & (WO_IN\|WO_ISNULL) ) continue;
	102197	+ if( pTerm->wtFlags & TERM_VNULL ) continue;
100249	102198	nTerm++;
100250	102199	}
100251	102200
100252	102201	/* If the ORDER BY clause contains only columns in the current
100253	102202	** virtual table then allocate space for the aOrderBy part of
		@@ -100294,10 +102243,11 @@
100294	102243	if( pTerm->leftCursor != pSrc->iCursor ) continue;
100295	102244	assert( (pTerm->eOperator&(pTerm->eOperator-1))==0 );
100296	102245	testcase( pTerm->eOperator==WO_IN );
100297	102246	testcase( pTerm->eOperator==WO_ISNULL );
100298	102247	if( pTerm->eOperator & (WO_IN\|WO_ISNULL) ) continue;
	102248	+ if( pTerm->wtFlags & TERM_VNULL ) continue;
100299	102249	pIdxCons[j].iColumn = pTerm->u.leftColumn;
100300	102250	pIdxCons[j].iTermOffset = i;
100301	102251	pIdxCons[j].op = (u8)pTerm->eOperator;
100302	102252	/* The direct assignment in the previous line is possible only because
100303	102253	** the WO_ and SQLITE_INDEX_CONSTRAINT_ codes are identical. The
		@@ -100959,10 +102909,11 @@
100959	102909	WhereClause pWC, / The WHERE clause */
100960	102910	struct SrcList_item pSrc, / The FROM clause term to search */
100961	102911	Bitmask notReady, /* Mask of cursors not available for indexing */
100962	102912	Bitmask notValid, /* Cursors not available for any purpose */
100963	102913	ExprList pOrderBy, / The ORDER BY clause */
	102914	+ ExprList pDistinct, / The select-list if query is DISTINCT */
100964	102915	WhereCost pCost / Lowest cost query plan */
100965	102916	){
100966	102917	int iCur = pSrc->iCursor; /* The cursor of the table to be accessed */
100967	102918	Index pProbe; / An index we are evaluating */
100968	102919	Index pIdx; / Copy of pProbe, or zero for IPK index */
		@@ -101099,11 +103050,12 @@
101099	103050	int nEq; /* Number of == or IN terms matching index */
101100	103051	int bInEst = 0; /* True if "x IN (SELECT...)" seen */
101101	103052	int nInMul = 1; /* Number of distinct equalities to lookup */
101102	103053	int estBound = 100; /* Estimated reduction in search space */
101103	103054	int nBound = 0; /* Number of range constraints seen */
101104		- int bSort = 0; /* True if external sort required */
	103055	+ int bSort = !!pOrderBy; /* True if external sort required */
	103056	+ int bDist = !!pDistinct; /* True if index cannot help with DISTINCT */
101105	103057	int bLookup = 0; /* True if not a covering index */
101106	103058	WhereTerm pTerm; / A single term of the WHERE clause */
101107	103059	#ifdef SQLITE_ENABLE_STAT2
101108	103060	WhereTerm pFirstTerm = 0; / First term matching the index */
101109	103061	#endif
		@@ -101163,21 +103115,24 @@
101163	103115
101164	103116	/* If there is an ORDER BY clause and the index being considered will
101165	103117	** naturally scan rows in the required order, set the appropriate flags
101166	103118	** in wsFlags. Otherwise, if there is an ORDER BY clause but the index
101167	103119	** will scan rows in a different order, set the bSort variable. */
101168		- if( pOrderBy ){
101169		- if( (wsFlags & WHERE_COLUMN_IN)==0
101170		- && pProbe->bUnordered==0
101171		- && isSortingIndex(pParse, pWC->pMaskSet, pProbe, iCur, pOrderBy,
101172		- nEq, wsFlags, &rev)
101173		- ){
101174		- wsFlags \|= WHERE_ROWID_RANGE\|WHERE_COLUMN_RANGE\|WHERE_ORDERBY;
101175		- wsFlags \|= (rev ? WHERE_REVERSE : 0);
101176		- }else{
101177		- bSort = 1;
101178		- }
	103120	+ if( isSortingIndex(
	103121	+ pParse, pWC->pMaskSet, pProbe, iCur, pOrderBy, nEq, wsFlags, &rev)
	103122	+ ){
	103123	+ bSort = 0;
	103124	+ wsFlags \|= WHERE_ROWID_RANGE\|WHERE_COLUMN_RANGE\|WHERE_ORDERBY;
	103125	+ wsFlags \|= (rev ? WHERE_REVERSE : 0);
	103126	+ }
	103127	+
	103128	+ /* If there is a DISTINCT qualifier and this index will scan rows in
	103129	+ ** order of the DISTINCT expressions, clear bDist and set the appropriate
	103130	+ ** flags in wsFlags. */
	103131	+ if( isDistinctIndex(pParse, pWC, pProbe, iCur, pDistinct, nEq) ){
	103132	+ bDist = 0;
	103133	+ wsFlags \|= WHERE_ROWID_RANGE\|WHERE_COLUMN_RANGE\|WHERE_DISTINCT;
101179	103134	}
101180	103135
101181	103136	/* If currently calculating the cost of using an index (not the IPK
101182	103137	** index), determine if all required column data may be obtained without
101183	103138	** using the main table (i.e. if the index is a covering
		@@ -101208,16 +103163,17 @@
101208	103163	nRow = aiRowEst[0]/2;
101209	103164	nInMul = (int)(nRow / aiRowEst[nEq]);
101210	103165	}
101211	103166
101212	103167	#ifdef SQLITE_ENABLE_STAT2
101213		- /* If the constraint is of the form x=VALUE and histogram
	103168	+ /* If the constraint is of the form x=VALUE or x IN (E1,E2,...)
	103169	+ ** and we do not think that values of x are unique and if histogram
101214	103170	** data is available for column x, then it might be possible
101215	103171	** to get a better estimate on the number of rows based on
101216	103172	** VALUE and how common that value is according to the histogram.
101217	103173	*/
101218		- if( nRow>(double)1 && nEq==1 && pFirstTerm!=0 ){
	103174	+ if( nRow>(double)1 && nEq==1 && pFirstTerm!=0 && aiRowEst[1]>1 ){
101219	103175	if( pFirstTerm->eOperator & (WO_EQ\|WO_ISNULL) ){
101220	103176	testcase( pFirstTerm->eOperator==WO_EQ );
101221	103177	testcase( pFirstTerm->eOperator==WO_ISNULL );
101222	103178	whereEqualScanEst(pParse, pProbe, pFirstTerm->pExpr->pRight, &nRow);
101223	103179	}else if( pFirstTerm->eOperator==WO_IN && bInEst==0 ){
		@@ -101290,10 +103246,13 @@
101290	103246	** difference and select C of 3.0.
101291	103247	*/
101292	103248	if( bSort ){
101293	103249	cost += nRowestLog(nRow)3;
101294	103250	}
	103251	+ if( bDist ){
	103252	+ cost += nRowestLog(nRow)3;
	103253	+ }
101295	103254
101296	103255	/** Cost of using this index has now been computed **/
101297	103256
101298	103257	/* If there are additional constraints on this table that cannot
101299	103258	** be used with the current index, but which might lower the number
		@@ -101435,11 +103394,11 @@
101435	103394	}
101436	103395	sqlite3DbFree(pParse->db, p);
101437	103396	}else
101438	103397	#endif
101439	103398	{
101440		- bestBtreeIndex(pParse, pWC, pSrc, notReady, notValid, pOrderBy, pCost);
	103399	+ bestBtreeIndex(pParse, pWC, pSrc, notReady, notValid, pOrderBy, 0, pCost);
101441	103400	}
101442	103401	}
101443	103402
101444	103403	/*
101445	103404	** Disable a term in the WHERE clause. Except, do not disable the term
		@@ -102397,11 +104356,11 @@
102397	104356	for(ii=0; ii<pOrWc->nTerm; ii++){
102398	104357	WhereTerm *pOrTerm = &pOrWc->a[ii];
102399	104358	if( pOrTerm->leftCursor==iCur \|\| pOrTerm->eOperator==WO_AND ){
102400	104359	WhereInfo pSubWInfo; / Info for single OR-term scan */
102401	104360	/* Loop through table entries that match term pOrTerm. */
102402		- pSubWInfo = sqlite3WhereBegin(pParse, pOrTab, pOrTerm->pExpr, 0,
	104361	+ pSubWInfo = sqlite3WhereBegin(pParse, pOrTab, pOrTerm->pExpr, 0, 0,
102403	104362	WHERE_OMIT_OPEN \| WHERE_OMIT_CLOSE \|
102404	104363	WHERE_FORCE_TABLE \| WHERE_ONETABLE_ONLY);
102405	104364	if( pSubWInfo ){
102406	104365	explainOneScan(
102407	104366	pParse, pOrTab, &pSubWInfo->a[0], iLevel, pLevel->iFrom, 0
		@@ -102638,10 +104597,11 @@
102638	104597	SQLITE_PRIVATE WhereInfo *sqlite3WhereBegin(
102639	104598	Parse pParse, / The parser context */
102640	104599	SrcList pTabList, / A list of all tables to be scanned */
102641	104600	Expr pWhere, / The WHERE clause */
102642	104601	ExprList *ppOrderBy, / An ORDER BY clause, or NULL */
	104602	+ ExprList pDistinct, / The select-list for DISTINCT queries - or NULL */
102643	104603	u16 wctrlFlags /* One of the WHERE_* flags defined in sqliteInt.h */
102644	104604	){
102645	104605	int i; /* Loop counter */
102646	104606	int nByteWInfo; /* Num. bytes allocated for WhereInfo struct */
102647	104607	int nTabList; /* Number of elements in pTabList */
		@@ -102697,10 +104657,14 @@
102697	104657	pWInfo->iBreak = sqlite3VdbeMakeLabel(v);
102698	104658	pWInfo->pWC = pWC = (WhereClause )&((u8 )pWInfo)[nByteWInfo];
102699	104659	pWInfo->wctrlFlags = wctrlFlags;
102700	104660	pWInfo->savedNQueryLoop = pParse->nQueryLoop;
102701	104661	pMaskSet = (WhereMaskSet*)&pWC[1];
	104662	+
	104663	+ /* Disable the DISTINCT optimization if SQLITE_DistinctOpt is set via
	104664	+ ** sqlite3_test_ctrl(SQLITE_TESTCTRL_OPTIMIZATIONS,...) */
	104665	+ if( db->flags & SQLITE_DistinctOpt ) pDistinct = 0;
102702	104666
102703	104667	/* Split the WHERE clause into separate subexpressions where each
102704	104668	** subexpression is separated by an AND operator.
102705	104669	*/
102706	104670	initMaskSet(pMaskSet);
		@@ -102764,10 +104728,19 @@
102764	104728	*/
102765	104729	exprAnalyzeAll(pTabList, pWC);
102766	104730	if( db->mallocFailed ){
102767	104731	goto whereBeginError;
102768	104732	}
	104733	+
	104734	+ /* Check if the DISTINCT qualifier, if there is one, is redundant.
	104735	+ ** If it is, then set pDistinct to NULL and WhereInfo.eDistinct to
	104736	+ ** WHERE_DISTINCT_UNIQUE to tell the caller to ignore the DISTINCT.
	104737	+ */
	104738	+ if( pDistinct && isDistinctRedundant(pParse, pTabList, pWC, pDistinct) ){
	104739	+ pDistinct = 0;
	104740	+ pWInfo->eDistinct = WHERE_DISTINCT_UNIQUE;
	104741	+ }
102769	104742
102770	104743	/* Chose the best index to use for each table in the FROM clause.
102771	104744	**
102772	104745	** This loop fills in the following fields:
102773	104746	**
		@@ -102848,10 +104821,11 @@
102848	104821	Bitmask mask; /* Mask of tables not yet ready */
102849	104822	for(j=iFrom, pTabItem=&pTabList->a[j]; j<nTabList; j++, pTabItem++){
102850	104823	int doNotReorder; /* True if this table should not be reordered */
102851	104824	WhereCost sCost; /* Cost information from best[Virtual]Index() */
102852	104825	ExprList pOrderBy; / ORDER BY clause for index to optimize */
	104826	+ ExprList pDist; / DISTINCT clause for index to optimize */
102853	104827
102854	104828	doNotReorder = (pTabItem->jointype & (JT_LEFT\|JT_CROSS))!=0;
102855	104829	if( j!=iFrom && doNotReorder ) break;
102856	104830	m = getMask(pMaskSet, pTabItem->iCursor);
102857	104831	if( (m & notReady)==0 ){
		@@ -102858,10 +104832,11 @@
102858	104832	if( j==iFrom ) iFrom++;
102859	104833	continue;
102860	104834	}
102861	104835	mask = (isOptimal ? m : notReady);
102862	104836	pOrderBy = ((i==0 && ppOrderBy )?*ppOrderBy:0);
	104837	+ pDist = (i==0 ? pDistinct : 0);
102863	104838	if( pTabItem->pIndex==0 ) nUnconstrained++;
102864	104839
102865	104840	WHERETRACE(("=== trying table %d with isOptimal=%d ===\n",
102866	104841	j, isOptimal));
102867	104842	assert( pTabItem->pTab );
		@@ -102872,11 +104847,11 @@
102872	104847	&sCost, pp);
102873	104848	}else
102874	104849	#endif
102875	104850	{
102876	104851	bestBtreeIndex(pParse, pWC, pTabItem, mask, notReady, pOrderBy,
102877		- &sCost);
	104852	+ pDist, &sCost);
102878	104853	}
102879	104854	assert( isOptimal \|\| (sCost.used&notReady)==0 );
102880	104855
102881	104856	/* If an INDEXED BY clause is present, then the plan must use that
102882	104857	** index if it uses any index at all */
		@@ -102932,10 +104907,14 @@
102932	104907	WHERETRACE(("*** Optimizer selects table %d for loop %d"
102933	104908	" with cost=%g and nRow=%g\n",
102934	104909	bestJ, pLevel-pWInfo->a, bestPlan.rCost, bestPlan.plan.nRow));
102935	104910	if( (bestPlan.plan.wsFlags & WHERE_ORDERBY)!=0 ){
102936	104911	*ppOrderBy = 0;
	104912	+ }
	104913	+ if( (bestPlan.plan.wsFlags & WHERE_DISTINCT)!=0 ){
	104914	+ assert( pWInfo->eDistinct==0 );
	104915	+ pWInfo->eDistinct = WHERE_DISTINCT_ORDERED;
102937	104916	}
102938	104917	andFlags &= bestPlan.plan.wsFlags;
102939	104918	pLevel->plan = bestPlan.plan;
102940	104919	testcase( bestPlan.plan.wsFlags & WHERE_INDEXED );
102941	104920	testcase( bestPlan.plan.wsFlags & WHERE_TEMP_INDEX );
		@@ -103267,10 +105246,11 @@
103267	105246	** LALR(1) grammar but which are always false in the
103268	105247	** specific grammar used by SQLite.
103269	105248	*/
103270	105249	/* First off, code is included that follows the "include" declaration
103271	105250	** in the input grammar file. */
	105251	+/* #include <stdio.h> */
103272	105252
103273	105253
103274	105254	/*
103275	105255	** Disable all error recovery processing in the parser push-down
103276	105256	** automaton.
		@@ -104127,10 +106107,11 @@
104127	106107	#endif
104128	106108	};
104129	106109	typedef struct yyParser yyParser;
104130	106110
104131	106111	#ifndef NDEBUG
	106112	+/* #include <stdio.h> */
104132	106113	static FILE *yyTraceFILE = 0;
104133	106114	static char *yyTracePrompt = 0;
104134	106115	#endif /* NDEBUG */
104135	106116
104136	106117	#ifndef NDEBUG
		@@ -106702,10 +108683,11 @@
106702	108683	**
106703	108684	** This file contains C code that splits an SQL input string up into
106704	108685	** individual tokens and sends those tokens one-by-one over to the
106705	108686	** parser for analysis.
106706	108687	*/
	108688	+/* #include <stdlib.h> */
106707	108689
106708	108690	/*
106709	108691	** The charMap() macro maps alphabetic characters into their
106710	108692	** lower-case ASCII equivalent. On ASCII machines, this is just
106711	108693	** an upper-to-lower case map. On EBCDIC machines we also need
		@@ -108093,10 +110075,20 @@
108093	110075	memcpy(&y, &x, 8);
108094	110076	assert( sqlite3IsNaN(y) );
108095	110077	}
108096	110078	#endif
108097	110079	#endif
	110080	+
	110081	+ /* Do extra initialization steps requested by the SQLITE_EXTRA_INIT
	110082	+ ** compile-time option.
	110083	+ */
	110084	+#ifdef SQLITE_EXTRA_INIT
	110085	+ if( rc==SQLITE_OK && sqlite3GlobalConfig.isInit ){
	110086	+ int SQLITE_EXTRA_INIT(void);
	110087	+ rc = SQLITE_EXTRA_INIT();
	110088	+ }
	110089	+#endif
108098	110090
108099	110091	return rc;
108100	110092	}
108101	110093
108102	110094	/*
		@@ -110763,10 +112755,21 @@
110763	112755	if( sz ) *ppNew = sqlite3ScratchMalloc(sz);
110764	112756	sqlite3ScratchFree(pFree);
110765	112757	break;
110766	112758	}
110767	112759
	112760	+ /* sqlite3_test_control(SQLITE_TESTCTRL_LOCALTIME_FAULT, int onoff);
	112761	+ **
	112762	+ ** If parameter onoff is non-zero, configure the wrappers so that all
	112763	+ ** subsequent calls to localtime() and variants fail. If onoff is zero,
	112764	+ ** undo this setting.
	112765	+ */
	112766	+ case SQLITE_TESTCTRL_LOCALTIME_FAULT: {
	112767	+ sqlite3GlobalConfig.bLocaltimeFault = va_arg(ap, int);
	112768	+ break;
	112769	+ }
	112770	+
110768	112771	}
110769	112772	va_end(ap);
110770	112773	#endif /* SQLITE_OMIT_BUILTIN_TEST */
110771	112774	return rc;
110772	112775	}
		@@ -111420,16 +113423,10 @@
111420	113423	** TODO(shess) Provide a VACUUM type operation to clear out all
111421	113424	** deletions and duplications. This would basically be a forced merge
111422	113425	** into a single segment.
111423	113426	*/
111424	113427
111425		-#if !defined(SQLITE_CORE) \|\| defined(SQLITE_ENABLE_FTS3)
111426		-
111427		-#if defined(SQLITE_ENABLE_FTS3) && !defined(SQLITE_CORE)
111428		-# define SQLITE_CORE 1
111429		-#endif
111430		-
111431	113428	/************ Include fts3Int.h in the middle of fts3.c ******************/
111432	113429	/************ Begin file fts3Int.h ***************************************/
111433	113430	/*
111434	113431	** 2009 Nov 12
111435	113432	**
		@@ -111441,18 +113438,33 @@
111441	113438	** May you share freely, never taking more than you give.
111442	113439	**
111443	113440	******************************************************************************
111444	113441	**
111445	113442	*/
111446		-
111447	113443	#ifndef _FTSINT_H
111448	113444	#define _FTSINT_H
111449	113445
111450	113446	#if !defined(NDEBUG) && !defined(SQLITE_DEBUG)
111451	113447	# define NDEBUG 1
111452	113448	#endif
111453	113449
	113450	+/*
	113451	+** FTS4 is really an extension for FTS3. It is enabled using the
	113452	+** SQLITE_ENABLE_FTS3 macro. But to avoid confusion we also all
	113453	+** the SQLITE_ENABLE_FTS4 macro to serve as an alisse for SQLITE_ENABLE_FTS3.
	113454	+*/
	113455	+#if defined(SQLITE_ENABLE_FTS4) && !defined(SQLITE_ENABLE_FTS3)
	113456	+# define SQLITE_ENABLE_FTS3
	113457	+#endif
	113458	+
	113459	+#if !defined(SQLITE_CORE) \|\| defined(SQLITE_ENABLE_FTS3)
	113460	+
	113461	+/* If not building as part of the core, include sqlite3ext.h. */
	113462	+#ifndef SQLITE_CORE
	113463	+SQLITE_API extern const sqlite3_api_routines *sqlite3_api;
	113464	+#endif
	113465	+
111454	113466	/************ Include fts3_tokenizer.h in the middle of fts3Int.h ********/
111455	113467	/************ Begin file fts3_tokenizer.h ********************************/
111456	113468	/*
111457	113469	** 2006 July 10
111458	113470	**
		@@ -111940,11 +113952,11 @@
111940	113952	char aDoclist; / List of docids for full-text queries */
111941	113953	int nDoclist; /* Size of buffer at aDoclist */
111942	113954	u8 bDesc; /* True to sort in descending order */
111943	113955	int eEvalmode; /* An FTS3_EVAL_XX constant */
111944	113956	int nRowAvg; /* Average size of database rows, in pages */
111945		- int nDoc; /* Documents in table */
	113957	+ sqlite3_int64 nDoc; /* Documents in table */
111946	113958
111947	113959	int isMatchinfoNeeded; /* True when aMatchinfo[] needs filling in */
111948	113960	u32 aMatchinfo; / Information about most recent match */
111949	113961	int nMatchinfo; /* Number of elements in aMatchinfo[] */
111950	113962	char zMatchinfo; / Matchinfo specification */
		@@ -111981,11 +113993,11 @@
111981	113993
111982	113994	sqlite3_int64 iDocid; /* Current docid (if pList!=0) */
111983	113995	int bFreeList; /* True if pList should be sqlite3_free()d */
111984	113996	char pList; / Pointer to position list following iDocid */
111985	113997	int nList; /* Length of position list */
111986		-} doclist;
	113998	+};
111987	113999
111988	114000	/*
111989	114001	** A "phrase" is a sequence of one or more tokens that must match in
111990	114002	** sequence. A single token is the base case and the most common case.
111991	114003	** For a sequence of tokens contained in double-quotes (i.e. "one two three")
		@@ -111997,19 +114009,19 @@
111997	114009	int isPrefix; /* True if token ends with a "" character /
111998	114010
111999	114011	/* Variables above this point are populated when the expression is
112000	114012	** parsed (by code in fts3_expr.c). Below this point the variables are
112001	114013	** used when evaluating the expression. */
112002		- int bFulltext; /* True if full-text index was used */
112003	114014	Fts3DeferredToken pDeferred; / Deferred token object for this token */
112004	114015	Fts3MultiSegReader pSegcsr; / Segment-reader for this token */
112005	114016	};
112006	114017
112007	114018	struct Fts3Phrase {
112008	114019	/* Cache of doclist for this phrase. */
112009	114020	Fts3Doclist doclist;
112010	114021	int bIncr; /* True if doclist is loaded incrementally */
	114022	+ int iDoclistToken;
112011	114023
112012	114024	/* Variables below this point are populated by fts3_expr.c when parsing
112013	114025	** a MATCH expression. Everything above is part of the evaluation phase.
112014	114026	*/
112015	114027	int nToken; /* Number of tokens in the phrase */
		@@ -112130,10 +114142,11 @@
112130	114142	Fts3SegFilter pFilter; / Pointer to filter object */
112131	114143	char aBuffer; / Buffer to merge doclists in */
112132	114144	int nBuffer; /* Allocated size of aBuffer[] in bytes */
112133	114145
112134	114146	int iColFilter; /* If >=0, filter for this column */
	114147	+ int bRestart;
112135	114148
112136	114149	/* Used by fts3.c only. */
112137	114150	int nCost; /* Cost of running iterator */
112138	114151	int bLookup; /* True if a lookup of a single entry. */
112139	114152
		@@ -112180,43 +114193,49 @@
112180	114193	#endif
112181	114194
112182	114195	/* fts3_aux.c */
112183	114196	SQLITE_PRIVATE int sqlite3Fts3InitAux(sqlite3 *db);
112184	114197
112185		-SQLITE_PRIVATE int sqlite3Fts3TermSegReaderCursor(
112186		- Fts3Cursor pCsr, / Virtual table cursor handle */
112187		- const char zTerm, / Term to query for */
112188		- int nTerm, /* Size of zTerm in bytes */
112189		- int isPrefix, /* True for a prefix search */
112190		- Fts3MultiSegReader *ppSegcsr / OUT: Allocated seg-reader cursor */
112191		-);
112192		-
112193	114198	SQLITE_PRIVATE void sqlite3Fts3EvalPhraseCleanup(Fts3Phrase *);
112194	114199
112195		-SQLITE_PRIVATE int sqlite3Fts3EvalStart(Fts3Cursor , Fts3Expr , int);
112196		-SQLITE_PRIVATE int sqlite3Fts3EvalNext(Fts3Cursor *pCsr);
112197		-
112198	114200	SQLITE_PRIVATE int sqlite3Fts3MsrIncrStart(
112199	114201	Fts3Table, Fts3MultiSegReader, int, const char*, int);
112200	114202	SQLITE_PRIVATE int sqlite3Fts3MsrIncrNext(
112201	114203	Fts3Table , Fts3MultiSegReader , sqlite3_int64 , char , int );
112202	114204	SQLITE_PRIVATE char sqlite3Fts3EvalPhrasePoslist(Fts3Cursor , Fts3Expr *, int iCol);
112203	114205	SQLITE_PRIVATE int sqlite3Fts3MsrOvfl(Fts3Cursor , Fts3MultiSegReader , int *);
	114206	+SQLITE_PRIVATE int sqlite3Fts3MsrIncrRestart(Fts3MultiSegReader *pCsr);
112204	114207
112205	114208	SQLITE_PRIVATE int sqlite3Fts3DeferredTokenList(Fts3DeferredToken , char , int );
112206	114209
112207		-
	114210	+#endif /* !SQLITE_CORE \|\| SQLITE_ENABLE_FTS3 */
112208	114211	#endif /* _FTSINT_H */
112209	114212
112210	114213	/************ End of fts3Int.h *******************************************/
112211	114214	/************ Continuing where we left off in fts3.c *********************/
	114215	+#if !defined(SQLITE_CORE) \|\| defined(SQLITE_ENABLE_FTS3)
112212	114216
	114217	+#if defined(SQLITE_ENABLE_FTS3) && !defined(SQLITE_CORE)
	114218	+# define SQLITE_CORE 1
	114219	+#endif
	114220	+
	114221	+/* #include <assert.h> */
	114222	+/* #include <stdlib.h> */
	114223	+/* #include <stddef.h> */
	114224	+/* #include <stdio.h> */
	114225	+/* #include <string.h> */
	114226	+/* #include <stdarg.h> */
112213	114227
112214	114228	#ifndef SQLITE_CORE
112215	114229	SQLITE_EXTENSION_INIT1
112216	114230	#endif
112217	114231
	114232	+static int fts3EvalNext(Fts3Cursor *pCsr);
	114233	+static int fts3EvalStart(Fts3Cursor *pCsr);
	114234	+static int fts3TermSegReaderCursor(
	114235	+ Fts3Cursor , const char , int, int, Fts3MultiSegReader **);
	114236	+
112218	114237	/*
112219	114238	** Write a 64-bit variable-length integer to memory starting at p[0].
112220	114239	** The length of data written will be between 1 and FTS3_VARINT_MAX bytes.
112221	114240	** The number of bytes written is returned.
112222	114241	*/
		@@ -112545,10 +114564,13 @@
112545	114564	rc = sqlite3_prepare(p->db, zSql, -1, &pStmt, 0);
112546	114565	if( rc==SQLITE_OK ){
112547	114566	sqlite3_step(pStmt);
112548	114567	p->nPgsz = sqlite3_column_int(pStmt, 0);
112549	114568	rc = sqlite3_finalize(pStmt);
	114569	+ }else if( rc==SQLITE_AUTH ){
	114570	+ p->nPgsz = 1024;
	114571	+ rc = SQLITE_OK;
112550	114572	}
112551	114573	}
112552	114574	assert( p->nPgsz>0 \|\| rc!=SQLITE_OK );
112553	114575	sqlite3_free(zSql);
112554	114576	*pRc = rc;
		@@ -112718,31 +114740,60 @@
112718	114740	}
112719	114741	sqlite3_free(zFree);
112720	114742	return zRet;
112721	114743	}
112722	114744
	114745	+/*
	114746	+** This function interprets the string at (*pp) as a non-negative integer
	114747	+** value. It reads the integer and sets *pnOut to the value read, then
	114748	+** sets *pp to point to the byte immediately following the last byte of
	114749	+** the integer value.
	114750	+**
	114751	+** Only decimal digits ('0'..'9') may be part of an integer value.
	114752	+**
	114753	+** If *pp does not being with a decimal digit SQLITE_ERROR is returned and
	114754	+** the output value undefined. Otherwise SQLITE_OK is returned.
	114755	+**
	114756	+** This function is used when parsing the "prefix=" FTS4 parameter.
	114757	+*/
112723	114758	static int fts3GobbleInt(const char *pp, int pnOut){
112724		- const char p = pp;
112725		- int nInt = 0;
	114759	+ const char p = pp; /* Iterator pointer */
	114760	+ int nInt = 0; /* Output value */
	114761	+
112726	114762	for(p=*pp; p[0]>='0' && p[0]<='9'; p++){
112727	114763	nInt = nInt * 10 + (p[0] - '0');
112728	114764	}
112729	114765	if( p==*pp ) return SQLITE_ERROR;
112730	114766	*pnOut = nInt;
112731	114767	*pp = p;
112732	114768	return SQLITE_OK;
112733	114769	}
112734	114770
112735		-
	114771	+/*
	114772	+** This function is called to allocate an array of Fts3Index structures
	114773	+** representing the indexes maintained by the current FTS table. FTS tables
	114774	+** always maintain the main "terms" index, but may also maintain one or
	114775	+** more "prefix" indexes, depending on the value of the "prefix=" parameter
	114776	+** (if any) specified as part of the CREATE VIRTUAL TABLE statement.
	114777	+**
	114778	+** Argument zParam is passed the value of the "prefix=" option if one was
	114779	+** specified, or NULL otherwise.
	114780	+**
	114781	+** If no error occurs, SQLITE_OK is returned and *apIndex set to point to
	114782	+** the allocated array. *pnIndex is set to the number of elements in the
	114783	+** array. If an error does occur, an SQLite error code is returned.
	114784	+**
	114785	+** Regardless of whether or not an error is returned, it is the responsibility
	114786	+** of the caller to call sqlite3_free() on the output array to free it.
	114787	+*/
112736	114788	static int fts3PrefixParameter(
112737	114789	const char zParam, / ABC in prefix=ABC parameter to parse */
112738	114790	int pnIndex, / OUT: size of apIndex[] array /
112739		- struct Fts3Index *apIndex, / OUT: Array of indexes for this table */
112740		- struct Fts3Index *apFree / OUT: Free this with sqlite3_free() */
	114791	+ struct Fts3Index *apIndex / OUT: Array of indexes for this table */
112741	114792	){
112742		- struct Fts3Index *aIndex;
112743		- int nIndex = 1;
	114793	+ struct Fts3Index aIndex; / Allocated array */
	114794	+ int nIndex = 1; /* Number of entries in array */
112744	114795
112745	114796	if( zParam && zParam[0] ){
112746	114797	const char *p;
112747	114798	nIndex++;
112748	114799	for(p=zParam; *p; p++){
		@@ -112749,11 +114800,11 @@
112749	114800	if( *p==',' ) nIndex++;
112750	114801	}
112751	114802	}
112752	114803
112753	114804	aIndex = sqlite3_malloc(sizeof(struct Fts3Index) * nIndex);
112754		- apIndex = apFree = aIndex;
	114805	+ *apIndex = aIndex;
112755	114806	*pnIndex = nIndex;
112756	114807	if( !aIndex ){
112757	114808	return SQLITE_NOMEM;
112758	114809	}
112759	114810
		@@ -112806,12 +114857,11 @@
112806	114857	int isFts4 = (argv[0][3]=='4'); /* True for FTS4, false for FTS3 */
112807	114858	const char *aCol; / Array of column names */
112808	114859	sqlite3_tokenizer pTokenizer = 0; / Tokenizer for this table */
112809	114860
112810	114861	int nIndex; /* Size of aIndex[] array */
112811		- struct Fts3Index aIndex; / Array of indexes for this table */
112812		- struct Fts3Index aFree = 0; / Free this before returning */
	114862	+ struct Fts3Index aIndex = 0; / Array of indexes for this table */
112813	114863
112814	114864	/* The results of parsing supported FTS4 key=value options: */
112815	114865	int bNoDocsize = 0; /* True to omit %_docsize table */
112816	114866	int bDescIdx = 0; /* True to store descending indexes */
112817	114867	char zPrefix = 0; / Prefix parameter value (or NULL) */
		@@ -112944,11 +114994,11 @@
112944	114994	rc = sqlite3Fts3InitTokenizer(pHash, "simple", &pTokenizer, pzErr);
112945	114995	if( rc!=SQLITE_OK ) goto fts3_init_out;
112946	114996	}
112947	114997	assert( pTokenizer );
112948	114998
112949		- rc = fts3PrefixParameter(zPrefix, &nIndex, &aIndex, &aFree);
	114999	+ rc = fts3PrefixParameter(zPrefix, &nIndex, &aIndex);
112950	115000	if( rc==SQLITE_ERROR ){
112951	115001	assert( zPrefix );
112952	115002	*pzErr = sqlite3_mprintf("error parsing prefix parameter: %s", zPrefix);
112953	115003	}
112954	115004	if( rc!=SQLITE_OK ) goto fts3_init_out;
		@@ -112995,11 +115045,11 @@
112995	115045	zCsr += nDb;
112996	115046
112997	115047	/* Fill in the azColumn array */
112998	115048	for(iCol=0; iCol<nCol; iCol++){
112999	115049	char *z;
113000		- int n;
	115050	+ int n = 0;
113001	115051	z = (char *)sqlite3Fts3NextToken(aCol[iCol], &n);
113002	115052	memcpy(zCsr, z, n);
113003	115053	zCsr[n] = '\0';
113004	115054	sqlite3Fts3Dequote(zCsr);
113005	115055	p->azColumn[iCol] = zCsr;
		@@ -113031,11 +115081,11 @@
113031	115081	/* Declare the table schema to SQLite. */
113032	115082	fts3DeclareVtab(&rc, p);
113033	115083
113034	115084	fts3_init_out:
113035	115085	sqlite3_free(zPrefix);
113036		- sqlite3_free(aFree);
	115086	+ sqlite3_free(aIndex);
113037	115087	sqlite3_free(zCompress);
113038	115088	sqlite3_free(zUncompress);
113039	115089	sqlite3_free((void *)aCol);
113040	115090	if( rc!=SQLITE_OK ){
113041	115091	if( p ){
		@@ -113622,12 +115672,10 @@
113622	115672	*pp1 = p1 + 1;
113623	115673	*pp2 = p2 + 1;
113624	115674	}
113625	115675
113626	115676	/*
113627		-** nToken==1 searches for adjacent positions.
113628		-**
113629	115677	** This function is used to merge two position lists into one. When it is
113630	115678	** called, pp1 and pp2 must both point to position lists. A position-list is
113631	115679	** the part of a doclist that follows each document id. For example, if a row
113632	115680	** contains:
113633	115681	**
		@@ -113643,10 +115691,12 @@
113643	115691	** If isSaveLeft is 0, an entry is added to the output position list for
113644	115692	** each position in *pp2 for which there exists one or more positions in
113645	115693	** pp1 so that (pos(pp2)>pos(pp1) && pos(pp2)-pos(*pp1)<=nToken). i.e.
113646	115694	** when the pp1 token appears before the pp2 token, but not more than nToken
113647	115695	** slots before it.
	115696	+**
	115697	+** e.g. nToken==1 searches for adjacent positions.
113648	115698	*/
113649	115699	static int fts3PoslistPhraseMerge(
113650	115700	char *pp, / IN/OUT: Preallocated output buffer */
113651	115701	int nToken, /* Maximum difference in token positions */
113652	115702	int isSaveLeft, /* Save the left position */
		@@ -113809,26 +115859,38 @@
113809	115859
113810	115860	return res;
113811	115861	}
113812	115862
113813	115863	/*
113814		-** A pointer to an instance of this structure is used as the context
113815		-** argument to sqlite3Fts3SegReaderIterate()
	115864	+** An instance of this function is used to merge together the (potentially
	115865	+** large number of) doclists for each term that matches a prefix query.
	115866	+** See function fts3TermSelectMerge() for details.
113816	115867	*/
113817	115868	typedef struct TermSelect TermSelect;
113818	115869	struct TermSelect {
113819		- int isReqPos;
113820		- char aaOutput[16]; / Malloc'd output buffer */
113821		- int anOutput[16]; /* Size of output in bytes */
	115870	+ char aaOutput[16]; / Malloc'd output buffers */
	115871	+ int anOutput[16]; /* Size each output buffer in bytes */
113822	115872	};
113823	115873
113824		-
	115874	+/*
	115875	+** This function is used to read a single varint from a buffer. Parameter
	115876	+** pEnd points 1 byte past the end of the buffer. When this function is
	115877	+** called, if *pp points to pEnd or greater, then the end of the buffer
	115878	+** has been reached. In this case *pp is set to 0 and the function returns.
	115879	+**
	115880	+** If *pp does not point to or past pEnd, then a single varint is read
	115881	+** from pp. pp is then set to point 1 byte past the end of the read varint.
	115882	+**
	115883	+** If bDescIdx is false, the value read is added to *pVal before returning.
	115884	+** If it is true, the value read is subtracted from *pVal before this
	115885	+** function returns.
	115886	+*/
113825	115887	static void fts3GetDeltaVarint3(
113826		- char **pp,
113827		- char *pEnd,
113828		- int bDescIdx,
113829		- sqlite3_int64 *pVal
	115888	+ char *pp, / IN/OUT: Point to read varint from */
	115889	+ char pEnd, / End of buffer */
	115890	+ int bDescIdx, /* True if docids are descending */
	115891	+ sqlite3_int64 pVal / IN/OUT: Integer value */
113830	115892	){
113831	115893	if( *pp>=pEnd ){
113832	115894	*pp = 0;
113833	115895	}else{
113834	115896	sqlite3_int64 iVal;
		@@ -113839,10 +115901,25 @@
113839	115901	*pVal += iVal;
113840	115902	}
113841	115903	}
113842	115904	}
113843	115905
	115906	+/*
	115907	+** This function is used to write a single varint to a buffer. The varint
	115908	+** is written to pp. Before returning, pp is set to point 1 byte past the
	115909	+** end of the value written.
	115910	+**
	115911	+** If *pbFirst is zero when this function is called, the value written to
	115912	+** the buffer is that of parameter iVal.
	115913	+**
	115914	+** If *pbFirst is non-zero when this function is called, then the value
	115915	+** written is either (iVal-piPrev) (if bDescIdx is zero) or (piPrev-iVal)
	115916	+** (if bDescIdx is non-zero).
	115917	+**
	115918	+** Before returning, this function always sets pbFirst to 1 and piPrev
	115919	+** to the value of parameter iVal.
	115920	+*/
113844	115921	static void fts3PutDeltaVarint3(
113845	115922	char *pp, / IN/OUT: Output pointer */
113846	115923	int bDescIdx, /* True for descending docids */
113847	115924	sqlite3_int64 piPrev, / IN/OUT: Previous value written to list */
113848	115925	int pbFirst, / IN/OUT: True after first int written */
		@@ -113859,14 +115936,38 @@
113859	115936	pp += sqlite3Fts3PutVarint(pp, iWrite);
113860	115937	*piPrev = iVal;
113861	115938	*pbFirst = 1;
113862	115939	}
113863	115940
113864		-#define COMPARE_DOCID(i1, i2) ((bDescIdx?-1:1) * (i1-i2))
113865	115941
	115942	+/*
	115943	+** This macro is used by various functions that merge doclists. The two
	115944	+** arguments are 64-bit docid values. If the value of the stack variable
	115945	+** bDescDoclist is 0 when this macro is invoked, then it returns (i1-i2).
	115946	+** Otherwise, (i2-i1).
	115947	+**
	115948	+** Using this makes it easier to write code that can merge doclists that are
	115949	+** sorted in either ascending or descending order.
	115950	+*/
	115951	+#define DOCID_CMP(i1, i2) ((bDescDoclist?-1:1) * (i1-i2))
	115952	+
	115953	+/*
	115954	+** This function does an "OR" merge of two doclists (output contains all
	115955	+** positions contained in either argument doclist). If the docids in the
	115956	+** input doclists are sorted in ascending order, parameter bDescDoclist
	115957	+** should be false. If they are sorted in ascending order, it should be
	115958	+** passed a non-zero value.
	115959	+**
	115960	+** If no error occurs, *paOut is set to point at an sqlite3_malloc'd buffer
	115961	+** containing the output doclist and SQLITE_OK is returned. In this case
	115962	+** *pnOut is set to the number of bytes in the output doclist.
	115963	+**
	115964	+** If an error occurs, an SQLite error code is returned. The output values
	115965	+** are undefined in this case.
	115966	+*/
113866	115967	static int fts3DoclistOrMerge(
113867		- int bDescIdx, /* True if arguments are desc */
	115968	+ int bDescDoclist, /* True if arguments are desc */
113868	115969	char a1, int n1, / First doclist */
113869	115970	char a2, int n2, / Second doclist */
113870	115971	char *paOut, int pnOut /* OUT: Malloc'd doclist */
113871	115972	){
113872	115973	sqlite3_int64 i1 = 0;
		@@ -113887,35 +115988,47 @@
113887	115988
113888	115989	p = aOut;
113889	115990	fts3GetDeltaVarint3(&p1, pEnd1, 0, &i1);
113890	115991	fts3GetDeltaVarint3(&p2, pEnd2, 0, &i2);
113891	115992	while( p1 \|\| p2 ){
113892		- sqlite3_int64 iDiff = COMPARE_DOCID(i1, i2);
	115993	+ sqlite3_int64 iDiff = DOCID_CMP(i1, i2);
113893	115994
113894	115995	if( p2 && p1 && iDiff==0 ){
113895		- fts3PutDeltaVarint3(&p, bDescIdx, &iPrev, &bFirstOut, i1);
	115996	+ fts3PutDeltaVarint3(&p, bDescDoclist, &iPrev, &bFirstOut, i1);
113896	115997	fts3PoslistMerge(&p, &p1, &p2);
113897		- fts3GetDeltaVarint3(&p1, pEnd1, bDescIdx, &i1);
113898		- fts3GetDeltaVarint3(&p2, pEnd2, bDescIdx, &i2);
	115998	+ fts3GetDeltaVarint3(&p1, pEnd1, bDescDoclist, &i1);
	115999	+ fts3GetDeltaVarint3(&p2, pEnd2, bDescDoclist, &i2);
113899	116000	}else if( !p2 \|\| (p1 && iDiff<0) ){
113900		- fts3PutDeltaVarint3(&p, bDescIdx, &iPrev, &bFirstOut, i1);
	116001	+ fts3PutDeltaVarint3(&p, bDescDoclist, &iPrev, &bFirstOut, i1);
113901	116002	fts3PoslistCopy(&p, &p1);
113902		- fts3GetDeltaVarint3(&p1, pEnd1, bDescIdx, &i1);
	116003	+ fts3GetDeltaVarint3(&p1, pEnd1, bDescDoclist, &i1);
113903	116004	}else{
113904		- fts3PutDeltaVarint3(&p, bDescIdx, &iPrev, &bFirstOut, i2);
	116005	+ fts3PutDeltaVarint3(&p, bDescDoclist, &iPrev, &bFirstOut, i2);
113905	116006	fts3PoslistCopy(&p, &p2);
113906		- fts3GetDeltaVarint3(&p2, pEnd2, bDescIdx, &i2);
	116007	+ fts3GetDeltaVarint3(&p2, pEnd2, bDescDoclist, &i2);
113907	116008	}
113908	116009	}
113909	116010
113910	116011	*paOut = aOut;
113911	116012	*pnOut = (p-aOut);
113912	116013	return SQLITE_OK;
113913	116014	}
113914	116015
	116016	+/*
	116017	+** This function does a "phrase" merge of two doclists. In a phrase merge,
	116018	+** the output contains a copy of each position from the right-hand input
	116019	+** doclist for which there is a position in the left-hand input doclist
	116020	+** exactly nDist tokens before it.
	116021	+**
	116022	+** If the docids in the input doclists are sorted in ascending order,
	116023	+** parameter bDescDoclist should be false. If they are sorted in ascending
	116024	+** order, it should be passed a non-zero value.
	116025	+**
	116026	+** The right-hand input doclist is overwritten by this function.
	116027	+*/
113915	116028	static void fts3DoclistPhraseMerge(
113916		- int bDescIdx, /* True if arguments are desc */
	116029	+ int bDescDoclist, /* True if arguments are desc */
113917	116030	int nDist, /* Distance from left to right (1=adjacent) */
113918	116031	char aLeft, int nLeft, / Left doclist */
113919	116032	char aRight, int pnRight /* IN/OUT: Right/output doclist */
113920	116033	){
113921	116034	sqlite3_int64 i1 = 0;
		@@ -113934,30 +116047,30 @@
113934	116047	p = aOut;
113935	116048	fts3GetDeltaVarint3(&p1, pEnd1, 0, &i1);
113936	116049	fts3GetDeltaVarint3(&p2, pEnd2, 0, &i2);
113937	116050
113938	116051	while( p1 && p2 ){
113939		- sqlite3_int64 iDiff = COMPARE_DOCID(i1, i2);
	116052	+ sqlite3_int64 iDiff = DOCID_CMP(i1, i2);
113940	116053	if( iDiff==0 ){
113941	116054	char *pSave = p;
113942	116055	sqlite3_int64 iPrevSave = iPrev;
113943	116056	int bFirstOutSave = bFirstOut;
113944	116057
113945		- fts3PutDeltaVarint3(&p, bDescIdx, &iPrev, &bFirstOut, i1);
	116058	+ fts3PutDeltaVarint3(&p, bDescDoclist, &iPrev, &bFirstOut, i1);
113946	116059	if( 0==fts3PoslistPhraseMerge(&p, nDist, 0, 1, &p1, &p2) ){
113947	116060	p = pSave;
113948	116061	iPrev = iPrevSave;
113949	116062	bFirstOut = bFirstOutSave;
113950	116063	}
113951		- fts3GetDeltaVarint3(&p1, pEnd1, bDescIdx, &i1);
113952		- fts3GetDeltaVarint3(&p2, pEnd2, bDescIdx, &i2);
	116064	+ fts3GetDeltaVarint3(&p1, pEnd1, bDescDoclist, &i1);
	116065	+ fts3GetDeltaVarint3(&p2, pEnd2, bDescDoclist, &i2);
113953	116066	}else if( iDiff<0 ){
113954	116067	fts3PoslistCopy(0, &p1);
113955		- fts3GetDeltaVarint3(&p1, pEnd1, bDescIdx, &i1);
	116068	+ fts3GetDeltaVarint3(&p1, pEnd1, bDescDoclist, &i1);
113956	116069	}else{
113957	116070	fts3PoslistCopy(0, &p2);
113958		- fts3GetDeltaVarint3(&p2, pEnd2, bDescIdx, &i2);
	116071	+ fts3GetDeltaVarint3(&p2, pEnd2, bDescDoclist, &i2);
113959	116072	}
113960	116073	}
113961	116074
113962	116075	*pnRight = p - aOut;
113963	116076	}
		@@ -113970,11 +116083,11 @@
113970	116083	**
113971	116084	** If an OOM error occurs, return SQLITE_NOMEM. In this case it is
113972	116085	** the responsibility of the caller to free any doclists left in the
113973	116086	** TermSelect.aaOutput[] array.
113974	116087	*/
113975		-static int fts3TermSelectMerge(Fts3Table p, TermSelect pTS){
	116088	+static int fts3TermSelectFinishMerge(Fts3Table p, TermSelect pTS){
113976	116089	char *aOut = 0;
113977	116090	int nOut = 0;
113978	116091	int i;
113979	116092
113980	116093	/* Loop through the doclists in the aaOutput[] array. Merge them all
		@@ -114011,28 +116124,29 @@
114011	116124	pTS->anOutput[0] = nOut;
114012	116125	return SQLITE_OK;
114013	116126	}
114014	116127
114015	116128	/*
114016		-** This function is used as the sqlite3Fts3SegReaderIterate() callback when
114017		-** querying the full-text index for a doclist associated with a term or
114018		-** term-prefix.
	116129	+** Merge the doclist aDoclist/nDoclist into the TermSelect object passed
	116130	+** as the first argument. The merge is an "OR" merge (see function
	116131	+** fts3DoclistOrMerge() for details).
	116132	+**
	116133	+** This function is called with the doclist for each term that matches
	116134	+** a queried prefix. It merges all these doclists into one, the doclist
	116135	+** for the specified prefix. Since there can be a very large number of
	116136	+** doclists to merge, the merging is done pair-wise using the TermSelect
	116137	+** object.
	116138	+**
	116139	+** This function returns SQLITE_OK if the merge is successful, or an
	116140	+** SQLite error code (SQLITE_NOMEM) if an error occurs.
114019	116141	*/
114020		-static int fts3TermSelectCb(
114021		- Fts3Table p, / Virtual table object */
114022		- void pContext, / Pointer to TermSelect structure */
114023		- char *zTerm,
114024		- int nTerm,
114025		- char *aDoclist,
114026		- int nDoclist
	116142	+static int fts3TermSelectMerge(
	116143	+ Fts3Table p, / FTS table handle */
	116144	+ TermSelect pTS, / TermSelect object to merge into */
	116145	+ char aDoclist, / Pointer to doclist */
	116146	+ int nDoclist /* Size of aDoclist in bytes */
114027	116147	){
114028		- TermSelect pTS = (TermSelect )pContext;
114029		-
114030		- UNUSED_PARAMETER(p);
114031		- UNUSED_PARAMETER(zTerm);
114032		- UNUSED_PARAMETER(nTerm);
114033		-
114034	116148	if( pTS->aaOutput[0]==0 ){
114035	116149	/* If this is the first term selected, copy the doclist to the output
114036	116150	** buffer using memcpy(). */
114037	116151	pTS->aaOutput[0] = sqlite3_malloc(nDoclist);
114038	116152	pTS->anOutput[0] = nDoclist;
		@@ -114099,23 +116213,30 @@
114099	116213	}
114100	116214	pCsr->apSegment[pCsr->nSegment++] = pNew;
114101	116215	return SQLITE_OK;
114102	116216	}
114103	116217
	116218	+/*
	116219	+** Add seg-reader objects to the Fts3MultiSegReader object passed as the
	116220	+** 8th argument.
	116221	+**
	116222	+** This function returns SQLITE_OK if successful, or an SQLite error code
	116223	+** otherwise.
	116224	+*/
114104	116225	static int fts3SegReaderCursor(
114105	116226	Fts3Table p, / FTS3 table handle */
114106	116227	int iIndex, /* Index to search (from 0 to p->nIndex-1) */
114107	116228	int iLevel, /* Level of segments to scan */
114108	116229	const char zTerm, / Term to query for */
114109	116230	int nTerm, /* Size of zTerm in bytes */
114110	116231	int isPrefix, /* True for a prefix search */
114111	116232	int isScan, /* True to scan from zTerm to EOF */
114112		- Fts3MultiSegReader pCsr / Cursor object to populate */
	116233	+ Fts3MultiSegReader pCsr / Cursor object to populate */
114113	116234	){
114114		- int rc = SQLITE_OK;
114115		- int rc2;
114116		- sqlite3_stmt *pStmt = 0;
	116235	+ int rc = SQLITE_OK; /* Error code */
	116236	+ sqlite3_stmt pStmt = 0; / Statement to iterate through segments */
	116237	+ int rc2; /* Result of sqlite3_reset() */
114117	116238
114118	116239	/* If iLevel is less than 0 and this is not a scan, include a seg-reader
114119	116240	** for the pending-terms. If this is a scan, then this call must be being
114120	116241	** made by an fts4aux module, not an FTS table. In this case calling
114121	116242	** Fts3SegReaderPending might segfault, as the data structures used by
		@@ -114200,28 +116321,46 @@
114200	116321	return fts3SegReaderCursor(
114201	116322	p, iIndex, iLevel, zTerm, nTerm, isPrefix, isScan, pCsr
114202	116323	);
114203	116324	}
114204	116325
	116326	+/*
	116327	+** In addition to its current configuration, have the Fts3MultiSegReader
	116328	+** passed as the 4th argument also scan the doclist for term zTerm/nTerm.
	116329	+**
	116330	+** SQLITE_OK is returned if no error occurs, otherwise an SQLite error code.
	116331	+*/
114205	116332	static int fts3SegReaderCursorAddZero(
114206		- Fts3Table *p,
114207		- const char *zTerm,
114208		- int nTerm,
114209		- Fts3MultiSegReader *pCsr
	116333	+ Fts3Table p, / FTS virtual table handle */
	116334	+ const char zTerm, / Term to scan doclist of */
	116335	+ int nTerm, /* Number of bytes in zTerm */
	116336	+ Fts3MultiSegReader pCsr / Fts3MultiSegReader to modify */
114210	116337	){
114211	116338	return fts3SegReaderCursor(p, 0, FTS3_SEGCURSOR_ALL, zTerm, nTerm, 0, 0,pCsr);
114212	116339	}
114213	116340
114214		-
114215		-SQLITE_PRIVATE int sqlite3Fts3TermSegReaderCursor(
	116341	+/*
	116342	+** Open an Fts3MultiSegReader to scan the doclist for term zTerm/nTerm. Or,
	116343	+** if isPrefix is true, to scan the doclist for all terms for which
	116344	+** zTerm/nTerm is a prefix. If successful, return SQLITE_OK and write
	116345	+** a pointer to the new Fts3MultiSegReader to *ppSegcsr. Otherwise, return
	116346	+** an SQLite error code.
	116347	+**
	116348	+** It is the responsibility of the caller to free this object by eventually
	116349	+** passing it to fts3SegReaderCursorFree()
	116350	+**
	116351	+** SQLITE_OK is returned if no error occurs, otherwise an SQLite error code.
	116352	+** Output parameter *ppSegcsr is set to 0 if an error occurs.
	116353	+*/
	116354	+static int fts3TermSegReaderCursor(
114216	116355	Fts3Cursor pCsr, / Virtual table cursor handle */
114217	116356	const char zTerm, / Term to query for */
114218	116357	int nTerm, /* Size of zTerm in bytes */
114219	116358	int isPrefix, /* True for a prefix search */
114220	116359	Fts3MultiSegReader *ppSegcsr / OUT: Allocated seg-reader cursor */
114221	116360	){
114222		- Fts3MultiSegReader pSegcsr; / Object to allocate and return */
	116361	+ Fts3MultiSegReader pSegcsr; / Object to allocate and return */
114223	116362	int rc = SQLITE_NOMEM; /* Return code */
114224	116363
114225	116364	pSegcsr = sqlite3_malloc(sizeof(Fts3MultiSegReader));
114226	116365	if( pSegcsr ){
114227	116366	int i;
		@@ -114261,62 +116400,53 @@
114261	116400
114262	116401	*ppSegcsr = pSegcsr;
114263	116402	return rc;
114264	116403	}
114265	116404
	116405	+/*
	116406	+** Free an Fts3MultiSegReader allocated by fts3TermSegReaderCursor().
	116407	+*/
114266	116408	static void fts3SegReaderCursorFree(Fts3MultiSegReader *pSegcsr){
114267	116409	sqlite3Fts3SegReaderFinish(pSegcsr);
114268	116410	sqlite3_free(pSegcsr);
114269	116411	}
114270	116412
114271	116413	/*
114272	116414	** This function retreives the doclist for the specified term (or term
114273		-** prefix) from the database.
114274		-**
114275		-** The returned doclist may be in one of two formats, depending on the
114276		-** value of parameter isReqPos. If isReqPos is zero, then the doclist is
114277		-** a sorted list of delta-compressed docids (a bare doclist). If isReqPos
114278		-** is non-zero, then the returned list is in the same format as is stored
114279		-** in the database without the found length specifier at the start of on-disk
114280		-** doclists.
	116415	+** prefix) from the database.
114281	116416	*/
114282	116417	static int fts3TermSelect(
114283	116418	Fts3Table p, / Virtual table handle */
114284	116419	Fts3PhraseToken pTok, / Token to query for */
114285	116420	int iColumn, /* Column to query (or -ve for all columns) */
114286		- int isReqPos, /* True to include position lists in output */
114287	116421	int pnOut, / OUT: Size of buffer at ppOut /
114288	116422	char *ppOut / OUT: Malloced result buffer */
114289	116423	){
114290	116424	int rc; /* Return code */
114291		- Fts3MultiSegReader pSegcsr; / Seg-reader cursor for this term */
114292		- TermSelect tsc; /* Context object for fts3TermSelectCb() */
	116425	+ Fts3MultiSegReader pSegcsr; / Seg-reader cursor for this term */
	116426	+ TermSelect tsc; /* Object for pair-wise doclist merging */
114293	116427	Fts3SegFilter filter; /* Segment term filter configuration */
114294	116428
114295	116429	pSegcsr = pTok->pSegcsr;
114296	116430	memset(&tsc, 0, sizeof(TermSelect));
114297		- tsc.isReqPos = isReqPos;
114298	116431
114299		- filter.flags = FTS3_SEGMENT_IGNORE_EMPTY
	116432	+ filter.flags = FTS3_SEGMENT_IGNORE_EMPTY \| FTS3_SEGMENT_REQUIRE_POS
114300	116433	\| (pTok->isPrefix ? FTS3_SEGMENT_PREFIX : 0)
114301		- \| (isReqPos ? FTS3_SEGMENT_REQUIRE_POS : 0)
114302	116434	\| (iColumn<p->nColumn ? FTS3_SEGMENT_COLUMN_FILTER : 0);
114303	116435	filter.iCol = iColumn;
114304	116436	filter.zTerm = pTok->z;
114305	116437	filter.nTerm = pTok->n;
114306	116438
114307	116439	rc = sqlite3Fts3SegReaderStart(p, pSegcsr, &filter);
114308	116440	while( SQLITE_OK==rc
114309	116441	&& SQLITE_ROW==(rc = sqlite3Fts3SegReaderStep(p, pSegcsr))
114310	116442	){
114311		- rc = fts3TermSelectCb(p, (void *)&tsc,
114312		- pSegcsr->zTerm, pSegcsr->nTerm, pSegcsr->aDoclist, pSegcsr->nDoclist
114313		- );
	116443	+ rc = fts3TermSelectMerge(p, &tsc, pSegcsr->aDoclist, pSegcsr->nDoclist);
114314	116444	}
114315	116445
114316	116446	if( rc==SQLITE_OK ){
114317		- rc = fts3TermSelectMerge(p, &tsc);
	116447	+ rc = fts3TermSelectFinishMerge(p, &tsc);
114318	116448	}
114319	116449	if( rc==SQLITE_OK ){
114320	116450	*ppOut = tsc.aaOutput[0];
114321	116451	*pnOut = tsc.anOutput[0];
114322	116452	}else{
		@@ -114338,28 +116468,19 @@
114338	116468	** If the isPoslist argument is true, then it is assumed that the doclist
114339	116469	** contains a position-list following each docid. Otherwise, it is assumed
114340	116470	** that the doclist is simply a list of docids stored as delta encoded
114341	116471	** varints.
114342	116472	*/
114343		-static int fts3DoclistCountDocids(int isPoslist, char *aList, int nList){
	116473	+static int fts3DoclistCountDocids(char *aList, int nList){
114344	116474	int nDoc = 0; /* Return value */
114345	116475	if( aList ){
114346	116476	char aEnd = &aList[nList]; / Pointer to one byte after EOF */
114347	116477	char p = aList; / Cursor */
114348		- if( !isPoslist ){
114349		- /* The number of docids in the list is the same as the number of
114350		- ** varints. In FTS3 a varint consists of a single byte with the 0x80
114351		- ** bit cleared and zero or more bytes with the 0x80 bit set. So to
114352		- ** count the varints in the buffer, just count the number of bytes
114353		- ** with the 0x80 bit clear. */
114354		- while( p<aEnd ) nDoc += (((*p++)&0x80)==0);
114355		- }else{
114356		- while( p<aEnd ){
114357		- nDoc++;
114358		- while( (p++)&0x80 ); / Skip docid varint */
114359		- fts3PoslistCopy(0, &p); /* Skip over position list */
114360		- }
	116478	+ while( p<aEnd ){
	116479	+ nDoc++;
	116480	+ while( (p++)&0x80 ); / Skip docid varint */
	116481	+ fts3PoslistCopy(0, &p); /* Skip over position list */
114361	116482	}
114362	116483	}
114363	116484
114364	116485	return nDoc;
114365	116486	}
		@@ -114385,11 +116506,11 @@
114385	116506	}else{
114386	116507	pCsr->iPrevId = sqlite3_column_int64(pCsr->pStmt, 0);
114387	116508	rc = SQLITE_OK;
114388	116509	}
114389	116510	}else{
114390		- rc = sqlite3Fts3EvalNext((Fts3Cursor *)pCursor);
	116511	+ rc = fts3EvalNext((Fts3Cursor *)pCursor);
114391	116512	}
114392	116513	assert( ((Fts3Table *)pCsr->base.pVtab)->pSegments==0 );
114393	116514	return rc;
114394	116515	}
114395	116516
		@@ -114462,11 +116583,11 @@
114462	116583	}
114463	116584
114464	116585	rc = sqlite3Fts3ReadLock(p);
114465	116586	if( rc!=SQLITE_OK ) return rc;
114466	116587
114467		- rc = sqlite3Fts3EvalStart(pCsr, pCsr->pExpr, 1);
	116588	+ rc = fts3EvalStart(pCsr);
114468	116589
114469	116590	sqlite3Fts3SegmentsClose(p);
114470	116591	if( rc!=SQLITE_OK ) return rc;
114471	116592	pCsr->pNextId = pCsr->aDoclist;
114472	116593	pCsr->iPrevId = 0;
		@@ -114581,12 +116702,12 @@
114581	116702
114582	116703	/*
114583	116704	** Implementation of xBegin() method. This is a no-op.
114584	116705	*/
114585	116706	static int fts3BeginMethod(sqlite3_vtab *pVtab){
114586		- UNUSED_PARAMETER(pVtab);
114587	116707	TESTONLY( Fts3Table p = (Fts3Table)pVtab );
	116708	+ UNUSED_PARAMETER(pVtab);
114588	116709	assert( p->pSegments==0 );
114589	116710	assert( p->nPendingData==0 );
114590	116711	assert( p->inTransaction!=1 );
114591	116712	TESTONLY( p->inTransaction = 1 );
114592	116713	TESTONLY( p->mxSavepoint = -1; );
		@@ -114597,12 +116718,12 @@
114597	116718	** Implementation of xCommit() method. This is a no-op. The contents of
114598	116719	** the pending-terms hash-table have already been flushed into the database
114599	116720	** by fts3SyncMethod().
114600	116721	*/
114601	116722	static int fts3CommitMethod(sqlite3_vtab *pVtab){
114602		- UNUSED_PARAMETER(pVtab);
114603	116723	TESTONLY( Fts3Table p = (Fts3Table)pVtab );
	116724	+ UNUSED_PARAMETER(pVtab);
114604	116725	assert( p->nPendingData==0 );
114605	116726	assert( p->inTransaction!=0 );
114606	116727	assert( p->pSegments==0 );
114607	116728	TESTONLY( p->inTransaction = 0 );
114608	116729	TESTONLY( p->mxSavepoint = -1; );
		@@ -114869,26 +116990,43 @@
114869	116990	p->zDb, p->zName, zName
114870	116991	);
114871	116992	return rc;
114872	116993	}
114873	116994
	116995	+/*
	116996	+** The xSavepoint() method.
	116997	+**
	116998	+** Flush the contents of the pending-terms table to disk.
	116999	+*/
114874	117000	static int fts3SavepointMethod(sqlite3_vtab *pVtab, int iSavepoint){
114875	117001	UNUSED_PARAMETER(iSavepoint);
114876	117002	assert( ((Fts3Table *)pVtab)->inTransaction );
114877	117003	assert( ((Fts3Table *)pVtab)->mxSavepoint < iSavepoint );
114878	117004	TESTONLY( ((Fts3Table *)pVtab)->mxSavepoint = iSavepoint );
114879	117005	return fts3SyncMethod(pVtab);
114880	117006	}
	117007	+
	117008	+/*
	117009	+** The xRelease() method.
	117010	+**
	117011	+** This is a no-op.
	117012	+*/
114881	117013	static int fts3ReleaseMethod(sqlite3_vtab *pVtab, int iSavepoint){
114882	117014	TESTONLY( Fts3Table p = (Fts3Table)pVtab );
114883	117015	UNUSED_PARAMETER(iSavepoint);
114884	117016	UNUSED_PARAMETER(pVtab);
114885	117017	assert( p->inTransaction );
114886	117018	assert( p->mxSavepoint >= iSavepoint );
114887	117019	TESTONLY( p->mxSavepoint = iSavepoint-1 );
114888	117020	return SQLITE_OK;
114889	117021	}
	117022	+
	117023	+/*
	117024	+** The xRollbackTo() method.
	117025	+**
	117026	+** Discard the contents of the pending terms table.
	117027	+*/
114890	117028	static int fts3RollbackToMethod(sqlite3_vtab *pVtab, int iSavepoint){
114891	117029	Fts3Table p = (Fts3Table)pVtab;
114892	117030	UNUSED_PARAMETER(iSavepoint);
114893	117031	assert( p->inTransaction );
114894	117032	assert( p->mxSavepoint >= iSavepoint );
		@@ -115033,22 +117171,10 @@
115033	117171	sqlite3Fts3HashClear(pHash);
115034	117172	sqlite3_free(pHash);
115035	117173	}
115036	117174	return rc;
115037	117175	}
115038		-
115039		-#if !SQLITE_CORE
115040		-SQLITE_API int sqlite3_extension_init(
115041		- sqlite3 *db,
115042		- char **pzErrMsg,
115043		- const sqlite3_api_routines *pApi
115044		-){
115045		- SQLITE_EXTENSION_INIT2(pApi)
115046		- return sqlite3Fts3Init(db);
115047		-}
115048		-#endif
115049		-
115050	117176
115051	117177	/*
115052	117178	** Allocate an Fts3MultiSegReader for each token in the expression headed
115053	117179	** by pExpr.
115054	117180	**
		@@ -115062,99 +117188,151 @@
115062	117188	** there exists prefix b-tree of the right length) then it may be traversed
115063	117189	** and merged incrementally. Otherwise, it has to be merged into an in-memory
115064	117190	** doclist and then traversed.
115065	117191	*/
115066	117192	static void fts3EvalAllocateReaders(
115067		- Fts3Cursor *pCsr,
115068		- Fts3Expr *pExpr,
	117193	+ Fts3Cursor pCsr, / FTS cursor handle */
	117194	+ Fts3Expr pExpr, / Allocate readers for this expression */
115069	117195	int pnToken, / OUT: Total number of tokens in phrase. */
115070	117196	int pnOr, / OUT: Total number of OR nodes in expr. */
115071		- int *pRc
	117197	+ int pRc / IN/OUT: Error code */
115072	117198	){
115073	117199	if( pExpr && SQLITE_OK==*pRc ){
115074	117200	if( pExpr->eType==FTSQUERY_PHRASE ){
115075	117201	int i;
115076	117202	int nToken = pExpr->pPhrase->nToken;
115077	117203	*pnToken += nToken;
115078	117204	for(i=0; i<nToken; i++){
115079	117205	Fts3PhraseToken *pToken = &pExpr->pPhrase->aToken[i];
115080		- int rc = sqlite3Fts3TermSegReaderCursor(pCsr,
	117206	+ int rc = fts3TermSegReaderCursor(pCsr,
115081	117207	pToken->z, pToken->n, pToken->isPrefix, &pToken->pSegcsr
115082	117208	);
115083	117209	if( rc!=SQLITE_OK ){
115084	117210	*pRc = rc;
115085	117211	return;
115086	117212	}
115087	117213	}
	117214	+ assert( pExpr->pPhrase->iDoclistToken==0 );
	117215	+ pExpr->pPhrase->iDoclistToken = -1;
115088	117216	}else{
115089	117217	*pnOr += (pExpr->eType==FTSQUERY_OR);
115090	117218	fts3EvalAllocateReaders(pCsr, pExpr->pLeft, pnToken, pnOr, pRc);
115091	117219	fts3EvalAllocateReaders(pCsr, pExpr->pRight, pnToken, pnOr, pRc);
115092	117220	}
115093	117221	}
115094	117222	}
115095	117223
	117224	+/*
	117225	+** Arguments pList/nList contain the doclist for token iToken of phrase p.
	117226	+** It is merged into the main doclist stored in p->doclist.aAll/nAll.
	117227	+**
	117228	+** This function assumes that pList points to a buffer allocated using
	117229	+** sqlite3_malloc(). This function takes responsibility for eventually
	117230	+** freeing the buffer.
	117231	+*/
	117232	+static void fts3EvalPhraseMergeToken(
	117233	+ Fts3Table pTab, / FTS Table pointer */
	117234	+ Fts3Phrase p, / Phrase to merge pList/nList into */
	117235	+ int iToken, /* Token pList/nList corresponds to */
	117236	+ char pList, / Pointer to doclist */
	117237	+ int nList /* Number of bytes in pList */
	117238	+){
	117239	+ assert( iToken!=p->iDoclistToken );
	117240	+
	117241	+ if( pList==0 ){
	117242	+ sqlite3_free(p->doclist.aAll);
	117243	+ p->doclist.aAll = 0;
	117244	+ p->doclist.nAll = 0;
	117245	+ }
	117246	+
	117247	+ else if( p->iDoclistToken<0 ){
	117248	+ p->doclist.aAll = pList;
	117249	+ p->doclist.nAll = nList;
	117250	+ }
	117251	+
	117252	+ else if( p->doclist.aAll==0 ){
	117253	+ sqlite3_free(pList);
	117254	+ }
	117255	+
	117256	+ else {
	117257	+ char *pLeft;
	117258	+ char *pRight;
	117259	+ int nLeft;
	117260	+ int nRight;
	117261	+ int nDiff;
	117262	+
	117263	+ if( p->iDoclistToken<iToken ){
	117264	+ pLeft = p->doclist.aAll;
	117265	+ nLeft = p->doclist.nAll;
	117266	+ pRight = pList;
	117267	+ nRight = nList;
	117268	+ nDiff = iToken - p->iDoclistToken;
	117269	+ }else{
	117270	+ pRight = p->doclist.aAll;
	117271	+ nRight = p->doclist.nAll;
	117272	+ pLeft = pList;
	117273	+ nLeft = nList;
	117274	+ nDiff = p->iDoclistToken - iToken;
	117275	+ }
	117276	+
	117277	+ fts3DoclistPhraseMerge(pTab->bDescIdx, nDiff, pLeft, nLeft, pRight,&nRight);
	117278	+ sqlite3_free(pLeft);
	117279	+ p->doclist.aAll = pRight;
	117280	+ p->doclist.nAll = nRight;
	117281	+ }
	117282	+
	117283	+ if( iToken>p->iDoclistToken ) p->iDoclistToken = iToken;
	117284	+}
	117285	+
	117286	+/*
	117287	+** Load the doclist for phrase p into p->doclist.aAll/nAll. The loaded doclist
	117288	+** does not take deferred tokens into account.
	117289	+**
	117290	+** SQLITE_OK is returned if no error occurs, otherwise an SQLite error code.
	117291	+*/
115096	117292	static int fts3EvalPhraseLoad(
115097		- Fts3Cursor *pCsr,
115098		- Fts3Phrase *p
	117293	+ Fts3Cursor pCsr, / FTS Cursor handle */
	117294	+ Fts3Phrase p / Phrase object */
115099	117295	){
115100	117296	Fts3Table pTab = (Fts3Table )pCsr->base.pVtab;
115101	117297	int iToken;
115102	117298	int rc = SQLITE_OK;
115103	117299
115104		- char *aDoclist = 0;
115105		- int nDoclist = 0;
115106		- int iPrev = -1;
115107		-
115108	117300	for(iToken=0; rc==SQLITE_OK && iToken<p->nToken; iToken++){
115109	117301	Fts3PhraseToken *pToken = &p->aToken[iToken];
115110		- assert( pToken->pSegcsr \|\| pToken->pDeferred );
	117302	+ assert( pToken->pDeferred==0 \|\| pToken->pSegcsr==0 );
115111	117303
115112		- if( pToken->pDeferred==0 ){
	117304	+ if( pToken->pSegcsr ){
115113	117305	int nThis = 0;
115114	117306	char *pThis = 0;
115115		- rc = fts3TermSelect(pTab, pToken, p->iColumn, 1, &nThis, &pThis);
115116		- if( rc==SQLITE_OK ){
115117		- if( pThis==0 ){
115118		- sqlite3_free(aDoclist);
115119		- aDoclist = 0;
115120		- nDoclist = 0;
115121		- break;
115122		- }else if( aDoclist==0 ){
115123		- aDoclist = pThis;
115124		- nDoclist = nThis;
115125		- }else{
115126		- assert( iPrev>=0 );
115127		- fts3DoclistPhraseMerge(pTab->bDescIdx,
115128		- iToken-iPrev, aDoclist, nDoclist, pThis, &nThis
115129		- );
115130		- sqlite3_free(aDoclist);
115131		- aDoclist = pThis;
115132		- nDoclist = nThis;
115133		- }
115134		- iPrev = iToken;
115135		- }
115136		- }
115137		- }
115138		-
115139		- if( rc==SQLITE_OK ){
115140		- p->doclist.aAll = aDoclist;
115141		- p->doclist.nAll = nDoclist;
115142		- }else{
115143		- sqlite3_free(aDoclist);
115144		- }
115145		- return rc;
115146		-}
115147		-
	117307	+ rc = fts3TermSelect(pTab, pToken, p->iColumn, &nThis, &pThis);
	117308	+ if( rc==SQLITE_OK ){
	117309	+ fts3EvalPhraseMergeToken(pTab, p, iToken, pThis, nThis);
	117310	+ }
	117311	+ }
	117312	+ assert( pToken->pSegcsr==0 );
	117313	+ }
	117314	+
	117315	+ return rc;
	117316	+}
	117317	+
	117318	+/*
	117319	+** This function is called on each phrase after the position lists for
	117320	+** any deferred tokens have been loaded into memory. It updates the phrases
	117321	+** current position list to include only those positions that are really
	117322	+** instances of the phrase (after considering deferred tokens). If this
	117323	+** means that the phrase does not appear in the current row, doclist.pList
	117324	+** and doclist.nList are both zeroed.
	117325	+**
	117326	+** SQLITE_OK is returned if no error occurs, otherwise an SQLite error code.
	117327	+*/
115148	117328	static int fts3EvalDeferredPhrase(Fts3Cursor pCsr, Fts3Phrase pPhrase){
115149		- int iToken;
115150		- int rc = SQLITE_OK;
115151		-
115152		- int nMaxUndeferred = -1;
115153		- char *aPoslist = 0;
115154		- int nPoslist = 0;
115155		- int iPrev = -1;
	117329	+ int iToken; /* Used to iterate through phrase tokens */
	117330	+ int rc = SQLITE_OK; /* Return code */
	117331	+ char aPoslist = 0; / Position list for deferred tokens */
	117332	+ int nPoslist = 0; /* Number of bytes in aPoslist */
	117333	+ int iPrev = -1; /* Token number of previous deferred token */
115156	117334
115157	117335	assert( pPhrase->doclist.bFreeList==0 );
115158	117336
115159	117337	for(iToken=0; rc==SQLITE_OK && iToken<pPhrase->nToken; iToken++){
115160	117338	Fts3PhraseToken *pToken = &pPhrase->aToken[iToken];
		@@ -115192,16 +117370,15 @@
115192	117370	pPhrase->doclist.nList = 0;
115193	117371	return SQLITE_OK;
115194	117372	}
115195	117373	}
115196	117374	iPrev = iToken;
115197		- }else{
115198		- nMaxUndeferred = iToken;
115199	117375	}
115200	117376	}
115201	117377
115202	117378	if( iPrev>=0 ){
	117379	+ int nMaxUndeferred = pPhrase->iDoclistToken;
115203	117380	if( nMaxUndeferred<0 ){
115204	117381	pPhrase->doclist.pList = aPoslist;
115205	117382	pPhrase->doclist.nList = nPoslist;
115206	117383	pPhrase->doclist.iDocid = pCsr->iPrevId;
115207	117384	pPhrase->doclist.bFreeList = 1;
		@@ -115246,19 +117423,27 @@
115246	117423	/*
115247	117424	** This function is called for each Fts3Phrase in a full-text query
115248	117425	** expression to initialize the mechanism for returning rows. Once this
115249	117426	** function has been called successfully on an Fts3Phrase, it may be
115250	117427	** used with fts3EvalPhraseNext() to iterate through the matching docids.
	117428	+**
	117429	+** If parameter bOptOk is true, then the phrase may (or may not) use the
	117430	+** incremental loading strategy. Otherwise, the entire doclist is loaded into
	117431	+** memory within this call.
	117432	+**
	117433	+** SQLITE_OK is returned if no error occurs, otherwise an SQLite error code.
115251	117434	*/
115252	117435	static int fts3EvalPhraseStart(Fts3Cursor pCsr, int bOptOk, Fts3Phrase p){
115253		- int rc;
	117436	+ int rc; /* Error code */
115254	117437	Fts3PhraseToken *pFirst = &p->aToken[0];
115255	117438	Fts3Table pTab = (Fts3Table )pCsr->base.pVtab;
115256	117439
115257		- assert( p->doclist.aAll==0 );
115258		- if( pCsr->bDesc==pTab->bDescIdx && bOptOk==1 && p->nToken==1
115259		- && pFirst->pSegcsr && pFirst->pSegcsr->bLookup
	117440	+ if( pCsr->bDesc==pTab->bDescIdx
	117441	+ && bOptOk==1
	117442	+ && p->nToken==1
	117443	+ && pFirst->pSegcsr
	117444	+ && pFirst->pSegcsr->bLookup
115260	117445	){
115261	117446	/* Use the incremental approach. */
115262	117447	int iCol = (p->iColumn >= pTab->nColumn ? -1 : p->iColumn);
115263	117448	rc = sqlite3Fts3MsrIncrStart(
115264	117449	pTab, pFirst->pSegcsr, iCol, pFirst->z, pFirst->n);
		@@ -115274,11 +117459,17 @@
115274	117459	return rc;
115275	117460	}
115276	117461
115277	117462	/*
115278	117463	** This function is used to iterate backwards (from the end to start)
115279		-** through doclists.
	117464	+** through doclists. It is used by this module to iterate through phrase
	117465	+** doclists in reverse and by the fts3_write.c module to iterate through
	117466	+** pending-terms lists when writing to databases with "order=desc".
	117467	+**
	117468	+** The doclist may be sorted in ascending (parameter bDescIdx==0) or
	117469	+** descending (parameter bDescIdx==1) order of docid. Regardless, this
	117470	+** function iterates from the end of the doclist to the beginning.
115280	117471	*/
115281	117472	SQLITE_PRIVATE void sqlite3Fts3DoclistPrev(
115282	117473	int bDescIdx, /* True if the doclist is desc */
115283	117474	char aDoclist, / Pointer to entire doclist */
115284	117475	int nDoclist, /* Length of aDoclist in bytes */
		@@ -115339,13 +117530,13 @@
115339	117530	** If there is no "next" entry and no error occurs, then *pbEof is set to
115340	117531	** 1 before returning. Otherwise, if no error occurs and the iterator is
115341	117532	** successfully advanced, *pbEof is set to 0.
115342	117533	*/
115343	117534	static int fts3EvalPhraseNext(
115344		- Fts3Cursor *pCsr,
115345		- Fts3Phrase *p,
115346		- u8 *pbEof
	117535	+ Fts3Cursor pCsr, / FTS Cursor handle */
	117536	+ Fts3Phrase p, / Phrase object to advance to next docid */
	117537	+ u8 pbEof / OUT: Set to 1 if EOF */
115347	117538	){
115348	117539	int rc = SQLITE_OK;
115349	117540	Fts3Doclist *pDL = &p->doclist;
115350	117541	Fts3Table pTab = (Fts3Table )pCsr->base.pVtab;
115351	117542
		@@ -115387,30 +117578,46 @@
115387	117578	fts3PoslistCopy(0, &pIter);
115388	117579	pDL->nList = (pIter - pDL->pList);
115389	117580
115390	117581	/* pIter now points just past the 0x00 that terminates the position-
115391	117582	** list for document pDL->iDocid. However, if this position-list was
115392		- ** edited in place by fts3EvalNearTrim2(), then pIter may not actually
	117583	+ ** edited in place by fts3EvalNearTrim(), then pIter may not actually
115393	117584	** point to the start of the next docid value. The following line deals
115394	117585	** with this case by advancing pIter past the zero-padding added by
115395		- ** fts3EvalNearTrim2(). */
	117586	+ ** fts3EvalNearTrim(). */
115396	117587	while( pIter<pEnd && *pIter==0 ) pIter++;
115397	117588
115398	117589	pDL->pNextDocid = pIter;
115399		- assert( *pIter \|\| pIter>=&pDL->aAll[pDL->nAll] );
	117590	+ assert( pIter>=&pDL->aAll[pDL->nAll] \|\| *pIter );
115400	117591	*pbEof = 0;
115401	117592	}
115402	117593	}
115403	117594
115404	117595	return rc;
115405	117596	}
115406	117597
	117598	+/*
	117599	+**
	117600	+** If *pRc is not SQLITE_OK when this function is called, it is a no-op.
	117601	+** Otherwise, fts3EvalPhraseStart() is called on all phrases within the
	117602	+** expression. Also the Fts3Expr.bDeferred variable is set to true for any
	117603	+** expressions for which all descendent tokens are deferred.
	117604	+**
	117605	+** If parameter bOptOk is zero, then it is guaranteed that the
	117606	+** Fts3Phrase.doclist.aAll/nAll variables contain the entire doclist for
	117607	+** each phrase in the expression (subject to deferred token processing).
	117608	+** Or, if bOptOk is non-zero, then one or more tokens within the expression
	117609	+** may be loaded incrementally, meaning doclist.aAll/nAll is not available.
	117610	+**
	117611	+** If an error occurs within this function, *pRc is set to an SQLite error
	117612	+** code before returning.
	117613	+*/
115407	117614	static void fts3EvalStartReaders(
115408		- Fts3Cursor *pCsr,
115409		- Fts3Expr *pExpr,
115410		- int bOptOk,
115411		- int *pRc
	117615	+ Fts3Cursor pCsr, / FTS Cursor handle */
	117616	+ Fts3Expr pExpr, / Expression to initialize phrases in */
	117617	+ int bOptOk, /* True to enable incremental loading */
	117618	+ int pRc / IN/OUT: Error code */
115412	117619	){
115413	117620	if( pExpr && SQLITE_OK==*pRc ){
115414	117621	if( pExpr->eType==FTSQUERY_PHRASE ){
115415	117622	int i;
115416	117623	int nToken = pExpr->pPhrase->nToken;
		@@ -115425,33 +117632,55 @@
115425	117632	pExpr->bDeferred = (pExpr->pLeft->bDeferred && pExpr->pRight->bDeferred);
115426	117633	}
115427	117634	}
115428	117635	}
115429	117636
115430		-
	117637	+/*
	117638	+** An array of the following structures is assembled as part of the process
	117639	+** of selecting tokens to defer before the query starts executing (as part
	117640	+** of the xFilter() method). There is one element in the array for each
	117641	+** token in the FTS expression.
	117642	+**
	117643	+** Tokens are divided into AND/NEAR clusters. All tokens in a cluster belong
	117644	+** to phrases that are connected only by AND and NEAR operators (not OR or
	117645	+** NOT). When determining tokens to defer, each AND/NEAR cluster is considered
	117646	+** separately. The root of a tokens AND/NEAR cluster is stored in
	117647	+** Fts3TokenAndCost.pRoot.
	117648	+*/
115431	117649	typedef struct Fts3TokenAndCost Fts3TokenAndCost;
115432	117650	struct Fts3TokenAndCost {
115433		- Fts3PhraseToken *pToken;
115434		- Fts3Expr *pRoot;
115435		- int nOvfl;
115436		- int iCol;
	117651	+ Fts3Phrase pPhrase; / The phrase the token belongs to */
	117652	+ int iToken; /* Position of token in phrase */
	117653	+ Fts3PhraseToken pToken; / The token itself */
	117654	+ Fts3Expr pRoot; / Root of NEAR/AND cluster */
	117655	+ int nOvfl; /* Number of overflow pages to load doclist */
	117656	+ int iCol; /* The column the token must match */
115437	117657	};
115438	117658
	117659	+/*
	117660	+** This function is used to populate an allocated Fts3TokenAndCost array.
	117661	+**
	117662	+** If *pRc is not SQLITE_OK when this function is called, it is a no-op.
	117663	+** Otherwise, if an error occurs during execution, *pRc is set to an
	117664	+** SQLite error code.
	117665	+*/
115439	117666	static void fts3EvalTokenCosts(
115440		- Fts3Cursor *pCsr,
115441		- Fts3Expr *pRoot,
115442		- Fts3Expr *pExpr,
115443		- Fts3TokenAndCost **ppTC,
115444		- Fts3Expr ***ppOr,
115445		- int *pRc
	117667	+ Fts3Cursor pCsr, / FTS Cursor handle */
	117668	+ Fts3Expr pRoot, / Root of current AND/NEAR cluster */
	117669	+ Fts3Expr pExpr, / Expression to consider */
	117670	+ Fts3TokenAndCost *ppTC, / Write new entries to (ppTC)++ */
	117671	+ Fts3Expr **ppOr, / Write new OR root to (ppOr)++ */
	117672	+ int pRc / IN/OUT: Error code */
115446	117673	){
115447	117674	if( *pRc==SQLITE_OK && pExpr ){
115448	117675	if( pExpr->eType==FTSQUERY_PHRASE ){
115449	117676	Fts3Phrase *pPhrase = pExpr->pPhrase;
115450	117677	int i;
115451	117678	for(i=0; *pRc==SQLITE_OK && i<pPhrase->nToken; i++){
115452	117679	Fts3TokenAndCost pTC = (ppTC)++;
	117680	+ pTC->pPhrase = pPhrase;
	117681	+ pTC->iToken = i;
115453	117682	pTC->pRoot = pRoot;
115454	117683	pTC->pToken = &pPhrase->aToken[i];
115455	117684	pTC->iCol = pPhrase->iColumn;
115456	117685	*pRc = sqlite3Fts3MsrOvfl(pCsr, pTC->pToken->pSegcsr, &pTC->nOvfl);
115457	117686	}
		@@ -115470,23 +117699,34 @@
115470	117699	fts3EvalTokenCosts(pCsr, pRoot, pExpr->pRight, ppTC, ppOr, pRc);
115471	117700	}
115472	117701	}
115473	117702	}
115474	117703
	117704	+/*
	117705	+** Determine the average document (row) size in pages. If successful,
	117706	+** write this value to *pnPage and return SQLITE_OK. Otherwise, return
	117707	+** an SQLite error code.
	117708	+**
	117709	+** The average document size in pages is calculated by first calculating
	117710	+** determining the average size in bytes, B. If B is less than the amount
	117711	+** of data that will fit on a single leaf page of an intkey table in
	117712	+** this database, then the average docsize is 1. Otherwise, it is 1 plus
	117713	+** the number of overflow pages consumed by a record B bytes in size.
	117714	+*/
115475	117715	static int fts3EvalAverageDocsize(Fts3Cursor pCsr, int pnPage){
115476	117716	if( pCsr->nRowAvg==0 ){
115477	117717	/* The average document size, which is required to calculate the cost
115478		- ** of each doclist, has not yet been determined. Read the required
115479		- ** data from the %_stat table to calculate it.
115480		- **
115481		- ** Entry 0 of the %_stat table is a blob containing (nCol+1) FTS3
115482		- ** varints, where nCol is the number of columns in the FTS3 table.
115483		- ** The first varint is the number of documents currently stored in
115484		- ** the table. The following nCol varints contain the total amount of
115485		- ** data stored in all rows of each column of the table, from left
115486		- ** to right.
115487		- */
	117718	+ ** of each doclist, has not yet been determined. Read the required
	117719	+ ** data from the %_stat table to calculate it.
	117720	+ **
	117721	+ ** Entry 0 of the %_stat table is a blob containing (nCol+1) FTS3
	117722	+ ** varints, where nCol is the number of columns in the FTS3 table.
	117723	+ ** The first varint is the number of documents currently stored in
	117724	+ ** the table. The following nCol varints contain the total amount of
	117725	+ ** data stored in all rows of each column of the table, from left
	117726	+ ** to right.
	117727	+ */
115488	117728	int rc;
115489	117729	Fts3Table p = (Fts3Table)pCsr->base.pVtab;
115490	117730	sqlite3_stmt *pStmt;
115491	117731	sqlite3_int64 nDoc = 0;
115492	117732	sqlite3_int64 nByte = 0;
		@@ -115517,113 +117757,151 @@
115517	117757
115518	117758	*pnPage = pCsr->nRowAvg;
115519	117759	return SQLITE_OK;
115520	117760	}
115521	117761
	117762	+/*
	117763	+** This function is called to select the tokens (if any) that will be
	117764	+** deferred. The array aTC[] has already been populated when this is
	117765	+** called.
	117766	+**
	117767	+** This function is called once for each AND/NEAR cluster in the
	117768	+** expression. Each invocation determines which tokens to defer within
	117769	+** the cluster with root node pRoot. See comments above the definition
	117770	+** of struct Fts3TokenAndCost for more details.
	117771	+**
	117772	+** If no error occurs, SQLITE_OK is returned and sqlite3Fts3DeferToken()
	117773	+** called on each token to defer. Otherwise, an SQLite error code is
	117774	+** returned.
	117775	+*/
115522	117776	static int fts3EvalSelectDeferred(
115523		- Fts3Cursor *pCsr,
115524		- Fts3Expr *pRoot,
115525		- Fts3TokenAndCost *aTC,
115526		- int nTC
	117777	+ Fts3Cursor pCsr, / FTS Cursor handle */
	117778	+ Fts3Expr pRoot, / Consider tokens with this root node */
	117779	+ Fts3TokenAndCost aTC, / Array of expression tokens and costs */
	117780	+ int nTC /* Number of entries in aTC[] */
115527	117781	){
115528		- int nDocSize = 0;
115529		- int nDocEst = 0;
115530		- int rc = SQLITE_OK;
115531	117782	Fts3Table pTab = (Fts3Table )pCsr->base.pVtab;
115532		- int ii;
	117783	+ int nDocSize = 0; /* Number of pages per doc loaded */
	117784	+ int rc = SQLITE_OK; /* Return code */
	117785	+ int ii; /* Iterator variable for various purposes */
	117786	+ int nOvfl = 0; /* Total overflow pages used by doclists */
	117787	+ int nToken = 0; /* Total number of tokens in cluster */
115533	117788
115534		- int nOvfl = 0;
115535		- int nTerm = 0;
	117789	+ int nMinEst = 0; /* The minimum count for any phrase so far. */
	117790	+ int nLoad4 = 1; /* (Phrases that will be loaded)^4. */
115536	117791
	117792	+ /* Count the tokens in this AND/NEAR cluster. If none of the doclists
	117793	+ ** associated with the tokens spill onto overflow pages, or if there is
	117794	+ ** only 1 token, exit early. No tokens to defer in this case. */
115537	117795	for(ii=0; ii<nTC; ii++){
115538	117796	if( aTC[ii].pRoot==pRoot ){
115539	117797	nOvfl += aTC[ii].nOvfl;
115540		- nTerm++;
	117798	+ nToken++;
115541	117799	}
115542	117800	}
115543		- if( nOvfl==0 \|\| nTerm<2 ) return SQLITE_OK;
	117801	+ if( nOvfl==0 \|\| nToken<2 ) return SQLITE_OK;
115544	117802
	117803	+ /* Obtain the average docsize (in pages). */
115545	117804	rc = fts3EvalAverageDocsize(pCsr, &nDocSize);
115546		-
115547		- for(ii=0; ii<nTerm && rc==SQLITE_OK; ii++){
115548		- int jj;
115549		- Fts3TokenAndCost *pTC = 0;
115550		-
115551		- for(jj=0; jj<nTC; jj++){
115552		- if( aTC[jj].pToken && aTC[jj].pRoot==pRoot
115553		- && (!pTC \|\| aTC[jj].nOvfl<pTC->nOvfl)
	117805	+ assert( rc!=SQLITE_OK \|\| nDocSize>0 );
	117806	+
	117807	+
	117808	+ /* Iterate through all tokens in this AND/NEAR cluster, in ascending order
	117809	+ ** of the number of overflow pages that will be loaded by the pager layer
	117810	+ ** to retrieve the entire doclist for the token from the full-text index.
	117811	+ ** Load the doclists for tokens that are either:
	117812	+ **
	117813	+ ** a. The cheapest token in the entire query (i.e. the one visited by the
	117814	+ ** first iteration of this loop), or
	117815	+ **
	117816	+ ** b. Part of a multi-token phrase.
	117817	+ **
	117818	+ ** After each token doclist is loaded, merge it with the others from the
	117819	+ ** same phrase and count the number of documents that the merged doclist
	117820	+ ** contains. Set variable "nMinEst" to the smallest number of documents in
	117821	+ ** any phrase doclist for which 1 or more token doclists have been loaded.
	117822	+ ** Let nOther be the number of other phrases for which it is certain that
	117823	+ ** one or more tokens will not be deferred.
	117824	+ **
	117825	+ ** Then, for each token, defer it if loading the doclist would result in
	117826	+ ** loading N or more overflow pages into memory, where N is computed as:
	117827	+ **
	117828	+ ** (nMinEst + 4^nOther - 1) / (4^nOther)
	117829	+ */
	117830	+ for(ii=0; ii<nToken && rc==SQLITE_OK; ii++){
	117831	+ int iTC; /* Used to iterate through aTC[] array. */
	117832	+ Fts3TokenAndCost pTC = 0; / Set to cheapest remaining token. */
	117833	+
	117834	+ /* Set pTC to point to the cheapest remaining token. */
	117835	+ for(iTC=0; iTC<nTC; iTC++){
	117836	+ if( aTC[iTC].pToken && aTC[iTC].pRoot==pRoot
	117837	+ && (!pTC \|\| aTC[iTC].nOvfl<pTC->nOvfl)
115554	117838	){
115555		- pTC = &aTC[jj];
	117839	+ pTC = &aTC[iTC];
115556	117840	}
115557	117841	}
115558	117842	assert( pTC );
115559	117843
115560		- /* At this point pTC points to the cheapest remaining token. */
115561		- if( ii==0 ){
115562		- if( pTC->nOvfl ){
115563		- nDocEst = (pTC->nOvfl * pTab->nPgsz + pTab->nPgsz) / 10;
115564		- }else{
115565		- /* TODO: Fix this so that the doclist need not be read twice. */
	117844	+ if( ii && pTC->nOvfl>=((nMinEst+(nLoad4/4)-1)/(nLoad4/4))*nDocSize ){
	117845	+ /* The number of overflow pages to load for this (and therefore all
	117846	+ ** subsequent) tokens is greater than the estimated number of pages
	117847	+ ** that will be loaded if all subsequent tokens are deferred.
	117848	+ */
	117849	+ Fts3PhraseToken *pToken = pTC->pToken;
	117850	+ rc = sqlite3Fts3DeferToken(pCsr, pToken, pTC->iCol);
	117851	+ fts3SegReaderCursorFree(pToken->pSegcsr);
	117852	+ pToken->pSegcsr = 0;
	117853	+ }else{
	117854	+ nLoad4 = nLoad4*4;
	117855	+ if( ii==0 \|\| pTC->pPhrase->nToken>1 ){
	117856	+ /* Either this is the cheapest token in the entire query, or it is
	117857	+ ** part of a multi-token phrase. Either way, the entire doclist will
	117858	+ ** (eventually) be loaded into memory. It may as well be now. */
115566	117859	Fts3PhraseToken *pToken = pTC->pToken;
115567	117860	int nList = 0;
115568	117861	char *pList = 0;
115569		- rc = fts3TermSelect(pTab, pToken, pTC->iCol, 1, &nList, &pList);
115570		- if( rc==SQLITE_OK ){
115571		- nDocEst = fts3DoclistCountDocids(1, pList, nList);
115572		- }
115573		- sqlite3_free(pList);
115574		- if( rc==SQLITE_OK ){
115575		- rc = sqlite3Fts3TermSegReaderCursor(pCsr,
115576		- pToken->z, pToken->n, pToken->isPrefix, &pToken->pSegcsr
	117862	+ rc = fts3TermSelect(pTab, pToken, pTC->iCol, &nList, &pList);
	117863	+ assert( rc==SQLITE_OK \|\| pList==0 );
	117864	+ if( rc==SQLITE_OK ){
	117865	+ int nCount;
	117866	+ fts3EvalPhraseMergeToken(pTab, pTC->pPhrase, pTC->iToken,pList,nList);
	117867	+ nCount = fts3DoclistCountDocids(
	117868	+ pTC->pPhrase->doclist.aAll, pTC->pPhrase->doclist.nAll
115577	117869	);
	117870	+ if( ii==0 \|\| nCount<nMinEst ) nMinEst = nCount;
115578	117871	}
115579	117872	}
115580		- }else{
115581		- if( pTC->nOvfl>=(nDocEst*nDocSize) ){
115582		- Fts3PhraseToken *pToken = pTC->pToken;
115583		- rc = sqlite3Fts3DeferToken(pCsr, pToken, pTC->iCol);
115584		- fts3SegReaderCursorFree(pToken->pSegcsr);
115585		- pToken->pSegcsr = 0;
115586		- }
115587		- nDocEst = 1 + (nDocEst/4);
115588	117873	}
115589	117874	pTC->pToken = 0;
115590	117875	}
115591	117876
115592	117877	return rc;
115593	117878	}
115594	117879
115595		-SQLITE_PRIVATE int sqlite3Fts3EvalStart(Fts3Cursor pCsr, Fts3Expr pExpr, int bOptOk){
	117880	+/*
	117881	+** This function is called from within the xFilter method. It initializes
	117882	+** the full-text query currently stored in pCsr->pExpr. To iterate through
	117883	+** the results of a query, the caller does:
	117884	+**
	117885	+** fts3EvalStart(pCsr);
	117886	+** while( 1 ){
	117887	+** fts3EvalNext(pCsr);
	117888	+** if( pCsr->bEof ) break;
	117889	+** ... return row pCsr->iPrevId to the caller ...
	117890	+** }
	117891	+*/
	117892	+static int fts3EvalStart(Fts3Cursor *pCsr){
115596	117893	Fts3Table pTab = (Fts3Table )pCsr->base.pVtab;
115597	117894	int rc = SQLITE_OK;
115598	117895	int nToken = 0;
115599	117896	int nOr = 0;
115600	117897
115601	117898	/* Allocate a MultiSegReader for each token in the expression. */
115602		- fts3EvalAllocateReaders(pCsr, pExpr, &nToken, &nOr, &rc);
115603		-
115604		- /* Call fts3EvalPhraseStart() on all phrases in the expression. TODO:
115605		- ** This call will eventually also be responsible for determining which
115606		- ** tokens are 'deferred' until the document text is loaded into memory.
115607		- **
115608		- ** Each token in each phrase is dealt with using one of the following
115609		- ** three strategies:
115610		- **
115611		- ** 1. Entire doclist loaded into memory as part of the
115612		- ** fts3EvalStartReaders() call.
115613		- **
115614		- ** 2. Doclist loaded into memory incrementally, as part of each
115615		- ** sqlite3Fts3EvalNext() call.
115616		- **
115617		- ** 3. Token doclist is never loaded. Instead, documents are loaded into
115618		- ** memory and scanned for the token as part of the sqlite3Fts3EvalNext()
115619		- ** call. This is known as a "deferred" token.
115620		- */
115621		-
115622		- /* If bOptOk is true, check if there are any tokens that should be deferred.
115623		- */
115624		- if( rc==SQLITE_OK && bOptOk && nToken>1 && pTab->bHasStat ){
	117899	+ fts3EvalAllocateReaders(pCsr, pCsr->pExpr, &nToken, &nOr, &rc);
	117900	+
	117901	+ /* Determine which, if any, tokens in the expression should be deferred. */
	117902	+ if( rc==SQLITE_OK && nToken>1 && pTab->bHasStat ){
115625	117903	Fts3TokenAndCost *aTC;
115626	117904	Fts3Expr **apOr;
115627	117905	aTC = (Fts3TokenAndCost *)sqlite3_malloc(
115628	117906	sizeof(Fts3TokenAndCost) * nToken
115629	117907	+ sizeof(Fts3Expr ) nOr * 2
		@@ -115635,11 +117913,11 @@
115635	117913	}else{
115636	117914	int ii;
115637	117915	Fts3TokenAndCost *pTC = aTC;
115638	117916	Fts3Expr **ppOr = apOr;
115639	117917
115640		- fts3EvalTokenCosts(pCsr, 0, pExpr, &pTC, &ppOr, &rc);
	117918	+ fts3EvalTokenCosts(pCsr, 0, pCsr->pExpr, &pTC, &ppOr, &rc);
115641	117919	nToken = pTC-aTC;
115642	117920	nOr = ppOr-apOr;
115643	117921
115644	117922	if( rc==SQLITE_OK ){
115645	117923	rc = fts3EvalSelectDeferred(pCsr, 0, aTC, nToken);
		@@ -115650,25 +117928,50 @@
115650	117928
115651	117929	sqlite3_free(aTC);
115652	117930	}
115653	117931	}
115654	117932
115655		- fts3EvalStartReaders(pCsr, pExpr, bOptOk, &rc);
	117933	+ fts3EvalStartReaders(pCsr, pCsr->pExpr, 1, &rc);
115656	117934	return rc;
115657	117935	}
115658	117936
115659		-static void fts3EvalZeroPoslist(Fts3Phrase *pPhrase){
	117937	+/*
	117938	+** Invalidate the current position list for phrase pPhrase.
	117939	+*/
	117940	+static void fts3EvalInvalidatePoslist(Fts3Phrase *pPhrase){
115660	117941	if( pPhrase->doclist.bFreeList ){
115661	117942	sqlite3_free(pPhrase->doclist.pList);
115662	117943	}
115663	117944	pPhrase->doclist.pList = 0;
115664	117945	pPhrase->doclist.nList = 0;
115665	117946	pPhrase->doclist.bFreeList = 0;
115666	117947	}
115667	117948
115668		-static int fts3EvalNearTrim2(
115669		- int nNear,
	117949	+/*
	117950	+** This function is called to edit the position list associated with
	117951	+** the phrase object passed as the fifth argument according to a NEAR
	117952	+** condition. For example:
	117953	+**
	117954	+** abc NEAR/5 "def ghi"
	117955	+**
	117956	+** Parameter nNear is passed the NEAR distance of the expression (5 in
	117957	+** the example above). When this function is called, *paPoslist points to
	117958	+** the position list, and *pnToken is the number of phrase tokens in, the
	117959	+** phrase on the other side of the NEAR operator to pPhrase. For example,
	117960	+** if pPhrase refers to the "def ghi" phrase, then *paPoslist points to
	117961	+** the position list associated with phrase "abc".
	117962	+**
	117963	+** All positions in the pPhrase position list that are not sufficiently
	117964	+** close to a position in the *paPoslist position list are removed. If this
	117965	+** leaves 0 positions, zero is returned. Otherwise, non-zero.
	117966	+**
	117967	+** Before returning, *paPoslist is set to point to the position lsit
	117968	+** associated with pPhrase. And *pnToken is set to the number of tokens in
	117969	+** pPhrase.
	117970	+*/
	117971	+static int fts3EvalNearTrim(
	117972	+ int nNear, /* NEAR distance. As in "NEAR/nNear". */
115670	117973	char aTmp, / Temporary space to use */
115671	117974	char *paPoslist, / IN/OUT: Position list */
115672	117975	int pnToken, / IN/OUT: Tokens in phrase of paPoslist /
115673	117976	Fts3Phrase pPhrase / The phrase object to trim the doclist of */
115674	117977	){
		@@ -115696,10 +117999,176 @@
115696	117999	}
115697	118000
115698	118001	return res;
115699	118002	}
115700	118003
	118004	+/*
	118005	+** This function is a no-op if *pRc is other than SQLITE_OK when it is called.
	118006	+** Otherwise, it advances the expression passed as the second argument to
	118007	+** point to the next matching row in the database. Expressions iterate through
	118008	+** matching rows in docid order. Ascending order if Fts3Cursor.bDesc is zero,
	118009	+** or descending if it is non-zero.
	118010	+**
	118011	+** If an error occurs, *pRc is set to an SQLite error code. Otherwise, if
	118012	+** successful, the following variables in pExpr are set:
	118013	+**
	118014	+** Fts3Expr.bEof (non-zero if EOF - there is no next row)
	118015	+** Fts3Expr.iDocid (valid if bEof==0. The docid of the next row)
	118016	+**
	118017	+** If the expression is of type FTSQUERY_PHRASE, and the expression is not
	118018	+** at EOF, then the following variables are populated with the position list
	118019	+** for the phrase for the visited row:
	118020	+**
	118021	+** FTs3Expr.pPhrase->doclist.nList (length of pList in bytes)
	118022	+** FTs3Expr.pPhrase->doclist.pList (pointer to position list)
	118023	+**
	118024	+** It says above that this function advances the expression to the next
	118025	+** matching row. This is usually true, but there are the following exceptions:
	118026	+**
	118027	+** 1. Deferred tokens are not taken into account. If a phrase consists
	118028	+** entirely of deferred tokens, it is assumed to match every row in
	118029	+** the db. In this case the position-list is not populated at all.
	118030	+**
	118031	+** Or, if a phrase contains one or more deferred tokens and one or
	118032	+** more non-deferred tokens, then the expression is advanced to the
	118033	+** next possible match, considering only non-deferred tokens. In other
	118034	+** words, if the phrase is "A B C", and "B" is deferred, the expression
	118035	+** is advanced to the next row that contains an instance of "A * C",
	118036	+** where "*" may match any single token. The position list in this case
	118037	+** is populated as for "A * C" before returning.
	118038	+**
	118039	+** 2. NEAR is treated as AND. If the expression is "x NEAR y", it is
	118040	+** advanced to point to the next row that matches "x AND y".
	118041	+**
	118042	+** See fts3EvalTestDeferredAndNear() for details on testing if a row is
	118043	+** really a match, taking into account deferred tokens and NEAR operators.
	118044	+*/
	118045	+static void fts3EvalNextRow(
	118046	+ Fts3Cursor pCsr, / FTS Cursor handle */
	118047	+ Fts3Expr pExpr, / Expr. to advance to next matching row */
	118048	+ int pRc / IN/OUT: Error code */
	118049	+){
	118050	+ if( *pRc==SQLITE_OK ){
	118051	+ int bDescDoclist = pCsr->bDesc; /* Used by DOCID_CMP() macro */
	118052	+ assert( pExpr->bEof==0 );
	118053	+ pExpr->bStart = 1;
	118054	+
	118055	+ switch( pExpr->eType ){
	118056	+ case FTSQUERY_NEAR:
	118057	+ case FTSQUERY_AND: {
	118058	+ Fts3Expr *pLeft = pExpr->pLeft;
	118059	+ Fts3Expr *pRight = pExpr->pRight;
	118060	+ assert( !pLeft->bDeferred \|\| !pRight->bDeferred );
	118061	+
	118062	+ if( pLeft->bDeferred ){
	118063	+ /* LHS is entirely deferred. So we assume it matches every row.
	118064	+ ** Advance the RHS iterator to find the next row visited. */
	118065	+ fts3EvalNextRow(pCsr, pRight, pRc);
	118066	+ pExpr->iDocid = pRight->iDocid;
	118067	+ pExpr->bEof = pRight->bEof;
	118068	+ }else if( pRight->bDeferred ){
	118069	+ /* RHS is entirely deferred. So we assume it matches every row.
	118070	+ ** Advance the LHS iterator to find the next row visited. */
	118071	+ fts3EvalNextRow(pCsr, pLeft, pRc);
	118072	+ pExpr->iDocid = pLeft->iDocid;
	118073	+ pExpr->bEof = pLeft->bEof;
	118074	+ }else{
	118075	+ /* Neither the RHS or LHS are deferred. */
	118076	+ fts3EvalNextRow(pCsr, pLeft, pRc);
	118077	+ fts3EvalNextRow(pCsr, pRight, pRc);
	118078	+ while( !pLeft->bEof && !pRight->bEof && *pRc==SQLITE_OK ){
	118079	+ sqlite3_int64 iDiff = DOCID_CMP(pLeft->iDocid, pRight->iDocid);
	118080	+ if( iDiff==0 ) break;
	118081	+ if( iDiff<0 ){
	118082	+ fts3EvalNextRow(pCsr, pLeft, pRc);
	118083	+ }else{
	118084	+ fts3EvalNextRow(pCsr, pRight, pRc);
	118085	+ }
	118086	+ }
	118087	+ pExpr->iDocid = pLeft->iDocid;
	118088	+ pExpr->bEof = (pLeft->bEof \|\| pRight->bEof);
	118089	+ }
	118090	+ break;
	118091	+ }
	118092	+
	118093	+ case FTSQUERY_OR: {
	118094	+ Fts3Expr *pLeft = pExpr->pLeft;
	118095	+ Fts3Expr *pRight = pExpr->pRight;
	118096	+ sqlite3_int64 iCmp = DOCID_CMP(pLeft->iDocid, pRight->iDocid);
	118097	+
	118098	+ assert( pLeft->bStart \|\| pLeft->iDocid==pRight->iDocid );
	118099	+ assert( pRight->bStart \|\| pLeft->iDocid==pRight->iDocid );
	118100	+
	118101	+ if( pRight->bEof \|\| (pLeft->bEof==0 && iCmp<0) ){
	118102	+ fts3EvalNextRow(pCsr, pLeft, pRc);
	118103	+ }else if( pLeft->bEof \|\| (pRight->bEof==0 && iCmp>0) ){
	118104	+ fts3EvalNextRow(pCsr, pRight, pRc);
	118105	+ }else{
	118106	+ fts3EvalNextRow(pCsr, pLeft, pRc);
	118107	+ fts3EvalNextRow(pCsr, pRight, pRc);
	118108	+ }
	118109	+
	118110	+ pExpr->bEof = (pLeft->bEof && pRight->bEof);
	118111	+ iCmp = DOCID_CMP(pLeft->iDocid, pRight->iDocid);
	118112	+ if( pRight->bEof \|\| (pLeft->bEof==0 && iCmp<0) ){
	118113	+ pExpr->iDocid = pLeft->iDocid;
	118114	+ }else{
	118115	+ pExpr->iDocid = pRight->iDocid;
	118116	+ }
	118117	+
	118118	+ break;
	118119	+ }
	118120	+
	118121	+ case FTSQUERY_NOT: {
	118122	+ Fts3Expr *pLeft = pExpr->pLeft;
	118123	+ Fts3Expr *pRight = pExpr->pRight;
	118124	+
	118125	+ if( pRight->bStart==0 ){
	118126	+ fts3EvalNextRow(pCsr, pRight, pRc);
	118127	+ assert( *pRc!=SQLITE_OK \|\| pRight->bStart );
	118128	+ }
	118129	+
	118130	+ fts3EvalNextRow(pCsr, pLeft, pRc);
	118131	+ if( pLeft->bEof==0 ){
	118132	+ while( !*pRc
	118133	+ && !pRight->bEof
	118134	+ && DOCID_CMP(pLeft->iDocid, pRight->iDocid)>0
	118135	+ ){
	118136	+ fts3EvalNextRow(pCsr, pRight, pRc);
	118137	+ }
	118138	+ }
	118139	+ pExpr->iDocid = pLeft->iDocid;
	118140	+ pExpr->bEof = pLeft->bEof;
	118141	+ break;
	118142	+ }
	118143	+
	118144	+ default: {
	118145	+ Fts3Phrase *pPhrase = pExpr->pPhrase;
	118146	+ fts3EvalInvalidatePoslist(pPhrase);
	118147	+ *pRc = fts3EvalPhraseNext(pCsr, pPhrase, &pExpr->bEof);
	118148	+ pExpr->iDocid = pPhrase->doclist.iDocid;
	118149	+ break;
	118150	+ }
	118151	+ }
	118152	+ }
	118153	+}
	118154	+
	118155	+/*
	118156	+** If *pRc is not SQLITE_OK, or if pExpr is not the root node of a NEAR
	118157	+** cluster, then this function returns 1 immediately.
	118158	+**
	118159	+** Otherwise, it checks if the current row really does match the NEAR
	118160	+** expression, using the data currently stored in the position lists
	118161	+** (Fts3Expr->pPhrase.doclist.pList/nList) for each phrase in the expression.
	118162	+**
	118163	+** If the current row is a match, the position list associated with each
	118164	+** phrase in the NEAR expression is edited in place to contain only those
	118165	+** phrase instances sufficiently close to their peers to satisfy all NEAR
	118166	+** constraints. In this case it returns 1. If the NEAR expression does not
	118167	+** match the current row, 0 is returned. The position lists may or may not
	118168	+** be edited if 0 is returned.
	118169	+*/
115701	118170	static int fts3EvalNearTest(Fts3Expr pExpr, int pRc){
115702	118171	int res = 1;
115703	118172
115704	118173	/* The following block runs if pExpr is the root of a NEAR query.
115705	118174	** For example, the query:
		@@ -115717,11 +118186,11 @@
115717	118186	** \| \|
115718	118187	** "w" "x"
115719	118188	**
115720	118189	** The right-hand child of a NEAR node is always a phrase. The
115721	118190	** left-hand child may be either a phrase or a NEAR node. There are
115722		- ** no exceptions to this.
	118191	+ ** no exceptions to this - it's the way the parser in fts3_expr.c works.
115723	118192	*/
115724	118193	if( *pRc==SQLITE_OK
115725	118194	&& pExpr->eType==FTSQUERY_NEAR
115726	118195	&& pExpr->bEof==0
115727	118196	&& (pExpr->pParent==0 \|\| pExpr->pParent->eType!=FTSQUERY_NEAR)
		@@ -115744,21 +118213,21 @@
115744	118213	int nToken = p->pPhrase->nToken;
115745	118214
115746	118215	for(p=p->pParent;res && p && p->eType==FTSQUERY_NEAR; p=p->pParent){
115747	118216	Fts3Phrase *pPhrase = p->pRight->pPhrase;
115748	118217	int nNear = p->nNear;
115749		- res = fts3EvalNearTrim2(nNear, aTmp, &aPoslist, &nToken, pPhrase);
	118218	+ res = fts3EvalNearTrim(nNear, aTmp, &aPoslist, &nToken, pPhrase);
115750	118219	}
115751	118220
115752	118221	aPoslist = pExpr->pRight->pPhrase->doclist.pList;
115753	118222	nToken = pExpr->pRight->pPhrase->nToken;
115754	118223	for(p=pExpr->pLeft; p && res; p=p->pLeft){
115755	118224	int nNear = p->pParent->nNear;
115756	118225	Fts3Phrase *pPhrase = (
115757	118226	p->eType==FTSQUERY_NEAR ? p->pRight->pPhrase : p->pPhrase
115758	118227	);
115759		- res = fts3EvalNearTrim2(nNear, aTmp, &aPoslist, &nToken, pPhrase);
	118228	+ res = fts3EvalNearTrim(nNear, aTmp, &aPoslist, &nToken, pPhrase);
115760	118229	}
115761	118230	}
115762	118231
115763	118232	sqlite3_free(aTmp);
115764	118233	}
		@@ -115765,132 +118234,33 @@
115765	118234
115766	118235	return res;
115767	118236	}
115768	118237
115769	118238	/*
115770		-** This macro is used by the fts3EvalNext() function. The two arguments are
115771		-** 64-bit docid values. If the current query is "ORDER BY docid ASC", then
115772		-** the macro returns (i1 - i2). Or if it is "ORDER BY docid DESC", then
115773		-** it returns (i2 - i1). This allows the same code to be used for merging
115774		-** doclists in ascending or descending order.
	118239	+** This function is a helper function for fts3EvalTestDeferredAndNear().
	118240	+** Assuming no error occurs or has occurred, It returns non-zero if the
	118241	+** expression passed as the second argument matches the row that pCsr
	118242	+** currently points to, or zero if it does not.
	118243	+**
	118244	+** If *pRc is not SQLITE_OK when this function is called, it is a no-op.
	118245	+** If an error occurs during execution of this function, *pRc is set to
	118246	+** the appropriate SQLite error code. In this case the returned value is
	118247	+** undefined.
115775	118248	*/
115776		-#define DOCID_CMP(i1, i2) ((pCsr->bDesc?-1:1) * (i1-i2))
115777		-
115778		-static void fts3EvalNext(
115779		- Fts3Cursor *pCsr,
115780		- Fts3Expr *pExpr,
115781		- int *pRc
115782		-){
115783		- if( *pRc==SQLITE_OK ){
115784		- assert( pExpr->bEof==0 );
115785		- pExpr->bStart = 1;
115786		-
115787		- switch( pExpr->eType ){
115788		- case FTSQUERY_NEAR:
115789		- case FTSQUERY_AND: {
115790		- Fts3Expr *pLeft = pExpr->pLeft;
115791		- Fts3Expr *pRight = pExpr->pRight;
115792		- assert( !pLeft->bDeferred \|\| !pRight->bDeferred );
115793		- if( pLeft->bDeferred ){
115794		- fts3EvalNext(pCsr, pRight, pRc);
115795		- pExpr->iDocid = pRight->iDocid;
115796		- pExpr->bEof = pRight->bEof;
115797		- }else if( pRight->bDeferred ){
115798		- fts3EvalNext(pCsr, pLeft, pRc);
115799		- pExpr->iDocid = pLeft->iDocid;
115800		- pExpr->bEof = pLeft->bEof;
115801		- }else{
115802		- fts3EvalNext(pCsr, pLeft, pRc);
115803		- fts3EvalNext(pCsr, pRight, pRc);
115804		-
115805		- while( !pLeft->bEof && !pRight->bEof && *pRc==SQLITE_OK ){
115806		- sqlite3_int64 iDiff = DOCID_CMP(pLeft->iDocid, pRight->iDocid);
115807		- if( iDiff==0 ) break;
115808		- if( iDiff<0 ){
115809		- fts3EvalNext(pCsr, pLeft, pRc);
115810		- }else{
115811		- fts3EvalNext(pCsr, pRight, pRc);
115812		- }
115813		- }
115814		-
115815		- pExpr->iDocid = pLeft->iDocid;
115816		- pExpr->bEof = (pLeft->bEof \|\| pRight->bEof);
115817		- }
115818		- break;
115819		- }
115820		-
115821		- case FTSQUERY_OR: {
115822		- Fts3Expr *pLeft = pExpr->pLeft;
115823		- Fts3Expr *pRight = pExpr->pRight;
115824		- sqlite3_int64 iCmp = DOCID_CMP(pLeft->iDocid, pRight->iDocid);
115825		-
115826		- assert( pLeft->bStart \|\| pLeft->iDocid==pRight->iDocid );
115827		- assert( pRight->bStart \|\| pLeft->iDocid==pRight->iDocid );
115828		-
115829		- if( pRight->bEof \|\| (pLeft->bEof==0 && iCmp<0) ){
115830		- fts3EvalNext(pCsr, pLeft, pRc);
115831		- }else if( pLeft->bEof \|\| (pRight->bEof==0 && iCmp>0) ){
115832		- fts3EvalNext(pCsr, pRight, pRc);
115833		- }else{
115834		- fts3EvalNext(pCsr, pLeft, pRc);
115835		- fts3EvalNext(pCsr, pRight, pRc);
115836		- }
115837		-
115838		- pExpr->bEof = (pLeft->bEof && pRight->bEof);
115839		- iCmp = DOCID_CMP(pLeft->iDocid, pRight->iDocid);
115840		- if( pRight->bEof \|\| (pLeft->bEof==0 && iCmp<0) ){
115841		- pExpr->iDocid = pLeft->iDocid;
115842		- }else{
115843		- pExpr->iDocid = pRight->iDocid;
115844		- }
115845		-
115846		- break;
115847		- }
115848		-
115849		- case FTSQUERY_NOT: {
115850		- Fts3Expr *pLeft = pExpr->pLeft;
115851		- Fts3Expr *pRight = pExpr->pRight;
115852		-
115853		- if( pRight->bStart==0 ){
115854		- fts3EvalNext(pCsr, pRight, pRc);
115855		- assert( *pRc!=SQLITE_OK \|\| pRight->bStart );
115856		- }
115857		-
115858		- fts3EvalNext(pCsr, pLeft, pRc);
115859		- if( pLeft->bEof==0 ){
115860		- while( !*pRc
115861		- && !pRight->bEof
115862		- && DOCID_CMP(pLeft->iDocid, pRight->iDocid)>0
115863		- ){
115864		- fts3EvalNext(pCsr, pRight, pRc);
115865		- }
115866		- }
115867		- pExpr->iDocid = pLeft->iDocid;
115868		- pExpr->bEof = pLeft->bEof;
115869		- break;
115870		- }
115871		-
115872		- default: {
115873		- Fts3Phrase *pPhrase = pExpr->pPhrase;
115874		- fts3EvalZeroPoslist(pPhrase);
115875		- *pRc = fts3EvalPhraseNext(pCsr, pPhrase, &pExpr->bEof);
115876		- pExpr->iDocid = pPhrase->doclist.iDocid;
115877		- break;
115878		- }
115879		- }
115880		- }
115881		-}
115882		-
115883		-static int fts3EvalDeferredTest(Fts3Cursor pCsr, Fts3Expr pExpr, int *pRc){
115884		- int bHit = 1;
	118249	+static int fts3EvalTestExpr(
	118250	+ Fts3Cursor pCsr, / FTS cursor handle */
	118251	+ Fts3Expr pExpr, / Expr to test. May or may not be root. */
	118252	+ int pRc / IN/OUT: Error code */
	118253	+){
	118254	+ int bHit = 1; /* Return value */
115885	118255	if( *pRc==SQLITE_OK ){
115886	118256	switch( pExpr->eType ){
115887	118257	case FTSQUERY_NEAR:
115888	118258	case FTSQUERY_AND:
115889	118259	bHit = (
115890		- fts3EvalDeferredTest(pCsr, pExpr->pLeft, pRc)
115891		- && fts3EvalDeferredTest(pCsr, pExpr->pRight, pRc)
	118260	+ fts3EvalTestExpr(pCsr, pExpr->pLeft, pRc)
	118261	+ && fts3EvalTestExpr(pCsr, pExpr->pRight, pRc)
115892	118262	&& fts3EvalNearTest(pExpr, pRc)
115893	118263	);
115894	118264
115895	118265	/* If the NEAR expression does not match any rows, zero the doclist for
115896	118266	** all phrases involved in the NEAR. This is because the snippet(),
		@@ -115912,31 +118282,31 @@
115912	118282	&& (pExpr->pParent==0 \|\| pExpr->pParent->eType!=FTSQUERY_NEAR)
115913	118283	){
115914	118284	Fts3Expr *p;
115915	118285	for(p=pExpr; p->pPhrase==0; p=p->pLeft){
115916	118286	if( p->pRight->iDocid==pCsr->iPrevId ){
115917		- fts3EvalZeroPoslist(p->pRight->pPhrase);
	118287	+ fts3EvalInvalidatePoslist(p->pRight->pPhrase);
115918	118288	}
115919	118289	}
115920	118290	if( p->iDocid==pCsr->iPrevId ){
115921		- fts3EvalZeroPoslist(p->pPhrase);
	118291	+ fts3EvalInvalidatePoslist(p->pPhrase);
115922	118292	}
115923	118293	}
115924	118294
115925	118295	break;
115926	118296
115927	118297	case FTSQUERY_OR: {
115928		- int bHit1 = fts3EvalDeferredTest(pCsr, pExpr->pLeft, pRc);
115929		- int bHit2 = fts3EvalDeferredTest(pCsr, pExpr->pRight, pRc);
	118298	+ int bHit1 = fts3EvalTestExpr(pCsr, pExpr->pLeft, pRc);
	118299	+ int bHit2 = fts3EvalTestExpr(pCsr, pExpr->pRight, pRc);
115930	118300	bHit = bHit1 \|\| bHit2;
115931	118301	break;
115932	118302	}
115933	118303
115934	118304	case FTSQUERY_NOT:
115935	118305	bHit = (
115936		- fts3EvalDeferredTest(pCsr, pExpr->pLeft, pRc)
115937		- && !fts3EvalDeferredTest(pCsr, pExpr->pRight, pRc)
	118306	+ fts3EvalTestExpr(pCsr, pExpr->pLeft, pRc)
	118307	+ && !fts3EvalTestExpr(pCsr, pExpr->pRight, pRc)
115938	118308	);
115939	118309	break;
115940	118310
115941	118311	default: {
115942	118312	if( pCsr->pDeferred
		@@ -115943,11 +118313,11 @@
115943	118313	&& (pExpr->iDocid==pCsr->iPrevId \|\| pExpr->bDeferred)
115944	118314	){
115945	118315	Fts3Phrase *pPhrase = pExpr->pPhrase;
115946	118316	assert( pExpr->bDeferred \|\| pPhrase->doclist.bFreeList==0 );
115947	118317	if( pExpr->bDeferred ){
115948		- fts3EvalZeroPoslist(pPhrase);
	118318	+ fts3EvalInvalidatePoslist(pPhrase);
115949	118319	}
115950	118320	*pRc = fts3EvalDeferredPhrase(pCsr, pPhrase);
115951	118321	bHit = (pPhrase->doclist.pList!=0);
115952	118322	pExpr->iDocid = pCsr->iPrevId;
115953	118323	}else{
		@@ -115959,31 +118329,53 @@
115959	118329	}
115960	118330	return bHit;
115961	118331	}
115962	118332
115963	118333	/*
115964		-** Return 1 if both of the following are true:
	118334	+** This function is called as the second part of each xNext operation when
	118335	+** iterating through the results of a full-text query. At this point the
	118336	+** cursor points to a row that matches the query expression, with the
	118337	+** following caveats:
	118338	+**
	118339	+** * Up until this point, "NEAR" operators in the expression have been
	118340	+** treated as "AND".
	118341	+**
	118342	+** * Deferred tokens have not yet been considered.
	118343	+**
	118344	+** If *pRc is not SQLITE_OK when this function is called, it immediately
	118345	+** returns 0. Otherwise, it tests whether or not after considering NEAR
	118346	+** operators and deferred tokens the current row is still a match for the
	118347	+** expression. It returns 1 if both of the following are true:
115965	118348	**
115966	118349	** 1. *pRc is SQLITE_OK when this function returns, and
115967	118350	**
115968	118351	** 2. After scanning the current FTS table row for the deferred tokens,
115969		-** it is determined that the row does not match the query.
	118352	+** it is determined that the row does not match the query.
115970	118353	**
115971	118354	** Or, if no error occurs and it seems the current row does match the FTS
115972	118355	** query, return 0.
115973	118356	*/
115974		-static int fts3EvalLoadDeferred(Fts3Cursor pCsr, int pRc){
	118357	+static int fts3EvalTestDeferredAndNear(Fts3Cursor pCsr, int pRc){
115975	118358	int rc = *pRc;
115976	118359	int bMiss = 0;
115977	118360	if( rc==SQLITE_OK ){
	118361	+
	118362	+ /* If there are one or more deferred tokens, load the current row into
	118363	+ ** memory and scan it to determine the position list for each deferred
	118364	+ ** token. Then, see if this row is really a match, considering deferred
	118365	+ ** tokens and NEAR operators (neither of which were taken into account
	118366	+ ** earlier, by fts3EvalNextRow()).
	118367	+ */
115978	118368	if( pCsr->pDeferred ){
115979	118369	rc = fts3CursorSeek(0, pCsr);
115980	118370	if( rc==SQLITE_OK ){
115981	118371	rc = sqlite3Fts3CacheDeferredDoclists(pCsr);
115982	118372	}
115983	118373	}
115984		- bMiss = (0==fts3EvalDeferredTest(pCsr, pCsr->pExpr, &rc));
	118374	+ bMiss = (0==fts3EvalTestExpr(pCsr, pCsr->pExpr, &rc));
	118375	+
	118376	+ /* Free the position-lists accumulated for each deferred token above. */
115985	118377	sqlite3Fts3FreeDeferredDoclists(pCsr);
115986	118378	*pRc = rc;
115987	118379	}
115988	118380	return (rc==SQLITE_OK && bMiss);
115989	118381	}
		@@ -115990,11 +118382,11 @@
115990	118382
115991	118383	/*
115992	118384	** Advance to the next document that matches the FTS expression in
115993	118385	** Fts3Cursor.pExpr.
115994	118386	*/
115995		-SQLITE_PRIVATE int sqlite3Fts3EvalNext(Fts3Cursor *pCsr){
	118387	+static int fts3EvalNext(Fts3Cursor *pCsr){
115996	118388	int rc = SQLITE_OK; /* Return Code */
115997	118389	Fts3Expr *pExpr = pCsr->pExpr;
115998	118390	assert( pCsr->isEof==0 );
115999	118391	if( pExpr==0 ){
116000	118392	pCsr->isEof = 1;
		@@ -116002,23 +118394,23 @@
116002	118394	do {
116003	118395	if( pCsr->isRequireSeek==0 ){
116004	118396	sqlite3_reset(pCsr->pStmt);
116005	118397	}
116006	118398	assert( sqlite3_data_count(pCsr->pStmt)==0 );
116007		- fts3EvalNext(pCsr, pExpr, &rc);
	118399	+ fts3EvalNextRow(pCsr, pExpr, &rc);
116008	118400	pCsr->isEof = pExpr->bEof;
116009	118401	pCsr->isRequireSeek = 1;
116010	118402	pCsr->isMatchinfoNeeded = 1;
116011	118403	pCsr->iPrevId = pExpr->iDocid;
116012		- }while( pCsr->isEof==0 && fts3EvalLoadDeferred(pCsr, &rc) );
	118404	+ }while( pCsr->isEof==0 && fts3EvalTestDeferredAndNear(pCsr, &rc) );
116013	118405	}
116014	118406	return rc;
116015	118407	}
116016	118408
116017	118409	/*
116018	118410	** Restart interation for expression pExpr so that the next call to
116019		-** sqlite3Fts3EvalNext() visits the first row. Do not allow incremental
	118411	+** fts3EvalNext() visits the first row. Do not allow incremental
116020	118412	** loading or merging of phrase doclists for this iteration.
116021	118413	**
116022	118414	** If *pRc is other than SQLITE_OK when this function is called, it is
116023	118415	** a no-op. If an error occurs within this function, *pRc is set to an
116024	118416	** SQLite error code before returning.
		@@ -116030,19 +118422,20 @@
116030	118422	){
116031	118423	if( pExpr && *pRc==SQLITE_OK ){
116032	118424	Fts3Phrase *pPhrase = pExpr->pPhrase;
116033	118425
116034	118426	if( pPhrase ){
116035		- fts3EvalZeroPoslist(pPhrase);
	118427	+ fts3EvalInvalidatePoslist(pPhrase);
116036	118428	if( pPhrase->bIncr ){
116037		- sqlite3Fts3EvalPhraseCleanup(pPhrase);
116038		- memset(&pPhrase->doclist, 0, sizeof(Fts3Doclist));
116039		- *pRc = sqlite3Fts3EvalStart(pCsr, pExpr, 0);
116040		- }else{
116041		- pPhrase->doclist.pNextDocid = 0;
116042		- pPhrase->doclist.iDocid = 0;
	118429	+ assert( pPhrase->nToken==1 );
	118430	+ assert( pPhrase->aToken[0].pSegcsr );
	118431	+ sqlite3Fts3MsrIncrRestart(pPhrase->aToken[0].pSegcsr);
	118432	+ *pRc = fts3EvalPhraseStart(pCsr, 0, pPhrase);
116043	118433	}
	118434	+
	118435	+ pPhrase->doclist.pNextDocid = 0;
	118436	+ pPhrase->doclist.iDocid = 0;
116044	118437	}
116045	118438
116046	118439	pExpr->iDocid = 0;
116047	118440	pExpr->bEof = 0;
116048	118441	pExpr->bStart = 0;
		@@ -116145,18 +118538,18 @@
116145	118538	/* Ensure the %_content statement is reset. */
116146	118539	if( pCsr->isRequireSeek==0 ) sqlite3_reset(pCsr->pStmt);
116147	118540	assert( sqlite3_data_count(pCsr->pStmt)==0 );
116148	118541
116149	118542	/* Advance to the next document */
116150		- fts3EvalNext(pCsr, pRoot, &rc);
	118543	+ fts3EvalNextRow(pCsr, pRoot, &rc);
116151	118544	pCsr->isEof = pRoot->bEof;
116152	118545	pCsr->isRequireSeek = 1;
116153	118546	pCsr->isMatchinfoNeeded = 1;
116154	118547	pCsr->iPrevId = pRoot->iDocid;
116155	118548	}while( pCsr->isEof==0
116156	118549	&& pRoot->eType==FTSQUERY_NEAR
116157		- && fts3EvalLoadDeferred(pCsr, &rc)
	118550	+ && fts3EvalTestDeferredAndNear(pCsr, &rc)
116158	118551	);
116159	118552
116160	118553	if( rc==SQLITE_OK && pCsr->isEof==0 ){
116161	118554	fts3EvalUpdateCounts(pRoot);
116162	118555	}
		@@ -116174,14 +118567,14 @@
116174	118567	**
116175	118568	** do {...} while( pRoot->iDocid<iDocid && rc==SQLITE_OK );
116176	118569	*/
116177	118570	fts3EvalRestart(pCsr, pRoot, &rc);
116178	118571	do {
116179		- fts3EvalNext(pCsr, pRoot, &rc);
	118572	+ fts3EvalNextRow(pCsr, pRoot, &rc);
116180	118573	assert( pRoot->bEof==0 );
116181	118574	}while( pRoot->iDocid!=iDocid && rc==SQLITE_OK );
116182		- fts3EvalLoadDeferred(pCsr, &rc);
	118575	+ fts3EvalTestDeferredAndNear(pCsr, &rc);
116183	118576	}
116184	118577	}
116185	118578	return rc;
116186	118579	}
116187	118580
		@@ -116225,12 +118618,12 @@
116225	118618	int iCol;
116226	118619
116227	118620	if( pExpr->bDeferred && pExpr->pParent->eType!=FTSQUERY_NEAR ){
116228	118621	assert( pCsr->nDoc>0 );
116229	118622	for(iCol=0; iCol<pTab->nColumn; iCol++){
116230		- aiOut[iCol*3 + 1] = pCsr->nDoc;
116231		- aiOut[iCol*3 + 2] = pCsr->nDoc;
	118623	+ aiOut[iCol*3 + 1] = (u32)pCsr->nDoc;
	118624	+ aiOut[iCol*3 + 2] = (u32)pCsr->nDoc;
116232	118625	}
116233	118626	}else{
116234	118627	rc = fts3EvalGatherStats(pCsr, pExpr);
116235	118628	if( rc==SQLITE_OK ){
116236	118629	assert( pExpr->aMI );
		@@ -116308,18 +118701,32 @@
116308	118701	*/
116309	118702	SQLITE_PRIVATE void sqlite3Fts3EvalPhraseCleanup(Fts3Phrase *pPhrase){
116310	118703	if( pPhrase ){
116311	118704	int i;
116312	118705	sqlite3_free(pPhrase->doclist.aAll);
116313		- fts3EvalZeroPoslist(pPhrase);
	118706	+ fts3EvalInvalidatePoslist(pPhrase);
116314	118707	memset(&pPhrase->doclist, 0, sizeof(Fts3Doclist));
116315	118708	for(i=0; i<pPhrase->nToken; i++){
116316	118709	fts3SegReaderCursorFree(pPhrase->aToken[i].pSegcsr);
116317	118710	pPhrase->aToken[i].pSegcsr = 0;
116318	118711	}
116319	118712	}
116320	118713	}
	118714	+
	118715	+#if !SQLITE_CORE
	118716	+/*
	118717	+** Initialize API pointer table, if required.
	118718	+*/
	118719	+SQLITE_API int sqlite3_extension_init(
	118720	+ sqlite3 *db,
	118721	+ char **pzErrMsg,
	118722	+ const sqlite3_api_routines *pApi
	118723	+){
	118724	+ SQLITE_EXTENSION_INIT2(pApi)
	118725	+ return sqlite3Fts3Init(db);
	118726	+}
	118727	+#endif
116321	118728
116322	118729	#endif
116323	118730
116324	118731	/************ End of fts3.c **********************************************/
116325	118732	/************ Begin file fts3_aux.c **************************************/
		@@ -116334,13 +118741,14 @@
116334	118741	** May you share freely, never taking more than you give.
116335	118742	**
116336	118743	******************************************************************************
116337	118744	**
116338	118745	*/
116339		-
116340	118746	#if !defined(SQLITE_CORE) \|\| defined(SQLITE_ENABLE_FTS3)
116341	118747
	118748	+/* #include <string.h> */
	118749	+/* #include <assert.h> */
116342	118750
116343	118751	typedef struct Fts3auxTable Fts3auxTable;
116344	118752	typedef struct Fts3auxCursor Fts3auxCursor;
116345	118753
116346	118754	struct Fts3auxTable {
		@@ -116875,10 +119283,12 @@
116875	119283	/*
116876	119284	** Default span for NEAR operators.
116877	119285	*/
116878	119286	#define SQLITE_FTS3_DEFAULT_NEAR_PARAM 10
116879	119287
	119288	+/* #include <string.h> */
	119289	+/* #include <assert.h> */
116880	119290
116881	119291	/*
116882	119292	** isNot:
116883	119293	** This variable is used by function getNextNode(). When getNextNode() is
116884	119294	** called, it sets ParseContext.isNot to true if the 'next node' is a
		@@ -117576,10 +119986,11 @@
117576	119986	** Everything after this point is just test code.
117577	119987	*/
117578	119988
117579	119989	#ifdef SQLITE_TEST
117580	119990
	119991	+/* #include <stdio.h> */
117581	119992
117582	119993	/*
117583	119994	** Function to query the hash-table of tokenizers (see README.tokenizers).
117584	119995	*/
117585	119996	static int queryTestTokenizer(
		@@ -117786,10 +120197,13 @@
117786	120197	** * The FTS3 module is being built into the core of
117787	120198	** SQLite (in which case SQLITE_ENABLE_FTS3 is defined).
117788	120199	*/
117789	120200	#if !defined(SQLITE_CORE) \|\| defined(SQLITE_ENABLE_FTS3)
117790	120201
	120202	+/* #include <assert.h> */
	120203	+/* #include <stdlib.h> */
	120204	+/* #include <string.h> */
117791	120205
117792	120206
117793	120207	/*
117794	120208	** Malloc and Free functions
117795	120209	*/
		@@ -118166,11 +120580,14 @@
118166	120580	** * The FTS3 module is being built into the core of
118167	120581	** SQLite (in which case SQLITE_ENABLE_FTS3 is defined).
118168	120582	*/
118169	120583	#if !defined(SQLITE_CORE) \|\| defined(SQLITE_ENABLE_FTS3)
118170	120584
118171		-
	120585	+/* #include <assert.h> */
	120586	+/* #include <stdlib.h> */
	120587	+/* #include <stdio.h> */
	120588	+/* #include <string.h> */
118172	120589
118173	120590
118174	120591	/*
118175	120592	** Class derived from sqlite3_tokenizer
118176	120593	*/
		@@ -118810,14 +121227,12 @@
118810	121227	** * The FTS3 module is being built into the core of
118811	121228	** SQLite (in which case SQLITE_ENABLE_FTS3 is defined).
118812	121229	*/
118813	121230	#if !defined(SQLITE_CORE) \|\| defined(SQLITE_ENABLE_FTS3)
118814	121231
118815		-#ifndef SQLITE_CORE
118816		- SQLITE_EXTENSION_INIT1
118817		-#endif
118818		-
	121232	+/* #include <assert.h> */
	121233	+/* #include <string.h> */
118819	121234
118820	121235	/*
118821	121236	** Implementation of the SQL scalar function for accessing the underlying
118822	121237	** hash table. This function may be called as follows:
118823	121238	**
		@@ -118937,11 +121352,11 @@
118937	121352	sqlite3_tokenizer *ppTok, / OUT: Tokenizer (if applicable) */
118938	121353	char *pzErr / OUT: Set to malloced error message */
118939	121354	){
118940	121355	int rc;
118941	121356	char z = (char )zArg;
118942		- int n;
	121357	+ int n = 0;
118943	121358	char *zCopy;
118944	121359	char zEnd; / Pointer to nul-term of zCopy */
118945	121360	sqlite3_tokenizer_module *m;
118946	121361
118947	121362	zCopy = sqlite3_mprintf("%s", zArg);
		@@ -118989,10 +121404,12 @@
118989	121404	}
118990	121405
118991	121406
118992	121407	#ifdef SQLITE_TEST
118993	121408
	121409	+/* #include <tcl.h> */
	121410	+/* #include <string.h> */
118994	121411
118995	121412	/*
118996	121413	** Implementation of a special SQL scalar function for testing tokenizers
118997	121414	** designed to be used in concert with the Tcl testing framework. This
118998	121415	** function must be called with two arguments:
		@@ -119300,11 +121717,14 @@
119300	121717	** * The FTS3 module is being built into the core of
119301	121718	** SQLite (in which case SQLITE_ENABLE_FTS3 is defined).
119302	121719	*/
119303	121720	#if !defined(SQLITE_CORE) \|\| defined(SQLITE_ENABLE_FTS3)
119304	121721
119305		-
	121722	+/* #include <assert.h> */
	121723	+/* #include <stdlib.h> */
	121724	+/* #include <stdio.h> */
	121725	+/* #include <string.h> */
119306	121726
119307	121727
119308	121728	typedef struct simple_tokenizer {
119309	121729	sqlite3_tokenizer base;
119310	121730	char delim[128]; /* flag ASCII delimiters */
		@@ -119526,10 +121946,13 @@
119526	121946	** code in fts3.c.
119527	121947	*/
119528	121948
119529	121949	#if !defined(SQLITE_CORE) \|\| defined(SQLITE_ENABLE_FTS3)
119530	121950
	121951	+/* #include <string.h> */
	121952	+/* #include <assert.h> */
	121953	+/* #include <stdlib.h> */
119531	121954
119532	121955	/*
119533	121956	** When full-text index nodes are loaded from disk, the buffer that they
119534	121957	** are loaded into has the following number of bytes of padding at the end
119535	121958	** of it. i.e. if a full-text index node is 900 bytes in size, then a buffer
		@@ -120750,10 +123173,11 @@
120750	123173	pReader->pOffsetList = 0;
120751	123174	}else{
120752	123175	pReader->pOffsetList = p;
120753	123176	}
120754	123177	}else{
	123178	+ char *pEnd = &pReader->aDoclist[pReader->nDoclist];
120755	123179
120756	123180	/* Pointer p currently points at the first byte of an offset list. The
120757	123181	** following block advances it to point one byte past the end of
120758	123182	** the same offset list. */
120759	123183	while( 1 ){
		@@ -120778,17 +123202,19 @@
120778	123202	*/
120779	123203	if( ppOffsetList ){
120780	123204	*ppOffsetList = pReader->pOffsetList;
120781	123205	*pnOffsetList = (int)(p - pReader->pOffsetList - 1);
120782	123206	}
	123207	+
	123208	+ while( p<pEnd && *p==0 ) p++;
120783	123209
120784	123210	/* If there are no more entries in the doclist, set pOffsetList to
120785	123211	** NULL. Otherwise, set Fts3SegReader.iDocid to the next docid and
120786	123212	** Fts3SegReader.pOffsetList to point to the next offset list before
120787	123213	** returning.
120788	123214	*/
120789		- if( p>=&pReader->aDoclist[pReader->nDoclist] ){
	123215	+ if( p>=pEnd ){
120790	123216	pReader->pOffsetList = 0;
120791	123217	}else{
120792	123218	rc = fts3SegReaderRequire(pReader, p, FTS3_VARINT_MAX);
120793	123219	if( rc==SQLITE_OK ){
120794	123220	sqlite3_int64 iDelta;
		@@ -120823,11 +123249,11 @@
120823	123249	for(ii=0; rc==SQLITE_OK && ii<pMsr->nSegment; ii++){
120824	123250	Fts3SegReader *pReader = pMsr->apSegment[ii];
120825	123251	if( !fts3SegReaderIsPending(pReader)
120826	123252	&& !fts3SegReaderIsRootOnly(pReader)
120827	123253	){
120828		- int jj;
	123254	+ sqlite3_int64 jj;
120829	123255	for(jj=pReader->iStartBlock; jj<=pReader->iLeafEndBlock; jj++){
120830	123256	int nBlob;
120831	123257	rc = sqlite3Fts3ReadBlock(p, jj, 0, &nBlob, 0);
120832	123258	if( rc!=SQLITE_OK ) break;
120833	123259	if( (nBlob+35)>pgsz ){
		@@ -121768,55 +124194,31 @@
121768	124194
121769	124195	*ppList = pList;
121770	124196	*pnList = nList;
121771	124197	}
121772	124198
121773		-SQLITE_PRIVATE int sqlite3Fts3MsrIncrStart(
121774		- Fts3Table p, / Virtual table handle */
121775		- Fts3MultiSegReader pCsr, / Cursor object */
121776		- int iCol, /* Column to match on. */
121777		- const char zTerm, / Term to iterate through a doclist for */
121778		- int nTerm /* Number of bytes in zTerm */
	124199	+/*
	124200	+** Cache data in the Fts3MultiSegReader.aBuffer[] buffer (overwriting any
	124201	+** existing data). Grow the buffer if required.
	124202	+**
	124203	+** If successful, return SQLITE_OK. Otherwise, if an OOM error is encountered
	124204	+** trying to resize the buffer, return SQLITE_NOMEM.
	124205	+*/
	124206	+static int fts3MsrBufferData(
	124207	+ Fts3MultiSegReader pMsr, / Multi-segment-reader handle */
	124208	+ char *pList,
	124209	+ int nList
121779	124210	){
121780		- int i;
121781		- int nSegment = pCsr->nSegment;
121782		- int (xCmp)(Fts3SegReader , Fts3SegReader *) = (
121783		- p->bDescIdx ? fts3SegReaderDoclistCmpRev : fts3SegReaderDoclistCmp
121784		- );
121785		-
121786		- assert( pCsr->pFilter==0 );
121787		- assert( zTerm && nTerm>0 );
121788		-
121789		- /* Advance each segment iterator until it points to the term zTerm/nTerm. */
121790		- for(i=0; i<nSegment; i++){
121791		- Fts3SegReader *pSeg = pCsr->apSegment[i];
121792		- do {
121793		- int rc = fts3SegReaderNext(p, pSeg, 1);
121794		- if( rc!=SQLITE_OK ) return rc;
121795		- }while( fts3SegReaderTermCmp(pSeg, zTerm, nTerm)<0 );
121796		- }
121797		- fts3SegReaderSort(pCsr->apSegment, nSegment, nSegment, fts3SegReaderCmp);
121798		-
121799		- /* Determine how many of the segments actually point to zTerm/nTerm. */
121800		- for(i=0; i<nSegment; i++){
121801		- Fts3SegReader *pSeg = pCsr->apSegment[i];
121802		- if( !pSeg->aNode \|\| fts3SegReaderTermCmp(pSeg, zTerm, nTerm) ){
121803		- break;
121804		- }
121805		- }
121806		- pCsr->nAdvance = i;
121807		-
121808		- /* Advance each of the segments to point to the first docid. */
121809		- for(i=0; i<pCsr->nAdvance; i++){
121810		- int rc = fts3SegReaderFirstDocid(p, pCsr->apSegment[i]);
121811		- if( rc!=SQLITE_OK ) return rc;
121812		- }
121813		- fts3SegReaderSort(pCsr->apSegment, i, i, xCmp);
121814		-
121815		- assert( iCol<0 \|\| iCol<p->nColumn );
121816		- pCsr->iColFilter = iCol;
121817		-
	124211	+ if( nList>pMsr->nBuffer ){
	124212	+ char *pNew;
	124213	+ pMsr->nBuffer = nList*2;
	124214	+ pNew = (char *)sqlite3_realloc(pMsr->aBuffer, pMsr->nBuffer);
	124215	+ if( !pNew ) return SQLITE_NOMEM;
	124216	+ pMsr->aBuffer = pNew;
	124217	+ }
	124218	+
	124219	+ memcpy(pMsr->aBuffer, pList, nList);
121818	124220	return SQLITE_OK;
121819	124221	}
121820	124222
121821	124223	SQLITE_PRIVATE int sqlite3Fts3MsrIncrNext(
121822	124224	Fts3Table p, / Virtual table handle */
		@@ -121866,52 +124268,138 @@
121866	124268	if( pMsr->iColFilter>=0 ){
121867	124269	fts3ColumnFilter(pMsr->iColFilter, &pList, &nList);
121868	124270	}
121869	124271
121870	124272	if( nList>0 ){
	124273	+ if( fts3SegReaderIsPending(apSegment[0]) ){
	124274	+ rc = fts3MsrBufferData(pMsr, pList, nList+1);
	124275	+ if( rc!=SQLITE_OK ) return rc;
	124276	+ *paPoslist = pMsr->aBuffer;
	124277	+ assert( (pMsr->aBuffer[nList] & 0xFE)==0x00 );
	124278	+ }else{
	124279	+ *paPoslist = pList;
	124280	+ }
121871	124281	*piDocid = iDocid;
121872		- *paPoslist = pList;
121873	124282	*pnPoslist = nList;
121874	124283	break;
121875	124284	}
121876	124285	}
121877		-
121878	124286	}
121879	124287
	124288	+ return SQLITE_OK;
	124289	+}
	124290	+
	124291	+static int fts3SegReaderStart(
	124292	+ Fts3Table p, / Virtual table handle */
	124293	+ Fts3MultiSegReader pCsr, / Cursor object */
	124294	+ const char zTerm, / Term searched for (or NULL) */
	124295	+ int nTerm /* Length of zTerm in bytes */
	124296	+){
	124297	+ int i;
	124298	+ int nSeg = pCsr->nSegment;
	124299	+
	124300	+ /* If the Fts3SegFilter defines a specific term (or term prefix) to search
	124301	+ ** for, then advance each segment iterator until it points to a term of
	124302	+ ** equal or greater value than the specified term. This prevents many
	124303	+ ** unnecessary merge/sort operations for the case where single segment
	124304	+ ** b-tree leaf nodes contain more than one term.
	124305	+ */
	124306	+ for(i=0; pCsr->bRestart==0 && i<pCsr->nSegment; i++){
	124307	+ Fts3SegReader *pSeg = pCsr->apSegment[i];
	124308	+ do {
	124309	+ int rc = fts3SegReaderNext(p, pSeg, 0);
	124310	+ if( rc!=SQLITE_OK ) return rc;
	124311	+ }while( zTerm && fts3SegReaderTermCmp(pSeg, zTerm, nTerm)<0 );
	124312	+ }
	124313	+ fts3SegReaderSort(pCsr->apSegment, nSeg, nSeg, fts3SegReaderCmp);
	124314	+
121880	124315	return SQLITE_OK;
121881	124316	}
121882	124317
121883	124318	SQLITE_PRIVATE int sqlite3Fts3SegReaderStart(
121884	124319	Fts3Table p, / Virtual table handle */
121885	124320	Fts3MultiSegReader pCsr, / Cursor object */
121886	124321	Fts3SegFilter pFilter / Restrictions on range of iteration */
121887	124322	){
121888		- int i;
121889		-
121890		- /* Initialize the cursor object */
121891	124323	pCsr->pFilter = pFilter;
121892		-
121893		- /* If the Fts3SegFilter defines a specific term (or term prefix) to search
121894		- ** for, then advance each segment iterator until it points to a term of
121895		- ** equal or greater value than the specified term. This prevents many
121896		- ** unnecessary merge/sort operations for the case where single segment
121897		- ** b-tree leaf nodes contain more than one term.
121898		- */
121899		- for(i=0; i<pCsr->nSegment; i++){
121900		- int nTerm = pFilter->nTerm;
121901		- const char *zTerm = pFilter->zTerm;
	124324	+ return fts3SegReaderStart(p, pCsr, pFilter->zTerm, pFilter->nTerm);
	124325	+}
	124326	+
	124327	+SQLITE_PRIVATE int sqlite3Fts3MsrIncrStart(
	124328	+ Fts3Table p, / Virtual table handle */
	124329	+ Fts3MultiSegReader pCsr, / Cursor object */
	124330	+ int iCol, /* Column to match on. */
	124331	+ const char zTerm, / Term to iterate through a doclist for */
	124332	+ int nTerm /* Number of bytes in zTerm */
	124333	+){
	124334	+ int i;
	124335	+ int rc;
	124336	+ int nSegment = pCsr->nSegment;
	124337	+ int (xCmp)(Fts3SegReader , Fts3SegReader *) = (
	124338	+ p->bDescIdx ? fts3SegReaderDoclistCmpRev : fts3SegReaderDoclistCmp
	124339	+ );
	124340	+
	124341	+ assert( pCsr->pFilter==0 );
	124342	+ assert( zTerm && nTerm>0 );
	124343	+
	124344	+ /* Advance each segment iterator until it points to the term zTerm/nTerm. */
	124345	+ rc = fts3SegReaderStart(p, pCsr, zTerm, nTerm);
	124346	+ if( rc!=SQLITE_OK ) return rc;
	124347	+
	124348	+ /* Determine how many of the segments actually point to zTerm/nTerm. */
	124349	+ for(i=0; i<nSegment; i++){
121902	124350	Fts3SegReader *pSeg = pCsr->apSegment[i];
121903		- do {
121904		- int rc = fts3SegReaderNext(p, pSeg, 0);
121905		- if( rc!=SQLITE_OK ) return rc;
121906		- }while( zTerm && fts3SegReaderTermCmp(pSeg, zTerm, nTerm)<0 );
	124351	+ if( !pSeg->aNode \|\| fts3SegReaderTermCmp(pSeg, zTerm, nTerm) ){
	124352	+ break;
	124353	+ }
121907	124354	}
121908		- fts3SegReaderSort(
121909		- pCsr->apSegment, pCsr->nSegment, pCsr->nSegment, fts3SegReaderCmp);
	124355	+ pCsr->nAdvance = i;
	124356	+
	124357	+ /* Advance each of the segments to point to the first docid. */
	124358	+ for(i=0; i<pCsr->nAdvance; i++){
	124359	+ rc = fts3SegReaderFirstDocid(p, pCsr->apSegment[i]);
	124360	+ if( rc!=SQLITE_OK ) return rc;
	124361	+ }
	124362	+ fts3SegReaderSort(pCsr->apSegment, i, i, xCmp);
	124363	+
	124364	+ assert( iCol<0 \|\| iCol<p->nColumn );
	124365	+ pCsr->iColFilter = iCol;
	124366	+
	124367	+ return SQLITE_OK;
	124368	+}
	124369	+
	124370	+/*
	124371	+** This function is called on a MultiSegReader that has been started using
	124372	+** sqlite3Fts3MsrIncrStart(). One or more calls to MsrIncrNext() may also
	124373	+** have been made. Calling this function puts the MultiSegReader in such
	124374	+** a state that if the next two calls are:
	124375	+**
	124376	+** sqlite3Fts3SegReaderStart()
	124377	+** sqlite3Fts3SegReaderStep()
	124378	+**
	124379	+** then the entire doclist for the term is available in
	124380	+** MultiSegReader.aDoclist/nDoclist.
	124381	+*/
	124382	+SQLITE_PRIVATE int sqlite3Fts3MsrIncrRestart(Fts3MultiSegReader *pCsr){
	124383	+ int i; /* Used to iterate through segment-readers */
	124384	+
	124385	+ assert( pCsr->zTerm==0 );
	124386	+ assert( pCsr->nTerm==0 );
	124387	+ assert( pCsr->aDoclist==0 );
	124388	+ assert( pCsr->nDoclist==0 );
	124389	+
	124390	+ pCsr->nAdvance = 0;
	124391	+ pCsr->bRestart = 1;
	124392	+ for(i=0; i<pCsr->nSegment; i++){
	124393	+ pCsr->apSegment[i]->pOffsetList = 0;
	124394	+ pCsr->apSegment[i]->nOffsetList = 0;
	124395	+ pCsr->apSegment[i]->iDocid = 0;
	124396	+ }
121910	124397
121911	124398	return SQLITE_OK;
121912	124399	}
	124400	+
121913	124401
121914	124402	SQLITE_PRIVATE int sqlite3Fts3SegReaderStep(
121915	124403	Fts3Table p, / Virtual table handle */
121916	124404	Fts3MultiSegReader pCsr / Cursor object */
121917	124405	){
		@@ -121981,13 +124469,18 @@
121981	124469	assert( isIgnoreEmpty \|\| (isRequirePos && !isColFilter) );
121982	124470	if( nMerge==1
121983	124471	&& !isIgnoreEmpty
121984	124472	&& (p->bDescIdx==0 \|\| fts3SegReaderIsPending(apSegment[0])==0)
121985	124473	){
121986		- pCsr->aDoclist = apSegment[0]->aDoclist;
121987	124474	pCsr->nDoclist = apSegment[0]->nDoclist;
121988		- rc = SQLITE_ROW;
	124475	+ if( fts3SegReaderIsPending(apSegment[0]) ){
	124476	+ rc = fts3MsrBufferData(pCsr, apSegment[0]->aDoclist, pCsr->nDoclist);
	124477	+ pCsr->aDoclist = pCsr->aBuffer;
	124478	+ }else{
	124479	+ pCsr->aDoclist = apSegment[0]->aDoclist;
	124480	+ }
	124481	+ if( rc==SQLITE_OK ) rc = SQLITE_ROW;
121989	124482	}else{
121990	124483	int nDoclist = 0; /* Size of doclist */
121991	124484	sqlite3_int64 iPrev = 0; /* Previous docid stored in doclist */
121992	124485
121993	124486	/* The current term of the first nMerge entries in the array
		@@ -122717,10 +125210,12 @@
122717	125210	******************************************************************************
122718	125211	*/
122719	125212
122720	125213	#if !defined(SQLITE_CORE) \|\| defined(SQLITE_ENABLE_FTS3)
122721	125214
	125215	+/* #include <string.h> */
	125216	+/* #include <assert.h> */
122722	125217
122723	125218	/*
122724	125219	** Characters that may appear in the second argument to matchinfo().
122725	125220	*/
122726	125221	#define FTS3_MATCHINFO_NPHRASE 'p' /* 1 value */
		@@ -123725,11 +126220,11 @@
123725	126220	if( bGlobal ) pInfo->aMatchinfo[0] = pInfo->nCol;
123726	126221	break;
123727	126222
123728	126223	case FTS3_MATCHINFO_NDOC:
123729	126224	if( bGlobal ){
123730		- sqlite3_int64 nDoc;
	126225	+ sqlite3_int64 nDoc = 0;
123731	126226	rc = fts3MatchinfoSelectDoctotal(pTab, &pSelect, &nDoc, 0);
123732	126227	pInfo->aMatchinfo[0] = (u32)nDoc;
123733	126228	}
123734	126229	break;
123735	126230
		@@ -124304,10 +126799,12 @@
124304	126799	#ifndef SQLITE_CORE
124305	126800	SQLITE_EXTENSION_INIT1
124306	126801	#else
124307	126802	#endif
124308	126803
	126804	+/* #include <string.h> */
	126805	+/* #include <assert.h> */
124309	126806
124310	126807	#ifndef SQLITE_AMALGAMATION
124311	126808	#include "sqlite3rtree.h"
124312	126809	typedef sqlite3_int64 i64;
124313	126810	typedef unsigned char u8;
		@@ -125620,11 +128117,11 @@
125620	128117	*/
125621	128118	static float cellArea(Rtree pRtree, RtreeCell p){
125622	128119	float area = 1.0;
125623	128120	int ii;
125624	128121	for(ii=0; ii<(pRtree->nDim*2); ii+=2){
125625		- area = area * (DCOORD(p->aCoord[ii+1]) - DCOORD(p->aCoord[ii]));
	128122	+ area = (float)(area * (DCOORD(p->aCoord[ii+1]) - DCOORD(p->aCoord[ii])));
125626	128123	}
125627	128124	return area;
125628	128125	}
125629	128126
125630	128127	/*
		@@ -125633,11 +128130,11 @@
125633	128130	*/
125634	128131	static float cellMargin(Rtree pRtree, RtreeCell p){
125635	128132	float margin = 0.0;
125636	128133	int ii;
125637	128134	for(ii=0; ii<(pRtree->nDim*2); ii+=2){
125638		- margin += (DCOORD(p->aCoord[ii+1]) - DCOORD(p->aCoord[ii]));
	128135	+ margin += (float)(DCOORD(p->aCoord[ii+1]) - DCOORD(p->aCoord[ii]));
125639	128136	}
125640	128137	return margin;
125641	128138	}
125642	128139
125643	128140	/*
		@@ -125718,11 +128215,11 @@
125718	128215
125719	128216	if( x2<x1 ){
125720	128217	o = 0.0;
125721	128218	break;
125722	128219	}else{
125723		- o = o * (x2-x1);
	128220	+ o = o * (float)(x2-x1);
125724	128221	}
125725	128222	}
125726	128223	overlap += o;
125727	128224	}
125728	128225	}
		@@ -125737,16 +128234,16 @@
125737	128234	RtreeCell *pInsert,
125738	128235	RtreeCell *aCell,
125739	128236	int nCell,
125740	128237	int iExclude
125741	128238	){
125742		- float before;
125743		- float after;
	128239	+ double before;
	128240	+ double after;
125744	128241	before = cellOverlap(pRtree, p, aCell, nCell, iExclude);
125745	128242	cellUnion(pRtree, p, pInsert);
125746	128243	after = cellOverlap(pRtree, p, aCell, nCell, iExclude);
125747		- return after-before;
	128244	+ return (float)(after-before);
125748	128245	}
125749	128246	#endif
125750	128247
125751	128248
125752	128249	/*
		@@ -125764,15 +128261,15 @@
125764	128261	RtreeNode *pNode;
125765	128262	rc = nodeAcquire(pRtree, 1, 0, &pNode);
125766	128263
125767	128264	for(ii=0; rc==SQLITE_OK && ii<(pRtree->iDepth-iHeight); ii++){
125768	128265	int iCell;
125769		- sqlite3_int64 iBest;
	128266	+ sqlite3_int64 iBest = 0;
125770	128267
125771		- float fMinGrowth;
125772		- float fMinArea;
125773		- float fMinOverlap;
	128268	+ float fMinGrowth = 0.0;
	128269	+ float fMinArea = 0.0;
	128270	+ float fMinOverlap = 0.0;
125774	128271
125775	128272	int nCell = NCELL(pNode);
125776	128273	RtreeCell cell;
125777	128274	RtreeNode *pChild;
125778	128275
		@@ -126198,13 +128695,13 @@
126198	128695	){
126199	128696	int **aaSorted;
126200	128697	int *aSpare;
126201	128698	int ii;
126202	128699
126203		- int iBestDim;
126204		- int iBestSplit;
126205		- float fBestMargin;
	128700	+ int iBestDim = 0;
	128701	+ int iBestSplit = 0;
	128702	+ float fBestMargin = 0.0;
126206	128703
126207	128704	int nByte = (pRtree->nDim+1)(sizeof(int)+nCell*sizeof(int));
126208	128705
126209	128706	aaSorted = (int **)sqlite3_malloc(nByte);
126210	128707	if( !aaSorted ){
		@@ -126222,13 +128719,13 @@
126222	128719	SortByDimension(pRtree, aaSorted[ii], nCell, ii, aCell, aSpare);
126223	128720	}
126224	128721
126225	128722	for(ii=0; ii<pRtree->nDim; ii++){
126226	128723	float margin = 0.0;
126227		- float fBestOverlap;
126228		- float fBestArea;
126229		- int iBestLeft;
	128724	+ float fBestOverlap = 0.0;
	128725	+ float fBestArea = 0.0;
	128726	+ int iBestLeft = 0;
126230	128727	int nLeft;
126231	128728
126232	128729	for(
126233	128730	nLeft=RTREE_MINCELLS(pRtree);
126234	128731	nLeft<=(nCell-RTREE_MINCELLS(pRtree));
		@@ -126539,11 +129036,11 @@
126539	129036	static int deleteCell(Rtree , RtreeNode , int, int);
126540	129037
126541	129038	static int removeNode(Rtree pRtree, RtreeNode pNode, int iHeight){
126542	129039	int rc;
126543	129040	int rc2;
126544		- RtreeNode *pParent;
	129041	+ RtreeNode *pParent = 0;
126545	129042	int iCell;
126546	129043
126547	129044	assert( pNode->nRef==1 );
126548	129045
126549	129046	/* Remove the entry in the parent cell. */
		@@ -126687,23 +129184,23 @@
126687	129184	}else{
126688	129185	nodeGetCell(pRtree, pNode, ii, &aCell[ii]);
126689	129186	}
126690	129187	aOrder[ii] = ii;
126691	129188	for(iDim=0; iDim<pRtree->nDim; iDim++){
126692		- aCenterCoord[iDim] += DCOORD(aCell[ii].aCoord[iDim*2]);
126693		- aCenterCoord[iDim] += DCOORD(aCell[ii].aCoord[iDim*2+1]);
	129189	+ aCenterCoord[iDim] += (float)DCOORD(aCell[ii].aCoord[iDim*2]);
	129190	+ aCenterCoord[iDim] += (float)DCOORD(aCell[ii].aCoord[iDim*2+1]);
126694	129191	}
126695	129192	}
126696	129193	for(iDim=0; iDim<pRtree->nDim; iDim++){
126697		- aCenterCoord[iDim] = aCenterCoord[iDim]/((float)nCell*2.0);
	129194	+ aCenterCoord[iDim] = (float)(aCenterCoord[iDim]/((float)nCell*2.0));
126698	129195	}
126699	129196
126700	129197	for(ii=0; ii<nCell; ii++){
126701	129198	aDistance[ii] = 0.0;
126702	129199	for(iDim=0; iDim<pRtree->nDim; iDim++){
126703		- float coord = DCOORD(aCell[ii].aCoord[iDim*2+1]) -
126704		- DCOORD(aCell[ii].aCoord[iDim*2]);
	129200	+ float coord = (float)(DCOORD(aCell[ii].aCoord[iDim*2+1]) -
	129201	+ DCOORD(aCell[ii].aCoord[iDim*2]));
126705	129202	aDistance[ii] += (coord-aCenterCoord[iDim])*(coord-aCenterCoord[iDim]);
126706	129203	}
126707	129204	}
126708	129205
126709	129206	SortByDistance(aOrder, nCell, aDistance, aSpare);
		@@ -126798,14 +129295,14 @@
126798	129295	nodeGetCell(pRtree, pNode, ii, &cell);
126799	129296
126800	129297	/* Find a node to store this cell in. pNode->iNode currently contains
126801	129298	** the height of the sub-tree headed by the cell.
126802	129299	*/
126803		- rc = ChooseLeaf(pRtree, &cell, pNode->iNode, &pInsert);
	129300	+ rc = ChooseLeaf(pRtree, &cell, (int)pNode->iNode, &pInsert);
126804	129301	if( rc==SQLITE_OK ){
126805	129302	int rc2;
126806		- rc = rtreeInsertCell(pRtree, pInsert, &cell, pNode->iNode);
	129303	+ rc = rtreeInsertCell(pRtree, pInsert, &cell, (int)pNode->iNode);
126807	129304	rc2 = nodeRelease(pRtree, pInsert);
126808	129305	if( rc==SQLITE_OK ){
126809	129306	rc = rc2;
126810	129307	}
126811	129308	}
		@@ -127190,11 +129687,11 @@
127190	129687	int isCreate /* True for xCreate, false for xConnect */
127191	129688	){
127192	129689	int rc;
127193	129690	char *zSql;
127194	129691	if( isCreate ){
127195		- int iPageSize;
	129692	+ int iPageSize = 0;
127196	129693	zSql = sqlite3_mprintf("PRAGMA %Q.page_size", pRtree->zDb);
127197	129694	rc = getIntFromStmt(db, zSql, &iPageSize);
127198	129695	if( rc==SQLITE_OK ){
127199	129696	pRtree->iNodeSize = iPageSize-64;
127200	129697	if( (4+pRtree->nBytesPerCell*RTREE_MAXCELLS)<pRtree->iNodeSize ){
		@@ -127518,10 +130015,11 @@
127518	130015	#include <unicode/utypes.h>
127519	130016	#include <unicode/uregex.h>
127520	130017	#include <unicode/ustring.h>
127521	130018	#include <unicode/ucol.h>
127522	130019
	130020	+/* #include <assert.h> */
127523	130021
127524	130022	#ifndef SQLITE_CORE
127525	130023	SQLITE_EXTENSION_INIT1
127526	130024	#else
127527	130025	#endif
		@@ -127993,19 +130491,20 @@
127993	130491	** May you find forgiveness for yourself and forgive others.
127994	130492	** May you share freely, never taking more than you give.
127995	130493	**
127996	130494	*************************************************************************
127997	130495	** This file implements a tokenizer for fts3 based on the ICU library.
127998		-**
127999		-** $Id: fts3_icu.c,v 1.3 2008/09/01 18:34:20 danielk1977 Exp $
128000	130496	*/
128001		-
128002	130497	#if !defined(SQLITE_CORE) \|\| defined(SQLITE_ENABLE_FTS3)
128003	130498	#ifdef SQLITE_ENABLE_ICU
128004	130499
	130500	+/* #include <assert.h> */
	130501	+/* #include <string.h> */
128005	130502
128006	130503	#include <unicode/ubrk.h>
	130504	+/* #include <unicode/ucol.h> */
	130505	+/* #include <unicode/ustring.h> */
128007	130506	#include <unicode/utf16.h>
128008	130507
128009	130508	typedef struct IcuTokenizer IcuTokenizer;
128010	130509	typedef struct IcuCursor IcuCursor;
128011	130510
128012	130511

	--- src/sqlite3.c
	+++ src/sqlite3.c
	@@ -1,8 +1,8 @@
1	/******************************************************************************
2	** This file is an amalgamation of many separate C source files from SQLite
3	** version 3.7.7. By combining all the individual C code files into this
4	** single large file, the entire code can be compiled as a single translation
5	** unit. This allows many compilers to do optimizations that would not be
6	** possible if the files were compiled separately. Performance improvements
7	** of 5% or more are commonly seen when SQLite is compiled as a single
8	** translation unit.
	@@ -385,23 +385,29 @@
385	/*
386	** Exactly one of the following macros must be defined in order to
387	** specify which memory allocation subsystem to use.
388	**
389	** SQLITE_SYSTEM_MALLOC // Use normal system malloc()

390	** SQLITE_MEMDEBUG // Debugging version of system malloc()





391	**
392	** (Historical note: There used to be several other options, but we've
393	** pared it down to just these two.)
394	**
395	** If none of the above are defined, then set SQLITE_SYSTEM_MALLOC as
396	** the default.
397	*/
398	#if defined(SQLITE_SYSTEM_MALLOC)+defined(SQLITE_MEMDEBUG)>1
399	# error "At most one of the following compile-time configuration options\
400	is allows: SQLITE_SYSTEM_MALLOC, SQLITE_MEMDEBUG"
401	#endif
402	#if defined(SQLITE_SYSTEM_MALLOC)+defined(SQLITE_MEMDEBUG)==0
403	# define SQLITE_SYSTEM_MALLOC 1
404	#endif
405
406	/*
407	** If SQLITE_MALLOC_SOFT_LIMIT is not zero, then try to keep the
	@@ -648,13 +654,13 @@
648	**
649	** See also: [sqlite3_libversion()],
650	** [sqlite3_libversion_number()], [sqlite3_sourceid()],
651	** [sqlite_version()] and [sqlite_source_id()].
652	*/
653	#define SQLITE_VERSION "3.7.7"
654	#define SQLITE_VERSION_NUMBER 3007007
655	#define SQLITE_SOURCE_ID "2011-06-15 13:11:06 f9750870ee04935f338e4d808900fee5a8b2b389"
656
657	/*
658	** CAPI3REF: Run-Time Library Version Numbers
659	** KEYWORDS: sqlite3_version, sqlite3_sourceid
660	**
	@@ -851,11 +857,11 @@
851	** semicolon-separate SQL statements passed into its 2nd argument,
852	** in the context of the [database connection] passed in as its 1st
853	** argument. ^If the callback function of the 3rd argument to
854	** sqlite3_exec() is not NULL, then it is invoked for each result row
855	** coming out of the evaluated SQL statements. ^The 4th argument to
856	** to sqlite3_exec() is relayed through to the 1st argument of each
857	** callback invocation. ^If the callback pointer to sqlite3_exec()
858	** is NULL, then no callback is ever invoked and result rows are
859	** ignored.
860	**
861	** ^If an error occurs while evaluating the SQL statements passed into
	@@ -1282,20 +1288,52 @@
1282	** when [PRAGMA synchronous \| PRAGMA synchronous=OFF] is set, but most
1283	** VFSes do not need this signal and should silently ignore this opcode.
1284	** Applications should not call [sqlite3_file_control()] with this
1285	** opcode as doing so may disrupt the operation of the specialized VFSes
1286	** that do require it.































1287	*/
1288	#define SQLITE_FCNTL_LOCKSTATE 1
1289	#define SQLITE_GET_LOCKPROXYFILE 2
1290	#define SQLITE_SET_LOCKPROXYFILE 3
1291	#define SQLITE_LAST_ERRNO 4
1292	#define SQLITE_FCNTL_SIZE_HINT 5
1293	#define SQLITE_FCNTL_CHUNK_SIZE 6
1294	#define SQLITE_FCNTL_FILE_POINTER 7
1295	#define SQLITE_FCNTL_SYNC_OMITTED 8
1296

1297
1298	/*
1299	** CAPI3REF: Mutex Handle
1300	**
1301	** The mutex module within SQLite defines [sqlite3_mutex] to be an
	@@ -1443,11 +1481,11 @@
1443	** The xSleep() method causes the calling thread to sleep for at
1444	** least the number of microseconds given. ^The xCurrentTime()
1445	** method returns a Julian Day Number for the current date and time as
1446	** a floating point value.
1447	** ^The xCurrentTimeInt64() method returns, as an integer, the Julian
1448	** Day Number multipled by 86400000 (the number of milliseconds in
1449	** a 24-hour day).
1450	** ^SQLite will use the xCurrentTimeInt64() method to get the current
1451	** date and time if that method is available (if iVersion is 2 or
1452	** greater and the function pointer is not NULL) and will fall back
1453	** to xCurrentTime() if xCurrentTimeInt64() is unavailable.
	@@ -1719,20 +1757,14 @@
1719	** also used during testing of SQLite in order to specify an alternative
1720	** memory allocator that simulates memory out-of-memory conditions in
1721	** order to verify that SQLite recovers gracefully from such
1722	** conditions.
1723	**
1724	** The xMalloc and xFree methods must work like the
1725	** malloc() and free() functions from the standard C library.
1726	** The xRealloc method must work like realloc() from the standard C library
1727	** with the exception that if the second argument to xRealloc is zero,
1728	** xRealloc must be a no-op - it must not perform any allocation or
1729	** deallocation. ^SQLite guarantees that the second argument to
1730	** xRealloc is always a value returned by a prior call to xRoundup.
1731	** And so in cases where xRoundup always returns a positive number,
1732	** xRealloc can perform exactly as the standard library realloc() and
1733	** still be in compliance with this specification.
1734	**
1735	** xSize should return the allocated size of a memory allocation
1736	** previously obtained from xMalloc or xRealloc. The allocated size
1737	** is always at least as big as the requested size but may be larger.
1738	**
	@@ -1881,11 +1913,11 @@
1881	** scratch memory beyond what is provided by this configuration option, then
1882	** [sqlite3_malloc()] will be used to obtain the memory needed.</dd>
1883	**
1884	** [[SQLITE_CONFIG_PAGECACHE]] <dt>SQLITE_CONFIG_PAGECACHE</dt>
1885	** <dd> ^This option specifies a static memory buffer that SQLite can use for
1886	** the database page cache with the default page cache implemenation.
1887	** This configuration should not be used if an application-define page
1888	** cache implementation is loaded using the SQLITE_CONFIG_PCACHE option.
1889	** There are three arguments to this option: A pointer to 8-byte aligned
1890	** memory, the size of each page buffer (sz), and the number of pages (N).
1891	** The sz argument should be the size of the largest database page
	@@ -2979,16 +3011,16 @@
2979	** [[URI filenames in sqlite3_open()]] <h3>URI Filenames</h3>
2980	**
2981	** ^If [URI filename] interpretation is enabled, and the filename argument
2982	** begins with "file:", then the filename is interpreted as a URI. ^URI
2983	** filename interpretation is enabled if the [SQLITE_OPEN_URI] flag is
2984	** is set in the fourth argument to sqlite3_open_v2(), or if it has
2985	** been enabled globally using the [SQLITE_CONFIG_URI] option with the
2986	** [sqlite3_config()] method or by the [SQLITE_USE_URI] compile-time option.
2987	** As of SQLite version 3.7.7, URI filename interpretation is turned off
2988	** by default, but future releases of SQLite might enable URI filename
2989	** intepretation by default. See "[URI filenames]" for additional
2990	** information.
2991	**
2992	** URI filenames are parsed according to RFC 3986. ^If the URI contains an
2993	** authority, then it must be either an empty string or the string
2994	** "localhost". ^If the authority is not an empty string or "localhost", an
	@@ -3803,11 +3835,11 @@
3803	** [extended result codes] might be returned as well.
3804	**
3805	** ^[SQLITE_BUSY] means that the database engine was unable to acquire the
3806	** database locks it needs to do its job. ^If the statement is a [COMMIT]
3807	** or occurs outside of an explicit transaction, then you can retry the
3808	** statement. If the statement is not a [COMMIT] and occurs within a
3809	** explicit transaction then you should rollback the transaction before
3810	** continuing.
3811	**
3812	** ^[SQLITE_DONE] means that the statement has finished executing
3813	** successfully. sqlite3_step() should not be called again on this virtual
	@@ -4082,11 +4114,11 @@
4082
4083	/*
4084	** CAPI3REF: Destroy A Prepared Statement Object
4085	**
4086	** ^The sqlite3_finalize() function is called to delete a [prepared statement].
4087	** ^If the most recent evaluation of the statement encountered no errors or
4088	** or if the statement is never been evaluated, then sqlite3_finalize() returns
4089	** SQLITE_OK. ^If the most recent evaluation of statement S failed, then
4090	** sqlite3_finalize(S) returns the appropriate [error code] or
4091	** [extended error code].
4092	**
	@@ -5996,11 +6028,11 @@
5996	** versions of these routines, it should at least provide stubs that always
5997	** return true so that one does not get spurious assertion failures.
5998	**
5999	** ^If the argument to sqlite3_mutex_held() is a NULL pointer then
6000	** the routine should return 1. This seems counter-intuitive since
6001	** clearly the mutex cannot be held if it does not exist. But the
6002	** the reason the mutex does not exist is because the build is not
6003	** using mutexes. And we do not want the assert() containing the
6004	** call to sqlite3_mutex_held() to fail, so a non-zero return is
6005	** the appropriate thing to do. ^The sqlite3_mutex_notheld()
6006	** interface should also return 1 when given a NULL pointer.
	@@ -6119,11 +6151,12 @@
6119	#define SQLITE_TESTCTRL_RESERVE 14
6120	#define SQLITE_TESTCTRL_OPTIMIZATIONS 15
6121	#define SQLITE_TESTCTRL_ISKEYWORD 16
6122	#define SQLITE_TESTCTRL_PGHDRSZ 17
6123	#define SQLITE_TESTCTRL_SCRATCHMALLOC 18
6124	#define SQLITE_TESTCTRL_LAST 18

6125
6126	/*
6127	** CAPI3REF: SQLite Runtime Status
6128	**
6129	** ^This interface is used to retrieve runtime status information
	@@ -6505,11 +6538,11 @@
6505	** [[the xFetch() page cache methods]]
6506	** The xFetch() method locates a page in the cache and returns a pointer to
6507	** the page, or a NULL pointer.
6508	** A "page", in this context, means a buffer of szPage bytes aligned at an
6509	** 8-byte boundary. The page to be fetched is determined by the key. ^The
6510	** mimimum key value is 1. After it has been retrieved using xFetch, the page
6511	** is considered to be "pinned".
6512	**
6513	** If the requested page is already in the page cache, then the page cache
6514	** implementation must return a pointer to the page buffer with its content
6515	** intact. If the requested page is not already in the cache, then the
	@@ -7599,10 +7632,18 @@
7599	*/
7600	#ifndef SQLITE_TEMP_STORE
7601	# define SQLITE_TEMP_STORE 1
7602	#endif
7603








7604	/*
7605	** GCC does not define the offsetof() macro so we'll have to do it
7606	** ourselves.
7607	*/
7608	#ifndef offsetof
	@@ -7941,10 +7982,11 @@
7941	#define BTREE_OMIT_JOURNAL 1 /* Do not create or use a rollback journal */
7942	#define BTREE_NO_READLOCK 2 /* Omit readlocks on readonly files */
7943	#define BTREE_MEMORY 4 /* This is an in-memory DB */
7944	#define BTREE_SINGLE 8 /* The file contains at most 1 b-tree */
7945	#define BTREE_UNORDERED 16 /* Use of a hash implementation is OK */

7946
7947	SQLITE_PRIVATE int sqlite3BtreeClose(Btree*);
7948	SQLITE_PRIVATE int sqlite3BtreeSetCacheSize(Btree*,int);
7949	SQLITE_PRIVATE int sqlite3BtreeSetSafetyLevel(Btree*,int,int,int);
7950	SQLITE_PRIVATE int sqlite3BtreeSyncDisabled(Btree*);
	@@ -8143,10 +8185,11 @@
8143	** or VDBE. The VDBE implements an abstract machine that runs a
8144	** simple program to access and modify the underlying database.
8145	*/
8146	#ifndef _SQLITE_VDBE_H_
8147	#define _SQLITE_VDBE_H_

8148
8149	/*
8150	** A single VDBE is an opaque structure named "Vdbe". Only routines
8151	** in the source file sqliteVdbe.c are allowed to see the insides
8152	** of this structure.
	@@ -8186,10 +8229,11 @@
8186	Mem pMem; / Used when p4type is P4_MEM */
8187	VTable pVtab; / Used when p4type is P4_VTAB */
8188	KeyInfo pKeyInfo; / Used when p4type is P4_KEYINFO */
8189	int ai; / Used when p4type is P4_INTARRAY */
8190	SubProgram pProgram; / Used when p4type is P4_SUBPROGRAM */

8191	} p4;
8192	#ifdef SQLITE_DEBUG
8193	char zComment; / Comment to improve readability */
8194	#endif
8195	#ifdef VDBE_PROFILE
	@@ -8241,10 +8285,11 @@
8241	#define P4_REAL (-12) /* P4 is a 64-bit floating point value */
8242	#define P4_INT64 (-13) /* P4 is a 64-bit signed integer */
8243	#define P4_INT32 (-14) /* P4 is a 32-bit signed integer */
8244	#define P4_INTARRAY (-15) /* P4 is a vector of 32-bit integers */
8245	#define P4_SUBPROGRAM (-18) /* P4 is a pointer to a SubProgram structure */

8246
8247	/* When adding a P4 argument using P4_KEYINFO, a copy of the KeyInfo structure
8248	** is made. That copy is freed when the Vdbe is finalized. But if the
8249	** argument is P4_KEYINFO_HANDOFF, the passed in pointer is used. It still
8250	** gets freed when the Vdbe is finalized so it still should be obtained
	@@ -8354,89 +8399,89 @@
8354	#define OP_ReadCookie 35
8355	#define OP_SetCookie 36
8356	#define OP_VerifyCookie 37
8357	#define OP_OpenRead 38
8358	#define OP_OpenWrite 39
8359	#define OP_OpenAutoindex 40
8360	#define OP_OpenEphemeral 41
8361	#define OP_OpenPseudo 42
8362	#define OP_Close 43
8363	#define OP_SeekLt 44
8364	#define OP_SeekLe 45
8365	#define OP_SeekGe 46
8366	#define OP_SeekGt 47
8367	#define OP_Seek 48
8368	#define OP_NotFound 49
8369	#define OP_Found 50
8370	#define OP_IsUnique 51
8371	#define OP_NotExists 52
8372	#define OP_Sequence 53
8373	#define OP_NewRowid 54
8374	#define OP_Insert 55
8375	#define OP_InsertInt 56
8376	#define OP_Delete 57
8377	#define OP_ResetCount 58
8378	#define OP_RowKey 59
8379	#define OP_RowData 60
8380	#define OP_Rowid 61
8381	#define OP_NullRow 62
8382	#define OP_Last 63
8383	#define OP_Sort 64
8384	#define OP_Rewind 65
8385	#define OP_Prev 66
8386	#define OP_Next 67
8387	#define OP_IdxInsert 70
8388	#define OP_IdxDelete 71
8389	#define OP_IdxRowid 72
8390	#define OP_IdxLT 81
8391	#define OP_IdxGE 92
8392	#define OP_Destroy 95
8393	#define OP_Clear 96
8394	#define OP_CreateIndex 97
8395	#define OP_CreateTable 98
8396	#define OP_ParseSchema 99
8397	#define OP_LoadAnalysis 100
8398	#define OP_DropTable 101
8399	#define OP_DropIndex 102
8400	#define OP_DropTrigger 103
8401	#define OP_IntegrityCk 104
8402	#define OP_RowSetAdd 105
8403	#define OP_RowSetRead 106
8404	#define OP_RowSetTest 107
8405	#define OP_Program 108
8406	#define OP_Param 109
8407	#define OP_FkCounter 110
8408	#define OP_FkIfZero 111
8409	#define OP_MemMax 112
8410	#define OP_IfPos 113
8411	#define OP_IfNeg 114
8412	#define OP_IfZero 115
8413	#define OP_AggStep 116
8414	#define OP_AggFinal 117
8415	#define OP_Checkpoint 118
8416	#define OP_JournalMode 119
8417	#define OP_Vacuum 120
8418	#define OP_IncrVacuum 121
8419	#define OP_Expire 122
8420	#define OP_TableLock 123
8421	#define OP_VBegin 124
8422	#define OP_VCreate 125
8423	#define OP_VDestroy 126
8424	#define OP_VOpen 127
8425	#define OP_VFilter 128
8426	#define OP_VColumn 129
8427	#define OP_VNext 131
8428	#define OP_VRename 132
8429	#define OP_VUpdate 133
8430	#define OP_Pagecount 134
8431	#define OP_MaxPgcnt 135
8432	#define OP_Trace 136
8433	#define OP_Noop 137
8434	#define OP_Explain 138
8435
8436	/* The following opcode values are never used */
8437	#define OP_NotUsed_139 139
8438	#define OP_NotUsed_140 140
8439
8440
8441	/* Properties such as "out2" or "jump" that are specified in
8442	** comments following the "case" for each opcode in the vdbe.c
	@@ -8453,23 +8498,23 @@
8453	/* 0 */ 0x00, 0x01, 0x05, 0x04, 0x04, 0x10, 0x00, 0x02,\
8454	/* 8 */ 0x02, 0x02, 0x02, 0x02, 0x02, 0x00, 0x24, 0x24,\
8455	/* 16 */ 0x00, 0x00, 0x00, 0x24, 0x04, 0x05, 0x04, 0x00,\
8456	/* 24 */ 0x00, 0x01, 0x05, 0x05, 0x00, 0x00, 0x00, 0x02,\
8457	/* 32 */ 0x00, 0x00, 0x00, 0x02, 0x10, 0x00, 0x00, 0x00,\
8458	/* 40 */ 0x00, 0x00, 0x00, 0x00, 0x11, 0x11, 0x11, 0x11,\
8459	/* 48 */ 0x08, 0x11, 0x11, 0x11, 0x11, 0x02, 0x02, 0x00,\
8460	/* 56 */ 0x00, 0x00, 0x00, 0x00, 0x00, 0x02, 0x00, 0x01,\
8461	/* 64 */ 0x01, 0x01, 0x01, 0x01, 0x4c, 0x4c, 0x08, 0x00,\
8462	/* 72 */ 0x02, 0x05, 0x05, 0x15, 0x15, 0x15, 0x15, 0x15,\
8463	/* 80 */ 0x15, 0x01, 0x4c, 0x4c, 0x4c, 0x4c, 0x4c, 0x4c,\
8464	/* 88 */ 0x4c, 0x4c, 0x4c, 0x4c, 0x01, 0x24, 0x02, 0x02,\
8465	/* 96 */ 0x00, 0x02, 0x02, 0x00, 0x00, 0x00, 0x00, 0x00,\
8466	/* 104 */ 0x00, 0x0c, 0x45, 0x15, 0x01, 0x02, 0x00, 0x01,\
8467	/* 112 */ 0x08, 0x05, 0x05, 0x05, 0x00, 0x00, 0x00, 0x02,\
8468	/* 120 */ 0x00, 0x01, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00,\
8469	/* 128 */ 0x01, 0x00, 0x02, 0x01, 0x00, 0x00, 0x02, 0x02,\
8470	/* 136 */ 0x00, 0x00, 0x00, 0x00, 0x00, 0x04, 0x04, 0x04,\
8471	/* 144 */ 0x04, 0x04,}
8472
8473	/************ End of opcodes.h *******************************************/
8474	/************ Continuing where we left off in vdbe.h *********************/
8475
	@@ -8484,13 +8529,13 @@
8484	SQLITE_PRIVATE int sqlite3VdbeAddOp3(Vdbe*,int,int,int,int);
8485	SQLITE_PRIVATE int sqlite3VdbeAddOp4(Vdbe,int,int,int,int,const char zP4,int);
8486	SQLITE_PRIVATE int sqlite3VdbeAddOp4Int(Vdbe*,int,int,int,int,int);
8487	SQLITE_PRIVATE int sqlite3VdbeAddOpList(Vdbe, int nOp, VdbeOpList const aOp);
8488	SQLITE_PRIVATE void sqlite3VdbeAddParseSchemaOp(Vdbe,int,char);
8489	SQLITE_PRIVATE void sqlite3VdbeChangeP1(Vdbe*, int addr, int P1);
8490	SQLITE_PRIVATE void sqlite3VdbeChangeP2(Vdbe*, int addr, int P2);
8491	SQLITE_PRIVATE void sqlite3VdbeChangeP3(Vdbe*, int addr, int P3);
8492	SQLITE_PRIVATE void sqlite3VdbeChangeP5(Vdbe*, u8 P5);
8493	SQLITE_PRIVATE void sqlite3VdbeJumpHere(Vdbe*, int addr);
8494	SQLITE_PRIVATE void sqlite3VdbeChangeToNoop(Vdbe*, int addr, int N);
8495	SQLITE_PRIVATE void sqlite3VdbeChangeP4(Vdbe, int addr, const char zP4, int N);
8496	SQLITE_PRIVATE void sqlite3VdbeUsesBtree(Vdbe*, int);
	@@ -8608,10 +8653,11 @@
8608	** NOTE: These values must match the corresponding BTREE_ values in btree.h.
8609	*/
8610	#define PAGER_OMIT_JOURNAL 0x0001 /* Do not use a rollback journal */
8611	#define PAGER_NO_READLOCK 0x0002 /* Omit readlocks on readonly files */
8612	#define PAGER_MEMORY 0x0004 /* In-memory database */

8613
8614	/*
8615	** Valid values for the second argument to sqlite3PagerLockingMode().
8616	*/
8617	#define PAGER_LOCKINGMODE_QUERY -1
	@@ -8703,10 +8749,13 @@
8703	SQLITE_PRIVATE sqlite3_file sqlite3PagerFile(Pager);
8704	SQLITE_PRIVATE const char sqlite3PagerJournalname(Pager);
8705	SQLITE_PRIVATE int sqlite3PagerNosync(Pager*);
8706	SQLITE_PRIVATE void sqlite3PagerTempSpace(Pager);
8707	SQLITE_PRIVATE int sqlite3PagerIsMemdb(Pager*);



8708
8709	/* Functions used to truncate the database file. */
8710	SQLITE_PRIVATE void sqlite3PagerTruncateImage(Pager*,Pgno);
8711
8712	#if defined(SQLITE_HAS_CODEC) && !defined(SQLITE_OMIT_WAL)
	@@ -9556,10 +9605,12 @@
9556	#define SQLITE_IndexSort 0x04 /* Disable indexes for sorting */
9557	#define SQLITE_IndexSearch 0x08 /* Disable indexes for searching */
9558	#define SQLITE_IndexCover 0x10 /* Disable index covering table */
9559	#define SQLITE_GroupByOrder 0x20 /* Disable GROUPBY cover of ORDERBY */
9560	#define SQLITE_FactorOutConst 0x40 /* Disable factoring out constants */


9561	#define SQLITE_OptMask 0xff /* Mask of all disablable opts */
9562
9563	/*
9564	** Possible values for the sqlite.magic field.
9565	** The numbers are obtained at random and have no special meaning, other
	@@ -10447,10 +10498,11 @@
10447	Table pTab; / An SQL table corresponding to zName */
10448	Select pSelect; / A SELECT statement used in place of a table name */
10449	u8 isPopulated; /* Temporary table associated with SELECT is populated */
10450	u8 jointype; /* Type of join between this able and the previous */
10451	u8 notIndexed; /* True if there is a NOT INDEXED clause */

10452	#ifndef SQLITE_OMIT_EXPLAIN
10453	u8 iSelectId; /* If pSelect!=0, the id of the sub-select in EQP */
10454	#endif
10455	int iCursor; /* The VDBE cursor number used to access this table */
10456	Expr pOn; / The ON clause of a join */
	@@ -10566,10 +10618,11 @@
10566	struct WhereInfo {
10567	Parse pParse; / Parsing and code generating context */
10568	u16 wctrlFlags; /* Flags originally passed to sqlite3WhereBegin() */
10569	u8 okOnePass; /* Ok to use one-pass algorithm for UPDATE or DELETE */
10570	u8 untestedTerms; /* Not all WHERE terms resolved by outer loop */

10571	SrcList pTabList; / List of tables in the join */
10572	int iTop; /* The very beginning of the WHERE loop */
10573	int iContinue; /* Jump here to continue with next record */
10574	int iBreak; /* Jump here to break out of the loop */
10575	int nLevel; /* Number of nested loop */
	@@ -10577,10 +10630,13 @@
10577	double savedNQueryLoop; /* pParse->nQueryLoop outside the WHERE loop */
10578	double nRowOut; /* Estimated number of output rows */
10579	WhereLevel a[1]; /* Information about each nest loop in WHERE */
10580	};
10581



10582	/*
10583	** A NameContext defines a context in which to resolve table and column
10584	** names. The context consists of a list of tables (the pSrcList) field and
10585	** a list of named expression (pEList). The named expression list may
10586	** be NULL. The pSrc corresponds to the FROM clause of a SELECT or
	@@ -11053,10 +11109,11 @@
11053	int isPCacheInit; /* True after malloc is initialized */
11054	sqlite3_mutex pInitMutex; / Mutex used by sqlite3_initialize() */
11055	int nRefInitMutex; /* Number of users of pInitMutex */
11056	void (xLog)(void,int,const char); / Function for logging */
11057	void pLogArg; / First argument to xLog() */

11058	};
11059
11060	/*
11061	** Context pointer passed down through the tree-walk.
11062	*/
	@@ -11337,11 +11394,11 @@
11337	#if defined(SQLITE_ENABLE_UPDATE_DELETE_LIMIT) && !defined(SQLITE_OMIT_SUBQUERY)
11338	SQLITE_PRIVATE Expr sqlite3LimitWhere(Parse , SrcList , Expr , ExprList , Expr , Expr , char );
11339	#endif
11340	SQLITE_PRIVATE void sqlite3DeleteFrom(Parse, SrcList, Expr*);
11341	SQLITE_PRIVATE void sqlite3Update(Parse, SrcList, ExprList, Expr, int);
11342	SQLITE_PRIVATE WhereInfo sqlite3WhereBegin(Parse, SrcList, Expr, ExprList**, u16);
11343	SQLITE_PRIVATE void sqlite3WhereEnd(WhereInfo*);
11344	SQLITE_PRIVATE int sqlite3ExprCodeGetColumn(Parse, Table, int, int, int);
11345	SQLITE_PRIVATE void sqlite3ExprCodeGetColumnOfTable(Vdbe, Table, int, int, int);
11346	SQLITE_PRIVATE void sqlite3ExprCodeMove(Parse*, int, int, int);
11347	SQLITE_PRIVATE void sqlite3ExprCodeCopy(Parse*, int, int, int);
	@@ -12013,10 +12070,11 @@
12013	0, /* isPCacheInit */
12014	0, /* pInitMutex */
12015	0, /* nRefInitMutex */
12016	0, /* xLog */
12017	0, /* pLogArg */

12018	};
12019
12020
12021	/*
12022	** Hash table for global functions - functions common to all
	@@ -12321,10 +12379,13 @@
12321	#ifdef SQLITE_OMIT_LOOKASIDE
12322	"OMIT_LOOKASIDE",
12323	#endif
12324	#ifdef SQLITE_OMIT_MEMORYDB
12325	"OMIT_MEMORYDB",



12326	#endif
12327	#ifdef SQLITE_OMIT_OR_OPTIMIZATION
12328	"OMIT_OR_OPTIMIZATION",
12329	#endif
12330	#ifdef SQLITE_OMIT_PAGER_PRAGMAS
	@@ -12387,10 +12448,13 @@
12387	#ifdef SQLITE_OMIT_WSD
12388	"OMIT_WSD",
12389	#endif
12390	#ifdef SQLITE_OMIT_XFER_OPT
12391	"OMIT_XFER_OPT",



12392	#endif
12393	#ifdef SQLITE_PERFORMANCE_TRACE
12394	"PERFORMANCE_TRACE",
12395	#endif
12396	#ifdef SQLITE_PROXY_DEBUG
	@@ -12508,10 +12572,13 @@
12508	/*
12509	** Boolean values
12510	*/
12511	typedef unsigned char Bool;
12512



12513	/*
12514	** A cursor is a pointer into a single BTree within a database file.
12515	** The cursor can seek to a BTree entry with a particular key, or
12516	** loop over all entries of the Btree. You can also insert new BTree
12517	** entries or retrieve the key or data from the entry that the cursor
	@@ -12539,10 +12606,11 @@
12539	sqlite3_vtab_cursor pVtabCursor; / The cursor for a virtual table */
12540	const sqlite3_module pModule; / Module for cursor pVtabCursor */
12541	i64 seqCount; /* Sequence counter */
12542	i64 movetoTarget; /* Argument to the deferred sqlite3BtreeMoveto() */
12543	i64 lastRowid; /* Last rowid from a Next or NextIdx operation */

12544
12545	/* Result of last sqlite3BtreeMoveto() done by an OP_NotExists or
12546	** OP_IsUnique opcode on this cursor. */
12547	int seekResult;
12548
	@@ -12858,17 +12926,36 @@
12858	SQLITE_PRIVATE int sqlite3VdbeMemRealify(Mem*);
12859	SQLITE_PRIVATE int sqlite3VdbeMemNumerify(Mem*);
12860	SQLITE_PRIVATE int sqlite3VdbeMemFromBtree(BtCursor,int,int,int,Mem);
12861	SQLITE_PRIVATE void sqlite3VdbeMemRelease(Mem *p);
12862	SQLITE_PRIVATE void sqlite3VdbeMemReleaseExternal(Mem *p);



12863	SQLITE_PRIVATE int sqlite3VdbeMemFinalize(Mem, FuncDef);
12864	SQLITE_PRIVATE const char *sqlite3OpcodeName(int);
12865	SQLITE_PRIVATE int sqlite3VdbeMemGrow(Mem *pMem, int n, int preserve);
12866	SQLITE_PRIVATE int sqlite3VdbeCloseStatement(Vdbe *, int);
12867	SQLITE_PRIVATE void sqlite3VdbeFrameDelete(VdbeFrame*);
12868	SQLITE_PRIVATE int sqlite3VdbeFrameRestore(VdbeFrame *);
12869	SQLITE_PRIVATE void sqlite3VdbeMemStoreType(Mem *pMem);
















12870
12871	#if !defined(SQLITE_OMIT_SHARED_CACHE) && SQLITE_THREADSAFE>0
12872	SQLITE_PRIVATE void sqlite3VdbeEnter(Vdbe*);
12873	SQLITE_PRIVATE void sqlite3VdbeLeave(Vdbe*);
12874	#else
	@@ -13159,30 +13246,16 @@
13159	** Astronomical Algorithms, 2nd Edition, 1998
13160	** ISBM 0-943396-61-1
13161	** Willmann-Bell, Inc
13162	** Richmond, Virginia (USA)
13163	*/


13164	#include <time.h>
13165
13166	#ifndef SQLITE_OMIT_DATETIME_FUNCS
13167
13168	/*
13169	** On recent Windows platforms, the localtime_s() function is available
13170	** as part of the "Secure CRT". It is essentially equivalent to
13171	** localtime_r() available under most POSIX platforms, except that the
13172	** order of the parameters is reversed.
13173	**
13174	** See http://msdn.microsoft.com/en-us/library/a442x3ye(VS.80).aspx.
13175	**
13176	** If the user has not indicated to use localtime_r() or localtime_s()
13177	** already, check for an MSVC build environment that provides
13178	** localtime_s().
13179	*/
13180	#if !defined(HAVE_LOCALTIME_R) && !defined(HAVE_LOCALTIME_S) && \
13181	defined(_MSC_VER) && defined(_CRT_INSECURE_DEPRECATE)
13182	#define HAVE_LOCALTIME_S 1
13183	#endif
13184
13185	/*
13186	** A structure for holding a single date and time.
13187	*/
13188	typedef struct DateTime DateTime;
	@@ -13523,20 +13596,90 @@
13523	static void clearYMD_HMS_TZ(DateTime *p){
13524	p->validYMD = 0;
13525	p->validHMS = 0;
13526	p->validTZ = 0;
13527	}


























































13528
13529	#ifndef SQLITE_OMIT_LOCALTIME
13530	/*
13531	** Compute the difference (in milliseconds)
13532	** between localtime and UTC (a.k.a. GMT)
13533	** for the time value p where p is in UTC.



13534	*/
13535	static sqlite3_int64 localtimeOffset(DateTime *p){




13536	DateTime x, y;
13537	time_t t;





13538	x = *p;
13539	computeYMD_HMS(&x);
13540	if( x.Y<1971 \|\| x.Y>=2038 ){
13541	x.Y = 2000;
13542	x.M = 1;
	@@ -13550,51 +13693,27 @@
13550	}
13551	x.tz = 0;
13552	x.validJD = 0;
13553	computeJD(&x);
13554	t = (time_t)(x.iJD/1000 - 21086676*(i64)10000);
13555	#ifdef HAVE_LOCALTIME_R
13556	{
13557	struct tm sLocal;
13558	localtime_r(&t, &sLocal);
13559	y.Y = sLocal.tm_year + 1900;
13560	y.M = sLocal.tm_mon + 1;
13561	y.D = sLocal.tm_mday;
13562	y.h = sLocal.tm_hour;
13563	y.m = sLocal.tm_min;
13564	y.s = sLocal.tm_sec;
13565	}
13566	#elif defined(HAVE_LOCALTIME_S) && HAVE_LOCALTIME_S
13567	{
13568	struct tm sLocal;
13569	localtime_s(&sLocal, &t);
13570	y.Y = sLocal.tm_year + 1900;
13571	y.M = sLocal.tm_mon + 1;
13572	y.D = sLocal.tm_mday;
13573	y.h = sLocal.tm_hour;
13574	y.m = sLocal.tm_min;
13575	y.s = sLocal.tm_sec;
13576	}
13577	#else
13578	{
13579	struct tm *pTm;
13580	sqlite3_mutex_enter(sqlite3MutexAlloc(SQLITE_MUTEX_STATIC_MASTER));
13581	pTm = localtime(&t);
13582	y.Y = pTm->tm_year + 1900;
13583	y.M = pTm->tm_mon + 1;
13584	y.D = pTm->tm_mday;
13585	y.h = pTm->tm_hour;
13586	y.m = pTm->tm_min;
13587	y.s = pTm->tm_sec;
13588	sqlite3_mutex_leave(sqlite3MutexAlloc(SQLITE_MUTEX_STATIC_MASTER));
13589	}
13590	#endif
13591	y.validYMD = 1;
13592	y.validHMS = 1;
13593	y.validJD = 0;
13594	y.validTZ = 0;
13595	computeJD(&y);

13596	return y.iJD - x.iJD;
13597	}
13598	#endif /* SQLITE_OMIT_LOCALTIME */
13599
13600	/*
	@@ -13614,13 +13733,16 @@
13614	** weekday N
13615	** unixepoch
13616	** localtime
13617	** utc
13618	**
13619	** Return 0 on success and 1 if there is any kind of error.



13620	*/
13621	static int parseModifier(const char zMod, DateTime p){
13622	int rc = 1;
13623	int n;
13624	double r;
13625	char *z, zBuf[30];
13626	z = zBuf;
	@@ -13636,13 +13758,12 @@
13636	** Assuming the current time value is UTC (a.k.a. GMT), shift it to
13637	** show local time.
13638	*/
13639	if( strcmp(z, "localtime")==0 ){
13640	computeJD(p);
13641	p->iJD += localtimeOffset(p);
13642	clearYMD_HMS_TZ(p);
13643	rc = 0;
13644	}
13645	break;
13646	}
13647	#endif
13648	case 'u': {
	@@ -13659,15 +13780,16 @@
13659	}
13660	#ifndef SQLITE_OMIT_LOCALTIME
13661	else if( strcmp(z, "utc")==0 ){
13662	sqlite3_int64 c1;
13663	computeJD(p);
13664	c1 = localtimeOffset(p);
13665	p->iJD -= c1;
13666	clearYMD_HMS_TZ(p);
13667	p->iJD += c1 - localtimeOffset(p);
13668	rc = 0;

13669	}
13670	#endif
13671	break;
13672	}
13673	case 'w': {
	@@ -13844,13 +13966,12 @@
13844	if( !z \|\| parseDateOrTime(context, (char*)z, p) ){
13845	return 1;
13846	}
13847	}
13848	for(i=1; i<argc; i++){
13849	if( (z = sqlite3_value_text(argv[i]))==0 \|\| parseModifier((char*)z, p) ){
13850	return 1;
13851	}
13852	}
13853	return 0;
13854	}
13855
13856
	@@ -14344,11 +14465,11 @@
14344	DO_OS_MALLOC_TEST(0);
14345	/* 0x87f3f is a mask of SQLITE_OPEN_ flags that are valid to be passed
14346	** down into the VFS layer. Some SQLITE_OPEN_ flags (for example,
14347	** SQLITE_OPEN_FULLMUTEX or SQLITE_OPEN_SHAREDCACHE) are blocked before
14348	** reaching the VFS. */
14349	rc = pVfs->xOpen(pVfs, zPath, pFile, flags & 0x87f3f, pFlagsOut);
14350	assert( rc==SQLITE_OK \|\| pFile->pMethods==0 );
14351	return rc;
14352	}
14353	SQLITE_PRIVATE int sqlite3OsDelete(sqlite3_vfs pVfs, const char zPath, int dirSync){
14354	return pVfs->xDelete(pVfs, zPath, dirSync);
	@@ -14880,10 +15001,11 @@
14880	extern void backtrace_symbols_fd(voidconst,int,int);
14881	#else
14882	# define backtrace(A,B) 1
14883	# define backtrace_symbols_fd(A,B,C)
14884	#endif

14885
14886	/*
14887	** Each memory allocation looks like this:
14888	**
14889	** ------------------------------------------------------------------------
	@@ -15805,11 +15927,11 @@
15805	** Free an outstanding memory allocation.
15806	**
15807	** This function assumes that the necessary mutexes, if any, are
15808	** already held by the caller. Hence "Unsafe".
15809	*/
15810	void memsys3FreeUnsafe(void *pOld){
15811	Mem3Block p = (Mem3Block)pOld;
15812	int i;
15813	u32 size, x;
15814	assert( sqlite3_mutex_held(mem3.mutex) );
15815	assert( p>mem3.aPool && p<&mem3.aPool[mem3.nPool] );
	@@ -15880,21 +16002,21 @@
15880	}
15881
15882	/*
15883	** Free memory.
15884	*/
15885	void memsys3Free(void *pPrior){
15886	assert( pPrior );
15887	memsys3Enter();
15888	memsys3FreeUnsafe(pPrior);
15889	memsys3Leave();
15890	}
15891
15892	/*
15893	** Change the size of an existing memory allocation
15894	*/
15895	void memsys3Realloc(void pPrior, int nBytes){
15896	int nOld;
15897	void *p;
15898	if( pPrior==0 ){
15899	return sqlite3_malloc(nBytes);
15900	}
	@@ -17983,10 +18105,11 @@
17983	**
17984	*************************************************************************
17985	**
17986	** Memory allocation functions used throughout sqlite.
17987	*/

17988
17989	/*
17990	** Attempt to release up to n bytes of non-essential memory currently
17991	** held by SQLite. An example of non-essential memory is memory used to
17992	** cache database pages that are not currently in use.
	@@ -19960,10 +20083,11 @@
19960	** BOM or Byte Order Mark:
19961	** 0xff 0xfe little-endian utf-16 follows
19962	** 0xfe 0xff big-endian utf-16 follows
19963	**
19964	*/

19965
19966	#ifndef SQLITE_AMALGAMATION
19967	/*
19968	** The following constant value is used by the SQLITE_BIGENDIAN and
19969	** SQLITE_LITTLEENDIAN macros.
	@@ -20502,10 +20626,11 @@
20502	**
20503	** This file contains functions for allocating memory, comparing
20504	** strings, and stuff like that.
20505	**
20506	*/

20507	#ifdef SQLITE_HAVE_ISNAN
20508	# include <math.h>
20509	#endif
20510
20511	/*
	@@ -21634,26 +21759,32 @@
21634	return -x;
21635	}
21636
21637	#ifdef SQLITE_ENABLE_8_3_NAMES
21638	/*
21639	** If SQLITE_ENABLE_8_3_NAME is set at compile-time and if the database
21640	** filename in zBaseFilename is a URI with the "8_3_names=1" parameter and
21641	** if filename in z[] has a suffix (a.k.a. "extension") that is longer than
21642	** three characters, then shorten the suffix on z[] to be the last three
21643	** characters of the original suffix.



21644	**
21645	** Examples:
21646	**
21647	** test.db-journal => test.nal
21648	** test.db-wal => test.wal
21649	** test.db-shm => test.shm
21650	*/
21651	SQLITE_PRIVATE void sqlite3FileSuffix3(const char zBaseFilename, char z){

21652	const char *zOk;
21653	zOk = sqlite3_uri_parameter(zBaseFilename, "8_3_names");
21654	if( zOk && sqlite3GetBoolean(zOk) ){


21655	int i, sz;
21656	sz = sqlite3Strlen30(z);
21657	for(i=sz-1; i>0 && z[i]!='/' && z[i]!='.'; i--){}
21658	if( z[i]=='.' && ALWAYS(sz>i+4) ) memcpy(&z[i+1], &z[sz-3], 4);
21659	}
	@@ -21674,10 +21805,11 @@
21674	**
21675	*************************************************************************
21676	** This is the implementation of generic hash-tables
21677	** used in SQLite.
21678	*/

21679
21680	/* Turn bulk memory into a hash table object by initializing the
21681	** fields of the Hash structure.
21682	**
21683	** "pNew" is a pointer to the hash table that is to be initialized.
	@@ -21982,52 +22114,52 @@
21982	/* 35 */ "ReadCookie",
21983	/* 36 */ "SetCookie",
21984	/* 37 */ "VerifyCookie",
21985	/* 38 */ "OpenRead",
21986	/* 39 */ "OpenWrite",
21987	/* 40 */ "OpenAutoindex",
21988	/* 41 */ "OpenEphemeral",
21989	/* 42 */ "OpenPseudo",
21990	/* 43 */ "Close",
21991	/* 44 */ "SeekLt",
21992	/* 45 */ "SeekLe",
21993	/* 46 */ "SeekGe",
21994	/* 47 */ "SeekGt",
21995	/* 48 */ "Seek",
21996	/* 49 */ "NotFound",
21997	/* 50 */ "Found",
21998	/* 51 */ "IsUnique",
21999	/* 52 */ "NotExists",
22000	/* 53 */ "Sequence",
22001	/* 54 */ "NewRowid",
22002	/* 55 */ "Insert",
22003	/* 56 */ "InsertInt",
22004	/* 57 */ "Delete",
22005	/* 58 */ "ResetCount",
22006	/* 59 */ "RowKey",
22007	/* 60 */ "RowData",
22008	/* 61 */ "Rowid",
22009	/* 62 */ "NullRow",
22010	/* 63 */ "Last",
22011	/* 64 */ "Sort",
22012	/* 65 */ "Rewind",
22013	/* 66 */ "Prev",
22014	/* 67 */ "Next",
22015	/* 68 */ "Or",
22016	/* 69 */ "And",
22017	/* 70 */ "IdxInsert",
22018	/* 71 */ "IdxDelete",
22019	/* 72 */ "IdxRowid",
22020	/* 73 */ "IsNull",
22021	/* 74 */ "NotNull",
22022	/* 75 */ "Ne",
22023	/* 76 */ "Eq",
22024	/* 77 */ "Gt",
22025	/* 78 */ "Le",
22026	/* 79 */ "Lt",
22027	/* 80 */ "Ge",
22028	/* 81 */ "IdxLT",
22029	/* 82 */ "BitAnd",
22030	/* 83 */ "BitOr",
22031	/* 84 */ "ShiftLeft",
22032	/* 85 */ "ShiftRight",
22033	/* 86 */ "Add",
	@@ -22034,58 +22166,58 @@
22034	/* 87 */ "Subtract",
22035	/* 88 */ "Multiply",
22036	/* 89 */ "Divide",
22037	/* 90 */ "Remainder",
22038	/* 91 */ "Concat",
22039	/* 92 */ "IdxGE",
22040	/* 93 */ "BitNot",
22041	/* 94 */ "String8",
22042	/* 95 */ "Destroy",
22043	/* 96 */ "Clear",
22044	/* 97 */ "CreateIndex",
22045	/* 98 */ "CreateTable",
22046	/* 99 */ "ParseSchema",
22047	/* 100 */ "LoadAnalysis",
22048	/* 101 */ "DropTable",
22049	/* 102 */ "DropIndex",
22050	/* 103 */ "DropTrigger",
22051	/* 104 */ "IntegrityCk",
22052	/* 105 */ "RowSetAdd",
22053	/* 106 */ "RowSetRead",
22054	/* 107 */ "RowSetTest",
22055	/* 108 */ "Program",
22056	/* 109 */ "Param",
22057	/* 110 */ "FkCounter",
22058	/* 111 */ "FkIfZero",
22059	/* 112 */ "MemMax",
22060	/* 113 */ "IfPos",
22061	/* 114 */ "IfNeg",
22062	/* 115 */ "IfZero",
22063	/* 116 */ "AggStep",
22064	/* 117 */ "AggFinal",
22065	/* 118 */ "Checkpoint",
22066	/* 119 */ "JournalMode",
22067	/* 120 */ "Vacuum",
22068	/* 121 */ "IncrVacuum",
22069	/* 122 */ "Expire",
22070	/* 123 */ "TableLock",
22071	/* 124 */ "VBegin",
22072	/* 125 */ "VCreate",
22073	/* 126 */ "VDestroy",
22074	/* 127 */ "VOpen",
22075	/* 128 */ "VFilter",
22076	/* 129 */ "VColumn",
22077	/* 130 */ "Real",
22078	/* 131 */ "VNext",
22079	/* 132 */ "VRename",
22080	/* 133 */ "VUpdate",
22081	/* 134 */ "Pagecount",
22082	/* 135 */ "MaxPgcnt",
22083	/* 136 */ "Trace",
22084	/* 137 */ "Noop",
22085	/* 138 */ "Explain",
22086	/* 139 */ "NotUsed_139",
22087	/* 140 */ "NotUsed_140",
22088	/* 141 */ "ToText",
22089	/* 142 */ "ToBlob",
22090	/* 143 */ "ToNumeric",
22091	/* 144 */ "ToInt",
	@@ -22183,14 +22315,17 @@
22183	#ifdef MEMORY_DEBUG
22184	# error "The MEMORY_DEBUG macro is obsolete. Use SQLITE_DEBUG instead."
22185	#endif
22186
22187	#ifdef SQLITE_DEBUG
22188	SQLITE_PRIVATE int sqlite3OSTrace = 0;
22189	#define OSTRACE(X) if( sqlite3OSTrace ) sqlite3DebugPrintf X



22190	#else
22191	#define OSTRACE(X)
22192	#endif
22193
22194	/*
22195	** Macros for performance tracing. Normally turned off. Only works
22196	** on i486 hardware.
	@@ -24343,10 +24478,11 @@
24343	*/
24344	#include <sys/types.h>
24345	#include <sys/stat.h>
24346	#include <fcntl.h>
24347	#include <unistd.h>

24348	#include <sys/time.h>
24349	#include <errno.h>
24350	#ifndef SQLITE_OMIT_WAL
24351	#include <sys/mman.h>
24352	#endif
	@@ -24364,10 +24500,14 @@
24364
24365	#if defined(__APPLE__) \|\| (SQLITE_ENABLE_LOCKING_STYLE && !OS_VXWORKS)
24366	# include <sys/mount.h>
24367	#endif
24368




24369	/*
24370	** Allowed values of unixFile.fsFlags
24371	*/
24372	#define SQLITE_FSFLAGS_IS_MSDOS 0x1
24373
	@@ -24374,10 +24514,11 @@
24374	/*
24375	** If we are to be thread-safe, include the pthreads header and define
24376	** the SQLITE_UNIX_THREADS macro.
24377	*/
24378	#if SQLITE_THREADSAFE

24379	# define SQLITE_UNIX_THREADS 1
24380	#endif
24381
24382	/*
24383	** Default permissions when creating a new file
	@@ -24429,11 +24570,10 @@
24429	typedef struct unixFile unixFile;
24430	struct unixFile {
24431	sqlite3_io_methods const pMethod; / Always the first entry */
24432	unixInodeInfo pInode; / Info about locks on this inode */
24433	int h; /* The file descriptor */
24434	int dirfd; /* File descriptor for the directory */
24435	unsigned char eFileLock; /* The type of lock held on this fd */
24436	unsigned char ctrlFlags; /* Behavioral bits. UNIXFILE_* flags */
24437	int lastErrno; /* The unix errno from last I/O error */
24438	void lockingContext; / Locking style specific state */
24439	UnixUnusedFd pUnused; / Pre-allocated UnixUnusedFd */
	@@ -24471,12 +24611,18 @@
24471	};
24472
24473	/*
24474	** Allowed values for the unixFile.ctrlFlags bitmask:
24475	*/
24476	#define UNIXFILE_EXCL 0x01 /* Connections from one process only */
24477	#define UNIXFILE_RDONLY 0x02 /* Connection is read only */






24478
24479	/*
24480	** Include code that is common to all os_*.c files
24481	*/
24482	/************ Include os_common.h in the middle of os_unix.c *************/
	@@ -24511,14 +24657,17 @@
24511	#ifdef MEMORY_DEBUG
24512	# error "The MEMORY_DEBUG macro is obsolete. Use SQLITE_DEBUG instead."
24513	#endif
24514
24515	#ifdef SQLITE_DEBUG
24516	SQLITE_PRIVATE int sqlite3OSTrace = 0;
24517	#define OSTRACE(X) if( sqlite3OSTrace ) sqlite3DebugPrintf X



24518	#else
24519	#define OSTRACE(X)
24520	#endif
24521
24522	/*
24523	** Macros for performance tracing. Normally turned off. Only works
24524	** on i486 hardware.
	@@ -24723,10 +24872,13 @@
24723	*/
24724	static int posixOpen(const char *zFile, int flags, int mode){
24725	return open(zFile, flags, mode);
24726	}
24727



24728	/*
24729	** Many system calls are accessed through pointer-to-functions so that
24730	** they may be overridden at runtime to facilitate fault injection during
24731	** testing and sandboxing. The following array holds the names and pointers
24732	** to all overrideable system calls.
	@@ -24819,10 +24971,16 @@
24819	#else
24820	{ "fallocate", (sqlite3_syscall_ptr)0, 0 },
24821	#endif
24822	#define osFallocate ((int(*)(int,off_t,off_t))aSyscall[15].pCurrent)
24823






24824	}; /* End of the overrideable system calls */
24825
24826	/*
24827	** This is the xSetSystemCall() method of sqlite3_vfs for all of the
24828	** "unix" VFSes. Return SQLITE_OK opon successfully updating the
	@@ -25103,11 +25261,13 @@
25103	case EINVAL:
25104	case ENOTCONN:
25105	case ENODEV:
25106	case ENXIO:
25107	case ENOENT:

25108	case ESTALE:

25109	case ENOSYS:
25110	/* these should force the client to close the file and reconnect */
25111
25112	default:
25113	return sqliteIOErr;
	@@ -26173,22 +26333,18 @@
26173	** even on VxWorks. A mutex will be acquired on VxWorks by the
26174	** vxworksReleaseFileId() routine.
26175	*/
26176	static int closeUnixFile(sqlite3_file *id){
26177	unixFile pFile = (unixFile)id;
26178	if( pFile->dirfd>=0 ){
26179	robust_close(pFile, pFile->dirfd, __LINE__);
26180	pFile->dirfd=-1;
26181	}
26182	if( pFile->h>=0 ){
26183	robust_close(pFile, pFile->h, __LINE__);
26184	pFile->h = -1;
26185	}
26186	#if OS_VXWORKS
26187	if( pFile->pId ){
26188	if( pFile->isDelete ){
26189	unlink(pFile->pId->zCanonicalName);
26190	}
26191	vxworksReleaseFileId(pFile->pId);
26192	pFile->pId = 0;
26193	}
26194	#endif
	@@ -26369,12 +26525,14 @@
26369	/* If we have any lock, then the lock file already exists. All we have
26370	** to do is adjust our internal record of the lock level.
26371	*/
26372	if( pFile->eFileLock > NO_LOCK ){
26373	pFile->eFileLock = eFileLock;
26374	#if !OS_VXWORKS
26375	/* Always update the timestamp on the old file */



26376	utimes(zLockFile, NULL);
26377	#endif
26378	return SQLITE_OK;
26379	}
26380
	@@ -26431,11 +26589,11 @@
26431	return SQLITE_OK;
26432	}
26433
26434	/* To fully unlock the database, delete the lock file */
26435	assert( eFileLock==NO_LOCK );
26436	if( unlink(zLockFile) ){
26437	int rc = 0;
26438	int tErrno = errno;
26439	if( ENOENT != tErrno ){
26440	rc = SQLITE_IOERR_UNLOCK;
26441	}
	@@ -26937,15 +27095,16 @@
26937	*/
26938	static int afpCheckReservedLock(sqlite3_file id, int pResOut){
26939	int rc = SQLITE_OK;
26940	int reserved = 0;
26941	unixFile pFile = (unixFile)id;

26942
26943	SimulateIOError( return SQLITE_IOERR_CHECKRESERVEDLOCK; );
26944
26945	assert( pFile );
26946	afpLockingContext context = (afpLockingContext ) pFile->lockingContext;
26947	if( context->reserved ){
26948	*pResOut = 1;
26949	return SQLITE_OK;
26950	}
26951	unixEnterMutex(); /* Because pFile->pInode is shared across threads */
	@@ -27081,11 +27240,11 @@
27081
27082	/* If control gets to this point, then actually go ahead and make
27083	** operating system calls for the specified lock.
27084	*/
27085	if( eFileLock==SHARED_LOCK ){
27086	int lrc1, lrc2, lrc1Errno;
27087	long lk, mask;
27088
27089	assert( pInode->nShared==0 );
27090	assert( pInode->eFileLock==0 );
27091
	@@ -27455,21 +27614,23 @@
27455	#if defined(USE_PREAD)
27456	do{ got = osPwrite(id->h, pBuf, cnt, offset); }while( got<0 && errno==EINTR );
27457	#elif defined(USE_PREAD64)
27458	do{ got = osPwrite64(id->h, pBuf, cnt, offset);}while( got<0 && errno==EINTR);
27459	#else
27460	newOffset = lseek(id->h, offset, SEEK_SET);
27461	SimulateIOError( newOffset-- );
27462	if( newOffset!=offset ){
27463	if( newOffset == -1 ){
27464	((unixFile*)id)->lastErrno = errno;
27465	}else{
27466	((unixFile*)id)->lastErrno = 0;
27467	}
27468	return -1;
27469	}
27470	do{ got = osWrite(id->h, pBuf, cnt); }while( got<0 && errno==EINTR );


27471	#endif
27472	TIMER_END;
27473	if( got<0 ){
27474	((unixFile*)id)->lastErrno = errno;
27475	}
	@@ -27666,10 +27827,54 @@
27666	if( OS_VXWORKS && rc!= -1 ){
27667	rc = 0;
27668	}
27669	return rc;
27670	}












































27671
27672	/*
27673	** Make sure all writes to a particular file are committed to disk.
27674	**
27675	** If dataOnly==0 then both the file itself and its metadata (file
	@@ -27707,32 +27912,27 @@
27707	SimulateIOError( rc=1 );
27708	if( rc ){
27709	pFile->lastErrno = errno;
27710	return unixLogError(SQLITE_IOERR_FSYNC, "full_fsync", pFile->zPath);
27711	}
27712	if( pFile->dirfd>=0 ){
27713	OSTRACE(("DIRSYNC %-3d (have_fullfsync=%d fullsync=%d)\n", pFile->dirfd,






27714	HAVE_FULLFSYNC, isFullsync));
27715	#ifndef SQLITE_DISABLE_DIRSYNC
27716	/* The directory sync is only attempted if full_fsync is
27717	** turned off or unavailable. If a full_fsync occurred above,
27718	** then the directory sync is superfluous.
27719	*/
27720	if( (!HAVE_FULLFSYNC \|\| !isFullsync) && full_fsync(pFile->dirfd,0,0) ){
27721	/*
27722	** We have received multiple reports of fsync() returning
27723	** errors when applied to directories on certain file systems.
27724	** A failed directory sync is not a big deal. So it seems
27725	** better to ignore the error. Ticket #1657
27726	*/
27727	/* pFile->lastErrno = errno; */
27728	/* return SQLITE_IOERR; */
27729	}
27730	#endif
27731	/* Only need to sync once, so close the directory when we are done */
27732	robust_close(pFile, pFile->dirfd, __LINE__);
27733	pFile->dirfd = -1;
27734	}
27735	return rc;
27736	}
27737
27738	/*
	@@ -27810,15 +28010,13 @@
27810	static int proxyFileControl(sqlite3_file,int,void);
27811	#endif
27812
27813	/*
27814	** This function is called to handle the SQLITE_FCNTL_SIZE_HINT
27815	** file-control operation.
27816	**
27817	** If the user has configured a chunk-size for this file, it could be
27818	** that the file needs to be extended at this point. Otherwise, the
27819	** SQLITE_FCNTL_SIZE_HINT operation is a no-op for Unix.
27820	*/
27821	static int fcntlSizeHint(unixFile *pFile, i64 nByte){
27822	if( pFile->szChunk ){
27823	i64 nSize; /* Required file size */
27824	struct stat buf; /* Used to hold return values of fstat() */
	@@ -27866,25 +28064,41 @@
27866
27867	/*
27868	** Information and control of an open file handle.
27869	*/
27870	static int unixFileControl(sqlite3_file id, int op, void pArg){

27871	switch( op ){
27872	case SQLITE_FCNTL_LOCKSTATE: {
27873	(int)pArg = ((unixFile*)id)->eFileLock;
27874	return SQLITE_OK;
27875	}
27876	case SQLITE_LAST_ERRNO: {
27877	(int)pArg = ((unixFile*)id)->lastErrno;
27878	return SQLITE_OK;
27879	}
27880	case SQLITE_FCNTL_CHUNK_SIZE: {
27881	((unixFile)id)->szChunk = (int *)pArg;
27882	return SQLITE_OK;
27883	}
27884	case SQLITE_FCNTL_SIZE_HINT: {
27885	return fcntlSizeHint((unixFile )id, (i64 *)pArg);















27886	}
27887	#ifndef NDEBUG
27888	/* The pager calls this method to signal that it has done
27889	** a rollback and that the database is therefore unchanged and
27890	** it hence it is OK for the transaction change counter to be
	@@ -28096,11 +28310,11 @@
28096	unixShmNode *p = pFd->pInode->pShmNode;
28097	assert( unixMutexHeld() );
28098	if( p && p->nRef==0 ){
28099	int i;
28100	assert( p->pInode==pFd->pInode );
28101	if( p->mutex ) sqlite3_mutex_free(p->mutex);
28102	for(i=0; i<p->nRegion; i++){
28103	if( p->h>=0 ){
28104	munmap(p->apRegion[i], p->szRegion);
28105	}else{
28106	sqlite3_free(p->apRegion[i]);
	@@ -28565,11 +28779,11 @@
28565	** shared-memory file, too */
28566	unixEnterMutex();
28567	assert( pShmNode->nRef>0 );
28568	pShmNode->nRef--;
28569	if( pShmNode->nRef==0 ){
28570	if( deleteFlag && pShmNode->h>=0 ) unlink(pShmNode->zFilename);
28571	unixShmPurge(pDbFd);
28572	}
28573	unixLeaveMutex();
28574
28575	return SQLITE_OK;
	@@ -28878,11 +29092,11 @@
28878	** Initialize the contents of the unixFile structure pointed to by pId.
28879	*/
28880	static int fillInUnixFile(
28881	sqlite3_vfs pVfs, / Pointer to vfs object */
28882	int h, /* Open file descriptor of file being opened */
28883	int dirfd, /* Directory file descriptor */
28884	sqlite3_file pId, / Write to the unixFile structure here */
28885	const char zFilename, / Name of the file being opened */
28886	int noLock, /* Omit locking if true */
28887	int isDelete, /* Delete on close if true */
28888	int isReadOnly /* True if the file is opened read-only */
	@@ -28909,20 +29123,22 @@
28909	assert( zFilename==0 \|\| zFilename[0]=='/' );
28910	#endif
28911
28912	OSTRACE(("OPEN %-3d %s\n", h, zFilename));
28913	pNew->h = h;
28914	pNew->dirfd = dirfd;
28915	pNew->zPath = zFilename;
28916	if( memcmp(pVfs->zName,"unix-excl",10)==0 ){
28917	pNew->ctrlFlags = UNIXFILE_EXCL;
28918	}else{
28919	pNew->ctrlFlags = 0;
28920	}
28921	if( isReadOnly ){
28922	pNew->ctrlFlags \|= UNIXFILE_RDONLY;
28923	}



28924
28925	#if OS_VXWORKS
28926	pNew->pId = vxworksFindFileId(zFilename);
28927	if( pNew->pId==0 ){
28928	noLock = 1;
	@@ -29045,56 +29261,24 @@
29045	pNew->lastErrno = 0;
29046	#if OS_VXWORKS
29047	if( rc!=SQLITE_OK ){
29048	if( h>=0 ) robust_close(pNew, h, __LINE__);
29049	h = -1;
29050	unlink(zFilename);
29051	isDelete = 0;
29052	}
29053	pNew->isDelete = isDelete;
29054	#endif
29055	if( rc!=SQLITE_OK ){
29056	if( dirfd>=0 ) robust_close(pNew, dirfd, __LINE__);
29057	if( h>=0 ) robust_close(pNew, h, __LINE__);
29058	}else{
29059	pNew->pMethod = pLockingStyle;
29060	OpenCounter(+1);
29061	}
29062	return rc;
29063	}
29064
29065	/*
29066	** Open a file descriptor to the directory containing file zFilename.
29067	** If successful, *pFd is set to the opened file descriptor and
29068	** SQLITE_OK is returned. If an error occurs, either SQLITE_NOMEM
29069	** or SQLITE_CANTOPEN is returned and *pFd is set to an undefined
29070	** value.
29071	**
29072	** If SQLITE_OK is returned, the caller is responsible for closing
29073	** the file descriptor *pFd using close().
29074	*/
29075	static int openDirectory(const char zFilename, int pFd){
29076	int ii;
29077	int fd = -1;
29078	char zDirname[MAX_PATHNAME+1];
29079
29080	sqlite3_snprintf(MAX_PATHNAME, zDirname, "%s", zFilename);
29081	for(ii=(int)strlen(zDirname); ii>1 && zDirname[ii]!='/'; ii--);
29082	if( ii>0 ){
29083	zDirname[ii] = '\0';
29084	fd = robust_open(zDirname, O_RDONLY\|O_BINARY, 0);
29085	if( fd>=0 ){
29086	#ifdef FD_CLOEXEC
29087	osFcntl(fd, F_SETFD, osFcntl(fd, F_GETFD, 0) \| FD_CLOEXEC);
29088	#endif
29089	OSTRACE(("OPENDIR %-3d %s\n", fd, zDirname));
29090	}
29091	}
29092	*pFd = fd;
29093	return (fd>=0?SQLITE_OK:unixLogError(SQLITE_CANTOPEN_BKPT, "open", zDirname));
29094	}
29095
29096	/*
29097	** Return the name of a directory in which to put temporary files.
29098	** If no suitable temporary file directory can be found, return NULL.
29099	*/
29100	static const char *unixTempFileDir(void){
	@@ -29205,11 +29389,11 @@
29205	** ignored and -1 is returned. The caller will try to open a new file
29206	** descriptor on the same path, fail, and return an error to SQLite.
29207	**
29208	** Even if a subsequent open() call does succeed, the consequences of
29209	** not searching for a resusable file descriptor are not dire. */
29210	if( 0==stat(zPath, &sStat) ){
29211	unixInodeInfo *pInode;
29212
29213	unixEnterMutex();
29214	pInode = inodeList;
29215	while( pInode && (pInode->fileId.dev!=sStat.st_dev
	@@ -29281,11 +29465,11 @@
29281	while( nDb>0 && zPath[nDb]!='-' ) nDb--;
29282	if( nDb==0 ) return SQLITE_OK;
29283	memcpy(zDb, zPath, nDb);
29284	zDb[nDb] = '\0';
29285
29286	if( 0==stat(zDb, &sStat) ){
29287	*pMode = sStat.st_mode & 0777;
29288	}else{
29289	rc = SQLITE_IOERR_FSTAT;
29290	}
29291	}else if( flags & SQLITE_OPEN_DELETEONCLOSE ){
	@@ -29323,11 +29507,10 @@
29323	int flags, /* Input flags to control the opening */
29324	int pOutFlags / Output flags returned to SQLite core */
29325	){
29326	unixFile p = (unixFile )pFile;
29327	int fd = -1; /* File descriptor returned by open() */
29328	int dirfd = -1; /* Directory file descriptor */
29329	int openFlags = 0; /* Flags to pass to open() */
29330	int eType = flags&0xFFFFFF00; /* Type of file to open */
29331	int noLock; /* True to omit locking primitives */
29332	int rc = SQLITE_OK; /* Function Return Code */
29333
	@@ -29337,16 +29520,19 @@
29337	int isReadonly = (flags & SQLITE_OPEN_READONLY);
29338	int isReadWrite = (flags & SQLITE_OPEN_READWRITE);
29339	#if SQLITE_ENABLE_LOCKING_STYLE
29340	int isAutoProxy = (flags & SQLITE_OPEN_AUTOPROXY);
29341	#endif



29342
29343	/* If creating a master or main-file journal, this function will open
29344	** a file-descriptor on the directory too. The first time unixSync()
29345	** is called the directory file descriptor will be fsync()ed and close()d.
29346	*/
29347	int isOpenDirectory = (isCreate && (
29348	eType==SQLITE_OPEN_MASTER_JOURNAL
29349	\|\| eType==SQLITE_OPEN_MAIN_JOURNAL
29350	\|\| eType==SQLITE_OPEN_WAL
29351	));
29352
	@@ -29396,11 +29582,11 @@
29396	}
29397	}
29398	p->pUnused = pUnused;
29399	}else if( !zName ){
29400	/* If zName is NULL, the upper layer is requesting a temp file. */
29401	assert(isDelete && !isOpenDirectory);
29402	rc = unixGetTempname(MAX_PATHNAME+1, zTmpname);
29403	if( rc!=SQLITE_OK ){
29404	return rc;
29405	}
29406	zName = zTmpname;
	@@ -29452,44 +29638,29 @@
29452
29453	if( isDelete ){
29454	#if OS_VXWORKS
29455	zPath = zName;
29456	#else
29457	unlink(zName);
29458	#endif
29459	}
29460	#if SQLITE_ENABLE_LOCKING_STYLE
29461	else{
29462	p->openFlags = openFlags;
29463	}
29464	#endif
29465
29466	if( isOpenDirectory ){
29467	rc = openDirectory(zPath, &dirfd);
29468	if( rc!=SQLITE_OK ){
29469	/* It is safe to close fd at this point, because it is guaranteed not
29470	** to be open on a database file. If it were open on a database file,
29471	** it would not be safe to close as this would release any locks held
29472	** on the file by this process. */
29473	assert( eType!=SQLITE_OPEN_MAIN_DB );
29474	robust_close(p, fd, __LINE__);
29475	goto open_finished;
29476	}
29477	}
29478
29479	#ifdef FD_CLOEXEC
29480	osFcntl(fd, F_SETFD, osFcntl(fd, F_GETFD, 0) \| FD_CLOEXEC);
29481	#endif
29482
29483	noLock = eType!=SQLITE_OPEN_MAIN_DB;
29484
29485
29486	#if defined(__APPLE__) \|\| SQLITE_ENABLE_LOCKING_STYLE
29487	struct statfs fsInfo;
29488	if( fstatfs(fd, &fsInfo) == -1 ){
29489	((unixFile*)pFile)->lastErrno = errno;
29490	if( dirfd>=0 ) robust_close(p, dirfd, __LINE__);
29491	robust_close(p, fd, __LINE__);
29492	return SQLITE_IOERR_ACCESS;
29493	}
29494	if (0 == strncmp("msdos", fsInfo.f_fstypename, 5)) {
29495	((unixFile*)pFile)->fsFlags \|= SQLITE_FSFLAGS_IS_MSDOS;
	@@ -29507,31 +29678,27 @@
29507	/* SQLITE_FORCE_PROXY_LOCKING==1 means force always use proxy, 0 means
29508	** never use proxy, NULL means use proxy for non-local files only. */
29509	if( envforce!=NULL ){
29510	useProxy = atoi(envforce)>0;
29511	}else{
29512	struct statfs fsInfo;
29513	if( statfs(zPath, &fsInfo) == -1 ){
29514	/* In theory, the close(fd) call is sub-optimal. If the file opened
29515	** with fd is a database file, and there are other connections open
29516	** on that file that are currently holding advisory locks on it,
29517	** then the call to close() will cancel those locks. In practice,
29518	** we're assuming that statfs() doesn't fail very often. At least
29519	** not while other file descriptors opened by the same process on
29520	** the same file are working. */
29521	p->lastErrno = errno;
29522	if( dirfd>=0 ){
29523	robust_close(p, dirfd, __LINE__);
29524	}
29525	robust_close(p, fd, __LINE__);
29526	rc = SQLITE_IOERR_ACCESS;
29527	goto open_finished;
29528	}
29529	useProxy = !(fsInfo.f_flags&MNT_LOCAL);
29530	}
29531	if( useProxy ){
29532	rc = fillInUnixFile(pVfs, fd, dirfd, pFile, zPath, noLock,
29533	isDelete, isReadonly);
29534	if( rc==SQLITE_OK ){
29535	rc = proxyTransformUnixFile((unixFile*)pFile, ":auto:");
29536	if( rc!=SQLITE_OK ){
29537	/* Use unixClose to clean up the resources added in fillInUnixFile
	@@ -29545,11 +29712,11 @@
29545	goto open_finished;
29546	}
29547	}
29548	#endif
29549
29550	rc = fillInUnixFile(pVfs, fd, dirfd, pFile, zPath, noLock,
29551	isDelete, isReadonly);
29552	open_finished:
29553	if( rc!=SQLITE_OK ){
29554	sqlite3_free(p->pUnused);
29555	}
	@@ -29567,17 +29734,17 @@
29567	int dirSync /* If true, fsync() directory after deleting file */
29568	){
29569	int rc = SQLITE_OK;
29570	UNUSED_PARAMETER(NotUsed);
29571	SimulateIOError(return SQLITE_IOERR_DELETE);
29572	if( unlink(zPath)==(-1) && errno!=ENOENT ){
29573	return unixLogError(SQLITE_IOERR_DELETE, "unlink", zPath);
29574	}
29575	#ifndef SQLITE_DISABLE_DIRSYNC
29576	if( dirSync ){
29577	int fd;
29578	rc = openDirectory(zPath, &fd);
29579	if( rc==SQLITE_OK ){
29580	#if OS_VXWORKS
29581	if( fsync(fd)==-1 )
29582	#else
29583	if( fsync(fd) )
	@@ -29584,10 +29751,12 @@
29584	#endif
29585	{
29586	rc = unixLogError(SQLITE_IOERR_DIR_FSYNC, "fsync", zPath);
29587	}
29588	robust_close(0, fd, __LINE__);


29589	}
29590	}
29591	#endif
29592	return rc;
29593	}
	@@ -29626,11 +29795,11 @@
29626	assert(!"Invalid flags argument");
29627	}
29628	*pResOut = (osAccess(zPath, amode)==0);
29629	if( flags==SQLITE_ACCESS_EXISTS && *pResOut ){
29630	struct stat buf;
29631	if( 0==stat(zPath, &buf) && buf.st_size==0 ){
29632	*pResOut = 0;
29633	}
29634	}
29635	return SQLITE_OK;
29636	}
	@@ -30145,11 +30314,10 @@
30145	const char path, / path for the new unixFile */
30146	unixFile *ppFile, / unixFile created and returned by ref */
30147	int islockfile /* if non zero missing dirs will be created */
30148	) {
30149	int fd = -1;
30150	int dirfd = -1;
30151	unixFile *pNew;
30152	int rc = SQLITE_OK;
30153	int openFlags = O_RDWR \| O_CREAT;
30154	sqlite3_vfs dummyVfs;
30155	int terrno = 0;
	@@ -30210,11 +30378,11 @@
30210	dummyVfs.zName = "dummy";
30211	pUnused->fd = fd;
30212	pUnused->flags = openFlags;
30213	pNew->pUnused = pUnused;
30214
30215	rc = fillInUnixFile(&dummyVfs, fd, dirfd, (sqlite3_file*)pNew, path, 0, 0, 0);
30216	if( rc==SQLITE_OK ){
30217	*ppFile = pNew;
30218	return SQLITE_OK;
30219	}
30220	end_create_proxy:
	@@ -30250,10 +30418,12 @@
30250	*pError = err;
30251	}
30252	return SQLITE_IOERR;
30253	}
30254	}


30255	#endif
30256	#ifdef SQLITE_TEST
30257	/* simulate multiple hosts by creating unique hostid file paths */
30258	if( sqlite3_hostid_num != 0){
30259	pHostID[0] = (char)(pHostID[0] + (char)(sqlite3_hostid_num & 0xFF));
	@@ -30324,11 +30494,11 @@
30324	conchFile->openFlags = O_RDWR \| O_CREAT;
30325
30326	end_breaklock:
30327	if( rc ){
30328	if( fd>=0 ){
30329	unlink(tPath);
30330	robust_close(pFile, fd, __LINE__);
30331	}
30332	fprintf(stderr, "failed to break stale lock on %s, %s\n", cPath, errmsg);
30333	}
30334	return rc;
	@@ -30342,10 +30512,11 @@
30342	unixFile *conchFile = pCtx->conchFile;
30343	int rc = SQLITE_OK;
30344	int nTries = 0;
30345	struct timespec conchModTime;
30346

30347	do {
30348	rc = conchFile->pMethod->xLock((sqlite3_file*)conchFile, lockType);
30349	nTries ++;
30350	if( rc==SQLITE_BUSY ){
30351	/* If the lock failed (busy):
	@@ -30573,15 +30744,16 @@
30573	conchFile->pMethod->xUnlock((sqlite3_file*)conchFile, SHARED_LOCK);
30574
30575	end_takeconch:
30576	OSTRACE(("TRANSPROXY: CLOSE %d\n", pFile->h));
30577	if( rc==SQLITE_OK && pFile->openFlags ){

30578	if( pFile->h>=0 ){
30579	robust_close(pFile, pFile->h, __LINE__);
30580	}
30581	pFile->h = -1;
30582	int fd = robust_open(pCtx->dbPath, pFile->openFlags,
30583	SQLITE_DEFAULT_FILE_PERMISSIONS);
30584	OSTRACE(("TRANSPROXY: OPEN %d\n", fd));
30585	if( fd>=0 ){
30586	pFile->h = fd;
30587	}else{
	@@ -31147,11 +31319,11 @@
31147	};
31148	unsigned int i; /* Loop counter */
31149
31150	/* Double-check that the aSyscall[] array has been constructed
31151	** correctly. See ticket [bb3a86e890c8e96ab] */
31152	assert( ArraySize(aSyscall)==16 );
31153
31154	/* Register all VFSes defined in the aVfs[] array */
31155	for(i=0; i<(sizeof(aVfs)/sizeof(sqlite3_vfs)); i++){
31156	sqlite3_vfs_register(&aVfs[i], i==0);
31157	}
	@@ -31264,14 +31436,17 @@
31264	#ifdef MEMORY_DEBUG
31265	# error "The MEMORY_DEBUG macro is obsolete. Use SQLITE_DEBUG instead."
31266	#endif
31267
31268	#ifdef SQLITE_DEBUG
31269	SQLITE_PRIVATE int sqlite3OSTrace = 0;
31270	#define OSTRACE(X) if( sqlite3OSTrace ) sqlite3DebugPrintf X



31271	#else
31272	#define OSTRACE(X)
31273	#endif
31274
31275	/*
31276	** Macros for performance tracing. Normally turned off. Only works
31277	** on i486 hardware.
	@@ -31479,12 +31654,13 @@
31479	typedef struct winFile winFile;
31480	struct winFile {
31481	const sqlite3_io_methods pMethod; / Must be first */
31482	sqlite3_vfs pVfs; / The VFS used to open this file */
31483	HANDLE h; /* Handle for accessing the file */
31484	unsigned char locktype; /* Type of lock currently held on this file */
31485	short sharedLockByte; /* Randomly chosen byte used as a shared lock */

31486	DWORD lastErrno; /* The Windows errno from the last I/O error */
31487	DWORD sectorSize; /* Sector size of the device file is on */
31488	winShm pShm; / Instance of shared memory on this file */
31489	const char zPath; / Full pathname of this file */
31490	int szChunk; /* Chunk size configured by FCNTL_CHUNK_SIZE */
	@@ -31495,10 +31671,80 @@
31495	winceLock local; /* Locks obtained by this instance of winFile */
31496	winceLock shared; / Global shared lock memory for the file */
31497	#endif
31498	};
31499






































































31500
31501	/*
31502	** Forward prototypes.
31503	*/
31504	static int getSectorSize(
	@@ -31547,10 +31793,192 @@
31547	}
31548	return sqlite3_os_type==2;
31549	}
31550	#endif /* SQLITE_OS_WINCE */
31551






















































































































































































31552	/*
31553	** Convert a UTF-8 string to microsoft unicode (UTF-16?).
31554	**
31555	** Space to hold the returned string is obtained from malloc.
31556	*/
	@@ -31778,19 +32206,68 @@
31778	iLine, iErrno, zFunc, zPath, zMsg
31779	);
31780
31781	return errcode;
31782	}
















































31783
31784	#if SQLITE_OS_WINCE
31785	/*************************************************************************
31786	** This section contains code for WinCE only.
31787	*/
31788	/*
31789	** WindowsCE does not have a localtime() function. So create a
31790	** substitute.
31791	*/

31792	struct tm __cdecl localtime(const time_t t)
31793	{
31794	static struct tm y;
31795	FILETIME uTm, lTm;
31796	SYSTEMTIME pTm;
	@@ -32196,22 +32673,25 @@
32196	int amt, /* Number of bytes to read */
32197	sqlite3_int64 offset /* Begin reading at this offset */
32198	){
32199	winFile pFile = (winFile)id; /* file handle */
32200	DWORD nRead; /* Number of bytes actually read from file */

32201
32202	assert( id!=0 );
32203	SimulateIOError(return SQLITE_IOERR_READ);
32204	OSTRACE(("READ %d lock=%d\n", pFile->h, pFile->locktype));
32205
32206	if( seekWinFile(pFile, offset) ){
32207	return SQLITE_FULL;
32208	}
32209	if( !ReadFile(pFile->h, pBuf, amt, &nRead, 0) ){

32210	pFile->lastErrno = GetLastError();
32211	return winLogError(SQLITE_IOERR_READ, "winRead", pFile->zPath);
32212	}

32213	if( nRead<(DWORD)amt ){
32214	/* Unread parts of the buffer must be zero-filled */
32215	memset(&((char*)pBuf)[nRead], 0, amt-nRead);
32216	return SQLITE_IOERR_SHORT_READ;
32217	}
	@@ -32229,10 +32709,11 @@
32229	int amt, /* Number of bytes to write */
32230	sqlite3_int64 offset /* Offset into the file to begin writing at */
32231	){
32232	int rc; /* True if error has occured, else false */
32233	winFile pFile = (winFile)id; /* File handle */

32234
32235	assert( amt>0 );
32236	assert( pFile );
32237	SimulateIOError(return SQLITE_IOERR_WRITE);
32238	SimulateDiskfullError(return SQLITE_FULL);
	@@ -32243,11 +32724,16 @@
32243	if( rc==0 ){
32244	u8 aRem = (u8 )pBuf; /* Data yet to be written */
32245	int nRem = amt; /* Number of bytes yet to be written */
32246	DWORD nWrite; /* Bytes written by each WriteFile() call */
32247
32248	while( nRem>0 && WriteFile(pFile->h, aRem, nRem, &nWrite, 0) && nWrite>0 ){





32249	aRem += nWrite;
32250	nRem -= nWrite;
32251	}
32252	if( nRem>0 ){
32253	pFile->lastErrno = GetLastError();
	@@ -32254,14 +32740,17 @@
32254	rc = 1;
32255	}
32256	}
32257
32258	if( rc ){
32259	if( pFile->lastErrno==ERROR_HANDLE_DISK_FULL ){

32260	return SQLITE_FULL;
32261	}
32262	return winLogError(SQLITE_IOERR_WRITE, "winWrite", pFile->zPath);


32263	}
32264	return SQLITE_OK;
32265	}
32266
32267	/*
	@@ -32651,32 +33140,63 @@
32651
32652	/*
32653	** Control and query of the open file handle.
32654	*/
32655	static int winFileControl(sqlite3_file id, int op, void pArg){

32656	switch( op ){
32657	case SQLITE_FCNTL_LOCKSTATE: {
32658	(int)pArg = ((winFile*)id)->locktype;
32659	return SQLITE_OK;
32660	}
32661	case SQLITE_LAST_ERRNO: {
32662	(int)pArg = (int)((winFile*)id)->lastErrno;
32663	return SQLITE_OK;
32664	}
32665	case SQLITE_FCNTL_CHUNK_SIZE: {
32666	((winFile)id)->szChunk = (int *)pArg;
32667	return SQLITE_OK;
32668	}
32669	case SQLITE_FCNTL_SIZE_HINT: {
32670	sqlite3_int64 sz = (sqlite3_int64)pArg;
32671	SimulateIOErrorBenign(1);
32672	winTruncate(id, sz);
32673	SimulateIOErrorBenign(0);
















32674	return SQLITE_OK;
32675	}
32676	case SQLITE_FCNTL_SYNC_OMITTED: {
32677	return SQLITE_OK;














32678	}
32679	}
32680	return SQLITE_NOTFOUND;
32681	}
32682
	@@ -33482,10 +34002,11 @@
33482	int isTemp = 0;
33483	#endif
33484	winFile pFile = (winFile)id;
33485	void zConverted; / Filename in OS encoding */
33486	const char zUtf8Name = zName; / Filename in UTF-8 encoding */

33487
33488	/* If argument zPath is a NULL pointer, this function is required to open
33489	** a temporary file. Use this buffer to store the file name in.
33490	*/
33491	char zTmpname[MAX_PATH+1]; /* Buffer used to create temp filename */
	@@ -33601,34 +34122,34 @@
33601	#if SQLITE_OS_WINCE
33602	dwFlagsAndAttributes \|= FILE_FLAG_RANDOM_ACCESS;
33603	#endif
33604
33605	if( isNT() ){
33606	h = CreateFileW((WCHAR*)zConverted,
33607	dwDesiredAccess,
33608	dwShareMode,
33609	NULL,
33610	dwCreationDisposition,
33611	dwFlagsAndAttributes,
33612	NULL
33613	);
33614	/* isNT() is 1 if SQLITE_OS_WINCE==1, so this else is never executed.
33615	** Since the ASCII version of these Windows API do not exist for WINCE,
33616	** it's important to not reference them for WINCE builds.
33617	*/
33618	#if SQLITE_OS_WINCE==0
33619	}else{
33620	h = CreateFileA((char*)zConverted,
33621	dwDesiredAccess,
33622	dwShareMode,
33623	NULL,
33624	dwCreationDisposition,
33625	dwFlagsAndAttributes,
33626	NULL
33627	);
33628	#endif
33629	}


33630
33631	OSTRACE(("OPEN %d %s 0x%lx %s\n",
33632	h, zName, dwDesiredAccess,
33633	h==INVALID_HANDLE_VALUE ? "failed" : "ok"));
33634
	@@ -33691,19 +34212,17 @@
33691	** file open, we will be unable to delete it. To work around this
33692	** problem, we delay 100 milliseconds and try to delete again. Up
33693	** to MX_DELETION_ATTEMPTs deletion attempts are run before giving
33694	** up and returning an error.
33695	*/
33696	#define MX_DELETION_ATTEMPTS 5
33697	static int winDelete(
33698	sqlite3_vfs pVfs, / Not used on win32 */
33699	const char zFilename, / Name of file to delete */
33700	int syncDir /* Not used on win32 */
33701	){
33702	int cnt = 0;
33703	DWORD rc;
33704	DWORD error = 0;
33705	void *zConverted;
33706	UNUSED_PARAMETER(pVfs);
33707	UNUSED_PARAMETER(syncDir);
33708
33709	SimulateIOError(return SQLITE_IOERR_DELETE);
	@@ -33710,38 +34229,34 @@
33710	zConverted = convertUtf8Filename(zFilename);
33711	if( zConverted==0 ){
33712	return SQLITE_NOMEM;
33713	}
33714	if( isNT() ){
33715	do{
33716	DeleteFileW(zConverted);
33717	}while( ( ((rc = GetFileAttributesW(zConverted)) != INVALID_FILE_ATTRIBUTES)
33718	\|\| ((error = GetLastError()) == ERROR_ACCESS_DENIED))
33719	&& (++cnt < MX_DELETION_ATTEMPTS)
33720	&& (Sleep(100), 1) );
33721	/* isNT() is 1 if SQLITE_OS_WINCE==1, so this else is never executed.
33722	** Since the ASCII version of these Windows API do not exist for WINCE,
33723	** it's important to not reference them for WINCE builds.
33724	*/
33725	#if SQLITE_OS_WINCE==0
33726	}else{
33727	do{
33728	DeleteFileA(zConverted);
33729	}while( ( ((rc = GetFileAttributesA(zConverted)) != INVALID_FILE_ATTRIBUTES)
33730	\|\| ((error = GetLastError()) == ERROR_ACCESS_DENIED))
33731	&& (++cnt < MX_DELETION_ATTEMPTS)
33732	&& (Sleep(100), 1) );
33733	#endif
33734	}





33735	free(zConverted);
33736	OSTRACE(("DELETE \"%s\" %s\n", zFilename,
33737	( (rc==INVALID_FILE_ATTRIBUTES) && (error==ERROR_FILE_NOT_FOUND)) ?
33738	"ok" : "failed" ));
33739
33740	return ( (rc == INVALID_FILE_ATTRIBUTES)
33741	&& (error == ERROR_FILE_NOT_FOUND)) ? SQLITE_OK :
33742	winLogError(SQLITE_IOERR_DELETE, "winDelete", zFilename);
33743	}
33744
33745	/*
33746	** Check the existance and status of a file.
33747	*/
	@@ -33760,15 +34275,17 @@
33760	zConverted = convertUtf8Filename(zFilename);
33761	if( zConverted==0 ){
33762	return SQLITE_NOMEM;
33763	}
33764	if( isNT() ){

33765	WIN32_FILE_ATTRIBUTE_DATA sAttrData;
33766	memset(&sAttrData, 0, sizeof(sAttrData));
33767	if( GetFileAttributesExW((WCHAR*)zConverted,
33768	GetFileExInfoStandard,
33769	&sAttrData) ){

33770	/* For an SQLITE_ACCESS_EXISTS query, treat a zero-length file
33771	** as if it does not exist.
33772	*/
33773	if( flags==SQLITE_ACCESS_EXISTS
33774	&& sAttrData.nFileSizeHigh==0
	@@ -33776,10 +34293,11 @@
33776	attr = INVALID_FILE_ATTRIBUTES;
33777	}else{
33778	attr = sAttrData.dwFileAttributes;
33779	}
33780	}else{

33781	if( GetLastError()!=ERROR_FILE_NOT_FOUND ){
33782	winLogError(SQLITE_IOERR_ACCESS, "winAccess", zFilename);
33783	free(zConverted);
33784	return SQLITE_IOERR_ACCESS;
33785	}else{
	@@ -33800,11 +34318,12 @@
33800	case SQLITE_ACCESS_READ:
33801	case SQLITE_ACCESS_EXISTS:
33802	rc = attr!=INVALID_FILE_ATTRIBUTES;
33803	break;
33804	case SQLITE_ACCESS_READWRITE:
33805	rc = (attr & FILE_ATTRIBUTE_READONLY)==0;

33806	break;
33807	default:
33808	assert(!"Invalid flags argument");
33809	}
33810	*pResOut = rc;
	@@ -35256,10 +35775,13 @@
35256	typedef struct PCache1 PCache1;
35257	typedef struct PgHdr1 PgHdr1;
35258	typedef struct PgFreeslot PgFreeslot;
35259	typedef struct PGroup PGroup;
35260



35261	/* Each page cache (or PCache) belongs to a PGroup. A PGroup is a set
35262	** of one or more PCaches that are able to recycle each others unpinned
35263	** pages when they are under memory pressure. A PGroup is an instance of
35264	** the following object.
35265	**
	@@ -35285,12 +35807,70 @@
35285	int nMaxPage; /* Sum of nMax for purgeable caches */
35286	int nMinPage; /* Sum of nMin for purgeable caches */
35287	int mxPinned; /* nMaxpage + 10 - nMinPage */
35288	int nCurrentPage; /* Number of purgeable pages allocated */
35289	PgHdr1 pLruHead, pLruTail; /* LRU list of unpinned pages */




35290	};
35291






















































35292	/* Each page cache is an instance of the following object. Every
35293	** open database file (including each in-memory database and each
35294	** temporary or transient database) has a single page cache which
35295	** is an instance of this object.
35296	**
	@@ -35389,10 +35969,21 @@
35389	**
35390	** assert( PGHDR1_TO_PAGE(PAGE_TO_PGHDR1(pCache, X))==X );
35391	*/
35392	#define PGHDR1_TO_PAGE(p) (void)(((char)p) - p->pCache->szPage)
35393	#define PAGE_TO_PGHDR1(c, p) (PgHdr1)(((char)p) + c->szPage)











35394
35395	/*
35396	** Macros to enter and leave the PCache LRU mutex.
35397	*/
35398	#define pcache1EnterMutex(X) sqlite3_mutex_enter((X)->mutex)
	@@ -35515,25 +36106,159 @@
35515	return iSize;
35516	}
35517	}
35518	#endif /* SQLITE_ENABLE_MEMORY_MANAGEMENT */
35519
































35520	/*
35521	** Allocate a new page object initially associated with cache pCache.
35522	*/
35523	static PgHdr1 pcache1AllocPage(PCache1 pCache){
35524	int nByte = sizeof(PgHdr1) + pCache->szPage;
35525	void *pPg = pcache1Alloc(nByte);
35526	PgHdr1 *p;





































































































35527	if( pPg ){
35528	p = PAGE_TO_PGHDR1(pCache, pPg);
35529	if( pCache->bPurgeable ){
35530	pCache->pGroup->nCurrentPage++;
35531	}
35532	}else{
35533	p = 0;
35534	}

35535	return p;
35536	}
35537
35538	/*
35539	** Free a page object allocated by pcache1AllocPage().
	@@ -35543,14 +36268,56 @@
35543	** with a NULL pointer, so we mark the NULL test with ALWAYS().
35544	*/
35545	static void pcache1FreePage(PgHdr1 *p){
35546	if( ALWAYS(p) ){
35547	PCache1 *pCache = p->pCache;











































35548	if( pCache->bPurgeable ){
35549	pCache->pGroup->nCurrentPage--;
35550	}
35551	pcache1Free(PGHDR1_TO_PAGE(p));
35552	}
35553	}
35554
35555	/*
35556	** Malloc function used by SQLite to obtain space from the buffer configured
	@@ -35984,13 +36751,11 @@
35984	/* Step 5. If a usable page buffer has still not been found,
35985	** attempt to allocate a new one.
35986	*/
35987	if( !pPage ){
35988	if( createFlag==1 ) sqlite3BeginBenignMalloc();
35989	pcache1LeaveMutex(pGroup);
35990	pPage = pcache1AllocPage(pCache);
35991	pcache1EnterMutex(pGroup);
35992	if( createFlag==1 ) sqlite3EndBenignMalloc();
35993	}
35994
35995	if( pPage ){
35996	unsigned int h = iKey % pCache->nHash;
	@@ -36156,10 +36921,13 @@
36156	** been released, the function returns. The return value is the total number
36157	** of bytes of memory released.
36158	*/
36159	SQLITE_PRIVATE int sqlite3PcacheReleaseMemory(int nReq){
36160	int nFree = 0;



36161	assert( sqlite3_mutex_notheld(pcache1.grp.mutex) );
36162	assert( sqlite3_mutex_notheld(pcache1.mutex) );
36163	if( pcache1.pStart==0 ){
36164	PgHdr1 *p;
36165	pcache1EnterMutex(&pcache1.grp);
	@@ -37368,10 +38136,12 @@
37368	u8 ckptSyncFlags; /* SYNC_NORMAL or SYNC_FULL for checkpoint */
37369	u8 syncFlags; /* SYNC_NORMAL or SYNC_FULL otherwise */
37370	u8 tempFile; /* zFilename is a temporary file */
37371	u8 readOnly; /* True for a read-only database */
37372	u8 memDb; /* True to inhibit all file I/O */


37373
37374	/**************************************************************************
37375	** The following block contains those class members that change during
37376	** routine opertion. Class members not in this block are either fixed
37377	** when the pager is first created or else only change when there is a
	@@ -37590,10 +38360,19 @@
37590	\|\| p->journalMode==PAGER_JOURNALMODE_MEMORY
37591	);
37592	assert( p->eState!=PAGER_ERROR && p->eState!=PAGER_OPEN );
37593	assert( pagerUseWal(p)==0 );
37594	}









37595
37596	/* If changeCountDone is set, a RESERVED lock or greater must be held
37597	** on the file.
37598	*/
37599	assert( pPager->changeCountDone==0 \|\| pPager->eLock>=RESERVED_LOCK );
	@@ -40487,10 +41266,11 @@
40487	** to the caller.
40488	*/
40489	SQLITE_PRIVATE int sqlite3PagerClose(Pager *pPager){
40490	u8 pTmp = (u8 )pPager->pTmpSpace;
40491

40492	disable_simulated_io_errors();
40493	sqlite3BeginBenignMalloc();
40494	/* pPager->errCode = 0; */
40495	pPager->exclusiveMode = 0;
40496	#ifndef SQLITE_OMIT_WAL
	@@ -40921,10 +41701,11 @@
40921	** is impossible for sqlite3PCacheFetch() to be called with createFlag==1
40922	** while in the error state, hence it is impossible for this routine to
40923	** be called in the error state. Nevertheless, we include a NEVER()
40924	** test for the error state as a safeguard against future changes.
40925	*/

40926	if( NEVER(pPager->errCode) ) return SQLITE_OK;
40927	if( pPager->doNotSpill ) return SQLITE_OK;
40928	if( pPager->doNotSyncSpill && (pPg->flags & PGHDR_NEED_SYNC)!=0 ){
40929	return SQLITE_OK;
40930	}
	@@ -41292,10 +42073,16 @@
41292	}
41293	/* pPager->xBusyHandler = 0; */
41294	/* pPager->pBusyHandlerArg = 0; */
41295	pPager->xReiniter = xReinit;
41296	/* memset(pPager->aHash, 0, sizeof(pPager->aHash)); */






41297
41298	*ppPager = pPager;
41299	return SQLITE_OK;
41300	}
41301
	@@ -42835,10 +43622,21 @@
42835	** Return true if this is an in-memory pager.
42836	*/
42837	SQLITE_PRIVATE int sqlite3PagerIsMemdb(Pager *pPager){
42838	return MEMDB;
42839	}











42840
42841	/*
42842	** Check that there are at least nSavepoint savepoints open. If there are
42843	** currently less than nSavepoints open, then open one or more savepoints
42844	** to make up the difference. If the number of savepoints is already
	@@ -45384,17 +46182,19 @@
45384	**
45385	** The EXCLUSIVE lock is not released before returning.
45386	*/
45387	rc = sqlite3OsLock(pWal->pDbFd, SQLITE_LOCK_EXCLUSIVE);
45388	if( rc==SQLITE_OK ){

45389	if( pWal->exclusiveMode==WAL_NORMAL_MODE ){
45390	pWal->exclusiveMode = WAL_EXCLUSIVE_MODE;
45391	}
45392	rc = sqlite3WalCheckpoint(
45393	pWal, SQLITE_CHECKPOINT_PASSIVE, 0, 0, sync_flags, nBuf, zBuf, 0, 0
45394	);
45395	if( rc==SQLITE_OK ){

45396	isDelete = 1;
45397	}
45398	}
45399
45400	walIndexClose(pWal, isDelete);
	@@ -49202,15 +50002,26 @@
49202	assert( (flags & BTREE_UNORDERED)==0 \|\| (flags & BTREE_SINGLE)!=0 );
49203
49204	/* A BTREE_SINGLE database is always a temporary and/or ephemeral */
49205	assert( (flags & BTREE_SINGLE)==0 \|\| isTempDb );
49206










49207	if( db->flags & SQLITE_NoReadlock ){
49208	flags \|= BTREE_NO_READLOCK;
49209	}
49210	if( isMemdb ){
49211	flags \|= BTREE_MEMORY;

49212	}
49213	if( (vfsFlags & SQLITE_OPEN_MAIN_DB)!=0 && (isMemdb \|\| isTempDb) ){
49214	vfsFlags = (vfsFlags & ~SQLITE_OPEN_MAIN_DB) \| SQLITE_OPEN_TEMP_DB;
49215	}
49216	p = sqlite3MallocZero(sizeof(Btree));
	@@ -50936,11 +51747,12 @@
50936
50937	if( NEVER(wrFlag && pBt->readOnly) ){
50938	return SQLITE_READONLY;
50939	}
50940	if( iTable==1 && btreePagecount(pBt)==0 ){
50941	return SQLITE_EMPTY;

50942	}
50943
50944	/* Now that no other errors can occur, finish filling in the BtCursor
50945	** variables and link the cursor into the BtShared list. */
50946	pCur->pgnoRoot = (Pgno)iTable;
	@@ -51690,10 +52502,13 @@
51690	int i;
51691	for(i=1; i<=pCur->iPage; i++){
51692	releasePage(pCur->apPage[i]);
51693	}
51694	pCur->iPage = 0;



51695	}else{
51696	rc = getAndInitPage(pBt, pCur->pgnoRoot, &pCur->apPage[0]);
51697	if( rc!=SQLITE_OK ){
51698	pCur->eState = CURSOR_INVALID;
51699	return rc;
	@@ -51799,11 +52614,11 @@
51799	assert( cursorHoldsMutex(pCur) );
51800	assert( sqlite3_mutex_held(pCur->pBtree->db->mutex) );
51801	rc = moveToRoot(pCur);
51802	if( rc==SQLITE_OK ){
51803	if( pCur->eState==CURSOR_INVALID ){
51804	assert( pCur->apPage[pCur->iPage]->nCell==0 );
51805	*pRes = 1;
51806	}else{
51807	assert( pCur->apPage[pCur->iPage]->nCell>0 );
51808	*pRes = 0;
51809	rc = moveToLeftmost(pCur);
	@@ -51838,11 +52653,11 @@
51838	}
51839
51840	rc = moveToRoot(pCur);
51841	if( rc==SQLITE_OK ){
51842	if( CURSOR_INVALID==pCur->eState ){
51843	assert( pCur->apPage[pCur->iPage]->nCell==0 );
51844	*pRes = 1;
51845	}else{
51846	assert( pCur->eState==CURSOR_VALID );
51847	*pRes = 0;
51848	rc = moveToRightmost(pCur);
	@@ -51911,16 +52726,16 @@
51911
51912	rc = moveToRoot(pCur);
51913	if( rc ){
51914	return rc;
51915	}
51916	assert( pCur->apPage[pCur->iPage] );
51917	assert( pCur->apPage[pCur->iPage]->isInit );
51918	assert( pCur->apPage[pCur->iPage]->nCell>0 \|\| pCur->eState==CURSOR_INVALID );
51919	if( pCur->eState==CURSOR_INVALID ){
51920	*pRes = -1;
51921	assert( pCur->apPage[pCur->iPage]->nCell==0 );
51922	return SQLITE_OK;
51923	}
51924	assert( pCur->apPage[0]->intKey \|\| pIdxKey );
51925	for(;;){
51926	int lwr, upr, idx;
	@@ -54745,13 +55560,20 @@
54745	releasePage(pPage);
54746	}
54747	return rc;
54748	}
54749	SQLITE_PRIVATE int sqlite3BtreeDropTable(Btree p, int iTable, int piMoved){

54750	int rc;
54751	sqlite3BtreeEnter(p);
54752	rc = btreeDropTable(p, iTable, piMoved);






54753	sqlite3BtreeLeave(p);
54754	return rc;
54755	}
54756
54757
	@@ -54826,10 +55648,15 @@
54826	** corruption) an SQLite error code is returned.
54827	*/
54828	SQLITE_PRIVATE int sqlite3BtreeCount(BtCursor pCur, i64 pnEntry){
54829	i64 nEntry = 0; /* Value to return in pnEntry /
54830	int rc; /* Return code */





54831	rc = moveToRoot(pCur);
54832
54833	/* Unless an error occurs, the following loop runs one iteration for each
54834	** page in the B-Tree structure (not including overflow pages).
54835	*/
	@@ -55610,11 +56437,10 @@
55610	*/
55611	SQLITE_PRIVATE int sqlite3BtreeSetVersion(Btree *pBtree, int iVersion){
55612	BtShared *pBt = pBtree->pBt;
55613	int rc; /* Return code */
55614
55615	assert( pBtree->inTrans==TRANS_NONE );
55616	assert( iVersion==1 \|\| iVersion==2 );
55617
55618	/* If setting the version fields to 1, do not automatically open the
55619	** WAL connection, even if the version fields are currently set to 2.
55620	*/
	@@ -56049,106 +56875,110 @@
56049	/* Update the schema version field in the destination database. This
56050	** is to make sure that the schema-version really does change in
56051	** the case where the source and destination databases have the
56052	** same schema version.
56053	*/
56054	if( rc==SQLITE_DONE
56055	&& (rc = sqlite3BtreeUpdateMeta(p->pDest,1,p->iDestSchema+1))==SQLITE_OK
56056	){
56057	int nDestTruncate;
56058
56059	if( p->pDestDb ){
56060	sqlite3ResetInternalSchema(p->pDestDb, -1);
56061	}
56062
56063	/* Set nDestTruncate to the final number of pages in the destination
56064	** database. The complication here is that the destination page
56065	** size may be different to the source page size.
56066	**
56067	** If the source page size is smaller than the destination page size,
56068	** round up. In this case the call to sqlite3OsTruncate() below will
56069	** fix the size of the file. However it is important to call
56070	** sqlite3PagerTruncateImage() here so that any pages in the
56071	** destination file that lie beyond the nDestTruncate page mark are
56072	** journalled by PagerCommitPhaseOne() before they are destroyed
56073	** by the file truncation.
56074	*/
56075	assert( pgszSrc==sqlite3BtreeGetPageSize(p->pSrc) );
56076	assert( pgszDest==sqlite3BtreeGetPageSize(p->pDest) );
56077	if( pgszSrc<pgszDest ){
56078	int ratio = pgszDest/pgszSrc;
56079	nDestTruncate = (nSrcPage+ratio-1)/ratio;
56080	if( nDestTruncate==(int)PENDING_BYTE_PAGE(p->pDest->pBt) ){
56081	nDestTruncate--;
56082	}
56083	}else{
56084	nDestTruncate = nSrcPage * (pgszSrc/pgszDest);
56085	}
56086	sqlite3PagerTruncateImage(pDestPager, nDestTruncate);
56087
56088	if( pgszSrc<pgszDest ){
56089	/* If the source page-size is smaller than the destination page-size,
56090	** two extra things may need to happen:
56091	**
56092	** * The destination may need to be truncated, and
56093	**
56094	** * Data stored on the pages immediately following the
56095	** pending-byte page in the source database may need to be
56096	** copied into the destination database.
56097	*/
56098	const i64 iSize = (i64)pgszSrc * (i64)nSrcPage;
56099	sqlite3_file * const pFile = sqlite3PagerFile(pDestPager);
56100	i64 iOff;
56101	i64 iEnd;
56102
56103	assert( pFile );
56104	assert( (i64)nDestTruncate*(i64)pgszDest >= iSize \|\| (
56105	nDestTruncate==(int)(PENDING_BYTE_PAGE(p->pDest->pBt)-1)
56106	&& iSize>=PENDING_BYTE && iSize<=PENDING_BYTE+pgszDest
56107	));
56108
56109	/* This call ensures that all data required to recreate the original
56110	** database has been stored in the journal for pDestPager and the
56111	** journal synced to disk. So at this point we may safely modify
56112	** the database file in any way, knowing that if a power failure
56113	** occurs, the original database will be reconstructed from the
56114	** journal file. */
56115	rc = sqlite3PagerCommitPhaseOne(pDestPager, 0, 1);
56116
56117	/* Write the extra pages and truncate the database file as required. */
56118	iEnd = MIN(PENDING_BYTE + pgszDest, iSize);
56119	for(
56120	iOff=PENDING_BYTE+pgszSrc;
56121	rc==SQLITE_OK && iOff<iEnd;
56122	iOff+=pgszSrc
56123	){
56124	PgHdr *pSrcPg = 0;
56125	const Pgno iSrcPg = (Pgno)((iOff/pgszSrc)+1);
56126	rc = sqlite3PagerGet(pSrcPager, iSrcPg, &pSrcPg);
56127	if( rc==SQLITE_OK ){
56128	u8 *zData = sqlite3PagerGetData(pSrcPg);
56129	rc = sqlite3OsWrite(pFile, zData, pgszSrc, iOff);
56130	}
56131	sqlite3PagerUnref(pSrcPg);
56132	}
56133	if( rc==SQLITE_OK ){
56134	rc = backupTruncateFile(pFile, iSize);
56135	}
56136
56137	/* Sync the database file to disk. */
56138	if( rc==SQLITE_OK ){
56139	rc = sqlite3PagerSync(pDestPager);
56140	}
56141	}else{
56142	rc = sqlite3PagerCommitPhaseOne(pDestPager, 0, 0);
56143	}
56144
56145	/* Finish committing the transaction to the destination database. */
56146	if( SQLITE_OK==rc
56147	&& SQLITE_OK==(rc = sqlite3BtreeCommitPhaseTwo(p->pDest, 0))
56148	){
56149	rc = SQLITE_DONE;




56150	}
56151	}
56152
56153	/* If bCloseTrans is true, then this function opened a read transaction
56154	** on the source database. Close the read transaction here. There is
	@@ -56612,38 +57442,32 @@
56612	** invoking an external callback, free it now. Calling this function
56613	** does not free any Mem.zMalloc buffer.
56614	*/
56615	SQLITE_PRIVATE void sqlite3VdbeMemReleaseExternal(Mem *p){
56616	assert( p->db==0 \|\| sqlite3_mutex_held(p->db->mutex) );
56617	testcase( p->flags & MEM_Agg );
56618	testcase( p->flags & MEM_Dyn );
56619	testcase( p->flags & MEM_RowSet );
56620	testcase( p->flags & MEM_Frame );
56621	if( p->flags&(MEM_Agg\|MEM_Dyn\|MEM_RowSet\|MEM_Frame) ){
56622	if( p->flags&MEM_Agg ){
56623	sqlite3VdbeMemFinalize(p, p->u.pDef);
56624	assert( (p->flags & MEM_Agg)==0 );
56625	sqlite3VdbeMemRelease(p);
56626	}else if( p->flags&MEM_Dyn && p->xDel ){
56627	assert( (p->flags&MEM_RowSet)==0 );
56628	p->xDel((void *)p->z);
56629	p->xDel = 0;
56630	}else if( p->flags&MEM_RowSet ){
56631	sqlite3RowSetClear(p->u.pRowSet);
56632	}else if( p->flags&MEM_Frame ){
56633	sqlite3VdbeMemSetNull(p);
56634	}
56635	}
56636	}
56637
56638	/*
56639	** Release any memory held by the Mem. This may leave the Mem in an
56640	** inconsistent state, for example with (Mem.z==0) and
56641	** (Mem.type==SQLITE_TEXT).
56642	*/
56643	SQLITE_PRIVATE void sqlite3VdbeMemRelease(Mem *p){
56644	sqlite3VdbeMemReleaseExternal(p);
56645	sqlite3DbFree(p->db, p->zMalloc);
56646	p->z = 0;
56647	p->zMalloc = 0;
56648	p->xDel = 0;
56649	}
	@@ -56961,11 +57785,11 @@
56961	** pFrom->z is used, then pTo->z points to the same thing as pFrom->z
56962	** and flags gets srcType (either MEM_Ephem or MEM_Static).
56963	*/
56964	SQLITE_PRIVATE void sqlite3VdbeMemShallowCopy(Mem pTo, const Mem pFrom, int srcType){
56965	assert( (pFrom->flags & MEM_RowSet)==0 );
56966	sqlite3VdbeMemReleaseExternal(pTo);
56967	memcpy(pTo, pFrom, MEMCELLSIZE);
56968	pTo->xDel = 0;
56969	if( (pFrom->flags&MEM_Static)==0 ){
56970	pTo->flags &= ~(MEM_Dyn\|MEM_Static\|MEM_Ephem);
56971	assert( srcType==MEM_Ephem \|\| srcType==MEM_Static );
	@@ -56979,11 +57803,11 @@
56979	*/
56980	SQLITE_PRIVATE int sqlite3VdbeMemCopy(Mem pTo, const Mem pFrom){
56981	int rc = SQLITE_OK;
56982
56983	assert( (pFrom->flags & MEM_RowSet)==0 );
56984	sqlite3VdbeMemReleaseExternal(pTo);
56985	memcpy(pTo, pFrom, MEMCELLSIZE);
56986	pTo->flags &= ~MEM_Dyn;
56987
56988	if( pTo->flags&(MEM_Str\|MEM_Blob) ){
56989	if( 0==(pFrom->flags&MEM_Static) ){
	@@ -57934,10 +58758,16 @@
57934	assert( p->nOp - i >= 3 );
57935	assert( pOp[-1].opcode==OP_Integer );
57936	n = pOp[-1].p1;
57937	if( n>nMaxArgs ) nMaxArgs = n;
57938	#endif






57939	}
57940
57941	if( (pOp->opflags & OPFLG_JUMP)!=0 && pOp->p2<0 ){
57942	assert( -1-pOp->p2<p->nLabel );
57943	pOp->p2 = aLabel[-1-pOp->p2];
	@@ -58025,37 +58855,34 @@
58025	** Change the value of the P1 operand for a specific instruction.
58026	** This routine is useful when a large program is loaded from a
58027	** static array using sqlite3VdbeAddOpList but we want to make a
58028	** few minor changes to the program.
58029	*/
58030	SQLITE_PRIVATE void sqlite3VdbeChangeP1(Vdbe *p, int addr, int val){
58031	assert( p!=0 );
58032	assert( addr>=0 );
58033	if( p->nOp>addr ){
58034	p->aOp[addr].p1 = val;
58035	}
58036	}
58037
58038	/*
58039	** Change the value of the P2 operand for a specific instruction.
58040	** This routine is useful for setting a jump destination.
58041	*/
58042	SQLITE_PRIVATE void sqlite3VdbeChangeP2(Vdbe *p, int addr, int val){
58043	assert( p!=0 );
58044	assert( addr>=0 );
58045	if( p->nOp>addr ){
58046	p->aOp[addr].p2 = val;
58047	}
58048	}
58049
58050	/*
58051	** Change the value of the P3 operand for a specific instruction.
58052	*/
58053	SQLITE_PRIVATE void sqlite3VdbeChangeP3(Vdbe *p, int addr, int val){
58054	assert( p!=0 );
58055	assert( addr>=0 );
58056	if( p->nOp>addr ){
58057	p->aOp[addr].p3 = val;
58058	}
58059	}
58060
58061	/*
	@@ -58442,10 +59269,14 @@
58442	break;
58443	}
58444	case P4_SUBPROGRAM: {
58445	sqlite3_snprintf(nTemp, zTemp, "program");
58446	break;




58447	}
58448	default: {
58449	zP4 = pOp->p4.z;
58450	if( zP4==0 ){
58451	zP4 = zTemp;
	@@ -59006,10 +59837,11 @@
59006	nMem = 10;
59007	}
59008	memset(zCsr, 0, zEnd-zCsr);
59009	zCsr += (zCsr - (u8*)0)&7;
59010	assert( EIGHT_BYTE_ALIGNMENT(zCsr) );

59011
59012	/* Memory for registers, parameters, cursor, etc, is allocated in two
59013	** passes. On the first pass, we try to reuse unused space at the
59014	** end of the opcode array. If we are unable to satisfy all memory
59015	** requirements by reusing the opcode array tail, then the second
	@@ -59065,10 +59897,11 @@
59065	*/
59066	SQLITE_PRIVATE void sqlite3VdbeFreeCursor(Vdbe p, VdbeCursor pCx){
59067	if( pCx==0 ){
59068	return;
59069	}

59070	if( pCx->pBt ){
59071	sqlite3BtreeClose(pCx->pBt);
59072	/* The pCx->pCursor will be close automatically, if it exists, by
59073	** the call above. */
59074	}else if( pCx->pCursor ){
	@@ -61205,10 +62038,18 @@
61205	rc = db->errCode = p->rc;
61206	}
61207	return (rc&db->errMask);
61208	}
61209








61210	/*
61211	** This is the top-level implementation of sqlite3_step(). Call
61212	** sqlite3Step() to do most of the work. If a schema error occurs,
61213	** call sqlite3Reprepare() and try again.
61214	*/
	@@ -61223,14 +62064,14 @@
61223	return SQLITE_MISUSE_BKPT;
61224	}
61225	db = v->db;
61226	sqlite3_mutex_enter(db->mutex);
61227	while( (rc = sqlite3Step(v))==SQLITE_SCHEMA
61228	&& cnt++ < 5
61229	&& (rc2 = rc = sqlite3Reprepare(v))==SQLITE_OK ){
61230	sqlite3_reset(pStmt);
61231	v->expired = 0;
61232	}
61233	if( rc2!=SQLITE_OK && ALWAYS(v->isPrepareV2) && ALWAYS(db->pErr) ){
61234	/* This case occurs after failing to recompile an sql statement.
61235	** The error message from the SQL compiler has already been loaded
61236	** into the database handle. This block copies the error message
	@@ -62357,10 +63198,17 @@
62357	** Call sqlite3VdbeMemExpandBlob() on the supplied value (type Mem*)
62358	** P if required.
62359	*/
62360	#define ExpandBlob(P) (((P)->flags&MEM_Zero)?sqlite3VdbeMemExpandBlob(P):0)
62361







62362	/*
62363	** Argument pMem points at a register that will be passed to a
62364	** user-defined function or returned to the user as the result of a query.
62365	** This routine sets the pMem->type variable used by the sqlite3_value_*()
62366	** routines.
	@@ -62951,10 +63799,11 @@
62951	u8 zEndHdr; / Pointer to first byte after the header */
62952	u32 offset; /* Offset into the data */
62953	u32 szField; /* Number of bytes in the content of a field */
62954	int szHdr; /* Size of the header size field at start of record */
62955	int avail; /* Number of bytes of available data */

62956	Mem pReg; / PseudoTable input register */
62957	} am;
62958	struct OP_Affinity_stack_vars {
62959	const char zAffinity; / The affinity to be applied */
62960	char cAff; /* A single character of affinity */
	@@ -63109,11 +63958,10 @@
63109	BtCursor *pCrsr;
63110	int res;
63111	} bl;
63112	struct OP_Next_stack_vars {
63113	VdbeCursor *pC;
63114	BtCursor *pCrsr;
63115	int res;
63116	} bm;
63117	struct OP_IdxInsert_stack_vars {
63118	VdbeCursor *pC;
63119	BtCursor *pCrsr;
	@@ -63363,11 +64211,11 @@
63363	if( pOp->opflags & OPFLG_OUT2_PRERELEASE ){
63364	assert( pOp->p2>0 );
63365	assert( pOp->p2<=p->nMem );
63366	pOut = &aMem[pOp->p2];
63367	memAboutToChange(p, pOut);
63368	sqlite3VdbeMemReleaseExternal(pOut);
63369	pOut->flags = MEM_Int;
63370	}
63371
63372	/* Sanity checking on other operands */
63373	#ifdef SQLITE_DEBUG
	@@ -63496,11 +64344,11 @@
63496	break;
63497	}
63498
63499	/* Opcode: HaltIfNull P1 P2 P3 P4 *
63500	**
63501	** Check the value in register P3. If is is NULL then Halt using
63502	** parameter P1, P2, and P4 as if this were a Halt instruction. If the
63503	** value in register P3 is not NULL, then this routine is a no-op.
63504	*/
63505	case OP_HaltIfNull: { /* in3 */
63506	pIn3 = &aMem[pOp->p3];
	@@ -63730,10 +64578,15 @@
63730	assert( memIsValid(pIn1) );
63731	memAboutToChange(p, pOut);
63732	u.ac.zMalloc = pOut->zMalloc;
63733	pOut->zMalloc = 0;
63734	sqlite3VdbeMemMove(pOut, pIn1);





63735	pIn1->zMalloc = u.ac.zMalloc;
63736	REGISTER_TRACE(u.ac.p2++, pOut);
63737	pIn1++;
63738	pOut++;
63739	}
	@@ -64433,11 +65286,11 @@
64433	** additional information.
64434	**
64435	** If SQLITE_NULLEQ is set in P5 then the result of comparison is always either
64436	** true or false and is never NULL. If both operands are NULL then the result
64437	** of comparison is false. If either operand is NULL then the result is true.
64438	** If neither operand is NULL the the result is the same as it would be if
64439	** the SQLITE_NULLEQ flag were omitted from P5.
64440	*/
64441	/* Opcode: Eq P1 P2 P3 P4 P5
64442	**
64443	** This works just like the Lt opcode except that the jump is taken if
	@@ -64445,11 +65298,11 @@
64445	** See the Lt opcode for additional information.
64446	**
64447	** If SQLITE_NULLEQ is set in P5 then the result of comparison is always either
64448	** true or false and is never NULL. If both operands are NULL then the result
64449	** of comparison is true. If either operand is NULL then the result is false.
64450	** If neither operand is NULL the the result is the same as it would be if
64451	** the SQLITE_NULLEQ flag were omitted from P5.
64452	*/
64453	/* Opcode: Le P1 P2 P3 P4 P5
64454	**
64455	** This works just like the Lt opcode except that the jump is taken if
	@@ -64730,17 +65583,17 @@
64730	break;
64731	}
64732
64733	/* Opcode: If P1 P2 P3 * *
64734	**
64735	** Jump to P2 if the value in register P1 is true. The value is
64736	** is considered true if it is numeric and non-zero. If the value
64737	** in P1 is NULL then take the jump if P3 is true.
64738	*/
64739	/* Opcode: IfNot P1 P2 P3 * *
64740	**
64741	** Jump to P2 if the value in register P1 is False. The value is
64742	** is considered true if it has a numeric value of zero. If the value
64743	** in P1 is NULL then take the jump if P3 is true.
64744	*/
64745	case OP_If: /* jump, in1 */
64746	case OP_IfNot: { /* jump, in1 */
	@@ -64828,10 +65681,11 @@
64828	u8 zEndHdr; / Pointer to first byte after the header */
64829	u32 offset; /* Offset into the data */
64830	u32 szField; /* Number of bytes in the content of a field */
64831	int szHdr; /* Size of the header size field at start of record */
64832	int avail; /* Number of bytes of available data */

64833	Mem pReg; / PseudoTable input register */
64834	#endif /* local variables moved into u.am */
64835
64836
64837	u.am.p1 = pOp->p1;
	@@ -64840,11 +65694,10 @@
64840	memset(&u.am.sMem, 0, sizeof(u.am.sMem));
64841	assert( u.am.p1<p->nCursor );
64842	assert( pOp->p3>0 && pOp->p3<=p->nMem );
64843	u.am.pDest = &aMem[pOp->p3];
64844	memAboutToChange(p, u.am.pDest);
64845	MemSetTypeFlag(u.am.pDest, MEM_Null);
64846	u.am.zRec = 0;
64847
64848	/* This block sets the variable u.am.payloadSize to be the total number of
64849	** bytes in the record.
64850	**
	@@ -64884,11 +65737,11 @@
64884	}else{
64885	assert( sqlite3BtreeCursorIsValid(u.am.pCrsr) );
64886	rc = sqlite3BtreeDataSize(u.am.pCrsr, &u.am.payloadSize);
64887	assert( rc==SQLITE_OK ); /* DataSize() cannot fail */
64888	}
64889	}else if( u.am.pC->pseudoTableReg>0 ){
64890	u.am.pReg = &aMem[u.am.pC->pseudoTableReg];
64891	assert( u.am.pReg->flags & MEM_Blob );
64892	assert( memIsValid(u.am.pReg) );
64893	u.am.payloadSize = u.am.pReg->n;
64894	u.am.zRec = u.am.pReg->z;
	@@ -64897,13 +65750,14 @@
64897	}else{
64898	/* Consider the row to be NULL */
64899	u.am.payloadSize = 0;
64900	}
64901
64902	/* If u.am.payloadSize is 0, then just store a NULL */

64903	if( u.am.payloadSize==0 ){
64904	assert( u.am.pDest->flags&MEM_Null );
64905	goto op_column_out;
64906	}
64907	assert( db->aLimit[SQLITE_LIMIT_LENGTH]>=0 );
64908	if( u.am.payloadSize > (u32)db->aLimit[SQLITE_LIMIT_LENGTH] ){
64909	goto too_big;
	@@ -65006,12 +65860,18 @@
65006	** of the record to the start of the data for the u.am.i-th column
65007	*/
65008	for(u.am.i=0; u.am.i<u.am.nField; u.am.i++){
65009	if( u.am.zIdx<u.am.zEndHdr ){
65010	u.am.aOffset[u.am.i] = u.am.offset;
65011	u.am.zIdx += getVarint32(u.am.zIdx, u.am.aType[u.am.i]);
65012	u.am.szField = sqlite3VdbeSerialTypeLen(u.am.aType[u.am.i]);






65013	u.am.offset += u.am.szField;
65014	if( u.am.offset<u.am.szField ){ /* True if u.am.offset overflows */
65015	u.am.zIdx = &u.am.zEndHdr[1]; /* Forces SQLITE_CORRUPT return below */
65016	break;
65017	}
	@@ -65048,11 +65908,11 @@
65048	** a pointer to a Mem object.
65049	*/
65050	if( u.am.aOffset[u.am.p2] ){
65051	assert( rc==SQLITE_OK );
65052	if( u.am.zRec ){
65053	sqlite3VdbeMemReleaseExternal(u.am.pDest);
65054	sqlite3VdbeSerialGet((u8 *)&u.am.zRec[u.am.aOffset[u.am.p2]], u.am.aType[u.am.p2], u.am.pDest);
65055	}else{
65056	u.am.len = sqlite3VdbeSerialTypeLen(u.am.aType[u.am.p2]);
65057	sqlite3VdbeMemMove(&u.am.sMem, u.am.pDest);
65058	rc = sqlite3VdbeMemFromBtree(u.am.pCrsr, u.am.aOffset[u.am.p2], u.am.len, u.am.pC->isIndex, &u.am.sMem);
	@@ -65065,11 +65925,11 @@
65065	u.am.pDest->enc = encoding;
65066	}else{
65067	if( pOp->p4type==P4_MEM ){
65068	sqlite3VdbeMemShallowCopy(u.am.pDest, pOp->p4.pMem, MEM_Static);
65069	}else{
65070	assert( u.am.pDest->flags&MEM_Null );
65071	}
65072	}
65073
65074	/* If we dynamically allocated space to hold the data (in the
65075	** sqlite3VdbeMemFromBtree() call above) then transfer control of that
	@@ -65267,11 +66127,11 @@
65267	i64 nEntry;
65268	BtCursor *pCrsr;
65269	#endif /* local variables moved into u.ap */
65270
65271	u.ap.pCrsr = p->apCsr[pOp->p1]->pCursor;
65272	if( u.ap.pCrsr ){
65273	rc = sqlite3BtreeCount(u.ap.pCrsr, &u.ap.nEntry);
65274	}else{
65275	u.ap.nEntry = 0;
65276	}
65277	pOut->u.i = u.ap.nEntry;
	@@ -65843,19 +66703,13 @@
65843	u.aw.pCur->nullRow = 1;
65844	u.aw.pCur->isOrdered = 1;
65845	rc = sqlite3BtreeCursor(u.aw.pX, u.aw.p2, u.aw.wrFlag, u.aw.pKeyInfo, u.aw.pCur->pCursor);
65846	u.aw.pCur->pKeyInfo = u.aw.pKeyInfo;
65847
65848	/* Since it performs no memory allocation or IO, the only values that
65849	** sqlite3BtreeCursor() may return are SQLITE_EMPTY and SQLITE_OK.
65850	** SQLITE_EMPTY is only returned when attempting to open the table
65851	** rooted at page 1 of a zero-byte database. */
65852	assert( rc==SQLITE_EMPTY \|\| rc==SQLITE_OK );
65853	if( rc==SQLITE_EMPTY ){
65854	u.aw.pCur->pCursor = 0;
65855	rc = SQLITE_OK;
65856	}
65857
65858	/* Set the VdbeCursor.isTable and isIndex variables. Previous versions of
65859	** SQLite used to check if the root-page flags were sane at this point
65860	** and report database corruption if they were not, but this check has
65861	** since moved into the btree layer. */
	@@ -65862,11 +66716,11 @@
65862	u.aw.pCur->isTable = pOp->p4type!=P4_KEYINFO;
65863	u.aw.pCur->isIndex = !u.aw.pCur->isTable;
65864	break;
65865	}
65866
65867	/* Opcode: OpenEphemeral P1 P2 * P4 *
65868	**
65869	** Open a new cursor P1 to a transient table.
65870	** The cursor is always opened read/write even if
65871	** the main database is read-only. The ephemeral
65872	** table is deleted automatically when the cursor is closed.
	@@ -65879,18 +66733,30 @@
65879	** This opcode was once called OpenTemp. But that created
65880	** confusion because the term "temp table", might refer either
65881	** to a TEMP table at the SQL level, or to a table opened by
65882	** this opcode. Then this opcode was call OpenVirtual. But
65883	** that created confusion with the whole virtual-table idea.





65884	*/
65885	/* Opcode: OpenAutoindex P1 P2 * P4 *
65886	**
65887	** This opcode works the same as OP_OpenEphemeral. It has a
65888	** different name to distinguish its use. Tables created using
65889	** by this opcode will be used for automatically created transient
65890	** indices in joins.
65891	*/







65892	case OP_OpenAutoindex:
65893	case OP_OpenEphemeral: {
65894	#if 0 /* local variables moved into u.ax */
65895	VdbeCursor *pCx;
65896	#endif /* local variables moved into u.ax */
	@@ -65900,10 +66766,11 @@
65900	SQLITE_OPEN_EXCLUSIVE \|
65901	SQLITE_OPEN_DELETEONCLOSE \|
65902	SQLITE_OPEN_TRANSIENT_DB;
65903
65904	assert( pOp->p1>=0 );

65905	u.ax.pCx = allocateCursor(p, pOp->p1, pOp->p2, -1, 1);
65906	if( u.ax.pCx==0 ) goto no_mem;
65907	u.ax.pCx->nullRow = 1;
65908	rc = sqlite3BtreeOpen(db->pVfs, 0, db, &u.ax.pCx->pBt,
65909	BTREE_OMIT_JOURNAL \| BTREE_SINGLE \| pOp->p5, vfsFlags);
	@@ -65917,11 +66784,11 @@
65917	** automatically created table with root-page 1 (an BLOB_INTKEY table).
65918	*/
65919	if( pOp->p4.pKeyInfo ){
65920	int pgno;
65921	assert( pOp->p4type==P4_KEYINFO );
65922	rc = sqlite3BtreeCreateTable(u.ax.pCx->pBt, &pgno, BTREE_BLOBKEY);
65923	if( rc==SQLITE_OK ){
65924	assert( pgno==MASTER_ROOT+1 );
65925	rc = sqlite3BtreeCursor(u.ax.pCx->pBt, pgno, 1,
65926	(KeyInfo*)pOp->p4.z, u.ax.pCx->pCursor);
65927	u.ax.pCx->pKeyInfo = pOp->p4.pKeyInfo;
	@@ -65933,10 +66800,15 @@
65933	u.ax.pCx->isTable = 1;
65934	}
65935	}
65936	u.ax.pCx->isOrdered = (pOp->p5!=BTREE_UNORDERED);
65937	u.ax.pCx->isIndex = !u.ax.pCx->isTable;





65938	break;
65939	}
65940
65941	/* Opcode: OpenPseudo P1 P2 P3 * *
65942	**
	@@ -66052,11 +66924,11 @@
66052	assert( u.az.pC->pseudoTableReg==0 );
66053	assert( OP_SeekLe == OP_SeekLt+1 );
66054	assert( OP_SeekGe == OP_SeekLt+2 );
66055	assert( OP_SeekGt == OP_SeekLt+3 );
66056	assert( u.az.pC->isOrdered );
66057	if( u.az.pC->pCursor!=0 ){
66058	u.az.oc = pOp->opcode;
66059	u.az.pC->nullRow = 0;
66060	if( u.az.pC->isTable ){
66061	/* The input value in P3 might be of any type: integer, real, string,
66062	** blob, or NULL. But it needs to be an integer before we can do
	@@ -66390,11 +67262,11 @@
66390	break;
66391	}
66392
66393	/* Opcode: NotExists P1 P2 P3 * *
66394	**
66395	** Use the content of register P3 as a integer key. If a record
66396	** with that key does not exist in table of P1, then jump to P2.
66397	** If the record does exist, then fall through. The cursor is left
66398	** pointing to the record if it exists.
66399	**
66400	** The difference between this operation and NotFound is that this
	@@ -66418,11 +67290,11 @@
66418	u.bd.pC = p->apCsr[pOp->p1];
66419	assert( u.bd.pC!=0 );
66420	assert( u.bd.pC->isTable );
66421	assert( u.bd.pC->pseudoTableReg==0 );
66422	u.bd.pCrsr = u.bd.pC->pCursor;
66423	if( u.bd.pCrsr!=0 ){
66424	u.bd.res = 0;
66425	u.bd.iKey = pIn3->u.i;
66426	rc = sqlite3BtreeMovetoUnpacked(u.bd.pCrsr, 0, u.bd.iKey, 0, &u.bd.res);
66427	u.bd.pC->lastRowid = pIn3->u.i;
66428	u.bd.pC->rowidIsValid = u.bd.res==0 ?1:0;
	@@ -66468,11 +67340,11 @@
66468	** written to register P2.
66469	**
66470	** If P3>0 then P3 is a register in the root frame of this VDBE that holds
66471	** the largest previously generated record number. No new record numbers are
66472	** allowed to be less than this value. When this value reaches its maximum,
66473	** a SQLITE_FULL error is generated. The P3 register is updated with the '
66474	** generated record number. This P3 mechanism is used to help implement the
66475	** AUTOINCREMENT feature.
66476	*/
66477	case OP_NewRowid: { /* out2-prerelease */
66478	#if 0 /* local variables moved into u.be */
	@@ -66842,10 +67714,17 @@
66842	assert( u.bh.pC->isTable \|\| pOp->opcode==OP_RowKey );
66843	assert( u.bh.pC->isIndex \|\| pOp->opcode==OP_RowData );
66844	assert( u.bh.pC!=0 );
66845	assert( u.bh.pC->nullRow==0 );
66846	assert( u.bh.pC->pseudoTableReg==0 );







66847	assert( u.bh.pC->pCursor!=0 );
66848	u.bh.pCrsr = u.bh.pC->pCursor;
66849	assert( sqlite3BtreeCursorIsValid(u.bh.pCrsr) );
66850
66851	/* The OP_RowKey and OP_RowData opcodes always follow OP_NotExists or
	@@ -66950,10 +67829,11 @@
66950	assert( pOp->p1>=0 && pOp->p1<p->nCursor );
66951	u.bj.pC = p->apCsr[pOp->p1];
66952	assert( u.bj.pC!=0 );
66953	u.bj.pC->nullRow = 1;
66954	u.bj.pC->rowidIsValid = 0;

66955	if( u.bj.pC->pCursor ){
66956	sqlite3BtreeClearCursor(u.bj.pC->pCursor);
66957	}
66958	break;
66959	}
	@@ -66975,11 +67855,11 @@
66975
66976	assert( pOp->p1>=0 && pOp->p1<p->nCursor );
66977	u.bk.pC = p->apCsr[pOp->p1];
66978	assert( u.bk.pC!=0 );
66979	u.bk.pCrsr = u.bk.pC->pCursor;
66980	if( u.bk.pCrsr==0 ){
66981	u.bk.res = 1;
66982	}else{
66983	rc = sqlite3BtreeLast(u.bk.pCrsr, &u.bk.res);
66984	}
66985	u.bk.pC->nullRow = (u8)u.bk.res;
	@@ -67030,11 +67910,15 @@
67030
67031	assert( pOp->p1>=0 && pOp->p1<p->nCursor );
67032	u.bl.pC = p->apCsr[pOp->p1];
67033	assert( u.bl.pC!=0 );
67034	u.bl.res = 1;
67035	if( (u.bl.pCrsr = u.bl.pC->pCursor)!=0 ){




67036	rc = sqlite3BtreeFirst(u.bl.pCrsr, &u.bl.res);
67037	u.bl.pC->atFirst = u.bl.res==0 ?1:0;
67038	u.bl.pC->deferredMoveto = 0;
67039	u.bl.pC->cacheStatus = CACHE_STALE;
67040	u.bl.pC->rowidIsValid = 0;
	@@ -67045,18 +67929,21 @@
67045	pc = pOp->p2 - 1;
67046	}
67047	break;
67048	}
67049
67050	/* Opcode: Next P1 P2 * * P5
67051	**
67052	** Advance cursor P1 so that it points to the next key/data pair in its
67053	** table or index. If there are no more key/value pairs then fall through
67054	** to the following instruction. But if the cursor advance was successful,
67055	** jump immediately to P2.
67056	**
67057	** The P1 cursor must be for a real table, not a pseudo-table.



67058	**
67059	** If P5 is positive and the jump is taken, then event counter
67060	** number P5-1 in the prepared statement is incremented.
67061	**
67062	** See also: Prev
	@@ -67067,19 +67954,21 @@
67067	** table or index. If there is no previous key/value pairs then fall through
67068	** to the following instruction. But if the cursor backup was successful,
67069	** jump immediately to P2.
67070	**
67071	** The P1 cursor must be for a real table, not a pseudo-table.



67072	**
67073	** If P5 is positive and the jump is taken, then event counter
67074	** number P5-1 in the prepared statement is incremented.
67075	*/
67076	case OP_Prev: /* jump */
67077	case OP_Next: { /* jump */
67078	#if 0 /* local variables moved into u.bm */
67079	VdbeCursor *pC;
67080	BtCursor *pCrsr;
67081	int res;
67082	#endif /* local variables moved into u.bm */
67083
67084	CHECK_FOR_INTERRUPT;
67085	assert( pOp->p1>=0 && pOp->p1<p->nCursor );
	@@ -67086,19 +67975,21 @@
67086	assert( pOp->p5<=ArraySize(p->aCounter) );
67087	u.bm.pC = p->apCsr[pOp->p1];
67088	if( u.bm.pC==0 ){
67089	break; /* See ticket #2273 */
67090	}
67091	u.bm.pCrsr = u.bm.pC->pCursor;
67092	if( u.bm.pCrsr==0 ){
67093	u.bm.pC->nullRow = 1;
67094	break;
67095	}
67096	u.bm.res = 1;
67097	assert( u.bm.pC->deferredMoveto==0 );
67098	rc = pOp->opcode==OP_Next ? sqlite3BtreeNext(u.bm.pCrsr, &u.bm.res) :
67099	sqlite3BtreePrevious(u.bm.pCrsr, &u.bm.res);


67100	u.bm.pC->nullRow = (u8)u.bm.res;
67101	u.bm.pC->cacheStatus = CACHE_STALE;
67102	if( u.bm.res==0 ){
67103	pc = pOp->p2 - 1;
67104	if( pOp->p5 ) p->aCounter[pOp->p5-1]++;
	@@ -67110,11 +68001,11 @@
67110	break;
67111	}
67112
67113	/* Opcode: IdxInsert P1 P2 P3 * P5
67114	**
67115	** Register P2 holds a SQL index key made using the
67116	** MakeRecord instructions. This opcode writes that key
67117	** into the index P1. Data for the entry is nil.
67118	**
67119	** P3 is a flag that provides a hint to the b-tree layer that this
67120	** insert is likely to be an append.
	@@ -67140,14 +68031,17 @@
67140	assert( u.bn.pC->isTable==0 );
67141	rc = ExpandBlob(pIn2);
67142	if( rc==SQLITE_OK ){
67143	u.bn.nKey = pIn2->n;
67144	u.bn.zKey = pIn2->z;
67145	rc = sqlite3BtreeInsert(u.bn.pCrsr, u.bn.zKey, u.bn.nKey, "", 0, 0, pOp->p3,
67146	((pOp->p5 & OPFLAG_USESEEKRESULT) ? u.bn.pC->seekResult : 0)
67147	);
67148	assert( u.bn.pC->deferredMoveto==0 );



67149	u.bn.pC->cacheStatus = CACHE_STALE;
67150	}
67151	}
67152	break;
67153	}
	@@ -69323,10 +70217,722 @@
69323	}
69324
69325	#endif /* #ifndef SQLITE_OMIT_INCRBLOB */
69326
69327	/************ End of vdbeblob.c ******************************************/








































































































































































































































































































































































































































































































































































































































































































































69328	/************ Begin file journal.c ***************************************/
69329	/*
69330	** 2007 August 22
69331	**
69332	** The author disclaims copyright to this source code. In place of
	@@ -69839,10 +71445,12 @@
69839	**
69840	*************************************************************************
69841	** This file contains routines used for walking the parser tree for
69842	** an SQL statement.
69843	*/


69844
69845
69846	/*
69847	** Walk an expression tree. Invoke the callback once for each node
69848	** of the expression, while decending. (In other words, the callback
	@@ -69977,10 +71585,12 @@
69977	**
69978	** This file contains routines used for walking the parser tree and
69979	** resolve all identifiers by associating them with a particular
69980	** table and column.
69981	*/


69982
69983	/*
69984	** Turn the pExpr expression into an alias for the iCol-th column of the
69985	** result set in pEList.
69986	**
	@@ -70956,15 +72566,29 @@
70956	/* Recursively resolve names in all subqueries
70957	*/
70958	for(i=0; i<p->pSrc->nSrc; i++){
70959	struct SrcList_item *pItem = &p->pSrc->a[i];
70960	if( pItem->pSelect ){


70961	const char *zSavedContext = pParse->zAuthContext;








70962	if( pItem->zName ) pParse->zAuthContext = pItem->zName;
70963	sqlite3ResolveSelectNames(pParse, pItem->pSelect, pOuterNC);
70964	pParse->zAuthContext = zSavedContext;
70965	if( pParse->nErr \|\| db->mallocFailed ) return WRC_Abort;




70966	}
70967	}
70968
70969	/* If there are no aggregate functions in the result-set, and no GROUP BY
70970	** expression, do not allow aggregates in any of the other expressions.
	@@ -72068,10 +73692,11 @@
72068	pNewItem->zName = sqlite3DbStrDup(db, pOldItem->zName);
72069	pNewItem->zAlias = sqlite3DbStrDup(db, pOldItem->zAlias);
72070	pNewItem->jointype = pOldItem->jointype;
72071	pNewItem->iCursor = pOldItem->iCursor;
72072	pNewItem->isPopulated = pOldItem->isPopulated;

72073	pNewItem->zIndex = sqlite3DbStrDup(db, pOldItem->zIndex);
72074	pNewItem->notIndexed = pOldItem->notIndexed;
72075	pNewItem->pIndex = pOldItem->pIndex;
72076	pTab = pNewItem->pTab = pOldItem->pTab;
72077	if( pTab ){
	@@ -78958,11 +80583,11 @@
78958	Table *p;
78959	int n;
78960	const char *z;
78961	Token sEnd;
78962	DbFixer sFix;
78963	Token *pName;
78964	int iDb;
78965	sqlite3 *db = pParse->db;
78966
78967	if( pParse->nVar>0 ){
78968	sqlite3ErrorMsg(pParse, "parameters are not allowed in views");
	@@ -79263,10 +80888,33 @@
79263	iDestroyed = iLargest;
79264	}
79265	}
79266	#endif
79267	}























79268
79269	/*
79270	** This routine is called to do the work of a DROP TABLE statement.
79271	** pName is the name of the table to be dropped.
79272	*/
	@@ -79403,18 +81051,11 @@
79403	** database.
79404	*/
79405	sqlite3NestedParse(pParse,
79406	"DELETE FROM %Q.%s WHERE tbl_name=%Q and type!='trigger'",
79407	pDb->zName, SCHEMA_TABLE(iDb), pTab->zName);
79408
79409	/* Drop any statistics from the sqlite_stat1 table, if it exists */
79410	if( sqlite3FindTable(db, "sqlite_stat1", db->aDb[iDb].zName) ){
79411	sqlite3NestedParse(pParse,
79412	"DELETE FROM %Q.sqlite_stat1 WHERE tbl=%Q", pDb->zName, pTab->zName
79413	);
79414	}
79415
79416	if( !isView && !IsVirtual(pTab) ){
79417	destroyTable(pParse, pTab);
79418	}
79419
79420	/* Remove the table entry from SQLite's internal schema and modify
	@@ -79592,18 +81233,28 @@
79592	*/
79593	static void sqlite3RefillIndex(Parse pParse, Index pIndex, int memRootPage){
79594	Table pTab = pIndex->pTable; / The table that is indexed */
79595	int iTab = pParse->nTab++; /* Btree cursor used for pTab */
79596	int iIdx = pParse->nTab++; /* Btree cursor used for pIndex */

79597	int addr1; /* Address of top of loop */
79598	int tnum; /* Root page of index */
79599	Vdbe v; / Generate code into this virtual machine */
79600	KeyInfo pKey; / KeyInfo for index */
79601	int regIdxKey; /* Registers containing the index key */
79602	int regRecord; /* Register holding assemblied index record */
79603	sqlite3 db = pParse->db; / The database connection */
79604	int iDb = sqlite3SchemaToIndex(db, pIndex->pSchema);









79605
79606	#ifndef SQLITE_OMIT_AUTHORIZATION
79607	if( sqlite3AuthCheck(pParse, SQLITE_REINDEX, pIndex->zName, 0,
79608	db->aDb[iDb].zName ) ){
79609	return;
	@@ -79625,14 +81276,33 @@
79625	sqlite3VdbeAddOp4(v, OP_OpenWrite, iIdx, tnum, iDb,
79626	(char *)pKey, P4_KEYINFO_HANDOFF);
79627	if( memRootPage>=0 ){
79628	sqlite3VdbeChangeP5(v, 1);
79629	}










79630	sqlite3OpenTable(pParse, iTab, iDb, pTab, OP_OpenRead);
79631	addr1 = sqlite3VdbeAddOp2(v, OP_Rewind, iTab, 0);
79632	regRecord = sqlite3GetTempReg(pParse);
79633	regIdxKey = sqlite3GenerateIndexKey(pParse, pIndex, iTab, regRecord, 1);









79634	if( pIndex->onError!=OE_None ){
79635	const int regRowid = regIdxKey + pIndex->nColumn;
79636	const int j2 = sqlite3VdbeCurrentAddr(v) + 2;
79637	void * const pRegKey = SQLITE_INT_TO_PTR(regIdxKey);
79638
	@@ -79647,17 +81317,19 @@
79647	*/
79648	sqlite3VdbeAddOp4(v, OP_IsUnique, iIdx, j2, regRowid, pRegKey, P4_INT32);
79649	sqlite3HaltConstraint(
79650	pParse, OE_Abort, "indexed columns are not unique", P4_STATIC);
79651	}
79652	sqlite3VdbeAddOp2(v, OP_IdxInsert, iIdx, regRecord);
79653	sqlite3VdbeChangeP5(v, OPFLAG_USESEEKRESULT);
79654	sqlite3ReleaseTempReg(pParse, regRecord);
79655	sqlite3VdbeAddOp2(v, OP_Next, iTab, addr1+1);
79656	sqlite3VdbeJumpHere(v, addr1);

79657	sqlite3VdbeAddOp1(v, OP_Close, iTab);
79658	sqlite3VdbeAddOp1(v, OP_Close, iIdx);

79659	}
79660
79661	/*
79662	** Create a new index for an SQL table. pName1.pName2 is the name of the index
79663	** and pTblList is the name of the table that is to be indexed. Both will
	@@ -80075,11 +81747,11 @@
80075	if( pStart ){
80076	assert( pEnd!=0 );
80077	/* A named index with an explicit CREATE INDEX statement */
80078	zStmt = sqlite3MPrintf(db, "CREATE%s INDEX %.*s",
80079	onError==OE_None ? "" : " UNIQUE",
80080	pEnd->z - pName->z + 1,
80081	pName->z);
80082	}else{
80083	/* An automatic index created by a PRIMARY KEY or UNIQUE constraint */
80084	/* zStmt = sqlite3MPrintf(""); */
80085	zStmt = 0;
	@@ -80233,19 +81905,13 @@
80233	v = sqlite3GetVdbe(pParse);
80234	if( v ){
80235	sqlite3BeginWriteOperation(pParse, 1, iDb);
80236	sqlite3NestedParse(pParse,
80237	"DELETE FROM %Q.%s WHERE name=%Q AND type='index'",
80238	db->aDb[iDb].zName, SCHEMA_TABLE(iDb),
80239	pIndex->zName
80240	);
80241	if( sqlite3FindTable(db, "sqlite_stat1", db->aDb[iDb].zName) ){
80242	sqlite3NestedParse(pParse,
80243	"DELETE FROM %Q.sqlite_stat1 WHERE idx=%Q",
80244	db->aDb[iDb].zName, pIndex->zName
80245	);
80246	}
80247	sqlite3ChangeCookie(pParse, iDb);
80248	destroyRootPage(pParse, pIndex->tnum, iDb);
80249	sqlite3VdbeAddOp4(v, OP_DropIndex, iDb, 0, 0, pIndex->zName, 0);
80250	}
80251
	@@ -80613,12 +82279,13 @@
80613	** The operator is "natural cross join". The A and B operands are stored
80614	** in p->a[0] and p->a[1], respectively. The parser initially stores the
80615	** operator with A. This routine shifts that operator over to B.
80616	*/
80617	SQLITE_PRIVATE void sqlite3SrcListShiftJoinType(SrcList *p){
80618	if( p && p->a ){
80619	int i;

80620	for(i=p->nSrc-1; i>0; i--){
80621	p->a[i].jointype = p->a[i-1].jointype;
80622	}
80623	p->a[0].jointype = 0;
80624	}
	@@ -81870,11 +83537,13 @@
81870	int regRowid; /* Actual register containing rowids */
81871
81872	/* Collect rowids of every row to be deleted.
81873	*/
81874	sqlite3VdbeAddOp2(v, OP_Null, 0, iRowSet);
81875	pWInfo = sqlite3WhereBegin(pParse, pTabList, pWhere,0,WHERE_DUPLICATES_OK);


81876	if( pWInfo==0 ) goto delete_from_cleanup;
81877	regRowid = sqlite3ExprCodeGetColumn(pParse, pTab, -1, iCur, iRowid);
81878	sqlite3VdbeAddOp2(v, OP_RowSetAdd, iRowSet, regRowid);
81879	if( db->flags & SQLITE_CountRows ){
81880	sqlite3VdbeAddOp2(v, OP_AddImm, memCnt, 1);
	@@ -82135,12 +83804,18 @@
82135	sqlite3VdbeAddOp3(v, OP_Column, iCur, idx, regBase+j);
82136	sqlite3ColumnDefault(v, pTab, idx, -1);
82137	}
82138	}
82139	if( doMakeRec ){






82140	sqlite3VdbeAddOp3(v, OP_MakeRecord, regBase, nCol+1, regOut);
82141	sqlite3VdbeChangeP4(v, -1, sqlite3IndexAffinityStr(v, pIdx), P4_TRANSIENT);
82142	}
82143	sqlite3ReleaseTempRange(pParse, regBase, nCol+1);
82144	return regBase;
82145	}
82146
	@@ -82162,10 +83837,12 @@
82162	**
82163	** There is only one exported symbol in this file - the function
82164	** sqliteRegisterBuildinFunctions() found at the bottom of the file.
82165	** All other code has file scope.
82166	*/


82167
82168	/*
82169	** Return the collating function associated with a function.
82170	*/
82171	static CollSeq sqlite3GetFuncCollSeq(sqlite3_context context){
	@@ -84311,11 +85988,11 @@
84311
84312	/* Create VDBE to loop through the entries in pSrc that match the WHERE
84313	** clause. If the constraint is not deferred, throw an exception for
84314	** each row found. Otherwise, for deferred constraints, increment the
84315	** deferred constraint counter by nIncr for each row selected. */
84316	pWInfo = sqlite3WhereBegin(pParse, pSrc, pWhere, 0, 0);
84317	if( nIncr>0 && pFKey->isDeferred==0 ){
84318	sqlite3ParseToplevel(pParse)->mayAbort = 1;
84319	}
84320	sqlite3VdbeAddOp2(v, OP_FkCounter, pFKey->isDeferred, nIncr);
84321	if( pWInfo ){
	@@ -84485,11 +86162,28 @@
84485	pTo = sqlite3FindTable(db, pFKey->zTo, zDb);
84486	}else{
84487	pTo = sqlite3LocateTable(pParse, 0, pFKey->zTo, zDb);
84488	}
84489	if( !pTo \|\| locateFkeyIndex(pParse, pTo, pFKey, &pIdx, &aiFree) ){

84490	if( !isIgnoreErrors \|\| db->mallocFailed ) return;
















84491	continue;
84492	}
84493	assert( pFKey->nCol==1 \|\| (aiFree && pIdx) );
84494
84495	if( aiFree ){
	@@ -87178,10 +88872,13 @@
87178	int (strnicmp)(const char,const char*,int);
87179	int (unlock_notify)(sqlite3,void()(void,int),void);
87180	int (wal_autocheckpoint)(sqlite3,int);
87181	int (wal_checkpoint)(sqlite3,const char*);
87182	void (wal_hook)(sqlite3,int()(void,sqlite3,const char,int),void);



87183	};
87184
87185	/*
87186	** The following macros redefine the API routines so that they are
87187	** redirected throught the global sqlite3_api structure.
	@@ -87378,19 +89075,23 @@
87378	#define sqlite3_strnicmp sqlite3_api->strnicmp
87379	#define sqlite3_unlock_notify sqlite3_api->unlock_notify
87380	#define sqlite3_wal_autocheckpoint sqlite3_api->wal_autocheckpoint
87381	#define sqlite3_wal_checkpoint sqlite3_api->wal_checkpoint
87382	#define sqlite3_wal_hook sqlite3_api->wal_hook



87383	#endif /* SQLITE_CORE */
87384
87385	#define SQLITE_EXTENSION_INIT1 const sqlite3_api_routines *sqlite3_api = 0;
87386	#define SQLITE_EXTENSION_INIT2(v) sqlite3_api = v;
87387
87388	#endif /* _SQLITE3EXT_H_ */
87389
87390	/************ End of sqlite3ext.h ****************************************/
87391	/************ Continuing where we left off in loadext.c ******************/

87392
87393	#ifndef SQLITE_OMIT_LOAD_EXTENSION
87394
87395	/*
87396	** Some API routines are omitted when various features are
	@@ -87452,10 +89153,12 @@
87452
87453	#ifdef SQLITE_OMIT_VIRTUALTABLE
87454	# define sqlite3_create_module 0
87455	# define sqlite3_create_module_v2 0
87456	# define sqlite3_declare_vtab 0


87457	#endif
87458
87459	#ifdef SQLITE_OMIT_SHARED_CACHE
87460	# define sqlite3_enable_shared_cache 0
87461	#endif
	@@ -87475,10 +89178,11 @@
87475	#define sqlite3_blob_bytes 0
87476	#define sqlite3_blob_close 0
87477	#define sqlite3_blob_open 0
87478	#define sqlite3_blob_read 0
87479	#define sqlite3_blob_write 0

87480	#endif
87481
87482	/*
87483	** The following structure contains pointers to all SQLite API routines.
87484	** A pointer to this structure is passed into extensions when they are
	@@ -87740,10 +89444,13 @@
87740	#else
87741	0,
87742	0,
87743	0,
87744	#endif



87745	};
87746
87747	/*
87748	** Attempt to load an SQLite extension library contained in the file
87749	** zFile. The entry point is zProc. zProc may be 0 in which case a
	@@ -94129,10 +95836,11 @@
94129	Expr pHaving; / The HAVING clause. May be NULL */
94130	int isDistinct; /* True if the DISTINCT keyword is present */
94131	int distinct; /* Table to use for the distinct set */
94132	int rc = 1; /* Value to return from this function */
94133	int addrSortIndex; /* Address of an OP_OpenEphemeral instruction */

94134	AggInfo sAggInfo; /* Information used by aggregate queries */
94135	int iEnd; /* Address of the end of the query */
94136	sqlite3 db; / The database connection */
94137
94138	#ifndef SQLITE_OMIT_EXPLAIN
	@@ -94255,20 +95963,10 @@
94255	explainSetInteger(pParse->iSelectId, iRestoreSelectId);
94256	return rc;
94257	}
94258	#endif
94259
94260	/* If possible, rewrite the query to use GROUP BY instead of DISTINCT.
94261	** GROUP BY might use an index, DISTINCT never does.
94262	*/
94263	assert( p->pGroupBy==0 \|\| (p->selFlags & SF_Aggregate)!=0 );
94264	if( (p->selFlags & (SF_Distinct\|SF_Aggregate))==SF_Distinct ){
94265	p->pGroupBy = sqlite3ExprListDup(db, p->pEList, 0);
94266	pGroupBy = p->pGroupBy;
94267	p->selFlags &= ~SF_Distinct;
94268	}
94269
94270	/* If there is both a GROUP BY and an ORDER BY clause and they are
94271	** identical, then disable the ORDER BY clause since the GROUP BY
94272	** will cause elements to come out in the correct order. This is
94273	** an optimization - the correct answer should result regardless.
94274	** Use the SQLITE_GroupByOrder flag with SQLITE_TESTCTRL_OPTIMIZER
	@@ -94276,10 +95974,34 @@
94276	*/
94277	if( sqlite3ExprListCompare(p->pGroupBy, pOrderBy)==0
94278	&& (db->flags & SQLITE_GroupByOrder)==0 ){
94279	pOrderBy = 0;
94280	}
























94281
94282	/* If there is an ORDER BY clause, then this sorting
94283	** index might end up being unused if the data can be
94284	** extracted in pre-sorted order. If that is the case, then the
94285	** OP_OpenEphemeral instruction will be changed to an OP_Noop once
	@@ -94312,26 +96034,25 @@
94312
94313	/* Open a virtual index to use for the distinct set.
94314	*/
94315	if( p->selFlags & SF_Distinct ){
94316	KeyInfo *pKeyInfo;
94317	assert( isAgg \|\| pGroupBy );
94318	distinct = pParse->nTab++;
94319	pKeyInfo = keyInfoFromExprList(pParse, p->pEList);
94320	sqlite3VdbeAddOp4(v, OP_OpenEphemeral, distinct, 0, 0,
94321	(char*)pKeyInfo, P4_KEYINFO_HANDOFF);
94322	sqlite3VdbeChangeP5(v, BTREE_UNORDERED);
94323	}else{
94324	distinct = -1;
94325	}
94326
94327	/* Aggregate and non-aggregate queries are handled differently */
94328	if( !isAgg && pGroupBy==0 ){
94329	/* This case is for non-aggregate queries
94330	** Begin the database scan
94331	*/
94332	pWInfo = sqlite3WhereBegin(pParse, pTabList, pWhere, &pOrderBy, 0);
94333	if( pWInfo==0 ) goto select_end;
94334	if( pWInfo->nRowOut < p->nSelectRow ) p->nSelectRow = pWInfo->nRowOut;
94335
94336	/* If sorting index that was created by a prior OP_OpenEphemeral
94337	** instruction ended up not being needed, then change the OP_OpenEphemeral
	@@ -94340,14 +96061,56 @@
94340	if( addrSortIndex>=0 && pOrderBy==0 ){
94341	sqlite3VdbeChangeToNoop(v, addrSortIndex, 1);
94342	p->addrOpenEphm[2] = -1;
94343	}
94344
94345	/* Use the standard inner loop
94346	*/
94347	assert(!isDistinct);
94348	selectInnerLoop(pParse, p, pEList, 0, 0, pOrderBy, -1, pDest,










































94349	pWInfo->iContinue, pWInfo->iBreak);
94350
94351	/* End the database scan loop.
94352	*/
94353	sqlite3WhereEnd(pWInfo);
	@@ -94453,11 +96216,11 @@
94453	** This might involve two separate loops with an OP_Sort in between, or
94454	** it might be a single loop that uses an index to extract information
94455	** in the right order to begin with.
94456	*/
94457	sqlite3VdbeAddOp2(v, OP_Gosub, regReset, addrReset);
94458	pWInfo = sqlite3WhereBegin(pParse, pTabList, pWhere, &pGroupBy, 0);
94459	if( pWInfo==0 ) goto select_end;
94460	if( pGroupBy==0 ){
94461	/* The optimizer is able to deliver rows in group by order so
94462	** we do not have to sort. The OP_OpenEphemeral table will be
94463	** cancelled later because we still need to use the pKeyInfo
	@@ -94715,11 +96478,11 @@
94715	/* This case runs if the aggregate has no GROUP BY clause. The
94716	** processing is much simpler since there is only a single row
94717	** of output.
94718	*/
94719	resetAccumulator(pParse, &sAggInfo);
94720	pWInfo = sqlite3WhereBegin(pParse, pTabList, pWhere, &pMinMax, flag);
94721	if( pWInfo==0 ){
94722	sqlite3ExprListDelete(db, pDel);
94723	goto select_end;
94724	}
94725	updateAccumulator(pParse, &sAggInfo);
	@@ -94894,10 +96657,12 @@
94894	** interface routine of sqlite3_exec().
94895	**
94896	** These routines are in a separate files so that they will not be linked
94897	** if they are not used.
94898	*/


94899
94900	#ifndef SQLITE_OMIT_GET_TABLE
94901
94902	/*
94903	** This structure is used to pass data from sqlite3_get_table() through
	@@ -95191,19 +96956,32 @@
95191	iDb = sqlite3TwoPartName(pParse, pName1, pName2, &pName);
95192	if( iDb<0 ){
95193	goto trigger_cleanup;
95194	}
95195	}
















95196
95197	/* If the trigger name was unqualified, and the table is a temp table,
95198	** then set iDb to 1 to create the trigger in the temporary database.
95199	** If sqlite3SrcListLookup() returns 0, indicating the table does not
95200	** exist, the error is caught by the block below.
95201	*/
95202	if( !pTableName \|\| db->mallocFailed ){
95203	goto trigger_cleanup;
95204	}
95205	pTab = sqlite3SrcListLookup(pParse, pTableName);
95206	if( db->init.busy==0 && pName2->n==0 && pTab
95207	&& pTab->pSchema==db->aDb[1].pSchema ){
95208	iDb = 1;
95209	}
	@@ -96497,11 +98275,13 @@
96497	}
96498
96499	/* Begin the database scan
96500	*/
96501	sqlite3VdbeAddOp2(v, OP_Null, 0, regOldRowid);
96502	pWInfo = sqlite3WhereBegin(pParse, pTabList, pWhere,0, WHERE_ONEPASS_DESIRED);


96503	if( pWInfo==0 ) goto update_cleanup;
96504	okOnePass = pWInfo->okOnePass;
96505
96506	/* Remember the rowid of every item to be updated.
96507	*/
	@@ -98523,10 +100303,11 @@
98523	#define WHERE_REVERSE 0x02000000 /* Scan in reverse order */
98524	#define WHERE_UNIQUE 0x04000000 /* Selects no more than one row */
98525	#define WHERE_VIRTUALTABLE 0x08000000 /* Use virtual-table processing */
98526	#define WHERE_MULTI_OR 0x10000000 /* OR using multiple indices */
98527	#define WHERE_TEMP_INDEX 0x20000000 /* Uses an ephemeral index */

98528
98529	/*
98530	** Initialize a preallocated WhereClause structure.
98531	*/
98532	static void whereClauseInit(
	@@ -99667,10 +101448,166 @@
99667	}
99668	}
99669	return 0;
99670	}
99671




























































































































































99672
99673	/*
99674	** This routine decides if pIdx can be used to satisfy the ORDER BY
99675	** clause. If it can, it returns 1. If pIdx cannot satisfy the
99676	** ORDER BY clause, this routine returns 0.
	@@ -99703,11 +101640,14 @@
99703	int sortOrder = 0; /* XOR of index and ORDER BY sort direction */
99704	int nTerm; /* Number of ORDER BY terms */
99705	struct ExprList_item pTerm; / A term of the ORDER BY clause */
99706	sqlite3 *db = pParse->db;
99707
99708	assert( pOrderBy!=0 );



99709	nTerm = pOrderBy->nExpr;
99710	assert( nTerm>0 );
99711
99712	/* Argument pIdx must either point to a 'real' named index structure,
99713	** or an index structure allocated on the stack by bestBtreeIndex() to
	@@ -100016,10 +101956,14 @@
100016	double costTempIdx; /* per-query cost of the transient index */
100017	WhereTerm pTerm; / A single term of the WHERE clause */
100018	WhereTerm pWCEnd; / End of pWC->a[] */
100019	Table pTable; / Table tht might be indexed */
100020




100021	if( (pParse->db->flags & SQLITE_AutoIndex)==0 ){
100022	/* Automatic indices are disabled at run-time */
100023	return;
100024	}
100025	if( (pCost->plan.wsFlags & WHERE_NOT_FULLSCAN)!=0 ){
	@@ -100027,10 +101971,14 @@
100027	return;
100028	}
100029	if( pSrc->notIndexed ){
100030	/* The NOT INDEXED clause appears in the SQL. */
100031	return;




100032	}
100033
100034	assert( pParse->nQueryLoop >= (double)1 );
100035	pTable = pSrc->pTab;
100036	nTableRow = pTable->nRowEst;
	@@ -100244,10 +102192,11 @@
100244	if( pTerm->leftCursor != pSrc->iCursor ) continue;
100245	assert( (pTerm->eOperator&(pTerm->eOperator-1))==0 );
100246	testcase( pTerm->eOperator==WO_IN );
100247	testcase( pTerm->eOperator==WO_ISNULL );
100248	if( pTerm->eOperator & (WO_IN\|WO_ISNULL) ) continue;

100249	nTerm++;
100250	}
100251
100252	/* If the ORDER BY clause contains only columns in the current
100253	** virtual table then allocate space for the aOrderBy part of
	@@ -100294,10 +102243,11 @@
100294	if( pTerm->leftCursor != pSrc->iCursor ) continue;
100295	assert( (pTerm->eOperator&(pTerm->eOperator-1))==0 );
100296	testcase( pTerm->eOperator==WO_IN );
100297	testcase( pTerm->eOperator==WO_ISNULL );
100298	if( pTerm->eOperator & (WO_IN\|WO_ISNULL) ) continue;

100299	pIdxCons[j].iColumn = pTerm->u.leftColumn;
100300	pIdxCons[j].iTermOffset = i;
100301	pIdxCons[j].op = (u8)pTerm->eOperator;
100302	/* The direct assignment in the previous line is possible only because
100303	** the WO_ and SQLITE_INDEX_CONSTRAINT_ codes are identical. The
	@@ -100959,10 +102909,11 @@
100959	WhereClause pWC, / The WHERE clause */
100960	struct SrcList_item pSrc, / The FROM clause term to search */
100961	Bitmask notReady, /* Mask of cursors not available for indexing */
100962	Bitmask notValid, /* Cursors not available for any purpose */
100963	ExprList pOrderBy, / The ORDER BY clause */

100964	WhereCost pCost / Lowest cost query plan */
100965	){
100966	int iCur = pSrc->iCursor; /* The cursor of the table to be accessed */
100967	Index pProbe; / An index we are evaluating */
100968	Index pIdx; / Copy of pProbe, or zero for IPK index */
	@@ -101099,11 +103050,12 @@
101099	int nEq; /* Number of == or IN terms matching index */
101100	int bInEst = 0; /* True if "x IN (SELECT...)" seen */
101101	int nInMul = 1; /* Number of distinct equalities to lookup */
101102	int estBound = 100; /* Estimated reduction in search space */
101103	int nBound = 0; /* Number of range constraints seen */
101104	int bSort = 0; /* True if external sort required */

101105	int bLookup = 0; /* True if not a covering index */
101106	WhereTerm pTerm; / A single term of the WHERE clause */
101107	#ifdef SQLITE_ENABLE_STAT2
101108	WhereTerm pFirstTerm = 0; / First term matching the index */
101109	#endif
	@@ -101163,21 +103115,24 @@
101163
101164	/* If there is an ORDER BY clause and the index being considered will
101165	** naturally scan rows in the required order, set the appropriate flags
101166	** in wsFlags. Otherwise, if there is an ORDER BY clause but the index
101167	** will scan rows in a different order, set the bSort variable. */
101168	if( pOrderBy ){
101169	if( (wsFlags & WHERE_COLUMN_IN)==0
101170	&& pProbe->bUnordered==0
101171	&& isSortingIndex(pParse, pWC->pMaskSet, pProbe, iCur, pOrderBy,
101172	nEq, wsFlags, &rev)
101173	){
101174	wsFlags \|= WHERE_ROWID_RANGE\|WHERE_COLUMN_RANGE\|WHERE_ORDERBY;
101175	wsFlags \|= (rev ? WHERE_REVERSE : 0);
101176	}else{
101177	bSort = 1;
101178	}



101179	}
101180
101181	/* If currently calculating the cost of using an index (not the IPK
101182	** index), determine if all required column data may be obtained without
101183	** using the main table (i.e. if the index is a covering
	@@ -101208,16 +103163,17 @@
101208	nRow = aiRowEst[0]/2;
101209	nInMul = (int)(nRow / aiRowEst[nEq]);
101210	}
101211
101212	#ifdef SQLITE_ENABLE_STAT2
101213	/* If the constraint is of the form x=VALUE and histogram

101214	** data is available for column x, then it might be possible
101215	** to get a better estimate on the number of rows based on
101216	** VALUE and how common that value is according to the histogram.
101217	*/
101218	if( nRow>(double)1 && nEq==1 && pFirstTerm!=0 ){
101219	if( pFirstTerm->eOperator & (WO_EQ\|WO_ISNULL) ){
101220	testcase( pFirstTerm->eOperator==WO_EQ );
101221	testcase( pFirstTerm->eOperator==WO_ISNULL );
101222	whereEqualScanEst(pParse, pProbe, pFirstTerm->pExpr->pRight, &nRow);
101223	}else if( pFirstTerm->eOperator==WO_IN && bInEst==0 ){
	@@ -101290,10 +103246,13 @@
101290	** difference and select C of 3.0.
101291	*/
101292	if( bSort ){
101293	cost += nRowestLog(nRow)3;
101294	}



101295
101296	/** Cost of using this index has now been computed **/
101297
101298	/* If there are additional constraints on this table that cannot
101299	** be used with the current index, but which might lower the number
	@@ -101435,11 +103394,11 @@
101435	}
101436	sqlite3DbFree(pParse->db, p);
101437	}else
101438	#endif
101439	{
101440	bestBtreeIndex(pParse, pWC, pSrc, notReady, notValid, pOrderBy, pCost);
101441	}
101442	}
101443
101444	/*
101445	** Disable a term in the WHERE clause. Except, do not disable the term
	@@ -102397,11 +104356,11 @@
102397	for(ii=0; ii<pOrWc->nTerm; ii++){
102398	WhereTerm *pOrTerm = &pOrWc->a[ii];
102399	if( pOrTerm->leftCursor==iCur \|\| pOrTerm->eOperator==WO_AND ){
102400	WhereInfo pSubWInfo; / Info for single OR-term scan */
102401	/* Loop through table entries that match term pOrTerm. */
102402	pSubWInfo = sqlite3WhereBegin(pParse, pOrTab, pOrTerm->pExpr, 0,
102403	WHERE_OMIT_OPEN \| WHERE_OMIT_CLOSE \|
102404	WHERE_FORCE_TABLE \| WHERE_ONETABLE_ONLY);
102405	if( pSubWInfo ){
102406	explainOneScan(
102407	pParse, pOrTab, &pSubWInfo->a[0], iLevel, pLevel->iFrom, 0
	@@ -102638,10 +104597,11 @@
102638	SQLITE_PRIVATE WhereInfo *sqlite3WhereBegin(
102639	Parse pParse, / The parser context */
102640	SrcList pTabList, / A list of all tables to be scanned */
102641	Expr pWhere, / The WHERE clause */
102642	ExprList *ppOrderBy, / An ORDER BY clause, or NULL */

102643	u16 wctrlFlags /* One of the WHERE_* flags defined in sqliteInt.h */
102644	){
102645	int i; /* Loop counter */
102646	int nByteWInfo; /* Num. bytes allocated for WhereInfo struct */
102647	int nTabList; /* Number of elements in pTabList */
	@@ -102697,10 +104657,14 @@
102697	pWInfo->iBreak = sqlite3VdbeMakeLabel(v);
102698	pWInfo->pWC = pWC = (WhereClause )&((u8 )pWInfo)[nByteWInfo];
102699	pWInfo->wctrlFlags = wctrlFlags;
102700	pWInfo->savedNQueryLoop = pParse->nQueryLoop;
102701	pMaskSet = (WhereMaskSet*)&pWC[1];




102702
102703	/* Split the WHERE clause into separate subexpressions where each
102704	** subexpression is separated by an AND operator.
102705	*/
102706	initMaskSet(pMaskSet);
	@@ -102764,10 +104728,19 @@
102764	*/
102765	exprAnalyzeAll(pTabList, pWC);
102766	if( db->mallocFailed ){
102767	goto whereBeginError;
102768	}









102769
102770	/* Chose the best index to use for each table in the FROM clause.
102771	**
102772	** This loop fills in the following fields:
102773	**
	@@ -102848,10 +104821,11 @@
102848	Bitmask mask; /* Mask of tables not yet ready */
102849	for(j=iFrom, pTabItem=&pTabList->a[j]; j<nTabList; j++, pTabItem++){
102850	int doNotReorder; /* True if this table should not be reordered */
102851	WhereCost sCost; /* Cost information from best[Virtual]Index() */
102852	ExprList pOrderBy; / ORDER BY clause for index to optimize */

102853
102854	doNotReorder = (pTabItem->jointype & (JT_LEFT\|JT_CROSS))!=0;
102855	if( j!=iFrom && doNotReorder ) break;
102856	m = getMask(pMaskSet, pTabItem->iCursor);
102857	if( (m & notReady)==0 ){
	@@ -102858,10 +104832,11 @@
102858	if( j==iFrom ) iFrom++;
102859	continue;
102860	}
102861	mask = (isOptimal ? m : notReady);
102862	pOrderBy = ((i==0 && ppOrderBy )?*ppOrderBy:0);

102863	if( pTabItem->pIndex==0 ) nUnconstrained++;
102864
102865	WHERETRACE(("=== trying table %d with isOptimal=%d ===\n",
102866	j, isOptimal));
102867	assert( pTabItem->pTab );
	@@ -102872,11 +104847,11 @@
102872	&sCost, pp);
102873	}else
102874	#endif
102875	{
102876	bestBtreeIndex(pParse, pWC, pTabItem, mask, notReady, pOrderBy,
102877	&sCost);
102878	}
102879	assert( isOptimal \|\| (sCost.used&notReady)==0 );
102880
102881	/* If an INDEXED BY clause is present, then the plan must use that
102882	** index if it uses any index at all */
	@@ -102932,10 +104907,14 @@
102932	WHERETRACE(("*** Optimizer selects table %d for loop %d"
102933	" with cost=%g and nRow=%g\n",
102934	bestJ, pLevel-pWInfo->a, bestPlan.rCost, bestPlan.plan.nRow));
102935	if( (bestPlan.plan.wsFlags & WHERE_ORDERBY)!=0 ){
102936	*ppOrderBy = 0;




102937	}
102938	andFlags &= bestPlan.plan.wsFlags;
102939	pLevel->plan = bestPlan.plan;
102940	testcase( bestPlan.plan.wsFlags & WHERE_INDEXED );
102941	testcase( bestPlan.plan.wsFlags & WHERE_TEMP_INDEX );
	@@ -103267,10 +105246,11 @@
103267	** LALR(1) grammar but which are always false in the
103268	** specific grammar used by SQLite.
103269	*/
103270	/* First off, code is included that follows the "include" declaration
103271	** in the input grammar file. */

103272
103273
103274	/*
103275	** Disable all error recovery processing in the parser push-down
103276	** automaton.
	@@ -104127,10 +106107,11 @@
104127	#endif
104128	};
104129	typedef struct yyParser yyParser;
104130
104131	#ifndef NDEBUG

104132	static FILE *yyTraceFILE = 0;
104133	static char *yyTracePrompt = 0;
104134	#endif /* NDEBUG */
104135
104136	#ifndef NDEBUG
	@@ -106702,10 +108683,11 @@
106702	**
106703	** This file contains C code that splits an SQL input string up into
106704	** individual tokens and sends those tokens one-by-one over to the
106705	** parser for analysis.
106706	*/

106707
106708	/*
106709	** The charMap() macro maps alphabetic characters into their
106710	** lower-case ASCII equivalent. On ASCII machines, this is just
106711	** an upper-to-lower case map. On EBCDIC machines we also need
	@@ -108093,10 +110075,20 @@
108093	memcpy(&y, &x, 8);
108094	assert( sqlite3IsNaN(y) );
108095	}
108096	#endif
108097	#endif










108098
108099	return rc;
108100	}
108101
108102	/*
	@@ -110763,10 +112755,21 @@
110763	if( sz ) *ppNew = sqlite3ScratchMalloc(sz);
110764	sqlite3ScratchFree(pFree);
110765	break;
110766	}
110767











110768	}
110769	va_end(ap);
110770	#endif /* SQLITE_OMIT_BUILTIN_TEST */
110771	return rc;
110772	}
	@@ -111420,16 +113423,10 @@
111420	** TODO(shess) Provide a VACUUM type operation to clear out all
111421	** deletions and duplications. This would basically be a forced merge
111422	** into a single segment.
111423	*/
111424
111425	#if !defined(SQLITE_CORE) \|\| defined(SQLITE_ENABLE_FTS3)
111426
111427	#if defined(SQLITE_ENABLE_FTS3) && !defined(SQLITE_CORE)
111428	# define SQLITE_CORE 1
111429	#endif
111430
111431	/************ Include fts3Int.h in the middle of fts3.c ******************/
111432	/************ Begin file fts3Int.h ***************************************/
111433	/*
111434	** 2009 Nov 12
111435	**
	@@ -111441,18 +113438,33 @@
111441	** May you share freely, never taking more than you give.
111442	**
111443	******************************************************************************
111444	**
111445	*/
111446
111447	#ifndef _FTSINT_H
111448	#define _FTSINT_H
111449
111450	#if !defined(NDEBUG) && !defined(SQLITE_DEBUG)
111451	# define NDEBUG 1
111452	#endif
111453
















111454	/************ Include fts3_tokenizer.h in the middle of fts3Int.h ********/
111455	/************ Begin file fts3_tokenizer.h ********************************/
111456	/*
111457	** 2006 July 10
111458	**
	@@ -111940,11 +113952,11 @@
111940	char aDoclist; / List of docids for full-text queries */
111941	int nDoclist; /* Size of buffer at aDoclist */
111942	u8 bDesc; /* True to sort in descending order */
111943	int eEvalmode; /* An FTS3_EVAL_XX constant */
111944	int nRowAvg; /* Average size of database rows, in pages */
111945	int nDoc; /* Documents in table */
111946
111947	int isMatchinfoNeeded; /* True when aMatchinfo[] needs filling in */
111948	u32 aMatchinfo; / Information about most recent match */
111949	int nMatchinfo; /* Number of elements in aMatchinfo[] */
111950	char zMatchinfo; / Matchinfo specification */
	@@ -111981,11 +113993,11 @@
111981
111982	sqlite3_int64 iDocid; /* Current docid (if pList!=0) */
111983	int bFreeList; /* True if pList should be sqlite3_free()d */
111984	char pList; / Pointer to position list following iDocid */
111985	int nList; /* Length of position list */
111986	} doclist;
111987
111988	/*
111989	** A "phrase" is a sequence of one or more tokens that must match in
111990	** sequence. A single token is the base case and the most common case.
111991	** For a sequence of tokens contained in double-quotes (i.e. "one two three")
	@@ -111997,19 +114009,19 @@
111997	int isPrefix; /* True if token ends with a "" character /
111998
111999	/* Variables above this point are populated when the expression is
112000	** parsed (by code in fts3_expr.c). Below this point the variables are
112001	** used when evaluating the expression. */
112002	int bFulltext; /* True if full-text index was used */
112003	Fts3DeferredToken pDeferred; / Deferred token object for this token */
112004	Fts3MultiSegReader pSegcsr; / Segment-reader for this token */
112005	};
112006
112007	struct Fts3Phrase {
112008	/* Cache of doclist for this phrase. */
112009	Fts3Doclist doclist;
112010	int bIncr; /* True if doclist is loaded incrementally */

112011
112012	/* Variables below this point are populated by fts3_expr.c when parsing
112013	** a MATCH expression. Everything above is part of the evaluation phase.
112014	*/
112015	int nToken; /* Number of tokens in the phrase */
	@@ -112130,10 +114142,11 @@
112130	Fts3SegFilter pFilter; / Pointer to filter object */
112131	char aBuffer; / Buffer to merge doclists in */
112132	int nBuffer; /* Allocated size of aBuffer[] in bytes */
112133
112134	int iColFilter; /* If >=0, filter for this column */

112135
112136	/* Used by fts3.c only. */
112137	int nCost; /* Cost of running iterator */
112138	int bLookup; /* True if a lookup of a single entry. */
112139
	@@ -112180,43 +114193,49 @@
112180	#endif
112181
112182	/* fts3_aux.c */
112183	SQLITE_PRIVATE int sqlite3Fts3InitAux(sqlite3 *db);
112184
112185	SQLITE_PRIVATE int sqlite3Fts3TermSegReaderCursor(
112186	Fts3Cursor pCsr, / Virtual table cursor handle */
112187	const char zTerm, / Term to query for */
112188	int nTerm, /* Size of zTerm in bytes */
112189	int isPrefix, /* True for a prefix search */
112190	Fts3MultiSegReader *ppSegcsr / OUT: Allocated seg-reader cursor */
112191	);
112192
112193	SQLITE_PRIVATE void sqlite3Fts3EvalPhraseCleanup(Fts3Phrase *);
112194
112195	SQLITE_PRIVATE int sqlite3Fts3EvalStart(Fts3Cursor , Fts3Expr , int);
112196	SQLITE_PRIVATE int sqlite3Fts3EvalNext(Fts3Cursor *pCsr);
112197
112198	SQLITE_PRIVATE int sqlite3Fts3MsrIncrStart(
112199	Fts3Table, Fts3MultiSegReader, int, const char*, int);
112200	SQLITE_PRIVATE int sqlite3Fts3MsrIncrNext(
112201	Fts3Table , Fts3MultiSegReader , sqlite3_int64 , char , int );
112202	SQLITE_PRIVATE char sqlite3Fts3EvalPhrasePoslist(Fts3Cursor , Fts3Expr *, int iCol);
112203	SQLITE_PRIVATE int sqlite3Fts3MsrOvfl(Fts3Cursor , Fts3MultiSegReader , int *);

112204
112205	SQLITE_PRIVATE int sqlite3Fts3DeferredTokenList(Fts3DeferredToken , char , int );
112206
112207
112208	#endif /* _FTSINT_H */
112209
112210	/************ End of fts3Int.h *******************************************/
112211	/************ Continuing where we left off in fts3.c *********************/

112212










112213
112214	#ifndef SQLITE_CORE
112215	SQLITE_EXTENSION_INIT1
112216	#endif
112217





112218	/*
112219	** Write a 64-bit variable-length integer to memory starting at p[0].
112220	** The length of data written will be between 1 and FTS3_VARINT_MAX bytes.
112221	** The number of bytes written is returned.
112222	*/
	@@ -112545,10 +114564,13 @@
112545	rc = sqlite3_prepare(p->db, zSql, -1, &pStmt, 0);
112546	if( rc==SQLITE_OK ){
112547	sqlite3_step(pStmt);
112548	p->nPgsz = sqlite3_column_int(pStmt, 0);
112549	rc = sqlite3_finalize(pStmt);



112550	}
112551	}
112552	assert( p->nPgsz>0 \|\| rc!=SQLITE_OK );
112553	sqlite3_free(zSql);
112554	*pRc = rc;
	@@ -112718,31 +114740,60 @@
112718	}
112719	sqlite3_free(zFree);
112720	return zRet;
112721	}
112722













112723	static int fts3GobbleInt(const char *pp, int pnOut){
112724	const char p = pp;
112725	int nInt = 0;

112726	for(p=*pp; p[0]>='0' && p[0]<='9'; p++){
112727	nInt = nInt * 10 + (p[0] - '0');
112728	}
112729	if( p==*pp ) return SQLITE_ERROR;
112730	*pnOut = nInt;
112731	*pp = p;
112732	return SQLITE_OK;
112733	}
112734
112735
















112736	static int fts3PrefixParameter(
112737	const char zParam, / ABC in prefix=ABC parameter to parse */
112738	int pnIndex, / OUT: size of apIndex[] array /
112739	struct Fts3Index *apIndex, / OUT: Array of indexes for this table */
112740	struct Fts3Index *apFree / OUT: Free this with sqlite3_free() */
112741	){
112742	struct Fts3Index *aIndex;
112743	int nIndex = 1;
112744
112745	if( zParam && zParam[0] ){
112746	const char *p;
112747	nIndex++;
112748	for(p=zParam; *p; p++){
	@@ -112749,11 +114800,11 @@
112749	if( *p==',' ) nIndex++;
112750	}
112751	}
112752
112753	aIndex = sqlite3_malloc(sizeof(struct Fts3Index) * nIndex);
112754	apIndex = apFree = aIndex;
112755	*pnIndex = nIndex;
112756	if( !aIndex ){
112757	return SQLITE_NOMEM;
112758	}
112759
	@@ -112806,12 +114857,11 @@
112806	int isFts4 = (argv[0][3]=='4'); /* True for FTS4, false for FTS3 */
112807	const char *aCol; / Array of column names */
112808	sqlite3_tokenizer pTokenizer = 0; / Tokenizer for this table */
112809
112810	int nIndex; /* Size of aIndex[] array */
112811	struct Fts3Index aIndex; / Array of indexes for this table */
112812	struct Fts3Index aFree = 0; / Free this before returning */
112813
112814	/* The results of parsing supported FTS4 key=value options: */
112815	int bNoDocsize = 0; /* True to omit %_docsize table */
112816	int bDescIdx = 0; /* True to store descending indexes */
112817	char zPrefix = 0; / Prefix parameter value (or NULL) */
	@@ -112944,11 +114994,11 @@
112944	rc = sqlite3Fts3InitTokenizer(pHash, "simple", &pTokenizer, pzErr);
112945	if( rc!=SQLITE_OK ) goto fts3_init_out;
112946	}
112947	assert( pTokenizer );
112948
112949	rc = fts3PrefixParameter(zPrefix, &nIndex, &aIndex, &aFree);
112950	if( rc==SQLITE_ERROR ){
112951	assert( zPrefix );
112952	*pzErr = sqlite3_mprintf("error parsing prefix parameter: %s", zPrefix);
112953	}
112954	if( rc!=SQLITE_OK ) goto fts3_init_out;
	@@ -112995,11 +115045,11 @@
112995	zCsr += nDb;
112996
112997	/* Fill in the azColumn array */
112998	for(iCol=0; iCol<nCol; iCol++){
112999	char *z;
113000	int n;
113001	z = (char *)sqlite3Fts3NextToken(aCol[iCol], &n);
113002	memcpy(zCsr, z, n);
113003	zCsr[n] = '\0';
113004	sqlite3Fts3Dequote(zCsr);
113005	p->azColumn[iCol] = zCsr;
	@@ -113031,11 +115081,11 @@
113031	/* Declare the table schema to SQLite. */
113032	fts3DeclareVtab(&rc, p);
113033
113034	fts3_init_out:
113035	sqlite3_free(zPrefix);
113036	sqlite3_free(aFree);
113037	sqlite3_free(zCompress);
113038	sqlite3_free(zUncompress);
113039	sqlite3_free((void *)aCol);
113040	if( rc!=SQLITE_OK ){
113041	if( p ){
	@@ -113622,12 +115672,10 @@
113622	*pp1 = p1 + 1;
113623	*pp2 = p2 + 1;
113624	}
113625
113626	/*
113627	** nToken==1 searches for adjacent positions.
113628	**
113629	** This function is used to merge two position lists into one. When it is
113630	** called, pp1 and pp2 must both point to position lists. A position-list is
113631	** the part of a doclist that follows each document id. For example, if a row
113632	** contains:
113633	**
	@@ -113643,10 +115691,12 @@
113643	** If isSaveLeft is 0, an entry is added to the output position list for
113644	** each position in *pp2 for which there exists one or more positions in
113645	** pp1 so that (pos(pp2)>pos(pp1) && pos(pp2)-pos(*pp1)<=nToken). i.e.
113646	** when the pp1 token appears before the pp2 token, but not more than nToken
113647	** slots before it.


113648	*/
113649	static int fts3PoslistPhraseMerge(
113650	char *pp, / IN/OUT: Preallocated output buffer */
113651	int nToken, /* Maximum difference in token positions */
113652	int isSaveLeft, /* Save the left position */
	@@ -113809,26 +115859,38 @@
113809
113810	return res;
113811	}
113812
113813	/*
113814	** A pointer to an instance of this structure is used as the context
113815	** argument to sqlite3Fts3SegReaderIterate()

113816	*/
113817	typedef struct TermSelect TermSelect;
113818	struct TermSelect {
113819	int isReqPos;
113820	char aaOutput[16]; / Malloc'd output buffer */
113821	int anOutput[16]; /* Size of output in bytes */
113822	};
113823
113824












113825	static void fts3GetDeltaVarint3(
113826	char **pp,
113827	char *pEnd,
113828	int bDescIdx,
113829	sqlite3_int64 *pVal
113830	){
113831	if( *pp>=pEnd ){
113832	*pp = 0;
113833	}else{
113834	sqlite3_int64 iVal;
	@@ -113839,10 +115901,25 @@
113839	*pVal += iVal;
113840	}
113841	}
113842	}
113843















113844	static void fts3PutDeltaVarint3(
113845	char *pp, / IN/OUT: Output pointer */
113846	int bDescIdx, /* True for descending docids */
113847	sqlite3_int64 piPrev, / IN/OUT: Previous value written to list */
113848	int pbFirst, / IN/OUT: True after first int written */
	@@ -113859,14 +115936,38 @@
113859	pp += sqlite3Fts3PutVarint(pp, iWrite);
113860	*piPrev = iVal;
113861	*pbFirst = 1;
113862	}
113863
113864	#define COMPARE_DOCID(i1, i2) ((bDescIdx?-1:1) * (i1-i2))
113865

























113866	static int fts3DoclistOrMerge(
113867	int bDescIdx, /* True if arguments are desc */
113868	char a1, int n1, / First doclist */
113869	char a2, int n2, / Second doclist */
113870	char *paOut, int pnOut /* OUT: Malloc'd doclist */
113871	){
113872	sqlite3_int64 i1 = 0;
	@@ -113887,35 +115988,47 @@
113887
113888	p = aOut;
113889	fts3GetDeltaVarint3(&p1, pEnd1, 0, &i1);
113890	fts3GetDeltaVarint3(&p2, pEnd2, 0, &i2);
113891	while( p1 \|\| p2 ){
113892	sqlite3_int64 iDiff = COMPARE_DOCID(i1, i2);
113893
113894	if( p2 && p1 && iDiff==0 ){
113895	fts3PutDeltaVarint3(&p, bDescIdx, &iPrev, &bFirstOut, i1);
113896	fts3PoslistMerge(&p, &p1, &p2);
113897	fts3GetDeltaVarint3(&p1, pEnd1, bDescIdx, &i1);
113898	fts3GetDeltaVarint3(&p2, pEnd2, bDescIdx, &i2);
113899	}else if( !p2 \|\| (p1 && iDiff<0) ){
113900	fts3PutDeltaVarint3(&p, bDescIdx, &iPrev, &bFirstOut, i1);
113901	fts3PoslistCopy(&p, &p1);
113902	fts3GetDeltaVarint3(&p1, pEnd1, bDescIdx, &i1);
113903	}else{
113904	fts3PutDeltaVarint3(&p, bDescIdx, &iPrev, &bFirstOut, i2);
113905	fts3PoslistCopy(&p, &p2);
113906	fts3GetDeltaVarint3(&p2, pEnd2, bDescIdx, &i2);
113907	}
113908	}
113909
113910	*paOut = aOut;
113911	*pnOut = (p-aOut);
113912	return SQLITE_OK;
113913	}
113914












113915	static void fts3DoclistPhraseMerge(
113916	int bDescIdx, /* True if arguments are desc */
113917	int nDist, /* Distance from left to right (1=adjacent) */
113918	char aLeft, int nLeft, / Left doclist */
113919	char aRight, int pnRight /* IN/OUT: Right/output doclist */
113920	){
113921	sqlite3_int64 i1 = 0;
	@@ -113934,30 +116047,30 @@
113934	p = aOut;
113935	fts3GetDeltaVarint3(&p1, pEnd1, 0, &i1);
113936	fts3GetDeltaVarint3(&p2, pEnd2, 0, &i2);
113937
113938	while( p1 && p2 ){
113939	sqlite3_int64 iDiff = COMPARE_DOCID(i1, i2);
113940	if( iDiff==0 ){
113941	char *pSave = p;
113942	sqlite3_int64 iPrevSave = iPrev;
113943	int bFirstOutSave = bFirstOut;
113944
113945	fts3PutDeltaVarint3(&p, bDescIdx, &iPrev, &bFirstOut, i1);
113946	if( 0==fts3PoslistPhraseMerge(&p, nDist, 0, 1, &p1, &p2) ){
113947	p = pSave;
113948	iPrev = iPrevSave;
113949	bFirstOut = bFirstOutSave;
113950	}
113951	fts3GetDeltaVarint3(&p1, pEnd1, bDescIdx, &i1);
113952	fts3GetDeltaVarint3(&p2, pEnd2, bDescIdx, &i2);
113953	}else if( iDiff<0 ){
113954	fts3PoslistCopy(0, &p1);
113955	fts3GetDeltaVarint3(&p1, pEnd1, bDescIdx, &i1);
113956	}else{
113957	fts3PoslistCopy(0, &p2);
113958	fts3GetDeltaVarint3(&p2, pEnd2, bDescIdx, &i2);
113959	}
113960	}
113961
113962	*pnRight = p - aOut;
113963	}
	@@ -113970,11 +116083,11 @@
113970	**
113971	** If an OOM error occurs, return SQLITE_NOMEM. In this case it is
113972	** the responsibility of the caller to free any doclists left in the
113973	** TermSelect.aaOutput[] array.
113974	*/
113975	static int fts3TermSelectMerge(Fts3Table p, TermSelect pTS){
113976	char *aOut = 0;
113977	int nOut = 0;
113978	int i;
113979
113980	/* Loop through the doclists in the aaOutput[] array. Merge them all
	@@ -114011,28 +116124,29 @@
114011	pTS->anOutput[0] = nOut;
114012	return SQLITE_OK;
114013	}
114014
114015	/*
114016	** This function is used as the sqlite3Fts3SegReaderIterate() callback when
114017	** querying the full-text index for a doclist associated with a term or
114018	** term-prefix.









114019	*/
114020	static int fts3TermSelectCb(
114021	Fts3Table p, / Virtual table object */
114022	void pContext, / Pointer to TermSelect structure */
114023	char *zTerm,
114024	int nTerm,
114025	char *aDoclist,
114026	int nDoclist
114027	){
114028	TermSelect pTS = (TermSelect )pContext;
114029
114030	UNUSED_PARAMETER(p);
114031	UNUSED_PARAMETER(zTerm);
114032	UNUSED_PARAMETER(nTerm);
114033
114034	if( pTS->aaOutput[0]==0 ){
114035	/* If this is the first term selected, copy the doclist to the output
114036	** buffer using memcpy(). */
114037	pTS->aaOutput[0] = sqlite3_malloc(nDoclist);
114038	pTS->anOutput[0] = nDoclist;
	@@ -114099,23 +116213,30 @@
114099	}
114100	pCsr->apSegment[pCsr->nSegment++] = pNew;
114101	return SQLITE_OK;
114102	}
114103







114104	static int fts3SegReaderCursor(
114105	Fts3Table p, / FTS3 table handle */
114106	int iIndex, /* Index to search (from 0 to p->nIndex-1) */
114107	int iLevel, /* Level of segments to scan */
114108	const char zTerm, / Term to query for */
114109	int nTerm, /* Size of zTerm in bytes */
114110	int isPrefix, /* True for a prefix search */
114111	int isScan, /* True to scan from zTerm to EOF */
114112	Fts3MultiSegReader pCsr / Cursor object to populate */
114113	){
114114	int rc = SQLITE_OK;
114115	int rc2;
114116	sqlite3_stmt *pStmt = 0;
114117
114118	/* If iLevel is less than 0 and this is not a scan, include a seg-reader
114119	** for the pending-terms. If this is a scan, then this call must be being
114120	** made by an fts4aux module, not an FTS table. In this case calling
114121	** Fts3SegReaderPending might segfault, as the data structures used by
	@@ -114200,28 +116321,46 @@
114200	return fts3SegReaderCursor(
114201	p, iIndex, iLevel, zTerm, nTerm, isPrefix, isScan, pCsr
114202	);
114203	}
114204






114205	static int fts3SegReaderCursorAddZero(
114206	Fts3Table *p,
114207	const char *zTerm,
114208	int nTerm,
114209	Fts3MultiSegReader *pCsr
114210	){
114211	return fts3SegReaderCursor(p, 0, FTS3_SEGCURSOR_ALL, zTerm, nTerm, 0, 0,pCsr);
114212	}
114213
114214
114215	SQLITE_PRIVATE int sqlite3Fts3TermSegReaderCursor(












114216	Fts3Cursor pCsr, / Virtual table cursor handle */
114217	const char zTerm, / Term to query for */
114218	int nTerm, /* Size of zTerm in bytes */
114219	int isPrefix, /* True for a prefix search */
114220	Fts3MultiSegReader *ppSegcsr / OUT: Allocated seg-reader cursor */
114221	){
114222	Fts3MultiSegReader pSegcsr; / Object to allocate and return */
114223	int rc = SQLITE_NOMEM; /* Return code */
114224
114225	pSegcsr = sqlite3_malloc(sizeof(Fts3MultiSegReader));
114226	if( pSegcsr ){
114227	int i;
	@@ -114261,62 +116400,53 @@
114261
114262	*ppSegcsr = pSegcsr;
114263	return rc;
114264	}
114265



114266	static void fts3SegReaderCursorFree(Fts3MultiSegReader *pSegcsr){
114267	sqlite3Fts3SegReaderFinish(pSegcsr);
114268	sqlite3_free(pSegcsr);
114269	}
114270
114271	/*
114272	** This function retreives the doclist for the specified term (or term
114273	** prefix) from the database.
114274	**
114275	** The returned doclist may be in one of two formats, depending on the
114276	** value of parameter isReqPos. If isReqPos is zero, then the doclist is
114277	** a sorted list of delta-compressed docids (a bare doclist). If isReqPos
114278	** is non-zero, then the returned list is in the same format as is stored
114279	** in the database without the found length specifier at the start of on-disk
114280	** doclists.
114281	*/
114282	static int fts3TermSelect(
114283	Fts3Table p, / Virtual table handle */
114284	Fts3PhraseToken pTok, / Token to query for */
114285	int iColumn, /* Column to query (or -ve for all columns) */
114286	int isReqPos, /* True to include position lists in output */
114287	int pnOut, / OUT: Size of buffer at ppOut /
114288	char *ppOut / OUT: Malloced result buffer */
114289	){
114290	int rc; /* Return code */
114291	Fts3MultiSegReader pSegcsr; / Seg-reader cursor for this term */
114292	TermSelect tsc; /* Context object for fts3TermSelectCb() */
114293	Fts3SegFilter filter; /* Segment term filter configuration */
114294
114295	pSegcsr = pTok->pSegcsr;
114296	memset(&tsc, 0, sizeof(TermSelect));
114297	tsc.isReqPos = isReqPos;
114298
114299	filter.flags = FTS3_SEGMENT_IGNORE_EMPTY
114300	\| (pTok->isPrefix ? FTS3_SEGMENT_PREFIX : 0)
114301	\| (isReqPos ? FTS3_SEGMENT_REQUIRE_POS : 0)
114302	\| (iColumn<p->nColumn ? FTS3_SEGMENT_COLUMN_FILTER : 0);
114303	filter.iCol = iColumn;
114304	filter.zTerm = pTok->z;
114305	filter.nTerm = pTok->n;
114306
114307	rc = sqlite3Fts3SegReaderStart(p, pSegcsr, &filter);
114308	while( SQLITE_OK==rc
114309	&& SQLITE_ROW==(rc = sqlite3Fts3SegReaderStep(p, pSegcsr))
114310	){
114311	rc = fts3TermSelectCb(p, (void *)&tsc,
114312	pSegcsr->zTerm, pSegcsr->nTerm, pSegcsr->aDoclist, pSegcsr->nDoclist
114313	);
114314	}
114315
114316	if( rc==SQLITE_OK ){
114317	rc = fts3TermSelectMerge(p, &tsc);
114318	}
114319	if( rc==SQLITE_OK ){
114320	*ppOut = tsc.aaOutput[0];
114321	*pnOut = tsc.anOutput[0];
114322	}else{
	@@ -114338,28 +116468,19 @@
114338	** If the isPoslist argument is true, then it is assumed that the doclist
114339	** contains a position-list following each docid. Otherwise, it is assumed
114340	** that the doclist is simply a list of docids stored as delta encoded
114341	** varints.
114342	*/
114343	static int fts3DoclistCountDocids(int isPoslist, char *aList, int nList){
114344	int nDoc = 0; /* Return value */
114345	if( aList ){
114346	char aEnd = &aList[nList]; / Pointer to one byte after EOF */
114347	char p = aList; / Cursor */
114348	if( !isPoslist ){
114349	/* The number of docids in the list is the same as the number of
114350	** varints. In FTS3 a varint consists of a single byte with the 0x80
114351	** bit cleared and zero or more bytes with the 0x80 bit set. So to
114352	** count the varints in the buffer, just count the number of bytes
114353	** with the 0x80 bit clear. */
114354	while( p<aEnd ) nDoc += (((*p++)&0x80)==0);
114355	}else{
114356	while( p<aEnd ){
114357	nDoc++;
114358	while( (p++)&0x80 ); / Skip docid varint */
114359	fts3PoslistCopy(0, &p); /* Skip over position list */
114360	}
114361	}
114362	}
114363
114364	return nDoc;
114365	}
	@@ -114385,11 +116506,11 @@
114385	}else{
114386	pCsr->iPrevId = sqlite3_column_int64(pCsr->pStmt, 0);
114387	rc = SQLITE_OK;
114388	}
114389	}else{
114390	rc = sqlite3Fts3EvalNext((Fts3Cursor *)pCursor);
114391	}
114392	assert( ((Fts3Table *)pCsr->base.pVtab)->pSegments==0 );
114393	return rc;
114394	}
114395
	@@ -114462,11 +116583,11 @@
114462	}
114463
114464	rc = sqlite3Fts3ReadLock(p);
114465	if( rc!=SQLITE_OK ) return rc;
114466
114467	rc = sqlite3Fts3EvalStart(pCsr, pCsr->pExpr, 1);
114468
114469	sqlite3Fts3SegmentsClose(p);
114470	if( rc!=SQLITE_OK ) return rc;
114471	pCsr->pNextId = pCsr->aDoclist;
114472	pCsr->iPrevId = 0;
	@@ -114581,12 +116702,12 @@
114581
114582	/*
114583	** Implementation of xBegin() method. This is a no-op.
114584	*/
114585	static int fts3BeginMethod(sqlite3_vtab *pVtab){
114586	UNUSED_PARAMETER(pVtab);
114587	TESTONLY( Fts3Table p = (Fts3Table)pVtab );

114588	assert( p->pSegments==0 );
114589	assert( p->nPendingData==0 );
114590	assert( p->inTransaction!=1 );
114591	TESTONLY( p->inTransaction = 1 );
114592	TESTONLY( p->mxSavepoint = -1; );
	@@ -114597,12 +116718,12 @@
114597	** Implementation of xCommit() method. This is a no-op. The contents of
114598	** the pending-terms hash-table have already been flushed into the database
114599	** by fts3SyncMethod().
114600	*/
114601	static int fts3CommitMethod(sqlite3_vtab *pVtab){
114602	UNUSED_PARAMETER(pVtab);
114603	TESTONLY( Fts3Table p = (Fts3Table)pVtab );

114604	assert( p->nPendingData==0 );
114605	assert( p->inTransaction!=0 );
114606	assert( p->pSegments==0 );
114607	TESTONLY( p->inTransaction = 0 );
114608	TESTONLY( p->mxSavepoint = -1; );
	@@ -114869,26 +116990,43 @@
114869	p->zDb, p->zName, zName
114870	);
114871	return rc;
114872	}
114873





114874	static int fts3SavepointMethod(sqlite3_vtab *pVtab, int iSavepoint){
114875	UNUSED_PARAMETER(iSavepoint);
114876	assert( ((Fts3Table *)pVtab)->inTransaction );
114877	assert( ((Fts3Table *)pVtab)->mxSavepoint < iSavepoint );
114878	TESTONLY( ((Fts3Table *)pVtab)->mxSavepoint = iSavepoint );
114879	return fts3SyncMethod(pVtab);
114880	}






114881	static int fts3ReleaseMethod(sqlite3_vtab *pVtab, int iSavepoint){
114882	TESTONLY( Fts3Table p = (Fts3Table)pVtab );
114883	UNUSED_PARAMETER(iSavepoint);
114884	UNUSED_PARAMETER(pVtab);
114885	assert( p->inTransaction );
114886	assert( p->mxSavepoint >= iSavepoint );
114887	TESTONLY( p->mxSavepoint = iSavepoint-1 );
114888	return SQLITE_OK;
114889	}






114890	static int fts3RollbackToMethod(sqlite3_vtab *pVtab, int iSavepoint){
114891	Fts3Table p = (Fts3Table)pVtab;
114892	UNUSED_PARAMETER(iSavepoint);
114893	assert( p->inTransaction );
114894	assert( p->mxSavepoint >= iSavepoint );
	@@ -115033,22 +117171,10 @@
115033	sqlite3Fts3HashClear(pHash);
115034	sqlite3_free(pHash);
115035	}
115036	return rc;
115037	}
115038
115039	#if !SQLITE_CORE
115040	SQLITE_API int sqlite3_extension_init(
115041	sqlite3 *db,
115042	char **pzErrMsg,
115043	const sqlite3_api_routines *pApi
115044	){
115045	SQLITE_EXTENSION_INIT2(pApi)
115046	return sqlite3Fts3Init(db);
115047	}
115048	#endif
115049
115050
115051	/*
115052	** Allocate an Fts3MultiSegReader for each token in the expression headed
115053	** by pExpr.
115054	**
	@@ -115062,99 +117188,151 @@
115062	** there exists prefix b-tree of the right length) then it may be traversed
115063	** and merged incrementally. Otherwise, it has to be merged into an in-memory
115064	** doclist and then traversed.
115065	*/
115066	static void fts3EvalAllocateReaders(
115067	Fts3Cursor *pCsr,
115068	Fts3Expr *pExpr,
115069	int pnToken, / OUT: Total number of tokens in phrase. */
115070	int pnOr, / OUT: Total number of OR nodes in expr. */
115071	int *pRc
115072	){
115073	if( pExpr && SQLITE_OK==*pRc ){
115074	if( pExpr->eType==FTSQUERY_PHRASE ){
115075	int i;
115076	int nToken = pExpr->pPhrase->nToken;
115077	*pnToken += nToken;
115078	for(i=0; i<nToken; i++){
115079	Fts3PhraseToken *pToken = &pExpr->pPhrase->aToken[i];
115080	int rc = sqlite3Fts3TermSegReaderCursor(pCsr,
115081	pToken->z, pToken->n, pToken->isPrefix, &pToken->pSegcsr
115082	);
115083	if( rc!=SQLITE_OK ){
115084	*pRc = rc;
115085	return;
115086	}
115087	}


115088	}else{
115089	*pnOr += (pExpr->eType==FTSQUERY_OR);
115090	fts3EvalAllocateReaders(pCsr, pExpr->pLeft, pnToken, pnOr, pRc);
115091	fts3EvalAllocateReaders(pCsr, pExpr->pRight, pnToken, pnOr, pRc);
115092	}
115093	}
115094	}
115095




































































115096	static int fts3EvalPhraseLoad(
115097	Fts3Cursor *pCsr,
115098	Fts3Phrase *p
115099	){
115100	Fts3Table pTab = (Fts3Table )pCsr->base.pVtab;
115101	int iToken;
115102	int rc = SQLITE_OK;
115103
115104	char *aDoclist = 0;
115105	int nDoclist = 0;
115106	int iPrev = -1;
115107
115108	for(iToken=0; rc==SQLITE_OK && iToken<p->nToken; iToken++){
115109	Fts3PhraseToken *pToken = &p->aToken[iToken];
115110	assert( pToken->pSegcsr \|\| pToken->pDeferred );
115111
115112	if( pToken->pDeferred==0 ){
115113	int nThis = 0;
115114	char *pThis = 0;
115115	rc = fts3TermSelect(pTab, pToken, p->iColumn, 1, &nThis, &pThis);
115116	if( rc==SQLITE_OK ){
115117	if( pThis==0 ){
115118	sqlite3_free(aDoclist);
115119	aDoclist = 0;
115120	nDoclist = 0;
115121	break;
115122	}else if( aDoclist==0 ){
115123	aDoclist = pThis;
115124	nDoclist = nThis;
115125	}else{
115126	assert( iPrev>=0 );
115127	fts3DoclistPhraseMerge(pTab->bDescIdx,
115128	iToken-iPrev, aDoclist, nDoclist, pThis, &nThis
115129	);
115130	sqlite3_free(aDoclist);
115131	aDoclist = pThis;
115132	nDoclist = nThis;
115133	}
115134	iPrev = iToken;
115135	}
115136	}
115137	}
115138
115139	if( rc==SQLITE_OK ){
115140	p->doclist.aAll = aDoclist;
115141	p->doclist.nAll = nDoclist;
115142	}else{
115143	sqlite3_free(aDoclist);
115144	}
115145	return rc;
115146	}
115147
115148	static int fts3EvalDeferredPhrase(Fts3Cursor pCsr, Fts3Phrase pPhrase){
115149	int iToken;
115150	int rc = SQLITE_OK;
115151
115152	int nMaxUndeferred = -1;
115153	char *aPoslist = 0;
115154	int nPoslist = 0;
115155	int iPrev = -1;
115156
115157	assert( pPhrase->doclist.bFreeList==0 );
115158
115159	for(iToken=0; rc==SQLITE_OK && iToken<pPhrase->nToken; iToken++){
115160	Fts3PhraseToken *pToken = &pPhrase->aToken[iToken];
	@@ -115192,16 +117370,15 @@
115192	pPhrase->doclist.nList = 0;
115193	return SQLITE_OK;
115194	}
115195	}
115196	iPrev = iToken;
115197	}else{
115198	nMaxUndeferred = iToken;
115199	}
115200	}
115201
115202	if( iPrev>=0 ){

115203	if( nMaxUndeferred<0 ){
115204	pPhrase->doclist.pList = aPoslist;
115205	pPhrase->doclist.nList = nPoslist;
115206	pPhrase->doclist.iDocid = pCsr->iPrevId;
115207	pPhrase->doclist.bFreeList = 1;
	@@ -115246,19 +117423,27 @@
115246	/*
115247	** This function is called for each Fts3Phrase in a full-text query
115248	** expression to initialize the mechanism for returning rows. Once this
115249	** function has been called successfully on an Fts3Phrase, it may be
115250	** used with fts3EvalPhraseNext() to iterate through the matching docids.






115251	*/
115252	static int fts3EvalPhraseStart(Fts3Cursor pCsr, int bOptOk, Fts3Phrase p){
115253	int rc;
115254	Fts3PhraseToken *pFirst = &p->aToken[0];
115255	Fts3Table pTab = (Fts3Table )pCsr->base.pVtab;
115256
115257	assert( p->doclist.aAll==0 );
115258	if( pCsr->bDesc==pTab->bDescIdx && bOptOk==1 && p->nToken==1
115259	&& pFirst->pSegcsr && pFirst->pSegcsr->bLookup


115260	){
115261	/* Use the incremental approach. */
115262	int iCol = (p->iColumn >= pTab->nColumn ? -1 : p->iColumn);
115263	rc = sqlite3Fts3MsrIncrStart(
115264	pTab, pFirst->pSegcsr, iCol, pFirst->z, pFirst->n);
	@@ -115274,11 +117459,17 @@
115274	return rc;
115275	}
115276
115277	/*
115278	** This function is used to iterate backwards (from the end to start)
115279	** through doclists.






115280	*/
115281	SQLITE_PRIVATE void sqlite3Fts3DoclistPrev(
115282	int bDescIdx, /* True if the doclist is desc */
115283	char aDoclist, / Pointer to entire doclist */
115284	int nDoclist, /* Length of aDoclist in bytes */
	@@ -115339,13 +117530,13 @@
115339	** If there is no "next" entry and no error occurs, then *pbEof is set to
115340	** 1 before returning. Otherwise, if no error occurs and the iterator is
115341	** successfully advanced, *pbEof is set to 0.
115342	*/
115343	static int fts3EvalPhraseNext(
115344	Fts3Cursor *pCsr,
115345	Fts3Phrase *p,
115346	u8 *pbEof
115347	){
115348	int rc = SQLITE_OK;
115349	Fts3Doclist *pDL = &p->doclist;
115350	Fts3Table pTab = (Fts3Table )pCsr->base.pVtab;
115351
	@@ -115387,30 +117578,46 @@
115387	fts3PoslistCopy(0, &pIter);
115388	pDL->nList = (pIter - pDL->pList);
115389
115390	/* pIter now points just past the 0x00 that terminates the position-
115391	** list for document pDL->iDocid. However, if this position-list was
115392	** edited in place by fts3EvalNearTrim2(), then pIter may not actually
115393	** point to the start of the next docid value. The following line deals
115394	** with this case by advancing pIter past the zero-padding added by
115395	** fts3EvalNearTrim2(). */
115396	while( pIter<pEnd && *pIter==0 ) pIter++;
115397
115398	pDL->pNextDocid = pIter;
115399	assert( *pIter \|\| pIter>=&pDL->aAll[pDL->nAll] );
115400	*pbEof = 0;
115401	}
115402	}
115403
115404	return rc;
115405	}
115406
















115407	static void fts3EvalStartReaders(
115408	Fts3Cursor *pCsr,
115409	Fts3Expr *pExpr,
115410	int bOptOk,
115411	int *pRc
115412	){
115413	if( pExpr && SQLITE_OK==*pRc ){
115414	if( pExpr->eType==FTSQUERY_PHRASE ){
115415	int i;
115416	int nToken = pExpr->pPhrase->nToken;
	@@ -115425,33 +117632,55 @@
115425	pExpr->bDeferred = (pExpr->pLeft->bDeferred && pExpr->pRight->bDeferred);
115426	}
115427	}
115428	}
115429
115430











115431	typedef struct Fts3TokenAndCost Fts3TokenAndCost;
115432	struct Fts3TokenAndCost {
115433	Fts3PhraseToken *pToken;
115434	Fts3Expr *pRoot;
115435	int nOvfl;
115436	int iCol;


115437	};
115438







115439	static void fts3EvalTokenCosts(
115440	Fts3Cursor *pCsr,
115441	Fts3Expr *pRoot,
115442	Fts3Expr *pExpr,
115443	Fts3TokenAndCost **ppTC,
115444	Fts3Expr ***ppOr,
115445	int *pRc
115446	){
115447	if( *pRc==SQLITE_OK && pExpr ){
115448	if( pExpr->eType==FTSQUERY_PHRASE ){
115449	Fts3Phrase *pPhrase = pExpr->pPhrase;
115450	int i;
115451	for(i=0; *pRc==SQLITE_OK && i<pPhrase->nToken; i++){
115452	Fts3TokenAndCost pTC = (ppTC)++;


115453	pTC->pRoot = pRoot;
115454	pTC->pToken = &pPhrase->aToken[i];
115455	pTC->iCol = pPhrase->iColumn;
115456	*pRc = sqlite3Fts3MsrOvfl(pCsr, pTC->pToken->pSegcsr, &pTC->nOvfl);
115457	}
	@@ -115470,23 +117699,34 @@
115470	fts3EvalTokenCosts(pCsr, pRoot, pExpr->pRight, ppTC, ppOr, pRc);
115471	}
115472	}
115473	}
115474











115475	static int fts3EvalAverageDocsize(Fts3Cursor pCsr, int pnPage){
115476	if( pCsr->nRowAvg==0 ){
115477	/* The average document size, which is required to calculate the cost
115478	** of each doclist, has not yet been determined. Read the required
115479	** data from the %_stat table to calculate it.
115480	**
115481	** Entry 0 of the %_stat table is a blob containing (nCol+1) FTS3
115482	** varints, where nCol is the number of columns in the FTS3 table.
115483	** The first varint is the number of documents currently stored in
115484	** the table. The following nCol varints contain the total amount of
115485	** data stored in all rows of each column of the table, from left
115486	** to right.
115487	*/
115488	int rc;
115489	Fts3Table p = (Fts3Table)pCsr->base.pVtab;
115490	sqlite3_stmt *pStmt;
115491	sqlite3_int64 nDoc = 0;
115492	sqlite3_int64 nByte = 0;
	@@ -115517,113 +117757,151 @@
115517
115518	*pnPage = pCsr->nRowAvg;
115519	return SQLITE_OK;
115520	}
115521














115522	static int fts3EvalSelectDeferred(
115523	Fts3Cursor *pCsr,
115524	Fts3Expr *pRoot,
115525	Fts3TokenAndCost *aTC,
115526	int nTC
115527	){
115528	int nDocSize = 0;
115529	int nDocEst = 0;
115530	int rc = SQLITE_OK;
115531	Fts3Table pTab = (Fts3Table )pCsr->base.pVtab;
115532	int ii;




115533
115534	int nOvfl = 0;
115535	int nTerm = 0;
115536



115537	for(ii=0; ii<nTC; ii++){
115538	if( aTC[ii].pRoot==pRoot ){
115539	nOvfl += aTC[ii].nOvfl;
115540	nTerm++;
115541	}
115542	}
115543	if( nOvfl==0 \|\| nTerm<2 ) return SQLITE_OK;
115544

115545	rc = fts3EvalAverageDocsize(pCsr, &nDocSize);
115546
115547	for(ii=0; ii<nTerm && rc==SQLITE_OK; ii++){
115548	int jj;
115549	Fts3TokenAndCost *pTC = 0;
115550
115551	for(jj=0; jj<nTC; jj++){
115552	if( aTC[jj].pToken && aTC[jj].pRoot==pRoot
115553	&& (!pTC \|\| aTC[jj].nOvfl<pTC->nOvfl)

























115554	){
115555	pTC = &aTC[jj];
115556	}
115557	}
115558	assert( pTC );
115559
115560	/* At this point pTC points to the cheapest remaining token. */
115561	if( ii==0 ){
115562	if( pTC->nOvfl ){
115563	nDocEst = (pTC->nOvfl * pTab->nPgsz + pTab->nPgsz) / 10;
115564	}else{
115565	/* TODO: Fix this so that the doclist need not be read twice. */









115566	Fts3PhraseToken *pToken = pTC->pToken;
115567	int nList = 0;
115568	char *pList = 0;
115569	rc = fts3TermSelect(pTab, pToken, pTC->iCol, 1, &nList, &pList);
115570	if( rc==SQLITE_OK ){
115571	nDocEst = fts3DoclistCountDocids(1, pList, nList);
115572	}
115573	sqlite3_free(pList);
115574	if( rc==SQLITE_OK ){
115575	rc = sqlite3Fts3TermSegReaderCursor(pCsr,
115576	pToken->z, pToken->n, pToken->isPrefix, &pToken->pSegcsr
115577	);

115578	}
115579	}
115580	}else{
115581	if( pTC->nOvfl>=(nDocEst*nDocSize) ){
115582	Fts3PhraseToken *pToken = pTC->pToken;
115583	rc = sqlite3Fts3DeferToken(pCsr, pToken, pTC->iCol);
115584	fts3SegReaderCursorFree(pToken->pSegcsr);
115585	pToken->pSegcsr = 0;
115586	}
115587	nDocEst = 1 + (nDocEst/4);
115588	}
115589	pTC->pToken = 0;
115590	}
115591
115592	return rc;
115593	}
115594
115595	SQLITE_PRIVATE int sqlite3Fts3EvalStart(Fts3Cursor pCsr, Fts3Expr pExpr, int bOptOk){












115596	Fts3Table pTab = (Fts3Table )pCsr->base.pVtab;
115597	int rc = SQLITE_OK;
115598	int nToken = 0;
115599	int nOr = 0;
115600
115601	/* Allocate a MultiSegReader for each token in the expression. */
115602	fts3EvalAllocateReaders(pCsr, pExpr, &nToken, &nOr, &rc);
115603
115604	/* Call fts3EvalPhraseStart() on all phrases in the expression. TODO:
115605	** This call will eventually also be responsible for determining which
115606	** tokens are 'deferred' until the document text is loaded into memory.
115607	**
115608	** Each token in each phrase is dealt with using one of the following
115609	** three strategies:
115610	**
115611	** 1. Entire doclist loaded into memory as part of the
115612	** fts3EvalStartReaders() call.
115613	**
115614	** 2. Doclist loaded into memory incrementally, as part of each
115615	** sqlite3Fts3EvalNext() call.
115616	**
115617	** 3. Token doclist is never loaded. Instead, documents are loaded into
115618	** memory and scanned for the token as part of the sqlite3Fts3EvalNext()
115619	** call. This is known as a "deferred" token.
115620	*/
115621
115622	/* If bOptOk is true, check if there are any tokens that should be deferred.
115623	*/
115624	if( rc==SQLITE_OK && bOptOk && nToken>1 && pTab->bHasStat ){
115625	Fts3TokenAndCost *aTC;
115626	Fts3Expr **apOr;
115627	aTC = (Fts3TokenAndCost *)sqlite3_malloc(
115628	sizeof(Fts3TokenAndCost) * nToken
115629	+ sizeof(Fts3Expr ) nOr * 2
	@@ -115635,11 +117913,11 @@
115635	}else{
115636	int ii;
115637	Fts3TokenAndCost *pTC = aTC;
115638	Fts3Expr **ppOr = apOr;
115639
115640	fts3EvalTokenCosts(pCsr, 0, pExpr, &pTC, &ppOr, &rc);
115641	nToken = pTC-aTC;
115642	nOr = ppOr-apOr;
115643
115644	if( rc==SQLITE_OK ){
115645	rc = fts3EvalSelectDeferred(pCsr, 0, aTC, nToken);
	@@ -115650,25 +117928,50 @@
115650
115651	sqlite3_free(aTC);
115652	}
115653	}
115654
115655	fts3EvalStartReaders(pCsr, pExpr, bOptOk, &rc);
115656	return rc;
115657	}
115658
115659	static void fts3EvalZeroPoslist(Fts3Phrase *pPhrase){



115660	if( pPhrase->doclist.bFreeList ){
115661	sqlite3_free(pPhrase->doclist.pList);
115662	}
115663	pPhrase->doclist.pList = 0;
115664	pPhrase->doclist.nList = 0;
115665	pPhrase->doclist.bFreeList = 0;
115666	}
115667
115668	static int fts3EvalNearTrim2(
115669	int nNear,






















115670	char aTmp, / Temporary space to use */
115671	char *paPoslist, / IN/OUT: Position list */
115672	int pnToken, / IN/OUT: Tokens in phrase of paPoslist /
115673	Fts3Phrase pPhrase / The phrase object to trim the doclist of */
115674	){
	@@ -115696,10 +117999,176 @@
115696	}
115697
115698	return res;
115699	}
115700






































































































































































115701	static int fts3EvalNearTest(Fts3Expr pExpr, int pRc){
115702	int res = 1;
115703
115704	/* The following block runs if pExpr is the root of a NEAR query.
115705	** For example, the query:
	@@ -115717,11 +118186,11 @@
115717	** \| \|
115718	** "w" "x"
115719	**
115720	** The right-hand child of a NEAR node is always a phrase. The
115721	** left-hand child may be either a phrase or a NEAR node. There are
115722	** no exceptions to this.
115723	*/
115724	if( *pRc==SQLITE_OK
115725	&& pExpr->eType==FTSQUERY_NEAR
115726	&& pExpr->bEof==0
115727	&& (pExpr->pParent==0 \|\| pExpr->pParent->eType!=FTSQUERY_NEAR)
	@@ -115744,21 +118213,21 @@
115744	int nToken = p->pPhrase->nToken;
115745
115746	for(p=p->pParent;res && p && p->eType==FTSQUERY_NEAR; p=p->pParent){
115747	Fts3Phrase *pPhrase = p->pRight->pPhrase;
115748	int nNear = p->nNear;
115749	res = fts3EvalNearTrim2(nNear, aTmp, &aPoslist, &nToken, pPhrase);
115750	}
115751
115752	aPoslist = pExpr->pRight->pPhrase->doclist.pList;
115753	nToken = pExpr->pRight->pPhrase->nToken;
115754	for(p=pExpr->pLeft; p && res; p=p->pLeft){
115755	int nNear = p->pParent->nNear;
115756	Fts3Phrase *pPhrase = (
115757	p->eType==FTSQUERY_NEAR ? p->pRight->pPhrase : p->pPhrase
115758	);
115759	res = fts3EvalNearTrim2(nNear, aTmp, &aPoslist, &nToken, pPhrase);
115760	}
115761	}
115762
115763	sqlite3_free(aTmp);
115764	}
	@@ -115765,132 +118234,33 @@
115765
115766	return res;
115767	}
115768
115769	/*
115770	** This macro is used by the fts3EvalNext() function. The two arguments are
115771	** 64-bit docid values. If the current query is "ORDER BY docid ASC", then
115772	** the macro returns (i1 - i2). Or if it is "ORDER BY docid DESC", then
115773	** it returns (i2 - i1). This allows the same code to be used for merging
115774	** doclists in ascending or descending order.




115775	*/
115776	#define DOCID_CMP(i1, i2) ((pCsr->bDesc?-1:1) * (i1-i2))
115777
115778	static void fts3EvalNext(
115779	Fts3Cursor *pCsr,
115780	Fts3Expr *pExpr,
115781	int *pRc
115782	){
115783	if( *pRc==SQLITE_OK ){
115784	assert( pExpr->bEof==0 );
115785	pExpr->bStart = 1;
115786
115787	switch( pExpr->eType ){
115788	case FTSQUERY_NEAR:
115789	case FTSQUERY_AND: {
115790	Fts3Expr *pLeft = pExpr->pLeft;
115791	Fts3Expr *pRight = pExpr->pRight;
115792	assert( !pLeft->bDeferred \|\| !pRight->bDeferred );
115793	if( pLeft->bDeferred ){
115794	fts3EvalNext(pCsr, pRight, pRc);
115795	pExpr->iDocid = pRight->iDocid;
115796	pExpr->bEof = pRight->bEof;
115797	}else if( pRight->bDeferred ){
115798	fts3EvalNext(pCsr, pLeft, pRc);
115799	pExpr->iDocid = pLeft->iDocid;
115800	pExpr->bEof = pLeft->bEof;
115801	}else{
115802	fts3EvalNext(pCsr, pLeft, pRc);
115803	fts3EvalNext(pCsr, pRight, pRc);
115804
115805	while( !pLeft->bEof && !pRight->bEof && *pRc==SQLITE_OK ){
115806	sqlite3_int64 iDiff = DOCID_CMP(pLeft->iDocid, pRight->iDocid);
115807	if( iDiff==0 ) break;
115808	if( iDiff<0 ){
115809	fts3EvalNext(pCsr, pLeft, pRc);
115810	}else{
115811	fts3EvalNext(pCsr, pRight, pRc);
115812	}
115813	}
115814
115815	pExpr->iDocid = pLeft->iDocid;
115816	pExpr->bEof = (pLeft->bEof \|\| pRight->bEof);
115817	}
115818	break;
115819	}
115820
115821	case FTSQUERY_OR: {
115822	Fts3Expr *pLeft = pExpr->pLeft;
115823	Fts3Expr *pRight = pExpr->pRight;
115824	sqlite3_int64 iCmp = DOCID_CMP(pLeft->iDocid, pRight->iDocid);
115825
115826	assert( pLeft->bStart \|\| pLeft->iDocid==pRight->iDocid );
115827	assert( pRight->bStart \|\| pLeft->iDocid==pRight->iDocid );
115828
115829	if( pRight->bEof \|\| (pLeft->bEof==0 && iCmp<0) ){
115830	fts3EvalNext(pCsr, pLeft, pRc);
115831	}else if( pLeft->bEof \|\| (pRight->bEof==0 && iCmp>0) ){
115832	fts3EvalNext(pCsr, pRight, pRc);
115833	}else{
115834	fts3EvalNext(pCsr, pLeft, pRc);
115835	fts3EvalNext(pCsr, pRight, pRc);
115836	}
115837
115838	pExpr->bEof = (pLeft->bEof && pRight->bEof);
115839	iCmp = DOCID_CMP(pLeft->iDocid, pRight->iDocid);
115840	if( pRight->bEof \|\| (pLeft->bEof==0 && iCmp<0) ){
115841	pExpr->iDocid = pLeft->iDocid;
115842	}else{
115843	pExpr->iDocid = pRight->iDocid;
115844	}
115845
115846	break;
115847	}
115848
115849	case FTSQUERY_NOT: {
115850	Fts3Expr *pLeft = pExpr->pLeft;
115851	Fts3Expr *pRight = pExpr->pRight;
115852
115853	if( pRight->bStart==0 ){
115854	fts3EvalNext(pCsr, pRight, pRc);
115855	assert( *pRc!=SQLITE_OK \|\| pRight->bStart );
115856	}
115857
115858	fts3EvalNext(pCsr, pLeft, pRc);
115859	if( pLeft->bEof==0 ){
115860	while( !*pRc
115861	&& !pRight->bEof
115862	&& DOCID_CMP(pLeft->iDocid, pRight->iDocid)>0
115863	){
115864	fts3EvalNext(pCsr, pRight, pRc);
115865	}
115866	}
115867	pExpr->iDocid = pLeft->iDocid;
115868	pExpr->bEof = pLeft->bEof;
115869	break;
115870	}
115871
115872	default: {
115873	Fts3Phrase *pPhrase = pExpr->pPhrase;
115874	fts3EvalZeroPoslist(pPhrase);
115875	*pRc = fts3EvalPhraseNext(pCsr, pPhrase, &pExpr->bEof);
115876	pExpr->iDocid = pPhrase->doclist.iDocid;
115877	break;
115878	}
115879	}
115880	}
115881	}
115882
115883	static int fts3EvalDeferredTest(Fts3Cursor pCsr, Fts3Expr pExpr, int *pRc){
115884	int bHit = 1;
115885	if( *pRc==SQLITE_OK ){
115886	switch( pExpr->eType ){
115887	case FTSQUERY_NEAR:
115888	case FTSQUERY_AND:
115889	bHit = (
115890	fts3EvalDeferredTest(pCsr, pExpr->pLeft, pRc)
115891	&& fts3EvalDeferredTest(pCsr, pExpr->pRight, pRc)
115892	&& fts3EvalNearTest(pExpr, pRc)
115893	);
115894
115895	/* If the NEAR expression does not match any rows, zero the doclist for
115896	** all phrases involved in the NEAR. This is because the snippet(),
	@@ -115912,31 +118282,31 @@
115912	&& (pExpr->pParent==0 \|\| pExpr->pParent->eType!=FTSQUERY_NEAR)
115913	){
115914	Fts3Expr *p;
115915	for(p=pExpr; p->pPhrase==0; p=p->pLeft){
115916	if( p->pRight->iDocid==pCsr->iPrevId ){
115917	fts3EvalZeroPoslist(p->pRight->pPhrase);
115918	}
115919	}
115920	if( p->iDocid==pCsr->iPrevId ){
115921	fts3EvalZeroPoslist(p->pPhrase);
115922	}
115923	}
115924
115925	break;
115926
115927	case FTSQUERY_OR: {
115928	int bHit1 = fts3EvalDeferredTest(pCsr, pExpr->pLeft, pRc);
115929	int bHit2 = fts3EvalDeferredTest(pCsr, pExpr->pRight, pRc);
115930	bHit = bHit1 \|\| bHit2;
115931	break;
115932	}
115933
115934	case FTSQUERY_NOT:
115935	bHit = (
115936	fts3EvalDeferredTest(pCsr, pExpr->pLeft, pRc)
115937	&& !fts3EvalDeferredTest(pCsr, pExpr->pRight, pRc)
115938	);
115939	break;
115940
115941	default: {
115942	if( pCsr->pDeferred
	@@ -115943,11 +118313,11 @@
115943	&& (pExpr->iDocid==pCsr->iPrevId \|\| pExpr->bDeferred)
115944	){
115945	Fts3Phrase *pPhrase = pExpr->pPhrase;
115946	assert( pExpr->bDeferred \|\| pPhrase->doclist.bFreeList==0 );
115947	if( pExpr->bDeferred ){
115948	fts3EvalZeroPoslist(pPhrase);
115949	}
115950	*pRc = fts3EvalDeferredPhrase(pCsr, pPhrase);
115951	bHit = (pPhrase->doclist.pList!=0);
115952	pExpr->iDocid = pCsr->iPrevId;
115953	}else{
	@@ -115959,31 +118329,53 @@
115959	}
115960	return bHit;
115961	}
115962
115963	/*
115964	** Return 1 if both of the following are true:













115965	**
115966	** 1. *pRc is SQLITE_OK when this function returns, and
115967	**
115968	** 2. After scanning the current FTS table row for the deferred tokens,
115969	** it is determined that the row does not match the query.
115970	**
115971	** Or, if no error occurs and it seems the current row does match the FTS
115972	** query, return 0.
115973	*/
115974	static int fts3EvalLoadDeferred(Fts3Cursor pCsr, int pRc){
115975	int rc = *pRc;
115976	int bMiss = 0;
115977	if( rc==SQLITE_OK ){







115978	if( pCsr->pDeferred ){
115979	rc = fts3CursorSeek(0, pCsr);
115980	if( rc==SQLITE_OK ){
115981	rc = sqlite3Fts3CacheDeferredDoclists(pCsr);
115982	}
115983	}
115984	bMiss = (0==fts3EvalDeferredTest(pCsr, pCsr->pExpr, &rc));


115985	sqlite3Fts3FreeDeferredDoclists(pCsr);
115986	*pRc = rc;
115987	}
115988	return (rc==SQLITE_OK && bMiss);
115989	}
	@@ -115990,11 +118382,11 @@
115990
115991	/*
115992	** Advance to the next document that matches the FTS expression in
115993	** Fts3Cursor.pExpr.
115994	*/
115995	SQLITE_PRIVATE int sqlite3Fts3EvalNext(Fts3Cursor *pCsr){
115996	int rc = SQLITE_OK; /* Return Code */
115997	Fts3Expr *pExpr = pCsr->pExpr;
115998	assert( pCsr->isEof==0 );
115999	if( pExpr==0 ){
116000	pCsr->isEof = 1;
	@@ -116002,23 +118394,23 @@
116002	do {
116003	if( pCsr->isRequireSeek==0 ){
116004	sqlite3_reset(pCsr->pStmt);
116005	}
116006	assert( sqlite3_data_count(pCsr->pStmt)==0 );
116007	fts3EvalNext(pCsr, pExpr, &rc);
116008	pCsr->isEof = pExpr->bEof;
116009	pCsr->isRequireSeek = 1;
116010	pCsr->isMatchinfoNeeded = 1;
116011	pCsr->iPrevId = pExpr->iDocid;
116012	}while( pCsr->isEof==0 && fts3EvalLoadDeferred(pCsr, &rc) );
116013	}
116014	return rc;
116015	}
116016
116017	/*
116018	** Restart interation for expression pExpr so that the next call to
116019	** sqlite3Fts3EvalNext() visits the first row. Do not allow incremental
116020	** loading or merging of phrase doclists for this iteration.
116021	**
116022	** If *pRc is other than SQLITE_OK when this function is called, it is
116023	** a no-op. If an error occurs within this function, *pRc is set to an
116024	** SQLite error code before returning.
	@@ -116030,19 +118422,20 @@
116030	){
116031	if( pExpr && *pRc==SQLITE_OK ){
116032	Fts3Phrase *pPhrase = pExpr->pPhrase;
116033
116034	if( pPhrase ){
116035	fts3EvalZeroPoslist(pPhrase);
116036	if( pPhrase->bIncr ){
116037	sqlite3Fts3EvalPhraseCleanup(pPhrase);
116038	memset(&pPhrase->doclist, 0, sizeof(Fts3Doclist));
116039	*pRc = sqlite3Fts3EvalStart(pCsr, pExpr, 0);
116040	}else{
116041	pPhrase->doclist.pNextDocid = 0;
116042	pPhrase->doclist.iDocid = 0;
116043	}



116044	}
116045
116046	pExpr->iDocid = 0;
116047	pExpr->bEof = 0;
116048	pExpr->bStart = 0;
	@@ -116145,18 +118538,18 @@
116145	/* Ensure the %_content statement is reset. */
116146	if( pCsr->isRequireSeek==0 ) sqlite3_reset(pCsr->pStmt);
116147	assert( sqlite3_data_count(pCsr->pStmt)==0 );
116148
116149	/* Advance to the next document */
116150	fts3EvalNext(pCsr, pRoot, &rc);
116151	pCsr->isEof = pRoot->bEof;
116152	pCsr->isRequireSeek = 1;
116153	pCsr->isMatchinfoNeeded = 1;
116154	pCsr->iPrevId = pRoot->iDocid;
116155	}while( pCsr->isEof==0
116156	&& pRoot->eType==FTSQUERY_NEAR
116157	&& fts3EvalLoadDeferred(pCsr, &rc)
116158	);
116159
116160	if( rc==SQLITE_OK && pCsr->isEof==0 ){
116161	fts3EvalUpdateCounts(pRoot);
116162	}
	@@ -116174,14 +118567,14 @@
116174	**
116175	** do {...} while( pRoot->iDocid<iDocid && rc==SQLITE_OK );
116176	*/
116177	fts3EvalRestart(pCsr, pRoot, &rc);
116178	do {
116179	fts3EvalNext(pCsr, pRoot, &rc);
116180	assert( pRoot->bEof==0 );
116181	}while( pRoot->iDocid!=iDocid && rc==SQLITE_OK );
116182	fts3EvalLoadDeferred(pCsr, &rc);
116183	}
116184	}
116185	return rc;
116186	}
116187
	@@ -116225,12 +118618,12 @@
116225	int iCol;
116226
116227	if( pExpr->bDeferred && pExpr->pParent->eType!=FTSQUERY_NEAR ){
116228	assert( pCsr->nDoc>0 );
116229	for(iCol=0; iCol<pTab->nColumn; iCol++){
116230	aiOut[iCol*3 + 1] = pCsr->nDoc;
116231	aiOut[iCol*3 + 2] = pCsr->nDoc;
116232	}
116233	}else{
116234	rc = fts3EvalGatherStats(pCsr, pExpr);
116235	if( rc==SQLITE_OK ){
116236	assert( pExpr->aMI );
	@@ -116308,18 +118701,32 @@
116308	*/
116309	SQLITE_PRIVATE void sqlite3Fts3EvalPhraseCleanup(Fts3Phrase *pPhrase){
116310	if( pPhrase ){
116311	int i;
116312	sqlite3_free(pPhrase->doclist.aAll);
116313	fts3EvalZeroPoslist(pPhrase);
116314	memset(&pPhrase->doclist, 0, sizeof(Fts3Doclist));
116315	for(i=0; i<pPhrase->nToken; i++){
116316	fts3SegReaderCursorFree(pPhrase->aToken[i].pSegcsr);
116317	pPhrase->aToken[i].pSegcsr = 0;
116318	}
116319	}
116320	}














116321
116322	#endif
116323
116324	/************ End of fts3.c **********************************************/
116325	/************ Begin file fts3_aux.c **************************************/
	@@ -116334,13 +118741,14 @@
116334	** May you share freely, never taking more than you give.
116335	**
116336	******************************************************************************
116337	**
116338	*/
116339
116340	#if !defined(SQLITE_CORE) \|\| defined(SQLITE_ENABLE_FTS3)
116341


116342
116343	typedef struct Fts3auxTable Fts3auxTable;
116344	typedef struct Fts3auxCursor Fts3auxCursor;
116345
116346	struct Fts3auxTable {
	@@ -116875,10 +119283,12 @@
116875	/*
116876	** Default span for NEAR operators.
116877	*/
116878	#define SQLITE_FTS3_DEFAULT_NEAR_PARAM 10
116879


116880
116881	/*
116882	** isNot:
116883	** This variable is used by function getNextNode(). When getNextNode() is
116884	** called, it sets ParseContext.isNot to true if the 'next node' is a
	@@ -117576,10 +119986,11 @@
117576	** Everything after this point is just test code.
117577	*/
117578
117579	#ifdef SQLITE_TEST
117580

117581
117582	/*
117583	** Function to query the hash-table of tokenizers (see README.tokenizers).
117584	*/
117585	static int queryTestTokenizer(
	@@ -117786,10 +120197,13 @@
117786	** * The FTS3 module is being built into the core of
117787	** SQLite (in which case SQLITE_ENABLE_FTS3 is defined).
117788	*/
117789	#if !defined(SQLITE_CORE) \|\| defined(SQLITE_ENABLE_FTS3)
117790



117791
117792
117793	/*
117794	** Malloc and Free functions
117795	*/
	@@ -118166,11 +120580,14 @@
118166	** * The FTS3 module is being built into the core of
118167	** SQLite (in which case SQLITE_ENABLE_FTS3 is defined).
118168	*/
118169	#if !defined(SQLITE_CORE) \|\| defined(SQLITE_ENABLE_FTS3)
118170
118171



118172
118173
118174	/*
118175	** Class derived from sqlite3_tokenizer
118176	*/
	@@ -118810,14 +121227,12 @@
118810	** * The FTS3 module is being built into the core of
118811	** SQLite (in which case SQLITE_ENABLE_FTS3 is defined).
118812	*/
118813	#if !defined(SQLITE_CORE) \|\| defined(SQLITE_ENABLE_FTS3)
118814
118815	#ifndef SQLITE_CORE
118816	SQLITE_EXTENSION_INIT1
118817	#endif
118818
118819
118820	/*
118821	** Implementation of the SQL scalar function for accessing the underlying
118822	** hash table. This function may be called as follows:
118823	**
	@@ -118937,11 +121352,11 @@
118937	sqlite3_tokenizer *ppTok, / OUT: Tokenizer (if applicable) */
118938	char *pzErr / OUT: Set to malloced error message */
118939	){
118940	int rc;
118941	char z = (char )zArg;
118942	int n;
118943	char *zCopy;
118944	char zEnd; / Pointer to nul-term of zCopy */
118945	sqlite3_tokenizer_module *m;
118946
118947	zCopy = sqlite3_mprintf("%s", zArg);
	@@ -118989,10 +121404,12 @@
118989	}
118990
118991
118992	#ifdef SQLITE_TEST
118993


118994
118995	/*
118996	** Implementation of a special SQL scalar function for testing tokenizers
118997	** designed to be used in concert with the Tcl testing framework. This
118998	** function must be called with two arguments:
	@@ -119300,11 +121717,14 @@
119300	** * The FTS3 module is being built into the core of
119301	** SQLite (in which case SQLITE_ENABLE_FTS3 is defined).
119302	*/
119303	#if !defined(SQLITE_CORE) \|\| defined(SQLITE_ENABLE_FTS3)
119304
119305



119306
119307
119308	typedef struct simple_tokenizer {
119309	sqlite3_tokenizer base;
119310	char delim[128]; /* flag ASCII delimiters */
	@@ -119526,10 +121946,13 @@
119526	** code in fts3.c.
119527	*/
119528
119529	#if !defined(SQLITE_CORE) \|\| defined(SQLITE_ENABLE_FTS3)
119530



119531
119532	/*
119533	** When full-text index nodes are loaded from disk, the buffer that they
119534	** are loaded into has the following number of bytes of padding at the end
119535	** of it. i.e. if a full-text index node is 900 bytes in size, then a buffer
	@@ -120750,10 +123173,11 @@
120750	pReader->pOffsetList = 0;
120751	}else{
120752	pReader->pOffsetList = p;
120753	}
120754	}else{

120755
120756	/* Pointer p currently points at the first byte of an offset list. The
120757	** following block advances it to point one byte past the end of
120758	** the same offset list. */
120759	while( 1 ){
	@@ -120778,17 +123202,19 @@
120778	*/
120779	if( ppOffsetList ){
120780	*ppOffsetList = pReader->pOffsetList;
120781	*pnOffsetList = (int)(p - pReader->pOffsetList - 1);
120782	}


120783
120784	/* If there are no more entries in the doclist, set pOffsetList to
120785	** NULL. Otherwise, set Fts3SegReader.iDocid to the next docid and
120786	** Fts3SegReader.pOffsetList to point to the next offset list before
120787	** returning.
120788	*/
120789	if( p>=&pReader->aDoclist[pReader->nDoclist] ){
120790	pReader->pOffsetList = 0;
120791	}else{
120792	rc = fts3SegReaderRequire(pReader, p, FTS3_VARINT_MAX);
120793	if( rc==SQLITE_OK ){
120794	sqlite3_int64 iDelta;
	@@ -120823,11 +123249,11 @@
120823	for(ii=0; rc==SQLITE_OK && ii<pMsr->nSegment; ii++){
120824	Fts3SegReader *pReader = pMsr->apSegment[ii];
120825	if( !fts3SegReaderIsPending(pReader)
120826	&& !fts3SegReaderIsRootOnly(pReader)
120827	){
120828	int jj;
120829	for(jj=pReader->iStartBlock; jj<=pReader->iLeafEndBlock; jj++){
120830	int nBlob;
120831	rc = sqlite3Fts3ReadBlock(p, jj, 0, &nBlob, 0);
120832	if( rc!=SQLITE_OK ) break;
120833	if( (nBlob+35)>pgsz ){
	@@ -121768,55 +124194,31 @@
121768
121769	*ppList = pList;
121770	*pnList = nList;
121771	}
121772
121773	SQLITE_PRIVATE int sqlite3Fts3MsrIncrStart(
121774	Fts3Table p, / Virtual table handle */
121775	Fts3MultiSegReader pCsr, / Cursor object */
121776	int iCol, /* Column to match on. */
121777	const char zTerm, / Term to iterate through a doclist for */
121778	int nTerm /* Number of bytes in zTerm */





121779	){
121780	int i;
121781	int nSegment = pCsr->nSegment;
121782	int (xCmp)(Fts3SegReader , Fts3SegReader *) = (
121783	p->bDescIdx ? fts3SegReaderDoclistCmpRev : fts3SegReaderDoclistCmp
121784	);
121785
121786	assert( pCsr->pFilter==0 );
121787	assert( zTerm && nTerm>0 );
121788
121789	/* Advance each segment iterator until it points to the term zTerm/nTerm. */
121790	for(i=0; i<nSegment; i++){
121791	Fts3SegReader *pSeg = pCsr->apSegment[i];
121792	do {
121793	int rc = fts3SegReaderNext(p, pSeg, 1);
121794	if( rc!=SQLITE_OK ) return rc;
121795	}while( fts3SegReaderTermCmp(pSeg, zTerm, nTerm)<0 );
121796	}
121797	fts3SegReaderSort(pCsr->apSegment, nSegment, nSegment, fts3SegReaderCmp);
121798
121799	/* Determine how many of the segments actually point to zTerm/nTerm. */
121800	for(i=0; i<nSegment; i++){
121801	Fts3SegReader *pSeg = pCsr->apSegment[i];
121802	if( !pSeg->aNode \|\| fts3SegReaderTermCmp(pSeg, zTerm, nTerm) ){
121803	break;
121804	}
121805	}
121806	pCsr->nAdvance = i;
121807
121808	/* Advance each of the segments to point to the first docid. */
121809	for(i=0; i<pCsr->nAdvance; i++){
121810	int rc = fts3SegReaderFirstDocid(p, pCsr->apSegment[i]);
121811	if( rc!=SQLITE_OK ) return rc;
121812	}
121813	fts3SegReaderSort(pCsr->apSegment, i, i, xCmp);
121814
121815	assert( iCol<0 \|\| iCol<p->nColumn );
121816	pCsr->iColFilter = iCol;
121817
121818	return SQLITE_OK;
121819	}
121820
121821	SQLITE_PRIVATE int sqlite3Fts3MsrIncrNext(
121822	Fts3Table p, / Virtual table handle */
	@@ -121866,52 +124268,138 @@
121866	if( pMsr->iColFilter>=0 ){
121867	fts3ColumnFilter(pMsr->iColFilter, &pList, &nList);
121868	}
121869
121870	if( nList>0 ){








121871	*piDocid = iDocid;
121872	*paPoslist = pList;
121873	*pnPoslist = nList;
121874	break;
121875	}
121876	}
121877
121878	}
121879



























121880	return SQLITE_OK;
121881	}
121882
121883	SQLITE_PRIVATE int sqlite3Fts3SegReaderStart(
121884	Fts3Table p, / Virtual table handle */
121885	Fts3MultiSegReader pCsr, / Cursor object */
121886	Fts3SegFilter pFilter / Restrictions on range of iteration */
121887	){
121888	int i;
121889
121890	/* Initialize the cursor object */
121891	pCsr->pFilter = pFilter;
121892
121893	/* If the Fts3SegFilter defines a specific term (or term prefix) to search
121894	** for, then advance each segment iterator until it points to a term of
121895	** equal or greater value than the specified term. This prevents many
121896	** unnecessary merge/sort operations for the case where single segment
121897	** b-tree leaf nodes contain more than one term.
121898	*/
121899	for(i=0; i<pCsr->nSegment; i++){
121900	int nTerm = pFilter->nTerm;
121901	const char *zTerm = pFilter->zTerm;
















121902	Fts3SegReader *pSeg = pCsr->apSegment[i];
121903	do {
121904	int rc = fts3SegReaderNext(p, pSeg, 0);
121905	if( rc!=SQLITE_OK ) return rc;
121906	}while( zTerm && fts3SegReaderTermCmp(pSeg, zTerm, nTerm)<0 );
121907	}
121908	fts3SegReaderSort(
121909	pCsr->apSegment, pCsr->nSegment, pCsr->nSegment, fts3SegReaderCmp);








































121910
121911	return SQLITE_OK;
121912	}

121913
121914	SQLITE_PRIVATE int sqlite3Fts3SegReaderStep(
121915	Fts3Table p, / Virtual table handle */
121916	Fts3MultiSegReader pCsr / Cursor object */
121917	){
	@@ -121981,13 +124469,18 @@
121981	assert( isIgnoreEmpty \|\| (isRequirePos && !isColFilter) );
121982	if( nMerge==1
121983	&& !isIgnoreEmpty
121984	&& (p->bDescIdx==0 \|\| fts3SegReaderIsPending(apSegment[0])==0)
121985	){
121986	pCsr->aDoclist = apSegment[0]->aDoclist;
121987	pCsr->nDoclist = apSegment[0]->nDoclist;
121988	rc = SQLITE_ROW;






121989	}else{
121990	int nDoclist = 0; /* Size of doclist */
121991	sqlite3_int64 iPrev = 0; /* Previous docid stored in doclist */
121992
121993	/* The current term of the first nMerge entries in the array
	@@ -122717,10 +125210,12 @@
122717	******************************************************************************
122718	*/
122719
122720	#if !defined(SQLITE_CORE) \|\| defined(SQLITE_ENABLE_FTS3)
122721


122722
122723	/*
122724	** Characters that may appear in the second argument to matchinfo().
122725	*/
122726	#define FTS3_MATCHINFO_NPHRASE 'p' /* 1 value */
	@@ -123725,11 +126220,11 @@
123725	if( bGlobal ) pInfo->aMatchinfo[0] = pInfo->nCol;
123726	break;
123727
123728	case FTS3_MATCHINFO_NDOC:
123729	if( bGlobal ){
123730	sqlite3_int64 nDoc;
123731	rc = fts3MatchinfoSelectDoctotal(pTab, &pSelect, &nDoc, 0);
123732	pInfo->aMatchinfo[0] = (u32)nDoc;
123733	}
123734	break;
123735
	@@ -124304,10 +126799,12 @@
124304	#ifndef SQLITE_CORE
124305	SQLITE_EXTENSION_INIT1
124306	#else
124307	#endif
124308


124309
124310	#ifndef SQLITE_AMALGAMATION
124311	#include "sqlite3rtree.h"
124312	typedef sqlite3_int64 i64;
124313	typedef unsigned char u8;
	@@ -125620,11 +128117,11 @@
125620	*/
125621	static float cellArea(Rtree pRtree, RtreeCell p){
125622	float area = 1.0;
125623	int ii;
125624	for(ii=0; ii<(pRtree->nDim*2); ii+=2){
125625	area = area * (DCOORD(p->aCoord[ii+1]) - DCOORD(p->aCoord[ii]));
125626	}
125627	return area;
125628	}
125629
125630	/*
	@@ -125633,11 +128130,11 @@
125633	*/
125634	static float cellMargin(Rtree pRtree, RtreeCell p){
125635	float margin = 0.0;
125636	int ii;
125637	for(ii=0; ii<(pRtree->nDim*2); ii+=2){
125638	margin += (DCOORD(p->aCoord[ii+1]) - DCOORD(p->aCoord[ii]));
125639	}
125640	return margin;
125641	}
125642
125643	/*
	@@ -125718,11 +128215,11 @@
125718
125719	if( x2<x1 ){
125720	o = 0.0;
125721	break;
125722	}else{
125723	o = o * (x2-x1);
125724	}
125725	}
125726	overlap += o;
125727	}
125728	}
	@@ -125737,16 +128234,16 @@
125737	RtreeCell *pInsert,
125738	RtreeCell *aCell,
125739	int nCell,
125740	int iExclude
125741	){
125742	float before;
125743	float after;
125744	before = cellOverlap(pRtree, p, aCell, nCell, iExclude);
125745	cellUnion(pRtree, p, pInsert);
125746	after = cellOverlap(pRtree, p, aCell, nCell, iExclude);
125747	return after-before;
125748	}
125749	#endif
125750
125751
125752	/*
	@@ -125764,15 +128261,15 @@
125764	RtreeNode *pNode;
125765	rc = nodeAcquire(pRtree, 1, 0, &pNode);
125766
125767	for(ii=0; rc==SQLITE_OK && ii<(pRtree->iDepth-iHeight); ii++){
125768	int iCell;
125769	sqlite3_int64 iBest;
125770
125771	float fMinGrowth;
125772	float fMinArea;
125773	float fMinOverlap;
125774
125775	int nCell = NCELL(pNode);
125776	RtreeCell cell;
125777	RtreeNode *pChild;
125778
	@@ -126198,13 +128695,13 @@
126198	){
126199	int **aaSorted;
126200	int *aSpare;
126201	int ii;
126202
126203	int iBestDim;
126204	int iBestSplit;
126205	float fBestMargin;
126206
126207	int nByte = (pRtree->nDim+1)(sizeof(int)+nCell*sizeof(int));
126208
126209	aaSorted = (int **)sqlite3_malloc(nByte);
126210	if( !aaSorted ){
	@@ -126222,13 +128719,13 @@
126222	SortByDimension(pRtree, aaSorted[ii], nCell, ii, aCell, aSpare);
126223	}
126224
126225	for(ii=0; ii<pRtree->nDim; ii++){
126226	float margin = 0.0;
126227	float fBestOverlap;
126228	float fBestArea;
126229	int iBestLeft;
126230	int nLeft;
126231
126232	for(
126233	nLeft=RTREE_MINCELLS(pRtree);
126234	nLeft<=(nCell-RTREE_MINCELLS(pRtree));
	@@ -126539,11 +129036,11 @@
126539	static int deleteCell(Rtree , RtreeNode , int, int);
126540
126541	static int removeNode(Rtree pRtree, RtreeNode pNode, int iHeight){
126542	int rc;
126543	int rc2;
126544	RtreeNode *pParent;
126545	int iCell;
126546
126547	assert( pNode->nRef==1 );
126548
126549	/* Remove the entry in the parent cell. */
	@@ -126687,23 +129184,23 @@
126687	}else{
126688	nodeGetCell(pRtree, pNode, ii, &aCell[ii]);
126689	}
126690	aOrder[ii] = ii;
126691	for(iDim=0; iDim<pRtree->nDim; iDim++){
126692	aCenterCoord[iDim] += DCOORD(aCell[ii].aCoord[iDim*2]);
126693	aCenterCoord[iDim] += DCOORD(aCell[ii].aCoord[iDim*2+1]);
126694	}
126695	}
126696	for(iDim=0; iDim<pRtree->nDim; iDim++){
126697	aCenterCoord[iDim] = aCenterCoord[iDim]/((float)nCell*2.0);
126698	}
126699
126700	for(ii=0; ii<nCell; ii++){
126701	aDistance[ii] = 0.0;
126702	for(iDim=0; iDim<pRtree->nDim; iDim++){
126703	float coord = DCOORD(aCell[ii].aCoord[iDim*2+1]) -
126704	DCOORD(aCell[ii].aCoord[iDim*2]);
126705	aDistance[ii] += (coord-aCenterCoord[iDim])*(coord-aCenterCoord[iDim]);
126706	}
126707	}
126708
126709	SortByDistance(aOrder, nCell, aDistance, aSpare);
	@@ -126798,14 +129295,14 @@
126798	nodeGetCell(pRtree, pNode, ii, &cell);
126799
126800	/* Find a node to store this cell in. pNode->iNode currently contains
126801	** the height of the sub-tree headed by the cell.
126802	*/
126803	rc = ChooseLeaf(pRtree, &cell, pNode->iNode, &pInsert);
126804	if( rc==SQLITE_OK ){
126805	int rc2;
126806	rc = rtreeInsertCell(pRtree, pInsert, &cell, pNode->iNode);
126807	rc2 = nodeRelease(pRtree, pInsert);
126808	if( rc==SQLITE_OK ){
126809	rc = rc2;
126810	}
126811	}
	@@ -127190,11 +129687,11 @@
127190	int isCreate /* True for xCreate, false for xConnect */
127191	){
127192	int rc;
127193	char *zSql;
127194	if( isCreate ){
127195	int iPageSize;
127196	zSql = sqlite3_mprintf("PRAGMA %Q.page_size", pRtree->zDb);
127197	rc = getIntFromStmt(db, zSql, &iPageSize);
127198	if( rc==SQLITE_OK ){
127199	pRtree->iNodeSize = iPageSize-64;
127200	if( (4+pRtree->nBytesPerCell*RTREE_MAXCELLS)<pRtree->iNodeSize ){
	@@ -127518,10 +130015,11 @@
127518	#include <unicode/utypes.h>
127519	#include <unicode/uregex.h>
127520	#include <unicode/ustring.h>
127521	#include <unicode/ucol.h>
127522

127523
127524	#ifndef SQLITE_CORE
127525	SQLITE_EXTENSION_INIT1
127526	#else
127527	#endif
	@@ -127993,19 +130491,20 @@
127993	** May you find forgiveness for yourself and forgive others.
127994	** May you share freely, never taking more than you give.
127995	**
127996	*************************************************************************
127997	** This file implements a tokenizer for fts3 based on the ICU library.
127998	**
127999	** $Id: fts3_icu.c,v 1.3 2008/09/01 18:34:20 danielk1977 Exp $
128000	*/
128001
128002	#if !defined(SQLITE_CORE) \|\| defined(SQLITE_ENABLE_FTS3)
128003	#ifdef SQLITE_ENABLE_ICU
128004


128005
128006	#include <unicode/ubrk.h>


128007	#include <unicode/utf16.h>
128008
128009	typedef struct IcuTokenizer IcuTokenizer;
128010	typedef struct IcuCursor IcuCursor;
128011
128012

	--- src/sqlite3.c
	+++ src/sqlite3.c
	@@ -1,8 +1,8 @@
1	/******************************************************************************
2	** This file is an amalgamation of many separate C source files from SQLite
3	** version 3.7.8. By combining all the individual C code files into this
4	** single large file, the entire code can be compiled as a single translation
5	** unit. This allows many compilers to do optimizations that would not be
6	** possible if the files were compiled separately. Performance improvements
7	** of 5% or more are commonly seen when SQLite is compiled as a single
8	** translation unit.
	@@ -385,23 +385,29 @@
385	/*
386	** Exactly one of the following macros must be defined in order to
387	** specify which memory allocation subsystem to use.
388	**
389	** SQLITE_SYSTEM_MALLOC // Use normal system malloc()
390	** SQLITE_WIN32_MALLOC // Use Win32 native heap API
391	** SQLITE_MEMDEBUG // Debugging version of system malloc()
392	**
393	** On Windows, if the SQLITE_WIN32_MALLOC_VALIDATE macro is defined and the
394	** assert() macro is enabled, each call into the Win32 native heap subsystem
395	** will cause HeapValidate to be called. If heap validation should fail, an
396	** assertion will be triggered.
397	**
398	** (Historical note: There used to be several other options, but we've
399	** pared it down to just these two.)
400	**
401	** If none of the above are defined, then set SQLITE_SYSTEM_MALLOC as
402	** the default.
403	*/
404	#if defined(SQLITE_SYSTEM_MALLOC)+defined(SQLITE_WIN32_MALLOC)+defined(SQLITE_MEMDEBUG)>1
405	# error "At most one of the following compile-time configuration options\
406	is allows: SQLITE_SYSTEM_MALLOC, SQLITE_WIN32_MALLOC, SQLITE_MEMDEBUG"
407	#endif
408	#if defined(SQLITE_SYSTEM_MALLOC)+defined(SQLITE_WIN32_MALLOC)+defined(SQLITE_MEMDEBUG)==0
409	# define SQLITE_SYSTEM_MALLOC 1
410	#endif
411
412	/*
413	** If SQLITE_MALLOC_SOFT_LIMIT is not zero, then try to keep the
	@@ -648,13 +654,13 @@
654	**
655	** See also: [sqlite3_libversion()],
656	** [sqlite3_libversion_number()], [sqlite3_sourceid()],
657	** [sqlite_version()] and [sqlite_source_id()].
658	*/
659	#define SQLITE_VERSION "3.7.8"
660	#define SQLITE_VERSION_NUMBER 3007008
661	#define SQLITE_SOURCE_ID "2011-08-29 11:56:14 639cc85a911454bffdcccb33f2976c683953ae64"
662
663	/*
664	** CAPI3REF: Run-Time Library Version Numbers
665	** KEYWORDS: sqlite3_version, sqlite3_sourceid
666	**
	@@ -851,11 +857,11 @@
857	** semicolon-separate SQL statements passed into its 2nd argument,
858	** in the context of the [database connection] passed in as its 1st
859	** argument. ^If the callback function of the 3rd argument to
860	** sqlite3_exec() is not NULL, then it is invoked for each result row
861	** coming out of the evaluated SQL statements. ^The 4th argument to
862	** sqlite3_exec() is relayed through to the 1st argument of each
863	** callback invocation. ^If the callback pointer to sqlite3_exec()
864	** is NULL, then no callback is ever invoked and result rows are
865	** ignored.
866	**
867	** ^If an error occurs while evaluating the SQL statements passed into
	@@ -1282,20 +1288,52 @@
1288	** when [PRAGMA synchronous \| PRAGMA synchronous=OFF] is set, but most
1289	** VFSes do not need this signal and should silently ignore this opcode.
1290	** Applications should not call [sqlite3_file_control()] with this
1291	** opcode as doing so may disrupt the operation of the specialized VFSes
1292	** that do require it.
1293	**
1294	** ^The [SQLITE_FCNTL_WIN32_AV_RETRY] opcode is used to configure automatic
1295	** retry counts and intervals for certain disk I/O operations for the
1296	** windows [VFS] in order to work to provide robustness against
1297	** anti-virus programs. By default, the windows VFS will retry file read,
1298	** file write, and file delete opertions up to 10 times, with a delay
1299	** of 25 milliseconds before the first retry and with the delay increasing
1300	** by an additional 25 milliseconds with each subsequent retry. This
1301	** opcode allows those to values (10 retries and 25 milliseconds of delay)
1302	** to be adjusted. The values are changed for all database connections
1303	** within the same process. The argument is a pointer to an array of two
1304	** integers where the first integer i the new retry count and the second
1305	** integer is the delay. If either integer is negative, then the setting
1306	** is not changed but instead the prior value of that setting is written
1307	** into the array entry, allowing the current retry settings to be
1308	** interrogated. The zDbName parameter is ignored.
1309	**
1310	** ^The [SQLITE_FCNTL_PERSIST_WAL] opcode is used to set or query the
1311	** persistent [WAL \| Write AHead Log] setting. By default, the auxiliary
1312	** write ahead log and shared memory files used for transaction control
1313	** are automatically deleted when the latest connection to the database
1314	** closes. Setting persistent WAL mode causes those files to persist after
1315	** close. Persisting the files is useful when other processes that do not
1316	** have write permission on the directory containing the database file want
1317	** to read the database file, as the WAL and shared memory files must exist
1318	** in order for the database to be readable. The fourth parameter to
1319	** [sqlite3_file_control()] for this opcode should be a pointer to an integer.
1320	** That integer is 0 to disable persistent WAL mode or 1 to enable persistent
1321	** WAL mode. If the integer is -1, then it is overwritten with the current
1322	** WAL persistence setting.
1323	**
1324	*/
1325	#define SQLITE_FCNTL_LOCKSTATE 1
1326	#define SQLITE_GET_LOCKPROXYFILE 2
1327	#define SQLITE_SET_LOCKPROXYFILE 3
1328	#define SQLITE_LAST_ERRNO 4
1329	#define SQLITE_FCNTL_SIZE_HINT 5
1330	#define SQLITE_FCNTL_CHUNK_SIZE 6
1331	#define SQLITE_FCNTL_FILE_POINTER 7
1332	#define SQLITE_FCNTL_SYNC_OMITTED 8
1333	#define SQLITE_FCNTL_WIN32_AV_RETRY 9
1334	#define SQLITE_FCNTL_PERSIST_WAL 10
1335
1336	/*
1337	** CAPI3REF: Mutex Handle
1338	**
1339	** The mutex module within SQLite defines [sqlite3_mutex] to be an
	@@ -1443,11 +1481,11 @@
1481	** The xSleep() method causes the calling thread to sleep for at
1482	** least the number of microseconds given. ^The xCurrentTime()
1483	** method returns a Julian Day Number for the current date and time as
1484	** a floating point value.
1485	** ^The xCurrentTimeInt64() method returns, as an integer, the Julian
1486	** Day Number multiplied by 86400000 (the number of milliseconds in
1487	** a 24-hour day).
1488	** ^SQLite will use the xCurrentTimeInt64() method to get the current
1489	** date and time if that method is available (if iVersion is 2 or
1490	** greater and the function pointer is not NULL) and will fall back
1491	** to xCurrentTime() if xCurrentTimeInt64() is unavailable.
	@@ -1719,20 +1757,14 @@
1757	** also used during testing of SQLite in order to specify an alternative
1758	** memory allocator that simulates memory out-of-memory conditions in
1759	** order to verify that SQLite recovers gracefully from such
1760	** conditions.
1761	**
1762	** The xMalloc, xRealloc, and xFree methods must work like the
1763	** malloc(), realloc() and free() functions from the standard C library.
1764	** ^SQLite guarantees that the second argument to



1765	** xRealloc is always a value returned by a prior call to xRoundup.



1766	**
1767	** xSize should return the allocated size of a memory allocation
1768	** previously obtained from xMalloc or xRealloc. The allocated size
1769	** is always at least as big as the requested size but may be larger.
1770	**
	@@ -1881,11 +1913,11 @@
1913	** scratch memory beyond what is provided by this configuration option, then
1914	** [sqlite3_malloc()] will be used to obtain the memory needed.</dd>
1915	**
1916	** [[SQLITE_CONFIG_PAGECACHE]] <dt>SQLITE_CONFIG_PAGECACHE</dt>
1917	** <dd> ^This option specifies a static memory buffer that SQLite can use for
1918	** the database page cache with the default page cache implementation.
1919	** This configuration should not be used if an application-define page
1920	** cache implementation is loaded using the SQLITE_CONFIG_PCACHE option.
1921	** There are three arguments to this option: A pointer to 8-byte aligned
1922	** memory, the size of each page buffer (sz), and the number of pages (N).
1923	** The sz argument should be the size of the largest database page
	@@ -2979,16 +3011,16 @@
3011	** [[URI filenames in sqlite3_open()]] <h3>URI Filenames</h3>
3012	**
3013	** ^If [URI filename] interpretation is enabled, and the filename argument
3014	** begins with "file:", then the filename is interpreted as a URI. ^URI
3015	** filename interpretation is enabled if the [SQLITE_OPEN_URI] flag is
3016	** set in the fourth argument to sqlite3_open_v2(), or if it has
3017	** been enabled globally using the [SQLITE_CONFIG_URI] option with the
3018	** [sqlite3_config()] method or by the [SQLITE_USE_URI] compile-time option.
3019	** As of SQLite version 3.7.7, URI filename interpretation is turned off
3020	** by default, but future releases of SQLite might enable URI filename
3021	** interpretation by default. See "[URI filenames]" for additional
3022	** information.
3023	**
3024	** URI filenames are parsed according to RFC 3986. ^If the URI contains an
3025	** authority, then it must be either an empty string or the string
3026	** "localhost". ^If the authority is not an empty string or "localhost", an
	@@ -3803,11 +3835,11 @@
3835	** [extended result codes] might be returned as well.
3836	**
3837	** ^[SQLITE_BUSY] means that the database engine was unable to acquire the
3838	** database locks it needs to do its job. ^If the statement is a [COMMIT]
3839	** or occurs outside of an explicit transaction, then you can retry the
3840	** statement. If the statement is not a [COMMIT] and occurs within an
3841	** explicit transaction then you should rollback the transaction before
3842	** continuing.
3843	**
3844	** ^[SQLITE_DONE] means that the statement has finished executing
3845	** successfully. sqlite3_step() should not be called again on this virtual
	@@ -4082,11 +4114,11 @@
4114
4115	/*
4116	** CAPI3REF: Destroy A Prepared Statement Object
4117	**
4118	** ^The sqlite3_finalize() function is called to delete a [prepared statement].
4119	** ^If the most recent evaluation of the statement encountered no errors
4120	** or if the statement is never been evaluated, then sqlite3_finalize() returns
4121	** SQLITE_OK. ^If the most recent evaluation of statement S failed, then
4122	** sqlite3_finalize(S) returns the appropriate [error code] or
4123	** [extended error code].
4124	**
	@@ -5996,11 +6028,11 @@
6028	** versions of these routines, it should at least provide stubs that always
6029	** return true so that one does not get spurious assertion failures.
6030	**
6031	** ^If the argument to sqlite3_mutex_held() is a NULL pointer then
6032	** the routine should return 1. This seems counter-intuitive since
6033	** clearly the mutex cannot be held if it does not exist. But
6034	** the reason the mutex does not exist is because the build is not
6035	** using mutexes. And we do not want the assert() containing the
6036	** call to sqlite3_mutex_held() to fail, so a non-zero return is
6037	** the appropriate thing to do. ^The sqlite3_mutex_notheld()
6038	** interface should also return 1 when given a NULL pointer.
	@@ -6119,11 +6151,12 @@
6151	#define SQLITE_TESTCTRL_RESERVE 14
6152	#define SQLITE_TESTCTRL_OPTIMIZATIONS 15
6153	#define SQLITE_TESTCTRL_ISKEYWORD 16
6154	#define SQLITE_TESTCTRL_PGHDRSZ 17
6155	#define SQLITE_TESTCTRL_SCRATCHMALLOC 18
6156	#define SQLITE_TESTCTRL_LOCALTIME_FAULT 19
6157	#define SQLITE_TESTCTRL_LAST 19
6158
6159	/*
6160	** CAPI3REF: SQLite Runtime Status
6161	**
6162	** ^This interface is used to retrieve runtime status information
	@@ -6505,11 +6538,11 @@
6538	** [[the xFetch() page cache methods]]
6539	** The xFetch() method locates a page in the cache and returns a pointer to
6540	** the page, or a NULL pointer.
6541	** A "page", in this context, means a buffer of szPage bytes aligned at an
6542	** 8-byte boundary. The page to be fetched is determined by the key. ^The
6543	** minimum key value is 1. After it has been retrieved using xFetch, the page
6544	** is considered to be "pinned".
6545	**
6546	** If the requested page is already in the page cache, then the page cache
6547	** implementation must return a pointer to the page buffer with its content
6548	** intact. If the requested page is not already in the cache, then the
	@@ -7599,10 +7632,18 @@
7632	*/
7633	#ifndef SQLITE_TEMP_STORE
7634	# define SQLITE_TEMP_STORE 1
7635	#endif
7636
7637	/*
7638	** If all temporary storage is in-memory, then omit the external merge-sort
7639	** logic since it is superfluous.
7640	*/
7641	#if SQLITE_TEMP_STORE==3 && !defined(SQLITE_OMIT_MERGE_SORT)
7642	# define SQLITE_OMIT_MERGE_SORT
7643	#endif
7644
7645	/*
7646	** GCC does not define the offsetof() macro so we'll have to do it
7647	** ourselves.
7648	*/
7649	#ifndef offsetof
	@@ -7941,10 +7982,11 @@
7982	#define BTREE_OMIT_JOURNAL 1 /* Do not create or use a rollback journal */
7983	#define BTREE_NO_READLOCK 2 /* Omit readlocks on readonly files */
7984	#define BTREE_MEMORY 4 /* This is an in-memory DB */
7985	#define BTREE_SINGLE 8 /* The file contains at most 1 b-tree */
7986	#define BTREE_UNORDERED 16 /* Use of a hash implementation is OK */
7987	#define BTREE_SORTER 32 /* Used as accumulator in external merge sort */
7988
7989	SQLITE_PRIVATE int sqlite3BtreeClose(Btree*);
7990	SQLITE_PRIVATE int sqlite3BtreeSetCacheSize(Btree*,int);
7991	SQLITE_PRIVATE int sqlite3BtreeSetSafetyLevel(Btree*,int,int,int);
7992	SQLITE_PRIVATE int sqlite3BtreeSyncDisabled(Btree*);
	@@ -8143,10 +8185,11 @@
8185	** or VDBE. The VDBE implements an abstract machine that runs a
8186	** simple program to access and modify the underlying database.
8187	*/
8188	#ifndef _SQLITE_VDBE_H_
8189	#define _SQLITE_VDBE_H_
8190	/* #include <stdio.h> */
8191
8192	/*
8193	** A single VDBE is an opaque structure named "Vdbe". Only routines
8194	** in the source file sqliteVdbe.c are allowed to see the insides
8195	** of this structure.
	@@ -8186,10 +8229,11 @@
8229	Mem pMem; / Used when p4type is P4_MEM */
8230	VTable pVtab; / Used when p4type is P4_VTAB */
8231	KeyInfo pKeyInfo; / Used when p4type is P4_KEYINFO */
8232	int ai; / Used when p4type is P4_INTARRAY */
8233	SubProgram pProgram; / Used when p4type is P4_SUBPROGRAM */
8234	int (xAdvance)(BtCursor , int *);
8235	} p4;
8236	#ifdef SQLITE_DEBUG
8237	char zComment; / Comment to improve readability */
8238	#endif
8239	#ifdef VDBE_PROFILE
	@@ -8241,10 +8285,11 @@
8285	#define P4_REAL (-12) /* P4 is a 64-bit floating point value */
8286	#define P4_INT64 (-13) /* P4 is a 64-bit signed integer */
8287	#define P4_INT32 (-14) /* P4 is a 32-bit signed integer */
8288	#define P4_INTARRAY (-15) /* P4 is a vector of 32-bit integers */
8289	#define P4_SUBPROGRAM (-18) /* P4 is a pointer to a SubProgram structure */
8290	#define P4_ADVANCE (-19) /* P4 is a pointer to BtreeNext() or BtreePrev() */
8291
8292	/* When adding a P4 argument using P4_KEYINFO, a copy of the KeyInfo structure
8293	** is made. That copy is freed when the Vdbe is finalized. But if the
8294	** argument is P4_KEYINFO_HANDOFF, the passed in pointer is used. It still
8295	** gets freed when the Vdbe is finalized so it still should be obtained
	@@ -8354,89 +8399,89 @@
8399	#define OP_ReadCookie 35
8400	#define OP_SetCookie 36
8401	#define OP_VerifyCookie 37
8402	#define OP_OpenRead 38
8403	#define OP_OpenWrite 39
8404	#define OP_OpenSorter 40
8405	#define OP_OpenAutoindex 41
8406	#define OP_OpenEphemeral 42
8407	#define OP_OpenPseudo 43
8408	#define OP_Close 44
8409	#define OP_SeekLt 45
8410	#define OP_SeekLe 46
8411	#define OP_SeekGe 47
8412	#define OP_SeekGt 48
8413	#define OP_Seek 49
8414	#define OP_NotFound 50
8415	#define OP_Found 51
8416	#define OP_IsUnique 52
8417	#define OP_NotExists 53
8418	#define OP_Sequence 54
8419	#define OP_NewRowid 55
8420	#define OP_Insert 56
8421	#define OP_InsertInt 57
8422	#define OP_Delete 58
8423	#define OP_ResetCount 59
8424	#define OP_RowKey 60
8425	#define OP_RowData 61
8426	#define OP_Rowid 62
8427	#define OP_NullRow 63
8428	#define OP_Last 64
8429	#define OP_Sort 65
8430	#define OP_Rewind 66
8431	#define OP_Prev 67
8432	#define OP_Next 70
8433	#define OP_IdxInsert 71
8434	#define OP_IdxDelete 72
8435	#define OP_IdxRowid 81
8436	#define OP_IdxLT 92
8437	#define OP_IdxGE 95
8438	#define OP_Destroy 96
8439	#define OP_Clear 97
8440	#define OP_CreateIndex 98
8441	#define OP_CreateTable 99
8442	#define OP_ParseSchema 100
8443	#define OP_LoadAnalysis 101
8444	#define OP_DropTable 102
8445	#define OP_DropIndex 103
8446	#define OP_DropTrigger 104
8447	#define OP_IntegrityCk 105
8448	#define OP_RowSetAdd 106
8449	#define OP_RowSetRead 107
8450	#define OP_RowSetTest 108
8451	#define OP_Program 109
8452	#define OP_Param 110
8453	#define OP_FkCounter 111
8454	#define OP_FkIfZero 112
8455	#define OP_MemMax 113
8456	#define OP_IfPos 114
8457	#define OP_IfNeg 115
8458	#define OP_IfZero 116
8459	#define OP_AggStep 117
8460	#define OP_AggFinal 118
8461	#define OP_Checkpoint 119
8462	#define OP_JournalMode 120
8463	#define OP_Vacuum 121
8464	#define OP_IncrVacuum 122
8465	#define OP_Expire 123
8466	#define OP_TableLock 124
8467	#define OP_VBegin 125
8468	#define OP_VCreate 126
8469	#define OP_VDestroy 127
8470	#define OP_VOpen 128
8471	#define OP_VFilter 129
8472	#define OP_VColumn 131
8473	#define OP_VNext 132
8474	#define OP_VRename 133
8475	#define OP_VUpdate 134
8476	#define OP_Pagecount 135
8477	#define OP_MaxPgcnt 136
8478	#define OP_Trace 137
8479	#define OP_Noop 138
8480	#define OP_Explain 139
8481
8482	/* The following opcode values are never used */
8483	#define OP_NotUsed_140 140
8484
8485
8486	/* Properties such as "out2" or "jump" that are specified in
8487	** comments following the "case" for each opcode in the vdbe.c
	@@ -8453,23 +8498,23 @@
8498	/* 0 */ 0x00, 0x01, 0x05, 0x04, 0x04, 0x10, 0x00, 0x02,\
8499	/* 8 */ 0x02, 0x02, 0x02, 0x02, 0x02, 0x00, 0x24, 0x24,\
8500	/* 16 */ 0x00, 0x00, 0x00, 0x24, 0x04, 0x05, 0x04, 0x00,\
8501	/* 24 */ 0x00, 0x01, 0x05, 0x05, 0x00, 0x00, 0x00, 0x02,\
8502	/* 32 */ 0x00, 0x00, 0x00, 0x02, 0x10, 0x00, 0x00, 0x00,\
8503	/* 40 */ 0x00, 0x00, 0x00, 0x00, 0x00, 0x11, 0x11, 0x11,\
8504	/* 48 */ 0x11, 0x08, 0x11, 0x11, 0x11, 0x11, 0x02, 0x02,\
8505	/* 56 */ 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x02, 0x00,\
8506	/* 64 */ 0x01, 0x01, 0x01, 0x01, 0x4c, 0x4c, 0x01, 0x08,\
8507	/* 72 */ 0x00, 0x05, 0x05, 0x15, 0x15, 0x15, 0x15, 0x15,\
8508	/* 80 */ 0x15, 0x02, 0x4c, 0x4c, 0x4c, 0x4c, 0x4c, 0x4c,\
8509	/* 88 */ 0x4c, 0x4c, 0x4c, 0x4c, 0x01, 0x24, 0x02, 0x01,\
8510	/* 96 */ 0x02, 0x00, 0x02, 0x02, 0x00, 0x00, 0x00, 0x00,\
8511	/* 104 */ 0x00, 0x00, 0x0c, 0x45, 0x15, 0x01, 0x02, 0x00,\
8512	/* 112 */ 0x01, 0x08, 0x05, 0x05, 0x05, 0x00, 0x00, 0x00,\
8513	/* 120 */ 0x02, 0x00, 0x01, 0x00, 0x00, 0x00, 0x00, 0x00,\
8514	/* 128 */ 0x00, 0x01, 0x02, 0x00, 0x01, 0x00, 0x00, 0x02,\
8515	/* 136 */ 0x02, 0x00, 0x00, 0x00, 0x00, 0x04, 0x04, 0x04,\
8516	/* 144 */ 0x04, 0x04,}
8517
8518	/************ End of opcodes.h *******************************************/
8519	/************ Continuing where we left off in vdbe.h *********************/
8520
	@@ -8484,13 +8529,13 @@
8529	SQLITE_PRIVATE int sqlite3VdbeAddOp3(Vdbe*,int,int,int,int);
8530	SQLITE_PRIVATE int sqlite3VdbeAddOp4(Vdbe,int,int,int,int,const char zP4,int);
8531	SQLITE_PRIVATE int sqlite3VdbeAddOp4Int(Vdbe*,int,int,int,int,int);
8532	SQLITE_PRIVATE int sqlite3VdbeAddOpList(Vdbe, int nOp, VdbeOpList const aOp);
8533	SQLITE_PRIVATE void sqlite3VdbeAddParseSchemaOp(Vdbe,int,char);
8534	SQLITE_PRIVATE void sqlite3VdbeChangeP1(Vdbe*, u32 addr, int P1);
8535	SQLITE_PRIVATE void sqlite3VdbeChangeP2(Vdbe*, u32 addr, int P2);
8536	SQLITE_PRIVATE void sqlite3VdbeChangeP3(Vdbe*, u32 addr, int P3);
8537	SQLITE_PRIVATE void sqlite3VdbeChangeP5(Vdbe*, u8 P5);
8538	SQLITE_PRIVATE void sqlite3VdbeJumpHere(Vdbe*, int addr);
8539	SQLITE_PRIVATE void sqlite3VdbeChangeToNoop(Vdbe*, int addr, int N);
8540	SQLITE_PRIVATE void sqlite3VdbeChangeP4(Vdbe, int addr, const char zP4, int N);
8541	SQLITE_PRIVATE void sqlite3VdbeUsesBtree(Vdbe*, int);
	@@ -8608,10 +8653,11 @@
8653	** NOTE: These values must match the corresponding BTREE_ values in btree.h.
8654	*/
8655	#define PAGER_OMIT_JOURNAL 0x0001 /* Do not use a rollback journal */
8656	#define PAGER_NO_READLOCK 0x0002 /* Omit readlocks on readonly files */
8657	#define PAGER_MEMORY 0x0004 /* In-memory database */
8658	#define PAGER_SORTER 0x0020 /* Accumulator in external merge sort */
8659
8660	/*
8661	** Valid values for the second argument to sqlite3PagerLockingMode().
8662	*/
8663	#define PAGER_LOCKINGMODE_QUERY -1
	@@ -8703,10 +8749,13 @@
8749	SQLITE_PRIVATE sqlite3_file sqlite3PagerFile(Pager);
8750	SQLITE_PRIVATE const char sqlite3PagerJournalname(Pager);
8751	SQLITE_PRIVATE int sqlite3PagerNosync(Pager*);
8752	SQLITE_PRIVATE void sqlite3PagerTempSpace(Pager);
8753	SQLITE_PRIVATE int sqlite3PagerIsMemdb(Pager*);
8754	#ifndef SQLITE_OMIT_MERGE_SORT
8755	SQLITE_PRIVATE int sqlite3PagerUnderStress(Pager*);
8756	#endif
8757
8758	/* Functions used to truncate the database file. */
8759	SQLITE_PRIVATE void sqlite3PagerTruncateImage(Pager*,Pgno);
8760
8761	#if defined(SQLITE_HAS_CODEC) && !defined(SQLITE_OMIT_WAL)
	@@ -9556,10 +9605,12 @@
9605	#define SQLITE_IndexSort 0x04 /* Disable indexes for sorting */
9606	#define SQLITE_IndexSearch 0x08 /* Disable indexes for searching */
9607	#define SQLITE_IndexCover 0x10 /* Disable index covering table */
9608	#define SQLITE_GroupByOrder 0x20 /* Disable GROUPBY cover of ORDERBY */
9609	#define SQLITE_FactorOutConst 0x40 /* Disable factoring out constants */
9610	#define SQLITE_IdxRealAsInt 0x80 /* Store REAL as INT in indices */
9611	#define SQLITE_DistinctOpt 0x80 /* DISTINCT using indexes */
9612	#define SQLITE_OptMask 0xff /* Mask of all disablable opts */
9613
9614	/*
9615	** Possible values for the sqlite.magic field.
9616	** The numbers are obtained at random and have no special meaning, other
	@@ -10447,10 +10498,11 @@
10498	Table pTab; / An SQL table corresponding to zName */
10499	Select pSelect; / A SELECT statement used in place of a table name */
10500	u8 isPopulated; /* Temporary table associated with SELECT is populated */
10501	u8 jointype; /* Type of join between this able and the previous */
10502	u8 notIndexed; /* True if there is a NOT INDEXED clause */
10503	u8 isCorrelated; /* True if sub-query is correlated */
10504	#ifndef SQLITE_OMIT_EXPLAIN
10505	u8 iSelectId; /* If pSelect!=0, the id of the sub-select in EQP */
10506	#endif
10507	int iCursor; /* The VDBE cursor number used to access this table */
10508	Expr pOn; / The ON clause of a join */
	@@ -10566,10 +10618,11 @@
10618	struct WhereInfo {
10619	Parse pParse; / Parsing and code generating context */
10620	u16 wctrlFlags; /* Flags originally passed to sqlite3WhereBegin() */
10621	u8 okOnePass; /* Ok to use one-pass algorithm for UPDATE or DELETE */
10622	u8 untestedTerms; /* Not all WHERE terms resolved by outer loop */
10623	u8 eDistinct;
10624	SrcList pTabList; / List of tables in the join */
10625	int iTop; /* The very beginning of the WHERE loop */
10626	int iContinue; /* Jump here to continue with next record */
10627	int iBreak; /* Jump here to break out of the loop */
10628	int nLevel; /* Number of nested loop */
	@@ -10577,10 +10630,13 @@
10630	double savedNQueryLoop; /* pParse->nQueryLoop outside the WHERE loop */
10631	double nRowOut; /* Estimated number of output rows */
10632	WhereLevel a[1]; /* Information about each nest loop in WHERE */
10633	};
10634
10635	#define WHERE_DISTINCT_UNIQUE 1
10636	#define WHERE_DISTINCT_ORDERED 2
10637
10638	/*
10639	** A NameContext defines a context in which to resolve table and column
10640	** names. The context consists of a list of tables (the pSrcList) field and
10641	** a list of named expression (pEList). The named expression list may
10642	** be NULL. The pSrc corresponds to the FROM clause of a SELECT or
	@@ -11053,10 +11109,11 @@
11109	int isPCacheInit; /* True after malloc is initialized */
11110	sqlite3_mutex pInitMutex; / Mutex used by sqlite3_initialize() */
11111	int nRefInitMutex; /* Number of users of pInitMutex */
11112	void (xLog)(void,int,const char); / Function for logging */
11113	void pLogArg; / First argument to xLog() */
11114	int bLocaltimeFault; /* True to fail localtime() calls */
11115	};
11116
11117	/*
11118	** Context pointer passed down through the tree-walk.
11119	*/
	@@ -11337,11 +11394,11 @@
11394	#if defined(SQLITE_ENABLE_UPDATE_DELETE_LIMIT) && !defined(SQLITE_OMIT_SUBQUERY)
11395	SQLITE_PRIVATE Expr sqlite3LimitWhere(Parse , SrcList , Expr , ExprList , Expr , Expr , char );
11396	#endif
11397	SQLITE_PRIVATE void sqlite3DeleteFrom(Parse, SrcList, Expr*);
11398	SQLITE_PRIVATE void sqlite3Update(Parse, SrcList, ExprList, Expr, int);
11399	SQLITE_PRIVATE WhereInfo sqlite3WhereBegin(Parse, SrcList, Expr, ExprList*,ExprList,u16);
11400	SQLITE_PRIVATE void sqlite3WhereEnd(WhereInfo*);
11401	SQLITE_PRIVATE int sqlite3ExprCodeGetColumn(Parse, Table, int, int, int);
11402	SQLITE_PRIVATE void sqlite3ExprCodeGetColumnOfTable(Vdbe, Table, int, int, int);
11403	SQLITE_PRIVATE void sqlite3ExprCodeMove(Parse*, int, int, int);
11404	SQLITE_PRIVATE void sqlite3ExprCodeCopy(Parse*, int, int, int);
	@@ -12013,10 +12070,11 @@
12070	0, /* isPCacheInit */
12071	0, /* pInitMutex */
12072	0, /* nRefInitMutex */
12073	0, /* xLog */
12074	0, /* pLogArg */
12075	0, /* bLocaltimeFault */
12076	};
12077
12078
12079	/*
12080	** Hash table for global functions - functions common to all
	@@ -12321,10 +12379,13 @@
12379	#ifdef SQLITE_OMIT_LOOKASIDE
12380	"OMIT_LOOKASIDE",
12381	#endif
12382	#ifdef SQLITE_OMIT_MEMORYDB
12383	"OMIT_MEMORYDB",
12384	#endif
12385	#ifdef SQLITE_OMIT_MERGE_SORT
12386	"OMIT_MERGE_SORT",
12387	#endif
12388	#ifdef SQLITE_OMIT_OR_OPTIMIZATION
12389	"OMIT_OR_OPTIMIZATION",
12390	#endif
12391	#ifdef SQLITE_OMIT_PAGER_PRAGMAS
	@@ -12387,10 +12448,13 @@
12448	#ifdef SQLITE_OMIT_WSD
12449	"OMIT_WSD",
12450	#endif
12451	#ifdef SQLITE_OMIT_XFER_OPT
12452	"OMIT_XFER_OPT",
12453	#endif
12454	#ifdef SQLITE_PAGECACHE_BLOCKALLOC
12455	"PAGECACHE_BLOCKALLOC",
12456	#endif
12457	#ifdef SQLITE_PERFORMANCE_TRACE
12458	"PERFORMANCE_TRACE",
12459	#endif
12460	#ifdef SQLITE_PROXY_DEBUG
	@@ -12508,10 +12572,13 @@
12572	/*
12573	** Boolean values
12574	*/
12575	typedef unsigned char Bool;
12576
12577	/* Opaque type used by code in vdbesort.c */
12578	typedef struct VdbeSorter VdbeSorter;
12579
12580	/*
12581	** A cursor is a pointer into a single BTree within a database file.
12582	** The cursor can seek to a BTree entry with a particular key, or
12583	** loop over all entries of the Btree. You can also insert new BTree
12584	** entries or retrieve the key or data from the entry that the cursor
	@@ -12539,10 +12606,11 @@
12606	sqlite3_vtab_cursor pVtabCursor; / The cursor for a virtual table */
12607	const sqlite3_module pModule; / Module for cursor pVtabCursor */
12608	i64 seqCount; /* Sequence counter */
12609	i64 movetoTarget; /* Argument to the deferred sqlite3BtreeMoveto() */
12610	i64 lastRowid; /* Last rowid from a Next or NextIdx operation */
12611	VdbeSorter pSorter; / Sorter object for OP_OpenSorter cursors */
12612
12613	/* Result of last sqlite3BtreeMoveto() done by an OP_NotExists or
12614	** OP_IsUnique opcode on this cursor. */
12615	int seekResult;
12616
	@@ -12858,17 +12926,36 @@
12926	SQLITE_PRIVATE int sqlite3VdbeMemRealify(Mem*);
12927	SQLITE_PRIVATE int sqlite3VdbeMemNumerify(Mem*);
12928	SQLITE_PRIVATE int sqlite3VdbeMemFromBtree(BtCursor,int,int,int,Mem);
12929	SQLITE_PRIVATE void sqlite3VdbeMemRelease(Mem *p);
12930	SQLITE_PRIVATE void sqlite3VdbeMemReleaseExternal(Mem *p);
12931	#define MemReleaseExt(X) \
12932	if((X)->flags&(MEM_Agg\|MEM_Dyn\|MEM_RowSet\|MEM_Frame)) \
12933	sqlite3VdbeMemReleaseExternal(X);
12934	SQLITE_PRIVATE int sqlite3VdbeMemFinalize(Mem, FuncDef);
12935	SQLITE_PRIVATE const char *sqlite3OpcodeName(int);
12936	SQLITE_PRIVATE int sqlite3VdbeMemGrow(Mem *pMem, int n, int preserve);
12937	SQLITE_PRIVATE int sqlite3VdbeCloseStatement(Vdbe *, int);
12938	SQLITE_PRIVATE void sqlite3VdbeFrameDelete(VdbeFrame*);
12939	SQLITE_PRIVATE int sqlite3VdbeFrameRestore(VdbeFrame *);
12940	SQLITE_PRIVATE void sqlite3VdbeMemStoreType(Mem *pMem);
12941
12942	#ifdef SQLITE_OMIT_MERGE_SORT
12943	# define sqlite3VdbeSorterInit(Y,Z) SQLITE_OK
12944	# define sqlite3VdbeSorterWrite(X,Y,Z) SQLITE_OK
12945	# define sqlite3VdbeSorterClose(Y,Z)
12946	# define sqlite3VdbeSorterRowkey(Y,Z) SQLITE_OK
12947	# define sqlite3VdbeSorterRewind(X,Y,Z) SQLITE_OK
12948	# define sqlite3VdbeSorterNext(X,Y,Z) SQLITE_OK
12949	#else
12950	SQLITE_PRIVATE int sqlite3VdbeSorterInit(sqlite3 , VdbeCursor );
12951	SQLITE_PRIVATE int sqlite3VdbeSorterWrite(sqlite3 , VdbeCursor , int);
12952	SQLITE_PRIVATE void sqlite3VdbeSorterClose(sqlite3 , VdbeCursor );
12953	SQLITE_PRIVATE int sqlite3VdbeSorterRowkey(VdbeCursor , Mem );
12954	SQLITE_PRIVATE int sqlite3VdbeSorterRewind(sqlite3 , VdbeCursor , int *);
12955	SQLITE_PRIVATE int sqlite3VdbeSorterNext(sqlite3 , VdbeCursor , int *);
12956	#endif
12957
12958	#if !defined(SQLITE_OMIT_SHARED_CACHE) && SQLITE_THREADSAFE>0
12959	SQLITE_PRIVATE void sqlite3VdbeEnter(Vdbe*);
12960	SQLITE_PRIVATE void sqlite3VdbeLeave(Vdbe*);
12961	#else
	@@ -13159,30 +13246,16 @@
13246	** Astronomical Algorithms, 2nd Edition, 1998
13247	** ISBM 0-943396-61-1
13248	** Willmann-Bell, Inc
13249	** Richmond, Virginia (USA)
13250	*/
13251	/* #include <stdlib.h> */
13252	/* #include <assert.h> */
13253	#include <time.h>
13254
13255	#ifndef SQLITE_OMIT_DATETIME_FUNCS
13256
















13257
13258	/*
13259	** A structure for holding a single date and time.
13260	*/
13261	typedef struct DateTime DateTime;
	@@ -13523,20 +13596,90 @@
13596	static void clearYMD_HMS_TZ(DateTime *p){
13597	p->validYMD = 0;
13598	p->validHMS = 0;
13599	p->validTZ = 0;
13600	}
13601
13602	/*
13603	** On recent Windows platforms, the localtime_s() function is available
13604	** as part of the "Secure CRT". It is essentially equivalent to
13605	** localtime_r() available under most POSIX platforms, except that the
13606	** order of the parameters is reversed.
13607	**
13608	** See http://msdn.microsoft.com/en-us/library/a442x3ye(VS.80).aspx.
13609	**
13610	** If the user has not indicated to use localtime_r() or localtime_s()
13611	** already, check for an MSVC build environment that provides
13612	** localtime_s().
13613	*/
13614	#if !defined(HAVE_LOCALTIME_R) && !defined(HAVE_LOCALTIME_S) && \
13615	defined(_MSC_VER) && defined(_CRT_INSECURE_DEPRECATE)
13616	#define HAVE_LOCALTIME_S 1
13617	#endif
13618
13619	#ifndef SQLITE_OMIT_LOCALTIME
13620	/*
13621	** The following routine implements the rough equivalent of localtime_r()
13622	** using whatever operating-system specific localtime facility that
13623	** is available. This routine returns 0 on success and
13624	** non-zero on any kind of error.
13625	**
13626	** If the sqlite3GlobalConfig.bLocaltimeFault variable is true then this
13627	** routine will always fail.
13628	*/
13629	static int osLocaltime(time_t t, struct tm pTm){
13630	int rc;
13631	#if (!defined(HAVE_LOCALTIME_R) \|\| !HAVE_LOCALTIME_R) \
13632	&& (!defined(HAVE_LOCALTIME_S) \|\| !HAVE_LOCALTIME_S)
13633	struct tm *pX;
13634	#if SQLITE_THREADSAFE>0
13635	sqlite3_mutex *mutex = sqlite3MutexAlloc(SQLITE_MUTEX_STATIC_MASTER);
13636	#endif
13637	sqlite3_mutex_enter(mutex);
13638	pX = localtime(t);
13639	#ifndef SQLITE_OMIT_BUILTIN_TEST
13640	if( sqlite3GlobalConfig.bLocaltimeFault ) pX = 0;
13641	#endif
13642	if( pX ) pTm = pX;
13643	sqlite3_mutex_leave(mutex);
13644	rc = pX==0;
13645	#else
13646	#ifndef SQLITE_OMIT_BUILTIN_TEST
13647	if( sqlite3GlobalConfig.bLocaltimeFault ) return 1;
13648	#endif
13649	#if defined(HAVE_LOCALTIME_R) && HAVE_LOCALTIME_R
13650	rc = localtime_r(t, pTm)==0;
13651	#else
13652	rc = localtime_s(pTm, t);
13653	#endif /* HAVE_LOCALTIME_R */
13654	#endif /* HAVE_LOCALTIME_R \|\| HAVE_LOCALTIME_S */
13655	return rc;
13656	}
13657	#endif /* SQLITE_OMIT_LOCALTIME */
13658
13659
13660	#ifndef SQLITE_OMIT_LOCALTIME
13661	/*
13662	** Compute the difference (in milliseconds) between localtime and UTC
13663	** (a.k.a. GMT) for the time value p where p is in UTC. If no error occurs,
13664	** return this value and set *pRc to SQLITE_OK.
13665	**
13666	** Or, if an error does occur, set *pRc to SQLITE_ERROR. The returned value
13667	** is undefined in this case.
13668	*/
13669	static sqlite3_int64 localtimeOffset(
13670	DateTime p, / Date at which to calculate offset */
13671	sqlite3_context pCtx, / Write error here if one occurs */
13672	int pRc / OUT: Error code. SQLITE_OK or ERROR */
13673	){
13674	DateTime x, y;
13675	time_t t;
13676	struct tm sLocal;
13677
13678	/* Initialize the contents of sLocal to avoid a compiler warning. */
13679	memset(&sLocal, 0, sizeof(sLocal));
13680
13681	x = *p;
13682	computeYMD_HMS(&x);
13683	if( x.Y<1971 \|\| x.Y>=2038 ){
13684	x.Y = 2000;
13685	x.M = 1;
	@@ -13550,51 +13693,27 @@
13693	}
13694	x.tz = 0;
13695	x.validJD = 0;
13696	computeJD(&x);
13697	t = (time_t)(x.iJD/1000 - 21086676*(i64)10000);
13698	if( osLocaltime(&t, &sLocal) ){
13699	sqlite3_result_error(pCtx, "local time unavailable", -1);
13700	*pRc = SQLITE_ERROR;
13701	return 0;
13702	}
13703	y.Y = sLocal.tm_year + 1900;
13704	y.M = sLocal.tm_mon + 1;
13705	y.D = sLocal.tm_mday;
13706	y.h = sLocal.tm_hour;
13707	y.m = sLocal.tm_min;
13708	y.s = sLocal.tm_sec;

























13709	y.validYMD = 1;
13710	y.validHMS = 1;
13711	y.validJD = 0;
13712	y.validTZ = 0;
13713	computeJD(&y);
13714	*pRc = SQLITE_OK;
13715	return y.iJD - x.iJD;
13716	}
13717	#endif /* SQLITE_OMIT_LOCALTIME */
13718
13719	/*
	@@ -13614,13 +13733,16 @@
13733	** weekday N
13734	** unixepoch
13735	** localtime
13736	** utc
13737	**
13738	** Return 0 on success and 1 if there is any kind of error. If the error
13739	** is in a system call (i.e. localtime()), then an error message is written
13740	** to context pCtx. If the error is an unrecognized modifier, no error is
13741	** written to pCtx.
13742	*/
13743	static int parseModifier(sqlite3_context pCtx, const char zMod, DateTime *p){
13744	int rc = 1;
13745	int n;
13746	double r;
13747	char *z, zBuf[30];
13748	z = zBuf;
	@@ -13636,13 +13758,12 @@
13758	** Assuming the current time value is UTC (a.k.a. GMT), shift it to
13759	** show local time.
13760	*/
13761	if( strcmp(z, "localtime")==0 ){
13762	computeJD(p);
13763	p->iJD += localtimeOffset(p, pCtx, &rc);
13764	clearYMD_HMS_TZ(p);

13765	}
13766	break;
13767	}
13768	#endif
13769	case 'u': {
	@@ -13659,15 +13780,16 @@
13780	}
13781	#ifndef SQLITE_OMIT_LOCALTIME
13782	else if( strcmp(z, "utc")==0 ){
13783	sqlite3_int64 c1;
13784	computeJD(p);
13785	c1 = localtimeOffset(p, pCtx, &rc);
13786	if( rc==SQLITE_OK ){
13787	p->iJD -= c1;
13788	clearYMD_HMS_TZ(p);
13789	p->iJD += c1 - localtimeOffset(p, pCtx, &rc);
13790	}
13791	}
13792	#endif
13793	break;
13794	}
13795	case 'w': {
	@@ -13844,13 +13966,12 @@
13966	if( !z \|\| parseDateOrTime(context, (char*)z, p) ){
13967	return 1;
13968	}
13969	}
13970	for(i=1; i<argc; i++){
13971	z = sqlite3_value_text(argv[i]);
13972	if( z==0 \|\| parseModifier(context, (char*)z, p) ) return 1;

13973	}
13974	return 0;
13975	}
13976
13977
	@@ -14344,11 +14465,11 @@
14465	DO_OS_MALLOC_TEST(0);
14466	/* 0x87f3f is a mask of SQLITE_OPEN_ flags that are valid to be passed
14467	** down into the VFS layer. Some SQLITE_OPEN_ flags (for example,
14468	** SQLITE_OPEN_FULLMUTEX or SQLITE_OPEN_SHAREDCACHE) are blocked before
14469	** reaching the VFS. */
14470	rc = pVfs->xOpen(pVfs, zPath, pFile, flags & 0x87f7f, pFlagsOut);
14471	assert( rc==SQLITE_OK \|\| pFile->pMethods==0 );
14472	return rc;
14473	}
14474	SQLITE_PRIVATE int sqlite3OsDelete(sqlite3_vfs pVfs, const char zPath, int dirSync){
14475	return pVfs->xDelete(pVfs, zPath, dirSync);
	@@ -14880,10 +15001,11 @@
15001	extern void backtrace_symbols_fd(voidconst,int,int);
15002	#else
15003	# define backtrace(A,B) 1
15004	# define backtrace_symbols_fd(A,B,C)
15005	#endif
15006	/* #include <stdio.h> */
15007
15008	/*
15009	** Each memory allocation looks like this:
15010	**
15011	** ------------------------------------------------------------------------
	@@ -15805,11 +15927,11 @@
15927	** Free an outstanding memory allocation.
15928	**
15929	** This function assumes that the necessary mutexes, if any, are
15930	** already held by the caller. Hence "Unsafe".
15931	*/
15932	static void memsys3FreeUnsafe(void *pOld){
15933	Mem3Block p = (Mem3Block)pOld;
15934	int i;
15935	u32 size, x;
15936	assert( sqlite3_mutex_held(mem3.mutex) );
15937	assert( p>mem3.aPool && p<&mem3.aPool[mem3.nPool] );
	@@ -15880,21 +16002,21 @@
16002	}
16003
16004	/*
16005	** Free memory.
16006	*/
16007	static void memsys3Free(void *pPrior){
16008	assert( pPrior );
16009	memsys3Enter();
16010	memsys3FreeUnsafe(pPrior);
16011	memsys3Leave();
16012	}
16013
16014	/*
16015	** Change the size of an existing memory allocation
16016	*/
16017	static void memsys3Realloc(void pPrior, int nBytes){
16018	int nOld;
16019	void *p;
16020	if( pPrior==0 ){
16021	return sqlite3_malloc(nBytes);
16022	}
	@@ -17983,10 +18105,11 @@
18105	**
18106	*************************************************************************
18107	**
18108	** Memory allocation functions used throughout sqlite.
18109	*/
18110	/* #include <stdarg.h> */
18111
18112	/*
18113	** Attempt to release up to n bytes of non-essential memory currently
18114	** held by SQLite. An example of non-essential memory is memory used to
18115	** cache database pages that are not currently in use.
	@@ -19960,10 +20083,11 @@
20083	** BOM or Byte Order Mark:
20084	** 0xff 0xfe little-endian utf-16 follows
20085	** 0xfe 0xff big-endian utf-16 follows
20086	**
20087	*/
20088	/* #include <assert.h> */
20089
20090	#ifndef SQLITE_AMALGAMATION
20091	/*
20092	** The following constant value is used by the SQLITE_BIGENDIAN and
20093	** SQLITE_LITTLEENDIAN macros.
	@@ -20502,10 +20626,11 @@
20626	**
20627	** This file contains functions for allocating memory, comparing
20628	** strings, and stuff like that.
20629	**
20630	*/
20631	/* #include <stdarg.h> */
20632	#ifdef SQLITE_HAVE_ISNAN
20633	# include <math.h>
20634	#endif
20635
20636	/*
	@@ -21634,26 +21759,32 @@
21759	return -x;
21760	}
21761
21762	#ifdef SQLITE_ENABLE_8_3_NAMES
21763	/*
21764	** If SQLITE_ENABLE_8_3_NAMES is set at compile-time and if the database
21765	** filename in zBaseFilename is a URI with the "8_3_names=1" parameter and
21766	** if filename in z[] has a suffix (a.k.a. "extension") that is longer than
21767	** three characters, then shorten the suffix on z[] to be the last three
21768	** characters of the original suffix.
21769	**
21770	** If SQLITE_ENABLE_8_3_NAMES is set to 2 at compile-time, then always
21771	** do the suffix shortening regardless of URI parameter.
21772	**
21773	** Examples:
21774	**
21775	** test.db-journal => test.nal
21776	** test.db-wal => test.wal
21777	** test.db-shm => test.shm
21778	*/
21779	SQLITE_PRIVATE void sqlite3FileSuffix3(const char zBaseFilename, char z){
21780	#if SQLITE_ENABLE_8_3_NAMES<2
21781	const char *zOk;
21782	zOk = sqlite3_uri_parameter(zBaseFilename, "8_3_names");
21783	if( zOk && sqlite3GetBoolean(zOk) )
21784	#endif
21785	{
21786	int i, sz;
21787	sz = sqlite3Strlen30(z);
21788	for(i=sz-1; i>0 && z[i]!='/' && z[i]!='.'; i--){}
21789	if( z[i]=='.' && ALWAYS(sz>i+4) ) memcpy(&z[i+1], &z[sz-3], 4);
21790	}
	@@ -21674,10 +21805,11 @@
21805	**
21806	*************************************************************************
21807	** This is the implementation of generic hash-tables
21808	** used in SQLite.
21809	*/
21810	/* #include <assert.h> */
21811
21812	/* Turn bulk memory into a hash table object by initializing the
21813	** fields of the Hash structure.
21814	**
21815	** "pNew" is a pointer to the hash table that is to be initialized.
	@@ -21982,52 +22114,52 @@
22114	/* 35 */ "ReadCookie",
22115	/* 36 */ "SetCookie",
22116	/* 37 */ "VerifyCookie",
22117	/* 38 */ "OpenRead",
22118	/* 39 */ "OpenWrite",
22119	/* 40 */ "OpenSorter",
22120	/* 41 */ "OpenAutoindex",
22121	/* 42 */ "OpenEphemeral",
22122	/* 43 */ "OpenPseudo",
22123	/* 44 */ "Close",
22124	/* 45 */ "SeekLt",
22125	/* 46 */ "SeekLe",
22126	/* 47 */ "SeekGe",
22127	/* 48 */ "SeekGt",
22128	/* 49 */ "Seek",
22129	/* 50 */ "NotFound",
22130	/* 51 */ "Found",
22131	/* 52 */ "IsUnique",
22132	/* 53 */ "NotExists",
22133	/* 54 */ "Sequence",
22134	/* 55 */ "NewRowid",
22135	/* 56 */ "Insert",
22136	/* 57 */ "InsertInt",
22137	/* 58 */ "Delete",
22138	/* 59 */ "ResetCount",
22139	/* 60 */ "RowKey",
22140	/* 61 */ "RowData",
22141	/* 62 */ "Rowid",
22142	/* 63 */ "NullRow",
22143	/* 64 */ "Last",
22144	/* 65 */ "Sort",
22145	/* 66 */ "Rewind",
22146	/* 67 */ "Prev",
22147	/* 68 */ "Or",
22148	/* 69 */ "And",
22149	/* 70 */ "Next",
22150	/* 71 */ "IdxInsert",
22151	/* 72 */ "IdxDelete",
22152	/* 73 */ "IsNull",
22153	/* 74 */ "NotNull",
22154	/* 75 */ "Ne",
22155	/* 76 */ "Eq",
22156	/* 77 */ "Gt",
22157	/* 78 */ "Le",
22158	/* 79 */ "Lt",
22159	/* 80 */ "Ge",
22160	/* 81 */ "IdxRowid",
22161	/* 82 */ "BitAnd",
22162	/* 83 */ "BitOr",
22163	/* 84 */ "ShiftLeft",
22164	/* 85 */ "ShiftRight",
22165	/* 86 */ "Add",
	@@ -22034,58 +22166,58 @@
22166	/* 87 */ "Subtract",
22167	/* 88 */ "Multiply",
22168	/* 89 */ "Divide",
22169	/* 90 */ "Remainder",
22170	/* 91 */ "Concat",
22171	/* 92 */ "IdxLT",
22172	/* 93 */ "BitNot",
22173	/* 94 */ "String8",
22174	/* 95 */ "IdxGE",
22175	/* 96 */ "Destroy",
22176	/* 97 */ "Clear",
22177	/* 98 */ "CreateIndex",
22178	/* 99 */ "CreateTable",
22179	/* 100 */ "ParseSchema",
22180	/* 101 */ "LoadAnalysis",
22181	/* 102 */ "DropTable",
22182	/* 103 */ "DropIndex",
22183	/* 104 */ "DropTrigger",
22184	/* 105 */ "IntegrityCk",
22185	/* 106 */ "RowSetAdd",
22186	/* 107 */ "RowSetRead",
22187	/* 108 */ "RowSetTest",
22188	/* 109 */ "Program",
22189	/* 110 */ "Param",
22190	/* 111 */ "FkCounter",
22191	/* 112 */ "FkIfZero",
22192	/* 113 */ "MemMax",
22193	/* 114 */ "IfPos",
22194	/* 115 */ "IfNeg",
22195	/* 116 */ "IfZero",
22196	/* 117 */ "AggStep",
22197	/* 118 */ "AggFinal",
22198	/* 119 */ "Checkpoint",
22199	/* 120 */ "JournalMode",
22200	/* 121 */ "Vacuum",
22201	/* 122 */ "IncrVacuum",
22202	/* 123 */ "Expire",
22203	/* 124 */ "TableLock",
22204	/* 125 */ "VBegin",
22205	/* 126 */ "VCreate",
22206	/* 127 */ "VDestroy",
22207	/* 128 */ "VOpen",
22208	/* 129 */ "VFilter",
22209	/* 130 */ "Real",
22210	/* 131 */ "VColumn",
22211	/* 132 */ "VNext",
22212	/* 133 */ "VRename",
22213	/* 134 */ "VUpdate",
22214	/* 135 */ "Pagecount",
22215	/* 136 */ "MaxPgcnt",
22216	/* 137 */ "Trace",
22217	/* 138 */ "Noop",
22218	/* 139 */ "Explain",
22219	/* 140 */ "NotUsed_140",
22220	/* 141 */ "ToText",
22221	/* 142 */ "ToBlob",
22222	/* 143 */ "ToNumeric",
22223	/* 144 */ "ToInt",
	@@ -22183,14 +22315,17 @@
22315	#ifdef MEMORY_DEBUG
22316	# error "The MEMORY_DEBUG macro is obsolete. Use SQLITE_DEBUG instead."
22317	#endif
22318
22319	#ifdef SQLITE_DEBUG
22320	# ifndef SQLITE_DEBUG_OS_TRACE
22321	# define SQLITE_DEBUG_OS_TRACE 0
22322	# endif
22323	int sqlite3OSTrace = SQLITE_DEBUG_OS_TRACE;
22324	# define OSTRACE(X) if( sqlite3OSTrace ) sqlite3DebugPrintf X
22325	#else
22326	# define OSTRACE(X)
22327	#endif
22328
22329	/*
22330	** Macros for performance tracing. Normally turned off. Only works
22331	** on i486 hardware.
	@@ -24343,10 +24478,11 @@
24478	*/
24479	#include <sys/types.h>
24480	#include <sys/stat.h>
24481	#include <fcntl.h>
24482	#include <unistd.h>
24483	/* #include <time.h> */
24484	#include <sys/time.h>
24485	#include <errno.h>
24486	#ifndef SQLITE_OMIT_WAL
24487	#include <sys/mman.h>
24488	#endif
	@@ -24364,10 +24500,14 @@
24500
24501	#if defined(__APPLE__) \|\| (SQLITE_ENABLE_LOCKING_STYLE && !OS_VXWORKS)
24502	# include <sys/mount.h>
24503	#endif
24504
24505	#ifdef HAVE_UTIME
24506	# include <utime.h>
24507	#endif
24508
24509	/*
24510	** Allowed values of unixFile.fsFlags
24511	*/
24512	#define SQLITE_FSFLAGS_IS_MSDOS 0x1
24513
	@@ -24374,10 +24514,11 @@
24514	/*
24515	** If we are to be thread-safe, include the pthreads header and define
24516	** the SQLITE_UNIX_THREADS macro.
24517	*/
24518	#if SQLITE_THREADSAFE
24519	/* # include <pthread.h> */
24520	# define SQLITE_UNIX_THREADS 1
24521	#endif
24522
24523	/*
24524	** Default permissions when creating a new file
	@@ -24429,11 +24570,10 @@
24570	typedef struct unixFile unixFile;
24571	struct unixFile {
24572	sqlite3_io_methods const pMethod; / Always the first entry */
24573	unixInodeInfo pInode; / Info about locks on this inode */
24574	int h; /* The file descriptor */

24575	unsigned char eFileLock; /* The type of lock held on this fd */
24576	unsigned char ctrlFlags; /* Behavioral bits. UNIXFILE_* flags */
24577	int lastErrno; /* The unix errno from last I/O error */
24578	void lockingContext; / Locking style specific state */
24579	UnixUnusedFd pUnused; / Pre-allocated UnixUnusedFd */
	@@ -24471,12 +24611,18 @@
24611	};
24612
24613	/*
24614	** Allowed values for the unixFile.ctrlFlags bitmask:
24615	*/
24616	#define UNIXFILE_EXCL 0x01 /* Connections from one process only */
24617	#define UNIXFILE_RDONLY 0x02 /* Connection is read only */
24618	#define UNIXFILE_PERSIST_WAL 0x04 /* Persistent WAL mode */
24619	#ifndef SQLITE_DISABLE_DIRSYNC
24620	# define UNIXFILE_DIRSYNC 0x08 /* Directory sync needed */
24621	#else
24622	# define UNIXFILE_DIRSYNC 0x00
24623	#endif
24624
24625	/*
24626	** Include code that is common to all os_*.c files
24627	*/
24628	/************ Include os_common.h in the middle of os_unix.c *************/
	@@ -24511,14 +24657,17 @@
24657	#ifdef MEMORY_DEBUG
24658	# error "The MEMORY_DEBUG macro is obsolete. Use SQLITE_DEBUG instead."
24659	#endif
24660
24661	#ifdef SQLITE_DEBUG
24662	# ifndef SQLITE_DEBUG_OS_TRACE
24663	# define SQLITE_DEBUG_OS_TRACE 0
24664	# endif
24665	int sqlite3OSTrace = SQLITE_DEBUG_OS_TRACE;
24666	# define OSTRACE(X) if( sqlite3OSTrace ) sqlite3DebugPrintf X
24667	#else
24668	# define OSTRACE(X)
24669	#endif
24670
24671	/*
24672	** Macros for performance tracing. Normally turned off. Only works
24673	** on i486 hardware.
	@@ -24723,10 +24872,13 @@
24872	*/
24873	static int posixOpen(const char *zFile, int flags, int mode){
24874	return open(zFile, flags, mode);
24875	}
24876
24877	/* Forward reference */
24878	static int openDirectory(const char, int);
24879
24880	/*
24881	** Many system calls are accessed through pointer-to-functions so that
24882	** they may be overridden at runtime to facilitate fault injection during
24883	** testing and sandboxing. The following array holds the names and pointers
24884	** to all overrideable system calls.
	@@ -24819,10 +24971,16 @@
24971	#else
24972	{ "fallocate", (sqlite3_syscall_ptr)0, 0 },
24973	#endif
24974	#define osFallocate ((int(*)(int,off_t,off_t))aSyscall[15].pCurrent)
24975
24976	{ "unlink", (sqlite3_syscall_ptr)unlink, 0 },
24977	#define osUnlink ((int()(const char))aSyscall[16].pCurrent)
24978
24979	{ "openDirectory", (sqlite3_syscall_ptr)openDirectory, 0 },
24980	#define osOpenDirectory ((int()(const char,int*))aSyscall[17].pCurrent)
24981
24982	}; /* End of the overrideable system calls */
24983
24984	/*
24985	** This is the xSetSystemCall() method of sqlite3_vfs for all of the
24986	** "unix" VFSes. Return SQLITE_OK opon successfully updating the
	@@ -25103,11 +25261,13 @@
25261	case EINVAL:
25262	case ENOTCONN:
25263	case ENODEV:
25264	case ENXIO:
25265	case ENOENT:
25266	#ifdef ESTALE /* ESTALE is not defined on Interix systems */
25267	case ESTALE:
25268	#endif
25269	case ENOSYS:
25270	/* these should force the client to close the file and reconnect */
25271
25272	default:
25273	return sqliteIOErr;
	@@ -26173,22 +26333,18 @@
26333	** even on VxWorks. A mutex will be acquired on VxWorks by the
26334	** vxworksReleaseFileId() routine.
26335	*/
26336	static int closeUnixFile(sqlite3_file *id){
26337	unixFile pFile = (unixFile)id;




26338	if( pFile->h>=0 ){
26339	robust_close(pFile, pFile->h, __LINE__);
26340	pFile->h = -1;
26341	}
26342	#if OS_VXWORKS
26343	if( pFile->pId ){
26344	if( pFile->isDelete ){
26345	osUnlink(pFile->pId->zCanonicalName);
26346	}
26347	vxworksReleaseFileId(pFile->pId);
26348	pFile->pId = 0;
26349	}
26350	#endif
	@@ -26369,12 +26525,14 @@
26525	/* If we have any lock, then the lock file already exists. All we have
26526	** to do is adjust our internal record of the lock level.
26527	*/
26528	if( pFile->eFileLock > NO_LOCK ){
26529	pFile->eFileLock = eFileLock;

26530	/* Always update the timestamp on the old file */
26531	#ifdef HAVE_UTIME
26532	utime(zLockFile, NULL);
26533	#else
26534	utimes(zLockFile, NULL);
26535	#endif
26536	return SQLITE_OK;
26537	}
26538
	@@ -26431,11 +26589,11 @@
26589	return SQLITE_OK;
26590	}
26591
26592	/* To fully unlock the database, delete the lock file */
26593	assert( eFileLock==NO_LOCK );
26594	if( osUnlink(zLockFile) ){
26595	int rc = 0;
26596	int tErrno = errno;
26597	if( ENOENT != tErrno ){
26598	rc = SQLITE_IOERR_UNLOCK;
26599	}
	@@ -26937,15 +27095,16 @@
27095	*/
27096	static int afpCheckReservedLock(sqlite3_file id, int pResOut){
27097	int rc = SQLITE_OK;
27098	int reserved = 0;
27099	unixFile pFile = (unixFile)id;
27100	afpLockingContext *context;
27101
27102	SimulateIOError( return SQLITE_IOERR_CHECKRESERVEDLOCK; );
27103
27104	assert( pFile );
27105	context = (afpLockingContext *) pFile->lockingContext;
27106	if( context->reserved ){
27107	*pResOut = 1;
27108	return SQLITE_OK;
27109	}
27110	unixEnterMutex(); /* Because pFile->pInode is shared across threads */
	@@ -27081,11 +27240,11 @@
27240
27241	/* If control gets to this point, then actually go ahead and make
27242	** operating system calls for the specified lock.
27243	*/
27244	if( eFileLock==SHARED_LOCK ){
27245	int lrc1, lrc2, lrc1Errno = 0;
27246	long lk, mask;
27247
27248	assert( pInode->nShared==0 );
27249	assert( pInode->eFileLock==0 );
27250
	@@ -27455,21 +27614,23 @@
27614	#if defined(USE_PREAD)
27615	do{ got = osPwrite(id->h, pBuf, cnt, offset); }while( got<0 && errno==EINTR );
27616	#elif defined(USE_PREAD64)
27617	do{ got = osPwrite64(id->h, pBuf, cnt, offset);}while( got<0 && errno==EINTR);
27618	#else
27619	do{
27620	newOffset = lseek(id->h, offset, SEEK_SET);
27621	SimulateIOError( newOffset-- );
27622	if( newOffset!=offset ){
27623	if( newOffset == -1 ){
27624	((unixFile*)id)->lastErrno = errno;
27625	}else{
27626	((unixFile*)id)->lastErrno = 0;
27627	}
27628	return -1;
27629	}
27630	got = osWrite(id->h, pBuf, cnt);
27631	}while( got<0 && errno==EINTR );
27632	#endif
27633	TIMER_END;
27634	if( got<0 ){
27635	((unixFile*)id)->lastErrno = errno;
27636	}
	@@ -27666,10 +27827,54 @@
27827	if( OS_VXWORKS && rc!= -1 ){
27828	rc = 0;
27829	}
27830	return rc;
27831	}
27832
27833	/*
27834	** Open a file descriptor to the directory containing file zFilename.
27835	** If successful, *pFd is set to the opened file descriptor and
27836	** SQLITE_OK is returned. If an error occurs, either SQLITE_NOMEM
27837	** or SQLITE_CANTOPEN is returned and *pFd is set to an undefined
27838	** value.
27839	**
27840	** The directory file descriptor is used for only one thing - to
27841	** fsync() a directory to make sure file creation and deletion events
27842	** are flushed to disk. Such fsyncs are not needed on newer
27843	** journaling filesystems, but are required on older filesystems.
27844	**
27845	** This routine can be overridden using the xSetSysCall interface.
27846	** The ability to override this routine was added in support of the
27847	** chromium sandbox. Opening a directory is a security risk (we are
27848	** told) so making it overrideable allows the chromium sandbox to
27849	** replace this routine with a harmless no-op. To make this routine
27850	** a no-op, replace it with a stub that returns SQLITE_OK but leaves
27851	** *pFd set to a negative number.
27852	**
27853	** If SQLITE_OK is returned, the caller is responsible for closing
27854	** the file descriptor *pFd using close().
27855	*/
27856	static int openDirectory(const char zFilename, int pFd){
27857	int ii;
27858	int fd = -1;
27859	char zDirname[MAX_PATHNAME+1];
27860
27861	sqlite3_snprintf(MAX_PATHNAME, zDirname, "%s", zFilename);
27862	for(ii=(int)strlen(zDirname); ii>1 && zDirname[ii]!='/'; ii--);
27863	if( ii>0 ){
27864	zDirname[ii] = '\0';
27865	fd = robust_open(zDirname, O_RDONLY\|O_BINARY, 0);
27866	if( fd>=0 ){
27867	#ifdef FD_CLOEXEC
27868	osFcntl(fd, F_SETFD, osFcntl(fd, F_GETFD, 0) \| FD_CLOEXEC);
27869	#endif
27870	OSTRACE(("OPENDIR %-3d %s\n", fd, zDirname));
27871	}
27872	}
27873	*pFd = fd;
27874	return (fd>=0?SQLITE_OK:unixLogError(SQLITE_CANTOPEN_BKPT, "open", zDirname));
27875	}
27876
27877	/*
27878	** Make sure all writes to a particular file are committed to disk.
27879	**
27880	** If dataOnly==0 then both the file itself and its metadata (file
	@@ -27707,32 +27912,27 @@
27912	SimulateIOError( rc=1 );
27913	if( rc ){
27914	pFile->lastErrno = errno;
27915	return unixLogError(SQLITE_IOERR_FSYNC, "full_fsync", pFile->zPath);
27916	}
27917
27918	/* Also fsync the directory containing the file if the DIRSYNC flag
27919	** is set. This is a one-time occurrance. Many systems (examples: AIX)
27920	** are unable to fsync a directory, so ignore errors on the fsync.
27921	*/
27922	if( pFile->ctrlFlags & UNIXFILE_DIRSYNC ){
27923	int dirfd;
27924	OSTRACE(("DIRSYNC %s (have_fullfsync=%d fullsync=%d)\n", pFile->zPath,
27925	HAVE_FULLFSYNC, isFullsync));
27926	rc = osOpenDirectory(pFile->zPath, &dirfd);
27927	if( rc==SQLITE_OK && dirfd>=0 ){
27928	full_fsync(dirfd, 0, 0);
27929	robust_close(pFile, dirfd, __LINE__);
27930	}else if( rc==SQLITE_CANTOPEN ){
27931	rc = SQLITE_OK;
27932	}
27933	pFile->ctrlFlags &= ~UNIXFILE_DIRSYNC;











27934	}
27935	return rc;
27936	}
27937
27938	/*
	@@ -27810,15 +28010,13 @@
28010	static int proxyFileControl(sqlite3_file,int,void);
28011	#endif
28012
28013	/*
28014	** This function is called to handle the SQLITE_FCNTL_SIZE_HINT
28015	** file-control operation. Enlarge the database to nBytes in size
28016	** (rounded up to the next chunk-size). If the database is already
28017	** nBytes or larger, this routine is a no-op.


28018	*/
28019	static int fcntlSizeHint(unixFile *pFile, i64 nByte){
28020	if( pFile->szChunk ){
28021	i64 nSize; /* Required file size */
28022	struct stat buf; /* Used to hold return values of fstat() */
	@@ -27866,25 +28064,41 @@
28064
28065	/*
28066	** Information and control of an open file handle.
28067	*/
28068	static int unixFileControl(sqlite3_file id, int op, void pArg){
28069	unixFile pFile = (unixFile)id;
28070	switch( op ){
28071	case SQLITE_FCNTL_LOCKSTATE: {
28072	(int)pArg = pFile->eFileLock;
28073	return SQLITE_OK;
28074	}
28075	case SQLITE_LAST_ERRNO: {
28076	(int)pArg = pFile->lastErrno;
28077	return SQLITE_OK;
28078	}
28079	case SQLITE_FCNTL_CHUNK_SIZE: {
28080	pFile->szChunk = (int )pArg;
28081	return SQLITE_OK;
28082	}
28083	case SQLITE_FCNTL_SIZE_HINT: {
28084	int rc;
28085	SimulateIOErrorBenign(1);
28086	rc = fcntlSizeHint(pFile, (i64 )pArg);
28087	SimulateIOErrorBenign(0);
28088	return rc;
28089	}
28090	case SQLITE_FCNTL_PERSIST_WAL: {
28091	int bPersist = (int)pArg;
28092	if( bPersist<0 ){
28093	(int)pArg = (pFile->ctrlFlags & UNIXFILE_PERSIST_WAL)!=0;
28094	}else if( bPersist==0 ){
28095	pFile->ctrlFlags &= ~UNIXFILE_PERSIST_WAL;
28096	}else{
28097	pFile->ctrlFlags \|= UNIXFILE_PERSIST_WAL;
28098	}
28099	return SQLITE_OK;
28100	}
28101	#ifndef NDEBUG
28102	/* The pager calls this method to signal that it has done
28103	** a rollback and that the database is therefore unchanged and
28104	** it hence it is OK for the transaction change counter to be
	@@ -28096,11 +28310,11 @@
28310	unixShmNode *p = pFd->pInode->pShmNode;
28311	assert( unixMutexHeld() );
28312	if( p && p->nRef==0 ){
28313	int i;
28314	assert( p->pInode==pFd->pInode );
28315	sqlite3_mutex_free(p->mutex);
28316	for(i=0; i<p->nRegion; i++){
28317	if( p->h>=0 ){
28318	munmap(p->apRegion[i], p->szRegion);
28319	}else{
28320	sqlite3_free(p->apRegion[i]);
	@@ -28565,11 +28779,11 @@
28779	** shared-memory file, too */
28780	unixEnterMutex();
28781	assert( pShmNode->nRef>0 );
28782	pShmNode->nRef--;
28783	if( pShmNode->nRef==0 ){
28784	if( deleteFlag && pShmNode->h>=0 ) osUnlink(pShmNode->zFilename);
28785	unixShmPurge(pDbFd);
28786	}
28787	unixLeaveMutex();
28788
28789	return SQLITE_OK;
	@@ -28878,11 +29092,11 @@
29092	** Initialize the contents of the unixFile structure pointed to by pId.
29093	*/
29094	static int fillInUnixFile(
29095	sqlite3_vfs pVfs, / Pointer to vfs object */
29096	int h, /* Open file descriptor of file being opened */
29097	int syncDir, /* True to sync directory on first sync */
29098	sqlite3_file pId, / Write to the unixFile structure here */
29099	const char zFilename, / Name of the file being opened */
29100	int noLock, /* Omit locking if true */
29101	int isDelete, /* Delete on close if true */
29102	int isReadOnly /* True if the file is opened read-only */
	@@ -28909,20 +29123,22 @@
29123	assert( zFilename==0 \|\| zFilename[0]=='/' );
29124	#endif
29125
29126	OSTRACE(("OPEN %-3d %s\n", h, zFilename));
29127	pNew->h = h;

29128	pNew->zPath = zFilename;
29129	if( memcmp(pVfs->zName,"unix-excl",10)==0 ){
29130	pNew->ctrlFlags = UNIXFILE_EXCL;
29131	}else{
29132	pNew->ctrlFlags = 0;
29133	}
29134	if( isReadOnly ){
29135	pNew->ctrlFlags \|= UNIXFILE_RDONLY;
29136	}
29137	if( syncDir ){
29138	pNew->ctrlFlags \|= UNIXFILE_DIRSYNC;
29139	}
29140
29141	#if OS_VXWORKS
29142	pNew->pId = vxworksFindFileId(zFilename);
29143	if( pNew->pId==0 ){
29144	noLock = 1;
	@@ -29045,56 +29261,24 @@
29261	pNew->lastErrno = 0;
29262	#if OS_VXWORKS
29263	if( rc!=SQLITE_OK ){
29264	if( h>=0 ) robust_close(pNew, h, __LINE__);
29265	h = -1;
29266	osUnlink(zFilename);
29267	isDelete = 0;
29268	}
29269	pNew->isDelete = isDelete;
29270	#endif
29271	if( rc!=SQLITE_OK ){

29272	if( h>=0 ) robust_close(pNew, h, __LINE__);
29273	}else{
29274	pNew->pMethod = pLockingStyle;
29275	OpenCounter(+1);
29276	}
29277	return rc;
29278	}
29279































29280	/*
29281	** Return the name of a directory in which to put temporary files.
29282	** If no suitable temporary file directory can be found, return NULL.
29283	*/
29284	static const char *unixTempFileDir(void){
	@@ -29205,11 +29389,11 @@
29389	** ignored and -1 is returned. The caller will try to open a new file
29390	** descriptor on the same path, fail, and return an error to SQLite.
29391	**
29392	** Even if a subsequent open() call does succeed, the consequences of
29393	** not searching for a resusable file descriptor are not dire. */
29394	if( 0==osStat(zPath, &sStat) ){
29395	unixInodeInfo *pInode;
29396
29397	unixEnterMutex();
29398	pInode = inodeList;
29399	while( pInode && (pInode->fileId.dev!=sStat.st_dev
	@@ -29281,11 +29465,11 @@
29465	while( nDb>0 && zPath[nDb]!='-' ) nDb--;
29466	if( nDb==0 ) return SQLITE_OK;
29467	memcpy(zDb, zPath, nDb);
29468	zDb[nDb] = '\0';
29469
29470	if( 0==osStat(zDb, &sStat) ){
29471	*pMode = sStat.st_mode & 0777;
29472	}else{
29473	rc = SQLITE_IOERR_FSTAT;
29474	}
29475	}else if( flags & SQLITE_OPEN_DELETEONCLOSE ){
	@@ -29323,11 +29507,10 @@
29507	int flags, /* Input flags to control the opening */
29508	int pOutFlags / Output flags returned to SQLite core */
29509	){
29510	unixFile p = (unixFile )pFile;
29511	int fd = -1; /* File descriptor returned by open() */

29512	int openFlags = 0; /* Flags to pass to open() */
29513	int eType = flags&0xFFFFFF00; /* Type of file to open */
29514	int noLock; /* True to omit locking primitives */
29515	int rc = SQLITE_OK; /* Function Return Code */
29516
	@@ -29337,16 +29520,19 @@
29520	int isReadonly = (flags & SQLITE_OPEN_READONLY);
29521	int isReadWrite = (flags & SQLITE_OPEN_READWRITE);
29522	#if SQLITE_ENABLE_LOCKING_STYLE
29523	int isAutoProxy = (flags & SQLITE_OPEN_AUTOPROXY);
29524	#endif
29525	#if defined(__APPLE__) \|\| SQLITE_ENABLE_LOCKING_STYLE
29526	struct statfs fsInfo;
29527	#endif
29528
29529	/* If creating a master or main-file journal, this function will open
29530	** a file-descriptor on the directory too. The first time unixSync()
29531	** is called the directory file descriptor will be fsync()ed and close()d.
29532	*/
29533	int syncDir = (isCreate && (
29534	eType==SQLITE_OPEN_MASTER_JOURNAL
29535	\|\| eType==SQLITE_OPEN_MAIN_JOURNAL
29536	\|\| eType==SQLITE_OPEN_WAL
29537	));
29538
	@@ -29396,11 +29582,11 @@
29582	}
29583	}
29584	p->pUnused = pUnused;
29585	}else if( !zName ){
29586	/* If zName is NULL, the upper layer is requesting a temp file. */
29587	assert(isDelete && !syncDir);
29588	rc = unixGetTempname(MAX_PATHNAME+1, zTmpname);
29589	if( rc!=SQLITE_OK ){
29590	return rc;
29591	}
29592	zName = zTmpname;
	@@ -29452,44 +29638,29 @@
29638
29639	if( isDelete ){
29640	#if OS_VXWORKS
29641	zPath = zName;
29642	#else
29643	osUnlink(zName);
29644	#endif
29645	}
29646	#if SQLITE_ENABLE_LOCKING_STYLE
29647	else{
29648	p->openFlags = openFlags;
29649	}
29650	#endif
29651













29652	#ifdef FD_CLOEXEC
29653	osFcntl(fd, F_SETFD, osFcntl(fd, F_GETFD, 0) \| FD_CLOEXEC);
29654	#endif
29655
29656	noLock = eType!=SQLITE_OPEN_MAIN_DB;
29657
29658
29659	#if defined(__APPLE__) \|\| SQLITE_ENABLE_LOCKING_STYLE

29660	if( fstatfs(fd, &fsInfo) == -1 ){
29661	((unixFile*)pFile)->lastErrno = errno;

29662	robust_close(p, fd, __LINE__);
29663	return SQLITE_IOERR_ACCESS;
29664	}
29665	if (0 == strncmp("msdos", fsInfo.f_fstypename, 5)) {
29666	((unixFile*)pFile)->fsFlags \|= SQLITE_FSFLAGS_IS_MSDOS;
	@@ -29507,31 +29678,27 @@
29678	/* SQLITE_FORCE_PROXY_LOCKING==1 means force always use proxy, 0 means
29679	** never use proxy, NULL means use proxy for non-local files only. */
29680	if( envforce!=NULL ){
29681	useProxy = atoi(envforce)>0;
29682	}else{

29683	if( statfs(zPath, &fsInfo) == -1 ){
29684	/* In theory, the close(fd) call is sub-optimal. If the file opened
29685	** with fd is a database file, and there are other connections open
29686	** on that file that are currently holding advisory locks on it,
29687	** then the call to close() will cancel those locks. In practice,
29688	** we're assuming that statfs() doesn't fail very often. At least
29689	** not while other file descriptors opened by the same process on
29690	** the same file are working. */
29691	p->lastErrno = errno;



29692	robust_close(p, fd, __LINE__);
29693	rc = SQLITE_IOERR_ACCESS;
29694	goto open_finished;
29695	}
29696	useProxy = !(fsInfo.f_flags&MNT_LOCAL);
29697	}
29698	if( useProxy ){
29699	rc = fillInUnixFile(pVfs, fd, syncDir, pFile, zPath, noLock,
29700	isDelete, isReadonly);
29701	if( rc==SQLITE_OK ){
29702	rc = proxyTransformUnixFile((unixFile*)pFile, ":auto:");
29703	if( rc!=SQLITE_OK ){
29704	/* Use unixClose to clean up the resources added in fillInUnixFile
	@@ -29545,11 +29712,11 @@
29712	goto open_finished;
29713	}
29714	}
29715	#endif
29716
29717	rc = fillInUnixFile(pVfs, fd, syncDir, pFile, zPath, noLock,
29718	isDelete, isReadonly);
29719	open_finished:
29720	if( rc!=SQLITE_OK ){
29721	sqlite3_free(p->pUnused);
29722	}
	@@ -29567,17 +29734,17 @@
29734	int dirSync /* If true, fsync() directory after deleting file */
29735	){
29736	int rc = SQLITE_OK;
29737	UNUSED_PARAMETER(NotUsed);
29738	SimulateIOError(return SQLITE_IOERR_DELETE);
29739	if( osUnlink(zPath)==(-1) && errno!=ENOENT ){
29740	return unixLogError(SQLITE_IOERR_DELETE, "unlink", zPath);
29741	}
29742	#ifndef SQLITE_DISABLE_DIRSYNC
29743	if( dirSync ){
29744	int fd;
29745	rc = osOpenDirectory(zPath, &fd);
29746	if( rc==SQLITE_OK ){
29747	#if OS_VXWORKS
29748	if( fsync(fd)==-1 )
29749	#else
29750	if( fsync(fd) )
	@@ -29584,10 +29751,12 @@
29751	#endif
29752	{
29753	rc = unixLogError(SQLITE_IOERR_DIR_FSYNC, "fsync", zPath);
29754	}
29755	robust_close(0, fd, __LINE__);
29756	}else if( rc==SQLITE_CANTOPEN ){
29757	rc = SQLITE_OK;
29758	}
29759	}
29760	#endif
29761	return rc;
29762	}
	@@ -29626,11 +29795,11 @@
29795	assert(!"Invalid flags argument");
29796	}
29797	*pResOut = (osAccess(zPath, amode)==0);
29798	if( flags==SQLITE_ACCESS_EXISTS && *pResOut ){
29799	struct stat buf;
29800	if( 0==osStat(zPath, &buf) && buf.st_size==0 ){
29801	*pResOut = 0;
29802	}
29803	}
29804	return SQLITE_OK;
29805	}
	@@ -30145,11 +30314,10 @@
30314	const char path, / path for the new unixFile */
30315	unixFile *ppFile, / unixFile created and returned by ref */
30316	int islockfile /* if non zero missing dirs will be created */
30317	) {
30318	int fd = -1;

30319	unixFile *pNew;
30320	int rc = SQLITE_OK;
30321	int openFlags = O_RDWR \| O_CREAT;
30322	sqlite3_vfs dummyVfs;
30323	int terrno = 0;
	@@ -30210,11 +30378,11 @@
30378	dummyVfs.zName = "dummy";
30379	pUnused->fd = fd;
30380	pUnused->flags = openFlags;
30381	pNew->pUnused = pUnused;
30382
30383	rc = fillInUnixFile(&dummyVfs, fd, 0, (sqlite3_file*)pNew, path, 0, 0, 0);
30384	if( rc==SQLITE_OK ){
30385	*ppFile = pNew;
30386	return SQLITE_OK;
30387	}
30388	end_create_proxy:
	@@ -30250,10 +30418,12 @@
30418	*pError = err;
30419	}
30420	return SQLITE_IOERR;
30421	}
30422	}
30423	#else
30424	UNUSED_PARAMETER(pError);
30425	#endif
30426	#ifdef SQLITE_TEST
30427	/* simulate multiple hosts by creating unique hostid file paths */
30428	if( sqlite3_hostid_num != 0){
30429	pHostID[0] = (char)(pHostID[0] + (char)(sqlite3_hostid_num & 0xFF));
	@@ -30324,11 +30494,11 @@
30494	conchFile->openFlags = O_RDWR \| O_CREAT;
30495
30496	end_breaklock:
30497	if( rc ){
30498	if( fd>=0 ){
30499	osUnlink(tPath);
30500	robust_close(pFile, fd, __LINE__);
30501	}
30502	fprintf(stderr, "failed to break stale lock on %s, %s\n", cPath, errmsg);
30503	}
30504	return rc;
	@@ -30342,10 +30512,11 @@
30512	unixFile *conchFile = pCtx->conchFile;
30513	int rc = SQLITE_OK;
30514	int nTries = 0;
30515	struct timespec conchModTime;
30516
30517	memset(&conchModTime, 0, sizeof(conchModTime));
30518	do {
30519	rc = conchFile->pMethod->xLock((sqlite3_file*)conchFile, lockType);
30520	nTries ++;
30521	if( rc==SQLITE_BUSY ){
30522	/* If the lock failed (busy):
	@@ -30573,15 +30744,16 @@
30744	conchFile->pMethod->xUnlock((sqlite3_file*)conchFile, SHARED_LOCK);
30745
30746	end_takeconch:
30747	OSTRACE(("TRANSPROXY: CLOSE %d\n", pFile->h));
30748	if( rc==SQLITE_OK && pFile->openFlags ){
30749	int fd;
30750	if( pFile->h>=0 ){
30751	robust_close(pFile, pFile->h, __LINE__);
30752	}
30753	pFile->h = -1;
30754	fd = robust_open(pCtx->dbPath, pFile->openFlags,
30755	SQLITE_DEFAULT_FILE_PERMISSIONS);
30756	OSTRACE(("TRANSPROXY: OPEN %d\n", fd));
30757	if( fd>=0 ){
30758	pFile->h = fd;
30759	}else{
	@@ -31147,11 +31319,11 @@
31319	};
31320	unsigned int i; /* Loop counter */
31321
31322	/* Double-check that the aSyscall[] array has been constructed
31323	** correctly. See ticket [bb3a86e890c8e96ab] */
31324	assert( ArraySize(aSyscall)==18 );
31325
31326	/* Register all VFSes defined in the aVfs[] array */
31327	for(i=0; i<(sizeof(aVfs)/sizeof(sqlite3_vfs)); i++){
31328	sqlite3_vfs_register(&aVfs[i], i==0);
31329	}
	@@ -31264,14 +31436,17 @@
31436	#ifdef MEMORY_DEBUG
31437	# error "The MEMORY_DEBUG macro is obsolete. Use SQLITE_DEBUG instead."
31438	#endif
31439
31440	#ifdef SQLITE_DEBUG
31441	# ifndef SQLITE_DEBUG_OS_TRACE
31442	# define SQLITE_DEBUG_OS_TRACE 0
31443	# endif
31444	int sqlite3OSTrace = SQLITE_DEBUG_OS_TRACE;
31445	# define OSTRACE(X) if( sqlite3OSTrace ) sqlite3DebugPrintf X
31446	#else
31447	# define OSTRACE(X)
31448	#endif
31449
31450	/*
31451	** Macros for performance tracing. Normally turned off. Only works
31452	** on i486 hardware.
	@@ -31479,12 +31654,13 @@
31654	typedef struct winFile winFile;
31655	struct winFile {
31656	const sqlite3_io_methods pMethod; / Must be first */
31657	sqlite3_vfs pVfs; / The VFS used to open this file */
31658	HANDLE h; /* Handle for accessing the file */
31659	u8 locktype; /* Type of lock currently held on this file */
31660	short sharedLockByte; /* Randomly chosen byte used as a shared lock */
31661	u8 bPersistWal; /* True to persist WAL files */
31662	DWORD lastErrno; /* The Windows errno from the last I/O error */
31663	DWORD sectorSize; /* Sector size of the device file is on */
31664	winShm pShm; / Instance of shared memory on this file */
31665	const char zPath; / Full pathname of this file */
31666	int szChunk; /* Chunk size configured by FCNTL_CHUNK_SIZE */
	@@ -31495,10 +31671,80 @@
31671	winceLock local; /* Locks obtained by this instance of winFile */
31672	winceLock shared; / Global shared lock memory for the file */
31673	#endif
31674	};
31675
31676	/*
31677	* If compiled with SQLITE_WIN32_MALLOC on Windows, we will use the
31678	* various Win32 API heap functions instead of our own.
31679	*/
31680	#ifdef SQLITE_WIN32_MALLOC
31681	/*
31682	* The initial size of the Win32-specific heap. This value may be zero.
31683	*/
31684	#ifndef SQLITE_WIN32_HEAP_INIT_SIZE
31685	# define SQLITE_WIN32_HEAP_INIT_SIZE ((SQLITE_DEFAULT_CACHE_SIZE) * \
31686	(SQLITE_DEFAULT_PAGE_SIZE) + 4194304)
31687	#endif
31688
31689	/*
31690	* The maximum size of the Win32-specific heap. This value may be zero.
31691	*/
31692	#ifndef SQLITE_WIN32_HEAP_MAX_SIZE
31693	# define SQLITE_WIN32_HEAP_MAX_SIZE (0)
31694	#endif
31695
31696	/*
31697	* The extra flags to use in calls to the Win32 heap APIs. This value may be
31698	* zero for the default behavior.
31699	*/
31700	#ifndef SQLITE_WIN32_HEAP_FLAGS
31701	# define SQLITE_WIN32_HEAP_FLAGS (0)
31702	#endif
31703
31704	/*
31705	** The winMemData structure stores information required by the Win32-specific
31706	** sqlite3_mem_methods implementation.
31707	*/
31708	typedef struct winMemData winMemData;
31709	struct winMemData {
31710	#ifndef NDEBUG
31711	u32 magic; /* Magic number to detect structure corruption. */
31712	#endif
31713	HANDLE hHeap; /* The handle to our heap. */
31714	BOOL bOwned; /* Do we own the heap (i.e. destroy it on shutdown)? */
31715	};
31716
31717	#ifndef NDEBUG
31718	#define WINMEM_MAGIC 0x42b2830b
31719	#endif
31720
31721	static struct winMemData win_mem_data = {
31722	#ifndef NDEBUG
31723	WINMEM_MAGIC,
31724	#endif
31725	NULL, FALSE
31726	};
31727
31728	#ifndef NDEBUG
31729	#define winMemAssertMagic() assert( win_mem_data.magic==WINMEM_MAGIC )
31730	#else
31731	#define winMemAssertMagic()
31732	#endif
31733
31734	#define winMemGetHeap() win_mem_data.hHeap
31735
31736	static void *winMemMalloc(int nBytes);
31737	static void winMemFree(void *pPrior);
31738	static void winMemRealloc(void pPrior, int nBytes);
31739	static int winMemSize(void *p);
31740	static int winMemRoundup(int n);
31741	static int winMemInit(void *pAppData);
31742	static void winMemShutdown(void *pAppData);
31743
31744	SQLITE_PRIVATE const sqlite3_mem_methods *sqlite3MemGetWin32(void);
31745	#endif /* SQLITE_WIN32_MALLOC */
31746
31747	/*
31748	** Forward prototypes.
31749	*/
31750	static int getSectorSize(
	@@ -31547,10 +31793,192 @@
31793	}
31794	return sqlite3_os_type==2;
31795	}
31796	#endif /* SQLITE_OS_WINCE */
31797
31798	#ifdef SQLITE_WIN32_MALLOC
31799	/*
31800	** Allocate nBytes of memory.
31801	*/
31802	static void *winMemMalloc(int nBytes){
31803	HANDLE hHeap;
31804	void *p;
31805
31806	winMemAssertMagic();
31807	hHeap = winMemGetHeap();
31808	assert( hHeap!=0 );
31809	assert( hHeap!=INVALID_HANDLE_VALUE );
31810	#ifdef SQLITE_WIN32_MALLOC_VALIDATE
31811	assert ( HeapValidate(hHeap, SQLITE_WIN32_HEAP_FLAGS, NULL) );
31812	#endif
31813	assert( nBytes>=0 );
31814	p = HeapAlloc(hHeap, SQLITE_WIN32_HEAP_FLAGS, (SIZE_T)nBytes);
31815	if( !p ){
31816	sqlite3_log(SQLITE_NOMEM, "failed to HeapAlloc %u bytes (%d), heap=%p",
31817	nBytes, GetLastError(), (void*)hHeap);
31818	}
31819	return p;
31820	}
31821
31822	/*
31823	** Free memory.
31824	*/
31825	static void winMemFree(void *pPrior){
31826	HANDLE hHeap;
31827
31828	winMemAssertMagic();
31829	hHeap = winMemGetHeap();
31830	assert( hHeap!=0 );
31831	assert( hHeap!=INVALID_HANDLE_VALUE );
31832	#ifdef SQLITE_WIN32_MALLOC_VALIDATE
31833	assert ( HeapValidate(hHeap, SQLITE_WIN32_HEAP_FLAGS, pPrior) );
31834	#endif
31835	if( !pPrior ) return; /* Passing NULL to HeapFree is undefined. */
31836	if( !HeapFree(hHeap, SQLITE_WIN32_HEAP_FLAGS, pPrior) ){
31837	sqlite3_log(SQLITE_NOMEM, "failed to HeapFree block %p (%d), heap=%p",
31838	pPrior, GetLastError(), (void*)hHeap);
31839	}
31840	}
31841
31842	/*
31843	** Change the size of an existing memory allocation
31844	*/
31845	static void winMemRealloc(void pPrior, int nBytes){
31846	HANDLE hHeap;
31847	void *p;
31848
31849	winMemAssertMagic();
31850	hHeap = winMemGetHeap();
31851	assert( hHeap!=0 );
31852	assert( hHeap!=INVALID_HANDLE_VALUE );
31853	#ifdef SQLITE_WIN32_MALLOC_VALIDATE
31854	assert ( HeapValidate(hHeap, SQLITE_WIN32_HEAP_FLAGS, pPrior) );
31855	#endif
31856	assert( nBytes>=0 );
31857	if( !pPrior ){
31858	p = HeapAlloc(hHeap, SQLITE_WIN32_HEAP_FLAGS, (SIZE_T)nBytes);
31859	}else{
31860	p = HeapReAlloc(hHeap, SQLITE_WIN32_HEAP_FLAGS, pPrior, (SIZE_T)nBytes);
31861	}
31862	if( !p ){
31863	sqlite3_log(SQLITE_NOMEM, "failed to %s %u bytes (%d), heap=%p",
31864	pPrior ? "HeapReAlloc" : "HeapAlloc", nBytes, GetLastError(),
31865	(void*)hHeap);
31866	}
31867	return p;
31868	}
31869
31870	/*
31871	** Return the size of an outstanding allocation, in bytes.
31872	*/
31873	static int winMemSize(void *p){
31874	HANDLE hHeap;
31875	SIZE_T n;
31876
31877	winMemAssertMagic();
31878	hHeap = winMemGetHeap();
31879	assert( hHeap!=0 );
31880	assert( hHeap!=INVALID_HANDLE_VALUE );
31881	#ifdef SQLITE_WIN32_MALLOC_VALIDATE
31882	assert ( HeapValidate(hHeap, SQLITE_WIN32_HEAP_FLAGS, NULL) );
31883	#endif
31884	if( !p ) return 0;
31885	n = HeapSize(hHeap, SQLITE_WIN32_HEAP_FLAGS, p);
31886	if( n==(SIZE_T)-1 ){
31887	sqlite3_log(SQLITE_NOMEM, "failed to HeapSize block %p (%d), heap=%p",
31888	p, GetLastError(), (void*)hHeap);
31889	return 0;
31890	}
31891	return (int)n;
31892	}
31893
31894	/*
31895	** Round up a request size to the next valid allocation size.
31896	*/
31897	static int winMemRoundup(int n){
31898	return n;
31899	}
31900
31901	/*
31902	** Initialize this module.
31903	*/
31904	static int winMemInit(void *pAppData){
31905	winMemData pWinMemData = (winMemData )pAppData;
31906
31907	if( !pWinMemData ) return SQLITE_ERROR;
31908	assert( pWinMemData->magic==WINMEM_MAGIC );
31909	if( !pWinMemData->hHeap ){
31910	pWinMemData->hHeap = HeapCreate(SQLITE_WIN32_HEAP_FLAGS,
31911	SQLITE_WIN32_HEAP_INIT_SIZE,
31912	SQLITE_WIN32_HEAP_MAX_SIZE);
31913	if( !pWinMemData->hHeap ){
31914	sqlite3_log(SQLITE_NOMEM,
31915	"failed to HeapCreate (%d), flags=%u, initSize=%u, maxSize=%u",
31916	GetLastError(), SQLITE_WIN32_HEAP_FLAGS, SQLITE_WIN32_HEAP_INIT_SIZE,
31917	SQLITE_WIN32_HEAP_MAX_SIZE);
31918	return SQLITE_NOMEM;
31919	}
31920	pWinMemData->bOwned = TRUE;
31921	}
31922	assert( pWinMemData->hHeap!=0 );
31923	assert( pWinMemData->hHeap!=INVALID_HANDLE_VALUE );
31924	#ifdef SQLITE_WIN32_MALLOC_VALIDATE
31925	assert( HeapValidate(pWinMemData->hHeap, SQLITE_WIN32_HEAP_FLAGS, NULL) );
31926	#endif
31927	return SQLITE_OK;
31928	}
31929
31930	/*
31931	** Deinitialize this module.
31932	*/
31933	static void winMemShutdown(void *pAppData){
31934	winMemData pWinMemData = (winMemData )pAppData;
31935
31936	if( !pWinMemData ) return;
31937	if( pWinMemData->hHeap ){
31938	assert( pWinMemData->hHeap!=INVALID_HANDLE_VALUE );
31939	#ifdef SQLITE_WIN32_MALLOC_VALIDATE
31940	assert( HeapValidate(pWinMemData->hHeap, SQLITE_WIN32_HEAP_FLAGS, NULL) );
31941	#endif
31942	if( pWinMemData->bOwned ){
31943	if( !HeapDestroy(pWinMemData->hHeap) ){
31944	sqlite3_log(SQLITE_NOMEM, "failed to HeapDestroy (%d), heap=%p",
31945	GetLastError(), (void*)pWinMemData->hHeap);
31946	}
31947	pWinMemData->bOwned = FALSE;
31948	}
31949	pWinMemData->hHeap = NULL;
31950	}
31951	}
31952
31953	/*
31954	** Populate the low-level memory allocation function pointers in
31955	** sqlite3GlobalConfig.m with pointers to the routines in this file. The
31956	** arguments specify the block of memory to manage.
31957	**
31958	** This routine is only called by sqlite3_config(), and therefore
31959	** is not required to be threadsafe (it is not).
31960	*/
31961	SQLITE_PRIVATE const sqlite3_mem_methods *sqlite3MemGetWin32(void){
31962	static const sqlite3_mem_methods winMemMethods = {
31963	winMemMalloc,
31964	winMemFree,
31965	winMemRealloc,
31966	winMemSize,
31967	winMemRoundup,
31968	winMemInit,
31969	winMemShutdown,
31970	&win_mem_data
31971	};
31972	return &winMemMethods;
31973	}
31974
31975	SQLITE_PRIVATE void sqlite3MemSetDefault(void){
31976	sqlite3_config(SQLITE_CONFIG_MALLOC, sqlite3MemGetWin32());
31977	}
31978	#endif /* SQLITE_WIN32_MALLOC */
31979
31980	/*
31981	** Convert a UTF-8 string to microsoft unicode (UTF-16?).
31982	**
31983	** Space to hold the returned string is obtained from malloc.
31984	*/
	@@ -31778,19 +32206,68 @@
32206	iLine, iErrno, zFunc, zPath, zMsg
32207	);
32208
32209	return errcode;
32210	}
32211
32212	/*
32213	** The number of times that a ReadFile(), WriteFile(), and DeleteFile()
32214	** will be retried following a locking error - probably caused by
32215	** antivirus software. Also the initial delay before the first retry.
32216	** The delay increases linearly with each retry.
32217	*/
32218	#ifndef SQLITE_WIN32_IOERR_RETRY
32219	# define SQLITE_WIN32_IOERR_RETRY 10
32220	#endif
32221	#ifndef SQLITE_WIN32_IOERR_RETRY_DELAY
32222	# define SQLITE_WIN32_IOERR_RETRY_DELAY 25
32223	#endif
32224	static int win32IoerrRetry = SQLITE_WIN32_IOERR_RETRY;
32225	static int win32IoerrRetryDelay = SQLITE_WIN32_IOERR_RETRY_DELAY;
32226
32227	/*
32228	** If a ReadFile() or WriteFile() error occurs, invoke this routine
32229	** to see if it should be retried. Return TRUE to retry. Return FALSE
32230	** to give up with an error.
32231	*/
32232	static int retryIoerr(int *pnRetry){
32233	DWORD e;
32234	if( *pnRetry>=win32IoerrRetry ){
32235	return 0;
32236	}
32237	e = GetLastError();
32238	if( e==ERROR_ACCESS_DENIED \|\|
32239	e==ERROR_LOCK_VIOLATION \|\|
32240	e==ERROR_SHARING_VIOLATION ){
32241	Sleep(win32IoerrRetryDelay(1+pnRetry));
32242	++*pnRetry;
32243	return 1;
32244	}
32245	return 0;
32246	}
32247
32248	/*
32249	** Log a I/O error retry episode.
32250	*/
32251	static void logIoerr(int nRetry){
32252	if( nRetry ){
32253	sqlite3_log(SQLITE_IOERR,
32254	"delayed %dms for lock/sharing conflict",
32255	win32IoerrRetryDelaynRetry(nRetry+1)/2
32256	);
32257	}
32258	}
32259
32260	#if SQLITE_OS_WINCE
32261	/*************************************************************************
32262	** This section contains code for WinCE only.
32263	*/
32264	/*
32265	** WindowsCE does not have a localtime() function. So create a
32266	** substitute.
32267	*/
32268	/* #include <time.h> */
32269	struct tm __cdecl localtime(const time_t t)
32270	{
32271	static struct tm y;
32272	FILETIME uTm, lTm;
32273	SYSTEMTIME pTm;
	@@ -32196,22 +32673,25 @@
32673	int amt, /* Number of bytes to read */
32674	sqlite3_int64 offset /* Begin reading at this offset */
32675	){
32676	winFile pFile = (winFile)id; /* file handle */
32677	DWORD nRead; /* Number of bytes actually read from file */
32678	int nRetry = 0; /* Number of retrys */
32679
32680	assert( id!=0 );
32681	SimulateIOError(return SQLITE_IOERR_READ);
32682	OSTRACE(("READ %d lock=%d\n", pFile->h, pFile->locktype));
32683
32684	if( seekWinFile(pFile, offset) ){
32685	return SQLITE_FULL;
32686	}
32687	while( !ReadFile(pFile->h, pBuf, amt, &nRead, 0) ){
32688	if( retryIoerr(&nRetry) ) continue;
32689	pFile->lastErrno = GetLastError();
32690	return winLogError(SQLITE_IOERR_READ, "winRead", pFile->zPath);
32691	}
32692	logIoerr(nRetry);
32693	if( nRead<(DWORD)amt ){
32694	/* Unread parts of the buffer must be zero-filled */
32695	memset(&((char*)pBuf)[nRead], 0, amt-nRead);
32696	return SQLITE_IOERR_SHORT_READ;
32697	}
	@@ -32229,10 +32709,11 @@
32709	int amt, /* Number of bytes to write */
32710	sqlite3_int64 offset /* Offset into the file to begin writing at */
32711	){
32712	int rc; /* True if error has occured, else false */
32713	winFile pFile = (winFile)id; /* File handle */
32714	int nRetry = 0; /* Number of retries */
32715
32716	assert( amt>0 );
32717	assert( pFile );
32718	SimulateIOError(return SQLITE_IOERR_WRITE);
32719	SimulateDiskfullError(return SQLITE_FULL);
	@@ -32243,11 +32724,16 @@
32724	if( rc==0 ){
32725	u8 aRem = (u8 )pBuf; /* Data yet to be written */
32726	int nRem = amt; /* Number of bytes yet to be written */
32727	DWORD nWrite; /* Bytes written by each WriteFile() call */
32728
32729	while( nRem>0 ){
32730	if( !WriteFile(pFile->h, aRem, nRem, &nWrite, 0) ){
32731	if( retryIoerr(&nRetry) ) continue;
32732	break;
32733	}
32734	if( nWrite<=0 ) break;
32735	aRem += nWrite;
32736	nRem -= nWrite;
32737	}
32738	if( nRem>0 ){
32739	pFile->lastErrno = GetLastError();
	@@ -32254,14 +32740,17 @@
32740	rc = 1;
32741	}
32742	}
32743
32744	if( rc ){
32745	if( ( pFile->lastErrno==ERROR_HANDLE_DISK_FULL )
32746	\|\| ( pFile->lastErrno==ERROR_DISK_FULL )){
32747	return SQLITE_FULL;
32748	}
32749	return winLogError(SQLITE_IOERR_WRITE, "winWrite", pFile->zPath);
32750	}else{
32751	logIoerr(nRetry);
32752	}
32753	return SQLITE_OK;
32754	}
32755
32756	/*
	@@ -32651,32 +33140,63 @@
33140
33141	/*
33142	** Control and query of the open file handle.
33143	*/
33144	static int winFileControl(sqlite3_file id, int op, void pArg){
33145	winFile pFile = (winFile)id;
33146	switch( op ){
33147	case SQLITE_FCNTL_LOCKSTATE: {
33148	(int)pArg = pFile->locktype;
33149	return SQLITE_OK;
33150	}
33151	case SQLITE_LAST_ERRNO: {
33152	(int)pArg = (int)pFile->lastErrno;
33153	return SQLITE_OK;
33154	}
33155	case SQLITE_FCNTL_CHUNK_SIZE: {
33156	pFile->szChunk = (int )pArg;
33157	return SQLITE_OK;
33158	}
33159	case SQLITE_FCNTL_SIZE_HINT: {
33160	winFile pFile = (winFile)id;
33161	sqlite3_int64 oldSz;
33162	int rc = winFileSize(id, &oldSz);
33163	if( rc==SQLITE_OK ){
33164	sqlite3_int64 newSz = (sqlite3_int64)pArg;
33165	if( newSz>oldSz ){
33166	SimulateIOErrorBenign(1);
33167	rc = winTruncate(id, newSz);
33168	SimulateIOErrorBenign(0);
33169	}
33170	}
33171	return rc;
33172	}
33173	case SQLITE_FCNTL_PERSIST_WAL: {
33174	int bPersist = (int)pArg;
33175	if( bPersist<0 ){
33176	(int)pArg = pFile->bPersistWal;
33177	}else{
33178	pFile->bPersistWal = bPersist!=0;
33179	}
33180	return SQLITE_OK;
33181	}
33182	case SQLITE_FCNTL_SYNC_OMITTED: {
33183	return SQLITE_OK;
33184	}
33185	case SQLITE_FCNTL_WIN32_AV_RETRY: {
33186	int a = (int)pArg;
33187	if( a[0]>0 ){
33188	win32IoerrRetry = a[0];
33189	}else{
33190	a[0] = win32IoerrRetry;
33191	}
33192	if( a[1]>0 ){
33193	win32IoerrRetryDelay = a[1];
33194	}else{
33195	a[1] = win32IoerrRetryDelay;
33196	}
33197	return SQLITE_OK;
33198	}
33199	}
33200	return SQLITE_NOTFOUND;
33201	}
33202
	@@ -33482,10 +34002,11 @@
34002	int isTemp = 0;
34003	#endif
34004	winFile pFile = (winFile)id;
34005	void zConverted; / Filename in OS encoding */
34006	const char zUtf8Name = zName; / Filename in UTF-8 encoding */
34007	int cnt = 0;
34008
34009	/* If argument zPath is a NULL pointer, this function is required to open
34010	** a temporary file. Use this buffer to store the file name in.
34011	*/
34012	char zTmpname[MAX_PATH+1]; /* Buffer used to create temp filename */
	@@ -33601,34 +34122,34 @@
34122	#if SQLITE_OS_WINCE
34123	dwFlagsAndAttributes \|= FILE_FLAG_RANDOM_ACCESS;
34124	#endif
34125
34126	if( isNT() ){
34127	while( (h = CreateFileW((WCHAR*)zConverted,
34128	dwDesiredAccess,
34129	dwShareMode, NULL,
34130	dwCreationDisposition,
34131	dwFlagsAndAttributes,
34132	NULL))==INVALID_HANDLE_VALUE &&
34133	retryIoerr(&cnt) ){}

34134	/* isNT() is 1 if SQLITE_OS_WINCE==1, so this else is never executed.
34135	** Since the ASCII version of these Windows API do not exist for WINCE,
34136	** it's important to not reference them for WINCE builds.
34137	*/
34138	#if SQLITE_OS_WINCE==0
34139	}else{
34140	while( (h = CreateFileA((char*)zConverted,
34141	dwDesiredAccess,
34142	dwShareMode, NULL,
34143	dwCreationDisposition,
34144	dwFlagsAndAttributes,
34145	NULL))==INVALID_HANDLE_VALUE &&
34146	retryIoerr(&cnt) ){}

34147	#endif
34148	}
34149
34150	logIoerr(cnt);
34151
34152	OSTRACE(("OPEN %d %s 0x%lx %s\n",
34153	h, zName, dwDesiredAccess,
34154	h==INVALID_HANDLE_VALUE ? "failed" : "ok"));
34155
	@@ -33691,19 +34212,17 @@
34212	** file open, we will be unable to delete it. To work around this
34213	** problem, we delay 100 milliseconds and try to delete again. Up
34214	** to MX_DELETION_ATTEMPTs deletion attempts are run before giving
34215	** up and returning an error.
34216	*/

34217	static int winDelete(
34218	sqlite3_vfs pVfs, / Not used on win32 */
34219	const char zFilename, / Name of file to delete */
34220	int syncDir /* Not used on win32 */
34221	){
34222	int cnt = 0;
34223	int rc;

34224	void *zConverted;
34225	UNUSED_PARAMETER(pVfs);
34226	UNUSED_PARAMETER(syncDir);
34227
34228	SimulateIOError(return SQLITE_IOERR_DELETE);
	@@ -33710,38 +34229,34 @@
34229	zConverted = convertUtf8Filename(zFilename);
34230	if( zConverted==0 ){
34231	return SQLITE_NOMEM;
34232	}
34233	if( isNT() ){
34234	rc = 1;
34235	while( GetFileAttributesW(zConverted)!=INVALID_FILE_ATTRIBUTES &&
34236	(rc = DeleteFileW(zConverted))==0 && retryIoerr(&cnt) ){}
34237	rc = rc ? SQLITE_OK : SQLITE_ERROR;


34238	/* isNT() is 1 if SQLITE_OS_WINCE==1, so this else is never executed.
34239	** Since the ASCII version of these Windows API do not exist for WINCE,
34240	** it's important to not reference them for WINCE builds.
34241	*/
34242	#if SQLITE_OS_WINCE==0
34243	}else{
34244	rc = 1;
34245	while( GetFileAttributesA(zConverted)!=INVALID_FILE_ATTRIBUTES &&
34246	(rc = DeleteFileA(zConverted))==0 && retryIoerr(&cnt) ){}
34247	rc = rc ? SQLITE_OK : SQLITE_ERROR;


34248	#endif
34249	}
34250	if( rc ){
34251	rc = winLogError(SQLITE_IOERR_DELETE, "winDelete", zFilename);
34252	}else{
34253	logIoerr(cnt);
34254	}
34255	free(zConverted);
34256	OSTRACE(("DELETE \"%s\" %s\n", zFilename, (rc ? "failed" : "ok" )));
34257	return rc;





34258	}
34259
34260	/*
34261	** Check the existance and status of a file.
34262	*/
	@@ -33760,15 +34275,17 @@
34275	zConverted = convertUtf8Filename(zFilename);
34276	if( zConverted==0 ){
34277	return SQLITE_NOMEM;
34278	}
34279	if( isNT() ){
34280	int cnt = 0;
34281	WIN32_FILE_ATTRIBUTE_DATA sAttrData;
34282	memset(&sAttrData, 0, sizeof(sAttrData));
34283	while( !(rc = GetFileAttributesExW((WCHAR*)zConverted,
34284	GetFileExInfoStandard,
34285	&sAttrData)) && retryIoerr(&cnt) ){}
34286	if( rc ){
34287	/* For an SQLITE_ACCESS_EXISTS query, treat a zero-length file
34288	** as if it does not exist.
34289	*/
34290	if( flags==SQLITE_ACCESS_EXISTS
34291	&& sAttrData.nFileSizeHigh==0
	@@ -33776,10 +34293,11 @@
34293	attr = INVALID_FILE_ATTRIBUTES;
34294	}else{
34295	attr = sAttrData.dwFileAttributes;
34296	}
34297	}else{
34298	logIoerr(cnt);
34299	if( GetLastError()!=ERROR_FILE_NOT_FOUND ){
34300	winLogError(SQLITE_IOERR_ACCESS, "winAccess", zFilename);
34301	free(zConverted);
34302	return SQLITE_IOERR_ACCESS;
34303	}else{
	@@ -33800,11 +34318,12 @@
34318	case SQLITE_ACCESS_READ:
34319	case SQLITE_ACCESS_EXISTS:
34320	rc = attr!=INVALID_FILE_ATTRIBUTES;
34321	break;
34322	case SQLITE_ACCESS_READWRITE:
34323	rc = attr!=INVALID_FILE_ATTRIBUTES &&
34324	(attr & FILE_ATTRIBUTE_READONLY)==0;
34325	break;
34326	default:
34327	assert(!"Invalid flags argument");
34328	}
34329	*pResOut = rc;
	@@ -35256,10 +35775,13 @@
35775	typedef struct PCache1 PCache1;
35776	typedef struct PgHdr1 PgHdr1;
35777	typedef struct PgFreeslot PgFreeslot;
35778	typedef struct PGroup PGroup;
35779
35780	typedef struct PGroupBlock PGroupBlock;
35781	typedef struct PGroupBlockList PGroupBlockList;
35782
35783	/* Each page cache (or PCache) belongs to a PGroup. A PGroup is a set
35784	** of one or more PCaches that are able to recycle each others unpinned
35785	** pages when they are under memory pressure. A PGroup is an instance of
35786	** the following object.
35787	**
	@@ -35285,12 +35807,70 @@
35807	int nMaxPage; /* Sum of nMax for purgeable caches */
35808	int nMinPage; /* Sum of nMin for purgeable caches */
35809	int mxPinned; /* nMaxpage + 10 - nMinPage */
35810	int nCurrentPage; /* Number of purgeable pages allocated */
35811	PgHdr1 pLruHead, pLruTail; /* LRU list of unpinned pages */
35812	#ifdef SQLITE_PAGECACHE_BLOCKALLOC
35813	int isBusy; /* Do not run ReleaseMemory() if true */
35814	PGroupBlockList pBlockList; / List of block-lists for this group */
35815	#endif
35816	};
35817
35818	/*
35819	** If SQLITE_PAGECACHE_BLOCKALLOC is defined when the library is built,
35820	** each PGroup structure has a linked list of the the following starting
35821	** at PGroup.pBlockList. There is one entry for each distinct page-size
35822	** currently used by members of the PGroup (i.e. 1024 bytes, 4096 bytes
35823	** etc.). Variable PGroupBlockList.nByte is set to the actual allocation
35824	** size requested by each pcache, which is the database page-size plus
35825	** the various header structures used by the pcache, pager and btree layers.
35826	** Usually around (pgsz+200) bytes.
35827	**
35828	** This size (pgsz+200) bytes is not allocated efficiently by some
35829	** implementations of malloc. In particular, some implementations are only
35830	** able to allocate blocks of memory chunks of 2^N bytes, where N is some
35831	** integer value. Since the page-size is a power of 2, this means we
35832	** end up wasting (pgsz-200) bytes in each allocation.
35833	**
35834	** If SQLITE_PAGECACHE_BLOCKALLOC is defined, the (pgsz+200) byte blocks
35835	** are not allocated directly. Instead, blocks of roughly M*(pgsz+200) bytes
35836	** are requested from malloc allocator. After a block is returned,
35837	** sqlite3MallocSize() is used to determine how many (pgsz+200) byte
35838	** allocations can fit in the space returned by malloc(). This value may
35839	** be more than M.
35840	**
35841	** The blocks are stored in a doubly-linked list. Variable PGroupBlock.nEntry
35842	** contains the number of allocations that will fit in the aData[] space.
35843	** nEntry is limited to the number of bits in bitmask mUsed. If a slot
35844	** within aData is in use, the corresponding bit in mUsed is set. Thus
35845	** when (mUsed+1==(1 << nEntry)) the block is completely full.
35846	**
35847	** Each time a slot within a block is freed, the block is moved to the start
35848	** of the linked-list. And if a block becomes completely full, then it is
35849	** moved to the end of the list. As a result, when searching for a free
35850	** slot, only the first block in the list need be examined. If it is full,
35851	** then it is guaranteed that all blocks are full.
35852	*/
35853	struct PGroupBlockList {
35854	int nByte; /* Size of each allocation in bytes */
35855	PGroupBlock pFirst; / First PGroupBlock in list */
35856	PGroupBlock pLast; / Last PGroupBlock in list */
35857	PGroupBlockList pNext; / Next block-list attached to group */
35858	};
35859
35860	struct PGroupBlock {
35861	Bitmask mUsed; /* Mask of used slots */
35862	int nEntry; /* Maximum number of allocations in aData[] */
35863	u8 aData; / Pointer to data block */
35864	PGroupBlock pNext; / Next PGroupBlock in list */
35865	PGroupBlock pPrev; / Previous PGroupBlock in list */
35866	PGroupBlockList pList; / Owner list */
35867	};
35868
35869	/* Minimum value for PGroupBlock.nEntry */
35870	#define PAGECACHE_BLOCKALLOC_MINENTRY 15
35871
35872	/* Each page cache is an instance of the following object. Every
35873	** open database file (including each in-memory database and each
35874	** temporary or transient database) has a single page cache which
35875	** is an instance of this object.
35876	**
	@@ -35389,10 +35969,21 @@
35969	**
35970	** assert( PGHDR1_TO_PAGE(PAGE_TO_PGHDR1(pCache, X))==X );
35971	*/
35972	#define PGHDR1_TO_PAGE(p) (void)(((char)p) - p->pCache->szPage)
35973	#define PAGE_TO_PGHDR1(c, p) (PgHdr1)(((char)p) + c->szPage)
35974
35975	/*
35976	** Blocks used by the SQLITE_PAGECACHE_BLOCKALLOC blocks to store/retrieve
35977	** a PGroupBlock pointer based on a pointer to a page buffer.
35978	*/
35979	#define PAGE_SET_BLOCKPTR(pCache, pPg, pBlock) \
35980	( (PGroupBlock )&(((u8)pPg)[sizeof(PgHdr1) + pCache->szPage]) = pBlock )
35981
35982	#define PAGE_GET_BLOCKPTR(pCache, pPg) \
35983	( (PGroupBlock )&(((u8)pPg)[sizeof(PgHdr1) + pCache->szPage]) )
35984
35985
35986	/*
35987	** Macros to enter and leave the PCache LRU mutex.
35988	*/
35989	#define pcache1EnterMutex(X) sqlite3_mutex_enter((X)->mutex)
	@@ -35515,25 +36106,159 @@
36106	return iSize;
36107	}
36108	}
36109	#endif /* SQLITE_ENABLE_MEMORY_MANAGEMENT */
36110
36111	#ifdef SQLITE_PAGECACHE_BLOCKALLOC
36112	/*
36113	** The block pBlock belongs to list pList but is not currently linked in.
36114	** Insert it into the start of the list.
36115	*/
36116	static void addBlockToList(PGroupBlockList pList, PGroupBlock pBlock){
36117	pBlock->pPrev = 0;
36118	pBlock->pNext = pList->pFirst;
36119	pList->pFirst = pBlock;
36120	if( pBlock->pNext ){
36121	pBlock->pNext->pPrev = pBlock;
36122	}else{
36123	assert( pList->pLast==0 );
36124	pList->pLast = pBlock;
36125	}
36126	}
36127
36128	/*
36129	** If there are no blocks in the list headed by pList, remove pList
36130	** from the pGroup->pBlockList list and free it with sqlite3_free().
36131	*/
36132	static void freeListIfEmpty(PGroup pGroup, PGroupBlockList pList){
36133	assert( sqlite3_mutex_held(pGroup->mutex) );
36134	if( pList->pFirst==0 ){
36135	PGroupBlockList **pp;
36136	for(pp=&pGroup->pBlockList; pp!=pList; pp=&(pp)->pNext);
36137	pp = (pp)->pNext;
36138	sqlite3_free(pList);
36139	}
36140	}
36141	#endif /* SQLITE_PAGECACHE_BLOCKALLOC */
36142
36143	/*
36144	** Allocate a new page object initially associated with cache pCache.
36145	*/
36146	static PgHdr1 pcache1AllocPage(PCache1 pCache){
36147	int nByte = sizeof(PgHdr1) + pCache->szPage;
36148	void *pPg = 0;
36149	PgHdr1 *p;
36150
36151	#ifdef SQLITE_PAGECACHE_BLOCKALLOC
36152	PGroup *pGroup = pCache->pGroup;
36153	PGroupBlockList *pList;
36154	PGroupBlock *pBlock;
36155	int i;
36156
36157	nByte += sizeof(PGroupBlockList *);
36158	nByte = ROUND8(nByte);
36159
36160	for(pList=pGroup->pBlockList; pList; pList=pList->pNext){
36161	if( pList->nByte==nByte ) break;
36162	}
36163	if( pList==0 ){
36164	PGroupBlockList *pNew;
36165	assert( pGroup->isBusy==0 );
36166	assert( sqlite3_mutex_held(pGroup->mutex) );
36167	pGroup->isBusy = 1; /* Disable sqlite3PcacheReleaseMemory() */
36168	pNew = (PGroupBlockList *)sqlite3MallocZero(sizeof(PGroupBlockList));
36169	pGroup->isBusy = 0; /* Reenable sqlite3PcacheReleaseMemory() */
36170	if( pNew==0 ){
36171	/* malloc() failure. Return early. */
36172	return 0;
36173	}
36174	#ifdef SQLITE_DEBUG
36175	for(pList=pGroup->pBlockList; pList; pList=pList->pNext){
36176	assert( pList->nByte!=nByte );
36177	}
36178	#endif
36179	pNew->nByte = nByte;
36180	pNew->pNext = pGroup->pBlockList;
36181	pGroup->pBlockList = pNew;
36182	pList = pNew;
36183	}
36184
36185	pBlock = pList->pFirst;
36186	if( pBlock==0 \|\| pBlock->mUsed==(((Bitmask)1<<pBlock->nEntry)-1) ){
36187	int sz;
36188
36189	/* Allocate a new block. Try to allocate enough space for the PGroupBlock
36190	** structure and MINENTRY allocations of nByte bytes each. If the
36191	** allocator returns more memory than requested, then more than MINENTRY
36192	** allocations may fit in it. */
36193	assert( sqlite3_mutex_held(pGroup->mutex) );
36194	pcache1LeaveMutex(pCache->pGroup);
36195	sz = sizeof(PGroupBlock) + PAGECACHE_BLOCKALLOC_MINENTRY * nByte;
36196	pBlock = (PGroupBlock *)sqlite3Malloc(sz);
36197	pcache1EnterMutex(pCache->pGroup);
36198
36199	if( !pBlock ){
36200	freeListIfEmpty(pGroup, pList);
36201	return 0;
36202	}
36203	pBlock->nEntry = (sqlite3MallocSize(pBlock) - sizeof(PGroupBlock)) / nByte;
36204	if( pBlock->nEntry>=BMS ){
36205	pBlock->nEntry = BMS-1;
36206	}
36207	pBlock->pList = pList;
36208	pBlock->mUsed = 0;
36209	pBlock->aData = (u8 *)&pBlock[1];
36210	addBlockToList(pList, pBlock);
36211
36212	sz = sqlite3MallocSize(pBlock);
36213	sqlite3_mutex_enter(pcache1.mutex);
36214	sqlite3StatusAdd(SQLITE_STATUS_PAGECACHE_OVERFLOW, sz);
36215	sqlite3_mutex_leave(pcache1.mutex);
36216	}
36217
36218	for(i=0; pPg==0 && ALWAYS(i<pBlock->nEntry); i++){
36219	if( 0==(pBlock->mUsed & ((Bitmask)1<<i)) ){
36220	pBlock->mUsed \|= ((Bitmask)1<<i);
36221	pPg = (void )&pBlock->aData[pList->nByte i];
36222	}
36223	}
36224	assert( pPg );
36225	PAGE_SET_BLOCKPTR(pCache, pPg, pBlock);
36226
36227	/* If the block is now full, shift it to the end of the list */
36228	if( pBlock->mUsed==(((Bitmask)1<<pBlock->nEntry)-1) && pList->pLast!=pBlock ){
36229	assert( pList->pFirst==pBlock );
36230	assert( pBlock->pPrev==0 );
36231	assert( pList->pLast->pNext==0 );
36232	pList->pFirst = pBlock->pNext;
36233	pList->pFirst->pPrev = 0;
36234	pBlock->pPrev = pList->pLast;
36235	pBlock->pNext = 0;
36236	pList->pLast->pNext = pBlock;
36237	pList->pLast = pBlock;
36238	}
36239	p = PAGE_TO_PGHDR1(pCache, pPg);
36240	if( pCache->bPurgeable ){
36241	pCache->pGroup->nCurrentPage++;
36242	}
36243	#else
36244	/* The group mutex must be released before pcache1Alloc() is called. This
36245	** is because it may call sqlite3_release_memory(), which assumes that
36246	** this mutex is not held. */
36247	assert( sqlite3_mutex_held(pCache->pGroup->mutex) );
36248	pcache1LeaveMutex(pCache->pGroup);
36249	pPg = pcache1Alloc(nByte);
36250	pcache1EnterMutex(pCache->pGroup);
36251	if( pPg ){
36252	p = PAGE_TO_PGHDR1(pCache, pPg);
36253	if( pCache->bPurgeable ){
36254	pCache->pGroup->nCurrentPage++;
36255	}
36256	}else{
36257	p = 0;
36258	}
36259	#endif
36260	return p;
36261	}
36262
36263	/*
36264	** Free a page object allocated by pcache1AllocPage().
	@@ -35543,14 +36268,56 @@
36268	** with a NULL pointer, so we mark the NULL test with ALWAYS().
36269	*/
36270	static void pcache1FreePage(PgHdr1 *p){
36271	if( ALWAYS(p) ){
36272	PCache1 *pCache = p->pCache;
36273	void *pPg = PGHDR1_TO_PAGE(p);
36274
36275	#ifdef SQLITE_PAGECACHE_BLOCKALLOC
36276	PGroupBlock *pBlock = PAGE_GET_BLOCKPTR(pCache, pPg);
36277	PGroupBlockList *pList = pBlock->pList;
36278	int i = ((u8 *)pPg - pBlock->aData) / pList->nByte;
36279
36280	assert( pPg==(void )&pBlock->aData[ipList->nByte] );
36281	assert( pBlock->mUsed & ((Bitmask)1<<i) );
36282	pBlock->mUsed &= ~((Bitmask)1<<i);
36283
36284	/* Remove the block from the list. If it is completely empty, free it.
36285	** Or if it is not completely empty, re-insert it at the start of the
36286	** list. */
36287	if( pList->pFirst==pBlock ){
36288	pList->pFirst = pBlock->pNext;
36289	if( pList->pFirst ) pList->pFirst->pPrev = 0;
36290	}else{
36291	pBlock->pPrev->pNext = pBlock->pNext;
36292	}
36293	if( pList->pLast==pBlock ){
36294	pList->pLast = pBlock->pPrev;
36295	if( pList->pLast ) pList->pLast->pNext = 0;
36296	}else{
36297	pBlock->pNext->pPrev = pBlock->pPrev;
36298	}
36299
36300	if( pBlock->mUsed==0 ){
36301	PGroup *pGroup = p->pCache->pGroup;
36302
36303	int sz = sqlite3MallocSize(pBlock);
36304	sqlite3_mutex_enter(pcache1.mutex);
36305	sqlite3StatusAdd(SQLITE_STATUS_PAGECACHE_OVERFLOW, -sz);
36306	sqlite3_mutex_leave(pcache1.mutex);
36307	freeListIfEmpty(pGroup, pList);
36308	sqlite3_free(pBlock);
36309	}else{
36310	addBlockToList(pList, pBlock);
36311	}
36312	#else
36313	assert( sqlite3_mutex_held(p->pCache->pGroup->mutex) );
36314	pcache1Free(pPg);
36315	#endif
36316	if( pCache->bPurgeable ){
36317	pCache->pGroup->nCurrentPage--;
36318	}

36319	}
36320	}
36321
36322	/*
36323	** Malloc function used by SQLite to obtain space from the buffer configured
	@@ -35984,13 +36751,11 @@
36751	/* Step 5. If a usable page buffer has still not been found,
36752	** attempt to allocate a new one.
36753	*/
36754	if( !pPage ){
36755	if( createFlag==1 ) sqlite3BeginBenignMalloc();

36756	pPage = pcache1AllocPage(pCache);

36757	if( createFlag==1 ) sqlite3EndBenignMalloc();
36758	}
36759
36760	if( pPage ){
36761	unsigned int h = iKey % pCache->nHash;
	@@ -36156,10 +36921,13 @@
36921	** been released, the function returns. The return value is the total number
36922	** of bytes of memory released.
36923	*/
36924	SQLITE_PRIVATE int sqlite3PcacheReleaseMemory(int nReq){
36925	int nFree = 0;
36926	#ifdef SQLITE_PAGECACHE_BLOCKALLOC
36927	if( pcache1.grp.isBusy ) return 0;
36928	#endif
36929	assert( sqlite3_mutex_notheld(pcache1.grp.mutex) );
36930	assert( sqlite3_mutex_notheld(pcache1.mutex) );
36931	if( pcache1.pStart==0 ){
36932	PgHdr1 *p;
36933	pcache1EnterMutex(&pcache1.grp);
	@@ -37368,10 +38136,12 @@
38136	u8 ckptSyncFlags; /* SYNC_NORMAL or SYNC_FULL for checkpoint */
38137	u8 syncFlags; /* SYNC_NORMAL or SYNC_FULL otherwise */
38138	u8 tempFile; /* zFilename is a temporary file */
38139	u8 readOnly; /* True for a read-only database */
38140	u8 memDb; /* True to inhibit all file I/O */
38141	u8 hasSeenStress; /* pagerStress() called one or more times */
38142	u8 isSorter; /* True for a PAGER_SORTER */
38143
38144	/**************************************************************************
38145	** The following block contains those class members that change during
38146	** routine opertion. Class members not in this block are either fixed
38147	** when the pager is first created or else only change when there is a
	@@ -37590,10 +38360,19 @@
38360	\|\| p->journalMode==PAGER_JOURNALMODE_MEMORY
38361	);
38362	assert( p->eState!=PAGER_ERROR && p->eState!=PAGER_OPEN );
38363	assert( pagerUseWal(p)==0 );
38364	}
38365
38366	/* A sorter is a temp file that never spills to disk and always has
38367	** the doNotSpill flag set
38368	*/
38369	if( p->isSorter ){
38370	assert( p->tempFile );
38371	assert( p->doNotSpill );
38372	assert( p->fd->pMethods==0 );
38373	}
38374
38375	/* If changeCountDone is set, a RESERVED lock or greater must be held
38376	** on the file.
38377	*/
38378	assert( pPager->changeCountDone==0 \|\| pPager->eLock>=RESERVED_LOCK );
	@@ -40487,10 +41266,11 @@
41266	** to the caller.
41267	*/
41268	SQLITE_PRIVATE int sqlite3PagerClose(Pager *pPager){
41269	u8 pTmp = (u8 )pPager->pTmpSpace;
41270
41271	assert( assert_pager_state(pPager) );
41272	disable_simulated_io_errors();
41273	sqlite3BeginBenignMalloc();
41274	/* pPager->errCode = 0; */
41275	pPager->exclusiveMode = 0;
41276	#ifndef SQLITE_OMIT_WAL
	@@ -40921,10 +41701,11 @@
41701	** is impossible for sqlite3PCacheFetch() to be called with createFlag==1
41702	** while in the error state, hence it is impossible for this routine to
41703	** be called in the error state. Nevertheless, we include a NEVER()
41704	** test for the error state as a safeguard against future changes.
41705	*/
41706	pPager->hasSeenStress = 1;
41707	if( NEVER(pPager->errCode) ) return SQLITE_OK;
41708	if( pPager->doNotSpill ) return SQLITE_OK;
41709	if( pPager->doNotSyncSpill && (pPg->flags & PGHDR_NEED_SYNC)!=0 ){
41710	return SQLITE_OK;
41711	}
	@@ -41292,10 +42073,16 @@
42073	}
42074	/* pPager->xBusyHandler = 0; */
42075	/* pPager->pBusyHandlerArg = 0; */
42076	pPager->xReiniter = xReinit;
42077	/* memset(pPager->aHash, 0, sizeof(pPager->aHash)); */
42078	#ifndef SQLITE_OMIT_MERGE_SORT
42079	if( flags & PAGER_SORTER ){
42080	pPager->doNotSpill = 1;
42081	pPager->isSorter = 1;
42082	}
42083	#endif
42084
42085	*ppPager = pPager;
42086	return SQLITE_OK;
42087	}
42088
	@@ -42835,10 +43622,21 @@
43622	** Return true if this is an in-memory pager.
43623	*/
43624	SQLITE_PRIVATE int sqlite3PagerIsMemdb(Pager *pPager){
43625	return MEMDB;
43626	}
43627
43628	#ifndef SQLITE_OMIT_MERGE_SORT
43629	/*
43630	** Return true if the pager has seen a pagerStress callback.
43631	*/
43632	SQLITE_PRIVATE int sqlite3PagerUnderStress(Pager *pPager){
43633	assert( pPager->isSorter );
43634	assert( pPager->doNotSpill );
43635	return pPager->hasSeenStress;
43636	}
43637	#endif
43638
43639	/*
43640	** Check that there are at least nSavepoint savepoints open. If there are
43641	** currently less than nSavepoints open, then open one or more savepoints
43642	** to make up the difference. If the number of savepoints is already
	@@ -45384,17 +46182,19 @@
46182	**
46183	** The EXCLUSIVE lock is not released before returning.
46184	*/
46185	rc = sqlite3OsLock(pWal->pDbFd, SQLITE_LOCK_EXCLUSIVE);
46186	if( rc==SQLITE_OK ){
46187	int bPersistWal = -1;
46188	if( pWal->exclusiveMode==WAL_NORMAL_MODE ){
46189	pWal->exclusiveMode = WAL_EXCLUSIVE_MODE;
46190	}
46191	rc = sqlite3WalCheckpoint(
46192	pWal, SQLITE_CHECKPOINT_PASSIVE, 0, 0, sync_flags, nBuf, zBuf, 0, 0
46193	);
46194	sqlite3OsFileControl(pWal->pDbFd, SQLITE_FCNTL_PERSIST_WAL, &bPersistWal);
46195	if( rc==SQLITE_OK && bPersistWal!=1 ){
46196	isDelete = 1;
46197	}
46198	}
46199
46200	walIndexClose(pWal, isDelete);
	@@ -49202,15 +50002,26 @@
50002	assert( (flags & BTREE_UNORDERED)==0 \|\| (flags & BTREE_SINGLE)!=0 );
50003
50004	/* A BTREE_SINGLE database is always a temporary and/or ephemeral */
50005	assert( (flags & BTREE_SINGLE)==0 \|\| isTempDb );
50006
50007	/* The BTREE_SORTER flag is only used if SQLITE_OMIT_MERGE_SORT is undef */
50008	#ifdef SQLITE_OMIT_MERGE_SORT
50009	assert( (flags & BTREE_SORTER)==0 );
50010	#endif
50011
50012	/* BTREE_SORTER is always on a BTREE_SINGLE, BTREE_OMIT_JOURNAL */
50013	assert( (flags & BTREE_SORTER)==0 \|\|
50014	(flags & (BTREE_SINGLE\|BTREE_OMIT_JOURNAL))
50015	==(BTREE_SINGLE\|BTREE_OMIT_JOURNAL) );
50016
50017	if( db->flags & SQLITE_NoReadlock ){
50018	flags \|= BTREE_NO_READLOCK;
50019	}
50020	if( isMemdb ){
50021	flags \|= BTREE_MEMORY;
50022	flags &= ~BTREE_SORTER;
50023	}
50024	if( (vfsFlags & SQLITE_OPEN_MAIN_DB)!=0 && (isMemdb \|\| isTempDb) ){
50025	vfsFlags = (vfsFlags & ~SQLITE_OPEN_MAIN_DB) \| SQLITE_OPEN_TEMP_DB;
50026	}
50027	p = sqlite3MallocZero(sizeof(Btree));
	@@ -50936,11 +51747,12 @@
51747
51748	if( NEVER(wrFlag && pBt->readOnly) ){
51749	return SQLITE_READONLY;
51750	}
51751	if( iTable==1 && btreePagecount(pBt)==0 ){
51752	assert( wrFlag==0 );
51753	iTable = 0;
51754	}
51755
51756	/* Now that no other errors can occur, finish filling in the BtCursor
51757	** variables and link the cursor into the BtShared list. */
51758	pCur->pgnoRoot = (Pgno)iTable;
	@@ -51690,10 +52502,13 @@
52502	int i;
52503	for(i=1; i<=pCur->iPage; i++){
52504	releasePage(pCur->apPage[i]);
52505	}
52506	pCur->iPage = 0;
52507	}else if( pCur->pgnoRoot==0 ){
52508	pCur->eState = CURSOR_INVALID;
52509	return SQLITE_OK;
52510	}else{
52511	rc = getAndInitPage(pBt, pCur->pgnoRoot, &pCur->apPage[0]);
52512	if( rc!=SQLITE_OK ){
52513	pCur->eState = CURSOR_INVALID;
52514	return rc;
	@@ -51799,11 +52614,11 @@
52614	assert( cursorHoldsMutex(pCur) );
52615	assert( sqlite3_mutex_held(pCur->pBtree->db->mutex) );
52616	rc = moveToRoot(pCur);
52617	if( rc==SQLITE_OK ){
52618	if( pCur->eState==CURSOR_INVALID ){
52619	assert( pCur->pgnoRoot==0 \|\| pCur->apPage[pCur->iPage]->nCell==0 );
52620	*pRes = 1;
52621	}else{
52622	assert( pCur->apPage[pCur->iPage]->nCell>0 );
52623	*pRes = 0;
52624	rc = moveToLeftmost(pCur);
	@@ -51838,11 +52653,11 @@
52653	}
52654
52655	rc = moveToRoot(pCur);
52656	if( rc==SQLITE_OK ){
52657	if( CURSOR_INVALID==pCur->eState ){
52658	assert( pCur->pgnoRoot==0 \|\| pCur->apPage[pCur->iPage]->nCell==0 );
52659	*pRes = 1;
52660	}else{
52661	assert( pCur->eState==CURSOR_VALID );
52662	*pRes = 0;
52663	rc = moveToRightmost(pCur);
	@@ -51911,16 +52726,16 @@
52726
52727	rc = moveToRoot(pCur);
52728	if( rc ){
52729	return rc;
52730	}
52731	assert( pCur->pgnoRoot==0 \|\| pCur->apPage[pCur->iPage] );
52732	assert( pCur->pgnoRoot==0 \|\| pCur->apPage[pCur->iPage]->isInit );
52733	assert( pCur->eState==CURSOR_INVALID \|\| pCur->apPage[pCur->iPage]->nCell>0 );
52734	if( pCur->eState==CURSOR_INVALID ){
52735	*pRes = -1;
52736	assert( pCur->pgnoRoot==0 \|\| pCur->apPage[pCur->iPage]->nCell==0 );
52737	return SQLITE_OK;
52738	}
52739	assert( pCur->apPage[0]->intKey \|\| pIdxKey );
52740	for(;;){
52741	int lwr, upr, idx;
	@@ -54745,13 +55560,20 @@
55560	releasePage(pPage);
55561	}
55562	return rc;
55563	}
55564	SQLITE_PRIVATE int sqlite3BtreeDropTable(Btree p, int iTable, int piMoved){
55565	BtShared *pBt = p->pBt;
55566	int rc;
55567	sqlite3BtreeEnter(p);
55568	if( (pBt->openFlags&BTREE_SINGLE) ){
55569	pBt->nPage = 0;
55570	sqlite3PagerTruncateImage(pBt->pPager, 1);
55571	rc = newDatabase(pBt);
55572	}else{
55573	rc = btreeDropTable(p, iTable, piMoved);
55574	}
55575	sqlite3BtreeLeave(p);
55576	return rc;
55577	}
55578
55579
	@@ -54826,10 +55648,15 @@
55648	** corruption) an SQLite error code is returned.
55649	*/
55650	SQLITE_PRIVATE int sqlite3BtreeCount(BtCursor pCur, i64 pnEntry){
55651	i64 nEntry = 0; /* Value to return in pnEntry /
55652	int rc; /* Return code */
55653
55654	if( pCur->pgnoRoot==0 ){
55655	*pnEntry = 0;
55656	return SQLITE_OK;
55657	}
55658	rc = moveToRoot(pCur);
55659
55660	/* Unless an error occurs, the following loop runs one iteration for each
55661	** page in the B-Tree structure (not including overflow pages).
55662	*/
	@@ -55610,11 +56437,10 @@
56437	*/
56438	SQLITE_PRIVATE int sqlite3BtreeSetVersion(Btree *pBtree, int iVersion){
56439	BtShared *pBt = pBtree->pBt;
56440	int rc; /* Return code */
56441

56442	assert( iVersion==1 \|\| iVersion==2 );
56443
56444	/* If setting the version fields to 1, do not automatically open the
56445	** WAL connection, even if the version fields are currently set to 2.
56446	*/
	@@ -56049,106 +56875,110 @@
56875	/* Update the schema version field in the destination database. This
56876	** is to make sure that the schema-version really does change in
56877	** the case where the source and destination databases have the
56878	** same schema version.
56879	*/
56880	if( rc==SQLITE_DONE ){
56881	rc = sqlite3BtreeUpdateMeta(p->pDest,1,p->iDestSchema+1);
56882	if( rc==SQLITE_OK ){
56883	if( p->pDestDb ){
56884	sqlite3ResetInternalSchema(p->pDestDb, -1);
56885	}
56886	if( destMode==PAGER_JOURNALMODE_WAL ){
56887	rc = sqlite3BtreeSetVersion(p->pDest, 2);
56888	}
56889	}
56890	if( rc==SQLITE_OK ){
56891	int nDestTruncate;
56892	/* Set nDestTruncate to the final number of pages in the destination
56893	** database. The complication here is that the destination page
56894	** size may be different to the source page size.
56895	**
56896	** If the source page size is smaller than the destination page size,
56897	** round up. In this case the call to sqlite3OsTruncate() below will
56898	** fix the size of the file. However it is important to call
56899	** sqlite3PagerTruncateImage() here so that any pages in the
56900	** destination file that lie beyond the nDestTruncate page mark are
56901	** journalled by PagerCommitPhaseOne() before they are destroyed
56902	** by the file truncation.
56903	*/
56904	assert( pgszSrc==sqlite3BtreeGetPageSize(p->pSrc) );
56905	assert( pgszDest==sqlite3BtreeGetPageSize(p->pDest) );
56906	if( pgszSrc<pgszDest ){
56907	int ratio = pgszDest/pgszSrc;
56908	nDestTruncate = (nSrcPage+ratio-1)/ratio;
56909	if( nDestTruncate==(int)PENDING_BYTE_PAGE(p->pDest->pBt) ){
56910	nDestTruncate--;
56911	}
56912	}else{
56913	nDestTruncate = nSrcPage * (pgszSrc/pgszDest);
56914	}
56915	sqlite3PagerTruncateImage(pDestPager, nDestTruncate);
56916
56917	if( pgszSrc<pgszDest ){
56918	/* If the source page-size is smaller than the destination page-size,
56919	** two extra things may need to happen:
56920	**
56921	** * The destination may need to be truncated, and
56922	**
56923	** * Data stored on the pages immediately following the
56924	** pending-byte page in the source database may need to be
56925	** copied into the destination database.
56926	*/
56927	const i64 iSize = (i64)pgszSrc * (i64)nSrcPage;
56928	sqlite3_file * const pFile = sqlite3PagerFile(pDestPager);
56929	i64 iOff;
56930	i64 iEnd;
56931
56932	assert( pFile );
56933	assert( (i64)nDestTruncate*(i64)pgszDest >= iSize \|\| (
56934	nDestTruncate==(int)(PENDING_BYTE_PAGE(p->pDest->pBt)-1)
56935	&& iSize>=PENDING_BYTE && iSize<=PENDING_BYTE+pgszDest
56936	));
56937
56938	/* This call ensures that all data required to recreate the original
56939	** database has been stored in the journal for pDestPager and the
56940	** journal synced to disk. So at this point we may safely modify
56941	** the database file in any way, knowing that if a power failure
56942	** occurs, the original database will be reconstructed from the
56943	** journal file. */
56944	rc = sqlite3PagerCommitPhaseOne(pDestPager, 0, 1);
56945
56946	/* Write the extra pages and truncate the database file as required */
56947	iEnd = MIN(PENDING_BYTE + pgszDest, iSize);
56948	for(
56949	iOff=PENDING_BYTE+pgszSrc;
56950	rc==SQLITE_OK && iOff<iEnd;
56951	iOff+=pgszSrc
56952	){
56953	PgHdr *pSrcPg = 0;
56954	const Pgno iSrcPg = (Pgno)((iOff/pgszSrc)+1);
56955	rc = sqlite3PagerGet(pSrcPager, iSrcPg, &pSrcPg);
56956	if( rc==SQLITE_OK ){
56957	u8 *zData = sqlite3PagerGetData(pSrcPg);
56958	rc = sqlite3OsWrite(pFile, zData, pgszSrc, iOff);
56959	}
56960	sqlite3PagerUnref(pSrcPg);
56961	}
56962	if( rc==SQLITE_OK ){
56963	rc = backupTruncateFile(pFile, iSize);
56964	}
56965
56966	/* Sync the database file to disk. */
56967	if( rc==SQLITE_OK ){
56968	rc = sqlite3PagerSync(pDestPager);
56969	}
56970	}else{
56971	rc = sqlite3PagerCommitPhaseOne(pDestPager, 0, 0);
56972	}
56973
56974	/* Finish committing the transaction to the destination database. */
56975	if( SQLITE_OK==rc
56976	&& SQLITE_OK==(rc = sqlite3BtreeCommitPhaseTwo(p->pDest, 0))
56977	){
56978	rc = SQLITE_DONE;
56979	}
56980	}
56981	}
56982
56983	/* If bCloseTrans is true, then this function opened a read transaction
56984	** on the source database. Close the read transaction here. There is
	@@ -56612,38 +57442,32 @@
57442	** invoking an external callback, free it now. Calling this function
57443	** does not free any Mem.zMalloc buffer.
57444	*/
57445	SQLITE_PRIVATE void sqlite3VdbeMemReleaseExternal(Mem *p){
57446	assert( p->db==0 \|\| sqlite3_mutex_held(p->db->mutex) );
57447	if( p->flags&MEM_Agg ){
57448	sqlite3VdbeMemFinalize(p, p->u.pDef);
57449	assert( (p->flags & MEM_Agg)==0 );
57450	sqlite3VdbeMemRelease(p);
57451	}else if( p->flags&MEM_Dyn && p->xDel ){
57452	assert( (p->flags&MEM_RowSet)==0 );
57453	p->xDel((void *)p->z);
57454	p->xDel = 0;
57455	}else if( p->flags&MEM_RowSet ){
57456	sqlite3RowSetClear(p->u.pRowSet);
57457	}else if( p->flags&MEM_Frame ){
57458	sqlite3VdbeMemSetNull(p);






57459	}
57460	}
57461
57462	/*
57463	** Release any memory held by the Mem. This may leave the Mem in an
57464	** inconsistent state, for example with (Mem.z==0) and
57465	** (Mem.type==SQLITE_TEXT).
57466	*/
57467	SQLITE_PRIVATE void sqlite3VdbeMemRelease(Mem *p){
57468	MemReleaseExt(p);
57469	sqlite3DbFree(p->db, p->zMalloc);
57470	p->z = 0;
57471	p->zMalloc = 0;
57472	p->xDel = 0;
57473	}
	@@ -56961,11 +57785,11 @@
57785	** pFrom->z is used, then pTo->z points to the same thing as pFrom->z
57786	** and flags gets srcType (either MEM_Ephem or MEM_Static).
57787	*/
57788	SQLITE_PRIVATE void sqlite3VdbeMemShallowCopy(Mem pTo, const Mem pFrom, int srcType){
57789	assert( (pFrom->flags & MEM_RowSet)==0 );
57790	MemReleaseExt(pTo);
57791	memcpy(pTo, pFrom, MEMCELLSIZE);
57792	pTo->xDel = 0;
57793	if( (pFrom->flags&MEM_Static)==0 ){
57794	pTo->flags &= ~(MEM_Dyn\|MEM_Static\|MEM_Ephem);
57795	assert( srcType==MEM_Ephem \|\| srcType==MEM_Static );
	@@ -56979,11 +57803,11 @@
57803	*/
57804	SQLITE_PRIVATE int sqlite3VdbeMemCopy(Mem pTo, const Mem pFrom){
57805	int rc = SQLITE_OK;
57806
57807	assert( (pFrom->flags & MEM_RowSet)==0 );
57808	MemReleaseExt(pTo);
57809	memcpy(pTo, pFrom, MEMCELLSIZE);
57810	pTo->flags &= ~MEM_Dyn;
57811
57812	if( pTo->flags&(MEM_Str\|MEM_Blob) ){
57813	if( 0==(pFrom->flags&MEM_Static) ){
	@@ -57934,10 +58758,16 @@
58758	assert( p->nOp - i >= 3 );
58759	assert( pOp[-1].opcode==OP_Integer );
58760	n = pOp[-1].p1;
58761	if( n>nMaxArgs ) nMaxArgs = n;
58762	#endif
58763	}else if( opcode==OP_Next ){
58764	pOp->p4.xAdvance = sqlite3BtreeNext;
58765	pOp->p4type = P4_ADVANCE;
58766	}else if( opcode==OP_Prev ){
58767	pOp->p4.xAdvance = sqlite3BtreePrevious;
58768	pOp->p4type = P4_ADVANCE;
58769	}
58770
58771	if( (pOp->opflags & OPFLG_JUMP)!=0 && pOp->p2<0 ){
58772	assert( -1-pOp->p2<p->nLabel );
58773	pOp->p2 = aLabel[-1-pOp->p2];
	@@ -58025,37 +58855,34 @@
58855	** Change the value of the P1 operand for a specific instruction.
58856	** This routine is useful when a large program is loaded from a
58857	** static array using sqlite3VdbeAddOpList but we want to make a
58858	** few minor changes to the program.
58859	*/
58860	SQLITE_PRIVATE void sqlite3VdbeChangeP1(Vdbe *p, u32 addr, int val){
58861	assert( p!=0 );
58862	if( ((u32)p->nOp)>addr ){

58863	p->aOp[addr].p1 = val;
58864	}
58865	}
58866
58867	/*
58868	** Change the value of the P2 operand for a specific instruction.
58869	** This routine is useful for setting a jump destination.
58870	*/
58871	SQLITE_PRIVATE void sqlite3VdbeChangeP2(Vdbe *p, u32 addr, int val){
58872	assert( p!=0 );
58873	if( ((u32)p->nOp)>addr ){

58874	p->aOp[addr].p2 = val;
58875	}
58876	}
58877
58878	/*
58879	** Change the value of the P3 operand for a specific instruction.
58880	*/
58881	SQLITE_PRIVATE void sqlite3VdbeChangeP3(Vdbe *p, u32 addr, int val){
58882	assert( p!=0 );
58883	if( ((u32)p->nOp)>addr ){

58884	p->aOp[addr].p3 = val;
58885	}
58886	}
58887
58888	/*
	@@ -58442,10 +59269,14 @@
59269	break;
59270	}
59271	case P4_SUBPROGRAM: {
59272	sqlite3_snprintf(nTemp, zTemp, "program");
59273	break;
59274	}
59275	case P4_ADVANCE: {
59276	zTemp[0] = 0;
59277	break;
59278	}
59279	default: {
59280	zP4 = pOp->p4.z;
59281	if( zP4==0 ){
59282	zP4 = zTemp;
	@@ -59006,10 +59837,11 @@
59837	nMem = 10;
59838	}
59839	memset(zCsr, 0, zEnd-zCsr);
59840	zCsr += (zCsr - (u8*)0)&7;
59841	assert( EIGHT_BYTE_ALIGNMENT(zCsr) );
59842	p->expired = 0;
59843
59844	/* Memory for registers, parameters, cursor, etc, is allocated in two
59845	** passes. On the first pass, we try to reuse unused space at the
59846	** end of the opcode array. If we are unable to satisfy all memory
59847	** requirements by reusing the opcode array tail, then the second
	@@ -59065,10 +59897,11 @@
59897	*/
59898	SQLITE_PRIVATE void sqlite3VdbeFreeCursor(Vdbe p, VdbeCursor pCx){
59899	if( pCx==0 ){
59900	return;
59901	}
59902	sqlite3VdbeSorterClose(p->db, pCx);
59903	if( pCx->pBt ){
59904	sqlite3BtreeClose(pCx->pBt);
59905	/* The pCx->pCursor will be close automatically, if it exists, by
59906	** the call above. */
59907	}else if( pCx->pCursor ){
	@@ -61205,10 +62038,18 @@
62038	rc = db->errCode = p->rc;
62039	}
62040	return (rc&db->errMask);
62041	}
62042
62043	/*
62044	** The maximum number of times that a statement will try to reparse
62045	** itself before giving up and returning SQLITE_SCHEMA.
62046	*/
62047	#ifndef SQLITE_MAX_SCHEMA_RETRY
62048	# define SQLITE_MAX_SCHEMA_RETRY 5
62049	#endif
62050
62051	/*
62052	** This is the top-level implementation of sqlite3_step(). Call
62053	** sqlite3Step() to do most of the work. If a schema error occurs,
62054	** call sqlite3Reprepare() and try again.
62055	*/
	@@ -61223,14 +62064,14 @@
62064	return SQLITE_MISUSE_BKPT;
62065	}
62066	db = v->db;
62067	sqlite3_mutex_enter(db->mutex);
62068	while( (rc = sqlite3Step(v))==SQLITE_SCHEMA
62069	&& cnt++ < SQLITE_MAX_SCHEMA_RETRY
62070	&& (rc2 = rc = sqlite3Reprepare(v))==SQLITE_OK ){
62071	sqlite3_reset(pStmt);
62072	assert( v->expired==0 );
62073	}
62074	if( rc2!=SQLITE_OK && ALWAYS(v->isPrepareV2) && ALWAYS(db->pErr) ){
62075	/* This case occurs after failing to recompile an sql statement.
62076	** The error message from the SQL compiler has already been loaded
62077	** into the database handle. This block copies the error message
	@@ -62357,10 +63198,17 @@
63198	** Call sqlite3VdbeMemExpandBlob() on the supplied value (type Mem*)
63199	** P if required.
63200	*/
63201	#define ExpandBlob(P) (((P)->flags&MEM_Zero)?sqlite3VdbeMemExpandBlob(P):0)
63202
63203	/* Return true if the cursor was opened using the OP_OpenSorter opcode. */
63204	#ifdef SQLITE_OMIT_MERGE_SORT
63205	# define isSorter(x) 0
63206	#else
63207	# define isSorter(x) ((x)->pSorter!=0)
63208	#endif
63209
63210	/*
63211	** Argument pMem points at a register that will be passed to a
63212	** user-defined function or returned to the user as the result of a query.
63213	** This routine sets the pMem->type variable used by the sqlite3_value_*()
63214	** routines.
	@@ -62951,10 +63799,11 @@
63799	u8 zEndHdr; / Pointer to first byte after the header */
63800	u32 offset; /* Offset into the data */
63801	u32 szField; /* Number of bytes in the content of a field */
63802	int szHdr; /* Size of the header size field at start of record */
63803	int avail; /* Number of bytes of available data */
63804	u32 t; /* A type code from the record header */
63805	Mem pReg; / PseudoTable input register */
63806	} am;
63807	struct OP_Affinity_stack_vars {
63808	const char zAffinity; / The affinity to be applied */
63809	char cAff; /* A single character of affinity */
	@@ -63109,11 +63958,10 @@
63958	BtCursor *pCrsr;
63959	int res;
63960	} bl;
63961	struct OP_Next_stack_vars {
63962	VdbeCursor *pC;

63963	int res;
63964	} bm;
63965	struct OP_IdxInsert_stack_vars {
63966	VdbeCursor *pC;
63967	BtCursor *pCrsr;
	@@ -63363,11 +64211,11 @@
64211	if( pOp->opflags & OPFLG_OUT2_PRERELEASE ){
64212	assert( pOp->p2>0 );
64213	assert( pOp->p2<=p->nMem );
64214	pOut = &aMem[pOp->p2];
64215	memAboutToChange(p, pOut);
64216	MemReleaseExt(pOut);
64217	pOut->flags = MEM_Int;
64218	}
64219
64220	/* Sanity checking on other operands */
64221	#ifdef SQLITE_DEBUG
	@@ -63496,11 +64344,11 @@
64344	break;
64345	}
64346
64347	/* Opcode: HaltIfNull P1 P2 P3 P4 *
64348	**
64349	** Check the value in register P3. If it is NULL then Halt using
64350	** parameter P1, P2, and P4 as if this were a Halt instruction. If the
64351	** value in register P3 is not NULL, then this routine is a no-op.
64352	*/
64353	case OP_HaltIfNull: { /* in3 */
64354	pIn3 = &aMem[pOp->p3];
	@@ -63730,10 +64578,15 @@
64578	assert( memIsValid(pIn1) );
64579	memAboutToChange(p, pOut);
64580	u.ac.zMalloc = pOut->zMalloc;
64581	pOut->zMalloc = 0;
64582	sqlite3VdbeMemMove(pOut, pIn1);
64583	#ifdef SQLITE_DEBUG
64584	if( pOut->pScopyFrom>=&aMem[u.ac.p1] && pOut->pScopyFrom<&aMem[u.ac.p1+pOp->p3] ){
64585	pOut->pScopyFrom += u.ac.p1 - pOp->p2;
64586	}
64587	#endif
64588	pIn1->zMalloc = u.ac.zMalloc;
64589	REGISTER_TRACE(u.ac.p2++, pOut);
64590	pIn1++;
64591	pOut++;
64592	}
	@@ -64433,11 +65286,11 @@
65286	** additional information.
65287	**
65288	** If SQLITE_NULLEQ is set in P5 then the result of comparison is always either
65289	** true or false and is never NULL. If both operands are NULL then the result
65290	** of comparison is false. If either operand is NULL then the result is true.
65291	** If neither operand is NULL the result is the same as it would be if
65292	** the SQLITE_NULLEQ flag were omitted from P5.
65293	*/
65294	/* Opcode: Eq P1 P2 P3 P4 P5
65295	**
65296	** This works just like the Lt opcode except that the jump is taken if
	@@ -64445,11 +65298,11 @@
65298	** See the Lt opcode for additional information.
65299	**
65300	** If SQLITE_NULLEQ is set in P5 then the result of comparison is always either
65301	** true or false and is never NULL. If both operands are NULL then the result
65302	** of comparison is true. If either operand is NULL then the result is false.
65303	** If neither operand is NULL the result is the same as it would be if
65304	** the SQLITE_NULLEQ flag were omitted from P5.
65305	*/
65306	/* Opcode: Le P1 P2 P3 P4 P5
65307	**
65308	** This works just like the Lt opcode except that the jump is taken if
	@@ -64730,17 +65583,17 @@
65583	break;
65584	}
65585
65586	/* Opcode: If P1 P2 P3 * *
65587	**
65588	** Jump to P2 if the value in register P1 is true. The value
65589	** is considered true if it is numeric and non-zero. If the value
65590	** in P1 is NULL then take the jump if P3 is true.
65591	*/
65592	/* Opcode: IfNot P1 P2 P3 * *
65593	**
65594	** Jump to P2 if the value in register P1 is False. The value
65595	** is considered true if it has a numeric value of zero. If the value
65596	** in P1 is NULL then take the jump if P3 is true.
65597	*/
65598	case OP_If: /* jump, in1 */
65599	case OP_IfNot: { /* jump, in1 */
	@@ -64828,10 +65681,11 @@
65681	u8 zEndHdr; / Pointer to first byte after the header */
65682	u32 offset; /* Offset into the data */
65683	u32 szField; /* Number of bytes in the content of a field */
65684	int szHdr; /* Size of the header size field at start of record */
65685	int avail; /* Number of bytes of available data */
65686	u32 t; /* A type code from the record header */
65687	Mem pReg; / PseudoTable input register */
65688	#endif /* local variables moved into u.am */
65689
65690
65691	u.am.p1 = pOp->p1;
	@@ -64840,11 +65694,10 @@
65694	memset(&u.am.sMem, 0, sizeof(u.am.sMem));
65695	assert( u.am.p1<p->nCursor );
65696	assert( pOp->p3>0 && pOp->p3<=p->nMem );
65697	u.am.pDest = &aMem[pOp->p3];
65698	memAboutToChange(p, u.am.pDest);

65699	u.am.zRec = 0;
65700
65701	/* This block sets the variable u.am.payloadSize to be the total number of
65702	** bytes in the record.
65703	**
	@@ -64884,11 +65737,11 @@
65737	}else{
65738	assert( sqlite3BtreeCursorIsValid(u.am.pCrsr) );
65739	rc = sqlite3BtreeDataSize(u.am.pCrsr, &u.am.payloadSize);
65740	assert( rc==SQLITE_OK ); /* DataSize() cannot fail */
65741	}
65742	}else if( ALWAYS(u.am.pC->pseudoTableReg>0) ){
65743	u.am.pReg = &aMem[u.am.pC->pseudoTableReg];
65744	assert( u.am.pReg->flags & MEM_Blob );
65745	assert( memIsValid(u.am.pReg) );
65746	u.am.payloadSize = u.am.pReg->n;
65747	u.am.zRec = u.am.pReg->z;
	@@ -64897,13 +65750,14 @@
65750	}else{
65751	/* Consider the row to be NULL */
65752	u.am.payloadSize = 0;
65753	}
65754
65755	/* If u.am.payloadSize is 0, then just store a NULL. This can happen because of
65756	** nullRow or because of a corrupt database. */
65757	if( u.am.payloadSize==0 ){
65758	MemSetTypeFlag(u.am.pDest, MEM_Null);
65759	goto op_column_out;
65760	}
65761	assert( db->aLimit[SQLITE_LIMIT_LENGTH]>=0 );
65762	if( u.am.payloadSize > (u32)db->aLimit[SQLITE_LIMIT_LENGTH] ){
65763	goto too_big;
	@@ -65006,12 +65860,18 @@
65860	** of the record to the start of the data for the u.am.i-th column
65861	*/
65862	for(u.am.i=0; u.am.i<u.am.nField; u.am.i++){
65863	if( u.am.zIdx<u.am.zEndHdr ){
65864	u.am.aOffset[u.am.i] = u.am.offset;
65865	if( u.am.zIdx[0]<0x80 ){
65866	u.am.t = u.am.zIdx[0];
65867	u.am.zIdx++;
65868	}else{
65869	u.am.zIdx += sqlite3GetVarint32(u.am.zIdx, &u.am.t);
65870	}
65871	u.am.aType[u.am.i] = u.am.t;
65872	u.am.szField = sqlite3VdbeSerialTypeLen(u.am.t);
65873	u.am.offset += u.am.szField;
65874	if( u.am.offset<u.am.szField ){ /* True if u.am.offset overflows */
65875	u.am.zIdx = &u.am.zEndHdr[1]; /* Forces SQLITE_CORRUPT return below */
65876	break;
65877	}
	@@ -65048,11 +65908,11 @@
65908	** a pointer to a Mem object.
65909	*/
65910	if( u.am.aOffset[u.am.p2] ){
65911	assert( rc==SQLITE_OK );
65912	if( u.am.zRec ){
65913	MemReleaseExt(u.am.pDest);
65914	sqlite3VdbeSerialGet((u8 *)&u.am.zRec[u.am.aOffset[u.am.p2]], u.am.aType[u.am.p2], u.am.pDest);
65915	}else{
65916	u.am.len = sqlite3VdbeSerialTypeLen(u.am.aType[u.am.p2]);
65917	sqlite3VdbeMemMove(&u.am.sMem, u.am.pDest);
65918	rc = sqlite3VdbeMemFromBtree(u.am.pCrsr, u.am.aOffset[u.am.p2], u.am.len, u.am.pC->isIndex, &u.am.sMem);
	@@ -65065,11 +65925,11 @@
65925	u.am.pDest->enc = encoding;
65926	}else{
65927	if( pOp->p4type==P4_MEM ){
65928	sqlite3VdbeMemShallowCopy(u.am.pDest, pOp->p4.pMem, MEM_Static);
65929	}else{
65930	MemSetTypeFlag(u.am.pDest, MEM_Null);
65931	}
65932	}
65933
65934	/* If we dynamically allocated space to hold the data (in the
65935	** sqlite3VdbeMemFromBtree() call above) then transfer control of that
	@@ -65267,11 +66127,11 @@
66127	i64 nEntry;
66128	BtCursor *pCrsr;
66129	#endif /* local variables moved into u.ap */
66130
66131	u.ap.pCrsr = p->apCsr[pOp->p1]->pCursor;
66132	if( ALWAYS(u.ap.pCrsr) ){
66133	rc = sqlite3BtreeCount(u.ap.pCrsr, &u.ap.nEntry);
66134	}else{
66135	u.ap.nEntry = 0;
66136	}
66137	pOut->u.i = u.ap.nEntry;
	@@ -65843,19 +66703,13 @@
66703	u.aw.pCur->nullRow = 1;
66704	u.aw.pCur->isOrdered = 1;
66705	rc = sqlite3BtreeCursor(u.aw.pX, u.aw.p2, u.aw.wrFlag, u.aw.pKeyInfo, u.aw.pCur->pCursor);
66706	u.aw.pCur->pKeyInfo = u.aw.pKeyInfo;
66707
66708	/* Since it performs no memory allocation or IO, the only value that
66709	** sqlite3BtreeCursor() may return is SQLITE_OK. */
66710	assert( rc==SQLITE_OK );






66711
66712	/* Set the VdbeCursor.isTable and isIndex variables. Previous versions of
66713	** SQLite used to check if the root-page flags were sane at this point
66714	** and report database corruption if they were not, but this check has
66715	** since moved into the btree layer. */
	@@ -65862,11 +66716,11 @@
66716	u.aw.pCur->isTable = pOp->p4type!=P4_KEYINFO;
66717	u.aw.pCur->isIndex = !u.aw.pCur->isTable;
66718	break;
66719	}
66720
66721	/* Opcode: OpenEphemeral P1 P2 * P4 P5
66722	**
66723	** Open a new cursor P1 to a transient table.
66724	** The cursor is always opened read/write even if
66725	** the main database is read-only. The ephemeral
66726	** table is deleted automatically when the cursor is closed.
	@@ -65879,18 +66733,30 @@
66733	** This opcode was once called OpenTemp. But that created
66734	** confusion because the term "temp table", might refer either
66735	** to a TEMP table at the SQL level, or to a table opened by
66736	** this opcode. Then this opcode was call OpenVirtual. But
66737	** that created confusion with the whole virtual-table idea.
66738	**
66739	** The P5 parameter can be a mask of the BTREE_* flags defined
66740	** in btree.h. These flags control aspects of the operation of
66741	** the btree. The BTREE_OMIT_JOURNAL and BTREE_SINGLE flags are
66742	** added automatically.
66743	*/
66744	/* Opcode: OpenAutoindex P1 P2 * P4 *
66745	**
66746	** This opcode works the same as OP_OpenEphemeral. It has a
66747	** different name to distinguish its use. Tables created using
66748	** by this opcode will be used for automatically created transient
66749	** indices in joins.
66750	*/
66751	/* Opcode: OpenSorter P1 P2 * P4 *
66752	**
66753	** This opcode works like OP_OpenEphemeral except that it opens
66754	** a transient index that is specifically designed to sort large
66755	** tables using an external merge-sort algorithm.
66756	*/
66757	case OP_OpenSorter:
66758	case OP_OpenAutoindex:
66759	case OP_OpenEphemeral: {
66760	#if 0 /* local variables moved into u.ax */
66761	VdbeCursor *pCx;
66762	#endif /* local variables moved into u.ax */
	@@ -65900,10 +66766,11 @@
66766	SQLITE_OPEN_EXCLUSIVE \|
66767	SQLITE_OPEN_DELETEONCLOSE \|
66768	SQLITE_OPEN_TRANSIENT_DB;
66769
66770	assert( pOp->p1>=0 );
66771	assert( (pOp->opcode==OP_OpenSorter)==((pOp->p5 & BTREE_SORTER)!=0) );
66772	u.ax.pCx = allocateCursor(p, pOp->p1, pOp->p2, -1, 1);
66773	if( u.ax.pCx==0 ) goto no_mem;
66774	u.ax.pCx->nullRow = 1;
66775	rc = sqlite3BtreeOpen(db->pVfs, 0, db, &u.ax.pCx->pBt,
66776	BTREE_OMIT_JOURNAL \| BTREE_SINGLE \| pOp->p5, vfsFlags);
	@@ -65917,11 +66784,11 @@
66784	** automatically created table with root-page 1 (an BLOB_INTKEY table).
66785	*/
66786	if( pOp->p4.pKeyInfo ){
66787	int pgno;
66788	assert( pOp->p4type==P4_KEYINFO );
66789	rc = sqlite3BtreeCreateTable(u.ax.pCx->pBt, &pgno, BTREE_BLOBKEY \| pOp->p5);
66790	if( rc==SQLITE_OK ){
66791	assert( pgno==MASTER_ROOT+1 );
66792	rc = sqlite3BtreeCursor(u.ax.pCx->pBt, pgno, 1,
66793	(KeyInfo*)pOp->p4.z, u.ax.pCx->pCursor);
66794	u.ax.pCx->pKeyInfo = pOp->p4.pKeyInfo;
	@@ -65933,10 +66800,15 @@
66800	u.ax.pCx->isTable = 1;
66801	}
66802	}
66803	u.ax.pCx->isOrdered = (pOp->p5!=BTREE_UNORDERED);
66804	u.ax.pCx->isIndex = !u.ax.pCx->isTable;
66805	#ifndef SQLITE_OMIT_MERGE_SORT
66806	if( rc==SQLITE_OK && pOp->opcode==OP_OpenSorter ){
66807	rc = sqlite3VdbeSorterInit(db, u.ax.pCx);
66808	}
66809	#endif
66810	break;
66811	}
66812
66813	/* Opcode: OpenPseudo P1 P2 P3 * *
66814	**
	@@ -66052,11 +66924,11 @@
66924	assert( u.az.pC->pseudoTableReg==0 );
66925	assert( OP_SeekLe == OP_SeekLt+1 );
66926	assert( OP_SeekGe == OP_SeekLt+2 );
66927	assert( OP_SeekGt == OP_SeekLt+3 );
66928	assert( u.az.pC->isOrdered );
66929	if( ALWAYS(u.az.pC->pCursor!=0) ){
66930	u.az.oc = pOp->opcode;
66931	u.az.pC->nullRow = 0;
66932	if( u.az.pC->isTable ){
66933	/* The input value in P3 might be of any type: integer, real, string,
66934	** blob, or NULL. But it needs to be an integer before we can do
	@@ -66390,11 +67262,11 @@
67262	break;
67263	}
67264
67265	/* Opcode: NotExists P1 P2 P3 * *
67266	**
67267	** Use the content of register P3 as an integer key. If a record
67268	** with that key does not exist in table of P1, then jump to P2.
67269	** If the record does exist, then fall through. The cursor is left
67270	** pointing to the record if it exists.
67271	**
67272	** The difference between this operation and NotFound is that this
	@@ -66418,11 +67290,11 @@
67290	u.bd.pC = p->apCsr[pOp->p1];
67291	assert( u.bd.pC!=0 );
67292	assert( u.bd.pC->isTable );
67293	assert( u.bd.pC->pseudoTableReg==0 );
67294	u.bd.pCrsr = u.bd.pC->pCursor;
67295	if( ALWAYS(u.bd.pCrsr!=0) ){
67296	u.bd.res = 0;
67297	u.bd.iKey = pIn3->u.i;
67298	rc = sqlite3BtreeMovetoUnpacked(u.bd.pCrsr, 0, u.bd.iKey, 0, &u.bd.res);
67299	u.bd.pC->lastRowid = pIn3->u.i;
67300	u.bd.pC->rowidIsValid = u.bd.res==0 ?1:0;
	@@ -66468,11 +67340,11 @@
67340	** written to register P2.
67341	**
67342	** If P3>0 then P3 is a register in the root frame of this VDBE that holds
67343	** the largest previously generated record number. No new record numbers are
67344	** allowed to be less than this value. When this value reaches its maximum,
67345	** an SQLITE_FULL error is generated. The P3 register is updated with the '
67346	** generated record number. This P3 mechanism is used to help implement the
67347	** AUTOINCREMENT feature.
67348	*/
67349	case OP_NewRowid: { /* out2-prerelease */
67350	#if 0 /* local variables moved into u.be */
	@@ -66842,10 +67714,17 @@
67714	assert( u.bh.pC->isTable \|\| pOp->opcode==OP_RowKey );
67715	assert( u.bh.pC->isIndex \|\| pOp->opcode==OP_RowData );
67716	assert( u.bh.pC!=0 );
67717	assert( u.bh.pC->nullRow==0 );
67718	assert( u.bh.pC->pseudoTableReg==0 );
67719
67720	if( isSorter(u.bh.pC) ){
67721	assert( pOp->opcode==OP_RowKey );
67722	rc = sqlite3VdbeSorterRowkey(u.bh.pC, pOut);
67723	break;
67724	}
67725
67726	assert( u.bh.pC->pCursor!=0 );
67727	u.bh.pCrsr = u.bh.pC->pCursor;
67728	assert( sqlite3BtreeCursorIsValid(u.bh.pCrsr) );
67729
67730	/* The OP_RowKey and OP_RowData opcodes always follow OP_NotExists or
	@@ -66950,10 +67829,11 @@
67829	assert( pOp->p1>=0 && pOp->p1<p->nCursor );
67830	u.bj.pC = p->apCsr[pOp->p1];
67831	assert( u.bj.pC!=0 );
67832	u.bj.pC->nullRow = 1;
67833	u.bj.pC->rowidIsValid = 0;
67834	assert( u.bj.pC->pCursor \|\| u.bj.pC->pVtabCursor );
67835	if( u.bj.pC->pCursor ){
67836	sqlite3BtreeClearCursor(u.bj.pC->pCursor);
67837	}
67838	break;
67839	}
	@@ -66975,11 +67855,11 @@
67855
67856	assert( pOp->p1>=0 && pOp->p1<p->nCursor );
67857	u.bk.pC = p->apCsr[pOp->p1];
67858	assert( u.bk.pC!=0 );
67859	u.bk.pCrsr = u.bk.pC->pCursor;
67860	if( NEVER(u.bk.pCrsr==0) ){
67861	u.bk.res = 1;
67862	}else{
67863	rc = sqlite3BtreeLast(u.bk.pCrsr, &u.bk.res);
67864	}
67865	u.bk.pC->nullRow = (u8)u.bk.res;
	@@ -67030,11 +67910,15 @@
67910
67911	assert( pOp->p1>=0 && pOp->p1<p->nCursor );
67912	u.bl.pC = p->apCsr[pOp->p1];
67913	assert( u.bl.pC!=0 );
67914	u.bl.res = 1;
67915	if( isSorter(u.bl.pC) ){
67916	rc = sqlite3VdbeSorterRewind(db, u.bl.pC, &u.bl.res);
67917	}else{
67918	u.bl.pCrsr = u.bl.pC->pCursor;
67919	assert( u.bl.pCrsr );
67920	rc = sqlite3BtreeFirst(u.bl.pCrsr, &u.bl.res);
67921	u.bl.pC->atFirst = u.bl.res==0 ?1:0;
67922	u.bl.pC->deferredMoveto = 0;
67923	u.bl.pC->cacheStatus = CACHE_STALE;
67924	u.bl.pC->rowidIsValid = 0;
	@@ -67045,18 +67929,21 @@
67929	pc = pOp->p2 - 1;
67930	}
67931	break;
67932	}
67933
67934	/* Opcode: Next P1 P2 * P4 P5
67935	**
67936	** Advance cursor P1 so that it points to the next key/data pair in its
67937	** table or index. If there are no more key/value pairs then fall through
67938	** to the following instruction. But if the cursor advance was successful,
67939	** jump immediately to P2.
67940	**
67941	** The P1 cursor must be for a real table, not a pseudo-table.
67942	**
67943	** P4 is always of type P4_ADVANCE. The function pointer points to
67944	** sqlite3BtreeNext().
67945	**
67946	** If P5 is positive and the jump is taken, then event counter
67947	** number P5-1 in the prepared statement is incremented.
67948	**
67949	** See also: Prev
	@@ -67067,19 +67954,21 @@
67954	** table or index. If there is no previous key/value pairs then fall through
67955	** to the following instruction. But if the cursor backup was successful,
67956	** jump immediately to P2.
67957	**
67958	** The P1 cursor must be for a real table, not a pseudo-table.
67959	**
67960	** P4 is always of type P4_ADVANCE. The function pointer points to
67961	** sqlite3BtreePrevious().
67962	**
67963	** If P5 is positive and the jump is taken, then event counter
67964	** number P5-1 in the prepared statement is incremented.
67965	*/
67966	case OP_Prev: /* jump */
67967	case OP_Next: { /* jump */
67968	#if 0 /* local variables moved into u.bm */
67969	VdbeCursor *pC;

67970	int res;
67971	#endif /* local variables moved into u.bm */
67972
67973	CHECK_FOR_INTERRUPT;
67974	assert( pOp->p1>=0 && pOp->p1<p->nCursor );
	@@ -67086,19 +67975,21 @@
67975	assert( pOp->p5<=ArraySize(p->aCounter) );
67976	u.bm.pC = p->apCsr[pOp->p1];
67977	if( u.bm.pC==0 ){
67978	break; /* See ticket #2273 */
67979	}
67980	if( isSorter(u.bm.pC) ){
67981	assert( pOp->opcode==OP_Next );
67982	rc = sqlite3VdbeSorterNext(db, u.bm.pC, &u.bm.res);
67983	}else{
67984	u.bm.res = 1;
67985	assert( u.bm.pC->deferredMoveto==0 );
67986	assert( u.bm.pC->pCursor );
67987	assert( pOp->opcode!=OP_Next \|\| pOp->p4.xAdvance==sqlite3BtreeNext );
67988	assert( pOp->opcode!=OP_Prev \|\| pOp->p4.xAdvance==sqlite3BtreePrevious );
67989	rc = pOp->p4.xAdvance(u.bm.pC->pCursor, &u.bm.res);
67990	}
67991	u.bm.pC->nullRow = (u8)u.bm.res;
67992	u.bm.pC->cacheStatus = CACHE_STALE;
67993	if( u.bm.res==0 ){
67994	pc = pOp->p2 - 1;
67995	if( pOp->p5 ) p->aCounter[pOp->p5-1]++;
	@@ -67110,11 +68001,11 @@
68001	break;
68002	}
68003
68004	/* Opcode: IdxInsert P1 P2 P3 * P5
68005	**
68006	** Register P2 holds an SQL index key made using the
68007	** MakeRecord instructions. This opcode writes that key
68008	** into the index P1. Data for the entry is nil.
68009	**
68010	** P3 is a flag that provides a hint to the b-tree layer that this
68011	** insert is likely to be an append.
	@@ -67140,14 +68031,17 @@
68031	assert( u.bn.pC->isTable==0 );
68032	rc = ExpandBlob(pIn2);
68033	if( rc==SQLITE_OK ){
68034	u.bn.nKey = pIn2->n;
68035	u.bn.zKey = pIn2->z;
68036	rc = sqlite3VdbeSorterWrite(db, u.bn.pC, u.bn.nKey);
68037	if( rc==SQLITE_OK ){
68038	rc = sqlite3BtreeInsert(u.bn.pCrsr, u.bn.zKey, u.bn.nKey, "", 0, 0, pOp->p3,
68039	((pOp->p5 & OPFLAG_USESEEKRESULT) ? u.bn.pC->seekResult : 0)
68040	);
68041	assert( u.bn.pC->deferredMoveto==0 );
68042	}
68043	u.bn.pC->cacheStatus = CACHE_STALE;
68044	}
68045	}
68046	break;
68047	}
	@@ -69323,10 +70217,722 @@
70217	}
70218
70219	#endif /* #ifndef SQLITE_OMIT_INCRBLOB */
70220
70221	/************ End of vdbeblob.c ******************************************/
70222	/************ Begin file vdbesort.c **************************************/
70223	/*
70224	** 2011 July 9
70225	**
70226	** The author disclaims copyright to this source code. In place of
70227	** a legal notice, here is a blessing:
70228	**
70229	** May you do good and not evil.
70230	** May you find forgiveness for yourself and forgive others.
70231	** May you share freely, never taking more than you give.
70232	**
70233	*************************************************************************
70234	** This file contains code for the VdbeSorter object, used in concert with
70235	** a VdbeCursor to sort large numbers of keys (as may be required, for
70236	** example, by CREATE INDEX statements on tables too large to fit in main
70237	** memory).
70238	*/
70239
70240
70241	#ifndef SQLITE_OMIT_MERGE_SORT
70242
70243	typedef struct VdbeSorterIter VdbeSorterIter;
70244
70245	/*
70246	** NOTES ON DATA STRUCTURE USED FOR N-WAY MERGES:
70247	**
70248	** As keys are added to the sorter, they are written to disk in a series
70249	** of sorted packed-memory-arrays (PMAs). The size of each PMA is roughly
70250	** the same as the cache-size allowed for temporary databases. In order
70251	** to allow the caller to extract keys from the sorter in sorted order,
70252	** all PMAs currently stored on disk must be merged together. This comment
70253	** describes the data structure used to do so. The structure supports
70254	** merging any number of arrays in a single pass with no redundant comparison
70255	** operations.
70256	**
70257	** The aIter[] array contains an iterator for each of the PMAs being merged.
70258	** An aIter[] iterator either points to a valid key or else is at EOF. For
70259	** the purposes of the paragraphs below, we assume that the array is actually
70260	** N elements in size, where N is the smallest power of 2 greater to or equal
70261	** to the number of iterators being merged. The extra aIter[] elements are
70262	** treated as if they are empty (always at EOF).
70263	**
70264	** The aTree[] array is also N elements in size. The value of N is stored in
70265	** the VdbeSorter.nTree variable.
70266	**
70267	** The final (N/2) elements of aTree[] contain the results of comparing
70268	** pairs of iterator keys together. Element i contains the result of
70269	** comparing aIter[2i-N] and aIter[2i-N+1]. Whichever key is smaller, the
70270	** aTree element is set to the index of it.
70271	**
70272	** For the purposes of this comparison, EOF is considered greater than any
70273	** other key value. If the keys are equal (only possible with two EOF
70274	** values), it doesn't matter which index is stored.
70275	**
70276	** The (N/4) elements of aTree[] that preceed the final (N/2) described
70277	** above contains the index of the smallest of each block of 4 iterators.
70278	** And so on. So that aTree[1] contains the index of the iterator that
70279	** currently points to the smallest key value. aTree[0] is unused.
70280	**
70281	** Example:
70282	**
70283	** aIter[0] -> Banana
70284	** aIter[1] -> Feijoa
70285	** aIter[2] -> Elderberry
70286	** aIter[3] -> Currant
70287	** aIter[4] -> Grapefruit
70288	** aIter[5] -> Apple
70289	** aIter[6] -> Durian
70290	** aIter[7] -> EOF
70291	**
70292	** aTree[] = { X, 5 0, 5 0, 3, 5, 6 }
70293	**
70294	** The current element is "Apple" (the value of the key indicated by
70295	** iterator 5). When the Next() operation is invoked, iterator 5 will
70296	** be advanced to the next key in its segment. Say the next key is
70297	** "Eggplant":
70298	**
70299	** aIter[5] -> Eggplant
70300	**
70301	** The contents of aTree[] are updated first by comparing the new iterator
70302	** 5 key to the current key of iterator 4 (still "Grapefruit"). The iterator
70303	** 5 value is still smaller, so aTree[6] is set to 5. And so on up the tree.
70304	** The value of iterator 6 - "Durian" - is now smaller than that of iterator
70305	** 5, so aTree[3] is set to 6. Key 0 is smaller than key 6 (Banana<Durian),
70306	** so the value written into element 1 of the array is 0. As follows:
70307	**
70308	** aTree[] = { X, 0 0, 6 0, 3, 5, 6 }
70309	**
70310	** In other words, each time we advance to the next sorter element, log2(N)
70311	** key comparison operations are required, where N is the number of segments
70312	** being merged (rounded up to the next power of 2).
70313	*/
70314	struct VdbeSorter {
70315	int nWorking; /* Start a new b-tree after this many pages */
70316	int nBtree; /* Current size of b-tree contents as PMA */
70317	int nTree; /* Used size of aTree/aIter (power of 2) */
70318	VdbeSorterIter aIter; / Array of iterators to merge */
70319	int aTree; / Current state of incremental merge */
70320	i64 iWriteOff; /* Current write offset within file pTemp1 */
70321	i64 iReadOff; /* Current read offset within file pTemp1 */
70322	sqlite3_file pTemp1; / PMA file 1 */
70323	int nPMA; /* Number of PMAs stored in pTemp1 */
70324	};
70325
70326	/*
70327	** The following type is an iterator for a PMA. It caches the current key in
70328	** variables nKey/aKey. If the iterator is at EOF, pFile==0.
70329	*/
70330	struct VdbeSorterIter {
70331	i64 iReadOff; /* Current read offset */
70332	i64 iEof; /* 1 byte past EOF for this iterator */
70333	sqlite3_file pFile; / File iterator is reading from */
70334	int nAlloc; /* Bytes of space at aAlloc */
70335	u8 aAlloc; / Allocated space */
70336	int nKey; /* Number of bytes in key */
70337	u8 aKey; / Pointer to current key */
70338	};
70339
70340	/* Minimum allowable value for the VdbeSorter.nWorking variable */
70341	#define SORTER_MIN_WORKING 10
70342
70343	/* Maximum number of segments to merge in a single pass. */
70344	#define SORTER_MAX_MERGE_COUNT 16
70345
70346	/*
70347	** Free all memory belonging to the VdbeSorterIter object passed as the second
70348	** argument. All structure fields are set to zero before returning.
70349	*/
70350	static void vdbeSorterIterZero(sqlite3 db, VdbeSorterIter pIter){
70351	sqlite3DbFree(db, pIter->aAlloc);
70352	memset(pIter, 0, sizeof(VdbeSorterIter));
70353	}
70354
70355	/*
70356	** Advance iterator pIter to the next key in its PMA. Return SQLITE_OK if
70357	** no error occurs, or an SQLite error code if one does.
70358	*/
70359	static int vdbeSorterIterNext(
70360	sqlite3 db, / Database handle (for sqlite3DbMalloc() ) */
70361	VdbeSorterIter pIter / Iterator to advance */
70362	){
70363	int rc; /* Return Code */
70364	int nRead; /* Number of bytes read */
70365	int nRec; /* Size of record in bytes */
70366	int iOff; /* Size of serialized size varint in bytes */
70367
70368	nRead = pIter->iEof - pIter->iReadOff;
70369	if( nRead>5 ) nRead = 5;
70370	if( nRead<=0 ){
70371	/* This is an EOF condition */
70372	vdbeSorterIterZero(db, pIter);
70373	return SQLITE_OK;
70374	}
70375
70376	rc = sqlite3OsRead(pIter->pFile, pIter->aAlloc, nRead, pIter->iReadOff);
70377	iOff = getVarint32(pIter->aAlloc, nRec);
70378
70379	if( rc==SQLITE_OK && (iOff+nRec)>nRead ){
70380	int nRead2; /* Number of extra bytes to read */
70381	if( (iOff+nRec)>pIter->nAlloc ){
70382	int nNew = pIter->nAlloc*2;
70383	while( (iOff+nRec)>nNew ) nNew = nNew*2;
70384	pIter->aAlloc = sqlite3DbReallocOrFree(db, pIter->aAlloc, nNew);
70385	if( !pIter->aAlloc ) return SQLITE_NOMEM;
70386	pIter->nAlloc = nNew;
70387	}
70388
70389	nRead2 = iOff + nRec - nRead;
70390	rc = sqlite3OsRead(
70391	pIter->pFile, &pIter->aAlloc[nRead], nRead2, pIter->iReadOff+nRead
70392	);
70393	}
70394
70395	assert( nRec>0 \|\| rc!=SQLITE_OK );
70396	pIter->iReadOff += iOff+nRec;
70397	pIter->nKey = nRec;
70398	pIter->aKey = &pIter->aAlloc[iOff];
70399	return rc;
70400	}
70401
70402	/*
70403	** Write a single varint, value iVal, to file-descriptor pFile. Return
70404	** SQLITE_OK if successful, or an SQLite error code if some error occurs.
70405	**
70406	** The value of *piOffset when this function is called is used as the byte
70407	** offset in file pFile to write to. Before returning, *piOffset is
70408	** incremented by the number of bytes written.
70409	*/
70410	static int vdbeSorterWriteVarint(
70411	sqlite3_file pFile, / File to write to */
70412	i64 iVal, /* Value to write as a varint */
70413	i64 piOffset / IN/OUT: Write offset in file pFile */
70414	){
70415	u8 aVarint[9]; /* Buffer large enough for a varint */
70416	int nVarint; /* Number of used bytes in varint */
70417	int rc; /* Result of write() call */
70418
70419	nVarint = sqlite3PutVarint(aVarint, iVal);
70420	rc = sqlite3OsWrite(pFile, aVarint, nVarint, *piOffset);
70421	*piOffset += nVarint;
70422
70423	return rc;
70424	}
70425
70426	/*
70427	** Read a single varint from file-descriptor pFile. Return SQLITE_OK if
70428	** successful, or an SQLite error code if some error occurs.
70429	**
70430	** The value of *piOffset when this function is called is used as the
70431	** byte offset in file pFile from whence to read the varint. If successful
70432	** (i.e. if no IO error occurs), then *piOffset is set to the offset of
70433	** the first byte past the end of the varint before returning. *piVal is
70434	** set to the integer value read. If an error occurs, the final values of
70435	** both piOffset and piVal are undefined.
70436	*/
70437	static int vdbeSorterReadVarint(
70438	sqlite3_file pFile, / File to read from */
70439	i64 iEof, /* Total number of bytes in file */
70440	i64 piOffset, / IN/OUT: Read offset in pFile */
70441	i64 piVal / OUT: Value read from file */
70442	){
70443	u8 aVarint[9]; /* Buffer large enough for a varint */
70444	i64 iOff = piOffset; / Offset in file to read from */
70445	int nRead = 9; /* Number of bytes to read from file */
70446	int rc; /* Return code */
70447
70448	assert( iEof>iOff );
70449	if( (iEof-iOff)<nRead ){
70450	nRead = iEof-iOff;
70451	}
70452
70453	rc = sqlite3OsRead(pFile, aVarint, nRead, iOff);
70454	if( rc==SQLITE_OK ){
70455	piOffset += getVarint(aVarint, (u64 )piVal);
70456	}
70457
70458	return rc;
70459	}
70460
70461	/*
70462	** Initialize iterator pIter to scan through the PMA stored in file pFile
70463	** starting at offset iStart and ending at offset iEof-1. This function
70464	** leaves the iterator pointing to the first key in the PMA (or EOF if the
70465	** PMA is empty).
70466	*/
70467	static int vdbeSorterIterInit(
70468	sqlite3 db, / Database handle */
70469	VdbeSorter pSorter, / Sorter object */
70470	i64 iStart, /* Start offset in pFile */
70471	VdbeSorterIter pIter, / Iterator to populate */
70472	i64 pnByte / IN/OUT: Increment this value by PMA size */
70473	){
70474	int rc;
70475
70476	assert( pSorter->iWriteOff>iStart );
70477	assert( pIter->aAlloc==0 );
70478	pIter->pFile = pSorter->pTemp1;
70479	pIter->iReadOff = iStart;
70480	pIter->nAlloc = 128;
70481	pIter->aAlloc = (u8 *)sqlite3DbMallocRaw(db, pIter->nAlloc);
70482	if( !pIter->aAlloc ){
70483	rc = SQLITE_NOMEM;
70484	}else{
70485	i64 iEof = pSorter->iWriteOff; /* EOF of file pSorter->pTemp1 */
70486	i64 nByte; /* Total size of PMA in bytes */
70487	rc = vdbeSorterReadVarint(pSorter->pTemp1, iEof, &pIter->iReadOff, &nByte);
70488	*pnByte += nByte;
70489	pIter->iEof = pIter->iReadOff + nByte;
70490	}
70491	if( rc==SQLITE_OK ){
70492	rc = vdbeSorterIterNext(db, pIter);
70493	}
70494	return rc;
70495	}
70496
70497	/*
70498	** This function is called to compare two iterator keys when merging
70499	** multiple b-tree segments. Parameter iOut is the index of the aTree[]
70500	** value to recalculate.
70501	*/
70502	static int vdbeSorterDoCompare(VdbeCursor *pCsr, int iOut){
70503	VdbeSorter *pSorter = pCsr->pSorter;
70504	int i1;
70505	int i2;
70506	int iRes;
70507	VdbeSorterIter *p1;
70508	VdbeSorterIter *p2;
70509
70510	assert( iOut<pSorter->nTree && iOut>0 );
70511
70512	if( iOut>=(pSorter->nTree/2) ){
70513	i1 = (iOut - pSorter->nTree/2) * 2;
70514	i2 = i1 + 1;
70515	}else{
70516	i1 = pSorter->aTree[iOut*2];
70517	i2 = pSorter->aTree[iOut*2+1];
70518	}
70519
70520	p1 = &pSorter->aIter[i1];
70521	p2 = &pSorter->aIter[i2];
70522
70523	if( p1->pFile==0 ){
70524	iRes = i2;
70525	}else if( p2->pFile==0 ){
70526	iRes = i1;
70527	}else{
70528	char aSpace[150];
70529	UnpackedRecord *r1;
70530
70531	r1 = sqlite3VdbeRecordUnpack(
70532	pCsr->pKeyInfo, p1->nKey, p1->aKey, aSpace, sizeof(aSpace)
70533	);
70534	if( r1==0 ) return SQLITE_NOMEM;
70535
70536	if( sqlite3VdbeRecordCompare(p2->nKey, p2->aKey, r1)>=0 ){
70537	iRes = i1;
70538	}else{
70539	iRes = i2;
70540	}
70541	sqlite3VdbeDeleteUnpackedRecord(r1);
70542	}
70543
70544	pSorter->aTree[iOut] = iRes;
70545	return SQLITE_OK;
70546	}
70547
70548	/*
70549	** Initialize the temporary index cursor just opened as a sorter cursor.
70550	*/
70551	SQLITE_PRIVATE int sqlite3VdbeSorterInit(sqlite3 db, VdbeCursor pCsr){
70552	assert( pCsr->pKeyInfo && pCsr->pBt );
70553	pCsr->pSorter = sqlite3DbMallocZero(db, sizeof(VdbeSorter));
70554	return (pCsr->pSorter ? SQLITE_OK : SQLITE_NOMEM);
70555	}
70556
70557	/*
70558	** Free any cursor components allocated by sqlite3VdbeSorterXXX routines.
70559	*/
70560	SQLITE_PRIVATE void sqlite3VdbeSorterClose(sqlite3 db, VdbeCursor pCsr){
70561	VdbeSorter *pSorter = pCsr->pSorter;
70562	if( pSorter ){
70563	if( pSorter->aIter ){
70564	int i;
70565	for(i=0; i<pSorter->nTree; i++){
70566	vdbeSorterIterZero(db, &pSorter->aIter[i]);
70567	}
70568	sqlite3DbFree(db, pSorter->aIter);
70569	}
70570	if( pSorter->pTemp1 ){
70571	sqlite3OsCloseFree(pSorter->pTemp1);
70572	}
70573	sqlite3DbFree(db, pSorter);
70574	pCsr->pSorter = 0;
70575	}
70576	}
70577
70578	/*
70579	** Allocate space for a file-handle and open a temporary file. If successful,
70580	** set *ppFile to point to the malloc'd file-handle and return SQLITE_OK.
70581	** Otherwise, set *ppFile to 0 and return an SQLite error code.
70582	*/
70583	static int vdbeSorterOpenTempFile(sqlite3 db, sqlite3_file *ppFile){
70584	int dummy;
70585	return sqlite3OsOpenMalloc(db->pVfs, 0, ppFile,
70586	SQLITE_OPEN_TEMP_JOURNAL \|
70587	SQLITE_OPEN_READWRITE \| SQLITE_OPEN_CREATE \|
70588	SQLITE_OPEN_EXCLUSIVE \| SQLITE_OPEN_DELETEONCLOSE, &dummy
70589	);
70590	}
70591
70592
70593	/*
70594	** Write the current contents of the b-tree to a PMA. Return SQLITE_OK
70595	** if successful, or an SQLite error code otherwise.
70596	**
70597	** The format of a PMA is:
70598	**
70599	** * A varint. This varint contains the total number of bytes of content
70600	** in the PMA (not including the varint itself).
70601	**
70602	** * One or more records packed end-to-end in order of ascending keys.
70603	** Each record consists of a varint followed by a blob of data (the
70604	** key). The varint is the number of bytes in the blob of data.
70605	*/
70606	static int vdbeSorterBtreeToPMA(sqlite3 db, VdbeCursor pCsr){
70607	int rc = SQLITE_OK; /* Return code */
70608	VdbeSorter *pSorter = pCsr->pSorter;
70609	int res = 0;
70610
70611	/* sqlite3BtreeFirst() cannot fail because sorter btrees are always held
70612	** in memory and so an I/O error is not possible. */
70613	rc = sqlite3BtreeFirst(pCsr->pCursor, &res);
70614	if( NEVER(rc!=SQLITE_OK) \|\| res ) return rc;
70615	assert( pSorter->nBtree>0 );
70616
70617	/* If the first temporary PMA file has not been opened, open it now. */
70618	if( pSorter->pTemp1==0 ){
70619	rc = vdbeSorterOpenTempFile(db, &pSorter->pTemp1);
70620	assert( rc!=SQLITE_OK \|\| pSorter->pTemp1 );
70621	assert( pSorter->iWriteOff==0 );
70622	assert( pSorter->nPMA==0 );
70623	}
70624
70625	if( rc==SQLITE_OK ){
70626	i64 iWriteOff = pSorter->iWriteOff;
70627	void aMalloc = 0; / Array used to hold a single record */
70628	int nMalloc = 0; /* Allocated size of aMalloc[] in bytes */
70629
70630	pSorter->nPMA++;
70631	for(
70632	rc = vdbeSorterWriteVarint(pSorter->pTemp1, pSorter->nBtree, &iWriteOff);
70633	rc==SQLITE_OK && res==0;
70634	rc = sqlite3BtreeNext(pCsr->pCursor, &res)
70635	){
70636	i64 nKey; /* Size of this key in bytes */
70637
70638	/* Write the size of the record in bytes to the output file */
70639	(void)sqlite3BtreeKeySize(pCsr->pCursor, &nKey);
70640	rc = vdbeSorterWriteVarint(pSorter->pTemp1, nKey, &iWriteOff);
70641
70642	/* Make sure the aMalloc[] buffer is large enough for the record */
70643	if( rc==SQLITE_OK && nKey>nMalloc ){
70644	aMalloc = sqlite3DbReallocOrFree(db, aMalloc, nKey);
70645	if( !aMalloc ){
70646	rc = SQLITE_NOMEM;
70647	}else{
70648	nMalloc = nKey;
70649	}
70650	}
70651
70652	/* Write the record itself to the output file */
70653	if( rc==SQLITE_OK ){
70654	/* sqlite3BtreeKey() cannot fail because sorter btrees held in memory */
70655	rc = sqlite3BtreeKey(pCsr->pCursor, 0, nKey, aMalloc);
70656	if( ALWAYS(rc==SQLITE_OK) ){
70657	rc = sqlite3OsWrite(pSorter->pTemp1, aMalloc, nKey, iWriteOff);
70658	iWriteOff += nKey;
70659	}
70660	}
70661
70662	if( rc!=SQLITE_OK ) break;
70663	}
70664
70665	/* This assert verifies that unless an error has occurred, the size of
70666	** the PMA on disk is the same as the expected size stored in
70667	** pSorter->nBtree. */
70668	assert( rc!=SQLITE_OK \|\| pSorter->nBtree==(
70669	iWriteOff-pSorter->iWriteOff-sqlite3VarintLen(pSorter->nBtree)
70670	));
70671
70672	pSorter->iWriteOff = iWriteOff;
70673	sqlite3DbFree(db, aMalloc);
70674	}
70675
70676	pSorter->nBtree = 0;
70677	return rc;
70678	}
70679
70680	/*
70681	** This function is called on a sorter cursor by the VDBE before each row
70682	** is inserted into VdbeCursor.pCsr. Argument nKey is the size of the key, in
70683	** bytes, about to be inserted.
70684	**
70685	** If it is determined that the temporary b-tree accessed via VdbeCursor.pCsr
70686	** is large enough, its contents are written to a sorted PMA on disk and the
70687	** tree emptied. This prevents the b-tree (which must be small enough to
70688	** fit entirely in the cache in order to support efficient inserts) from
70689	** growing too large.
70690	**
70691	** An SQLite error code is returned if an error occurs. Otherwise, SQLITE_OK.
70692	*/
70693	SQLITE_PRIVATE int sqlite3VdbeSorterWrite(sqlite3 db, VdbeCursor pCsr, int nKey){
70694	int rc = SQLITE_OK; /* Return code */
70695	VdbeSorter *pSorter = pCsr->pSorter;
70696	if( pSorter ){
70697	Pager *pPager = sqlite3BtreePager(pCsr->pBt);
70698	int nPage; /* Current size of temporary file in pages */
70699
70700	/* Sorters never spill to disk */
70701	assert( sqlite3PagerFile(pPager)->pMethods==0 );
70702
70703	/* Determine how many pages the temporary b-tree has grown to */
70704	sqlite3PagerPagecount(pPager, &nPage);
70705
70706	/* If pSorter->nWorking is still zero, but the temporary file has been
70707	** created in the file-system, then the most recent insert into the
70708	** current b-tree segment probably caused the cache to overflow (it is
70709	** also possible that sqlite3_release_memory() was called). So set the
70710	** size of the working set to a little less than the current size of the
70711	** file in pages. */
70712	if( pSorter->nWorking==0 && sqlite3PagerUnderStress(pPager) ){
70713	pSorter->nWorking = nPage-5;
70714	if( pSorter->nWorking<SORTER_MIN_WORKING ){
70715	pSorter->nWorking = SORTER_MIN_WORKING;
70716	}
70717	}
70718
70719	/* If the number of pages used by the current b-tree segment is greater
70720	** than the size of the working set (VdbeSorter.nWorking), start a new
70721	** segment b-tree. */
70722	if( pSorter->nWorking && nPage>=pSorter->nWorking ){
70723	BtCursor p = pCsr->pCursor;/ Cursor structure to close and reopen */
70724	int iRoot; /* Root page of new tree */
70725
70726	/* Copy the current contents of the b-tree into a PMA in sorted order.
70727	** Close the currently open b-tree cursor. */
70728	rc = vdbeSorterBtreeToPMA(db, pCsr);
70729	sqlite3BtreeCloseCursor(p);
70730
70731	if( rc==SQLITE_OK ){
70732	rc = sqlite3BtreeDropTable(pCsr->pBt, 2, 0);
70733	#ifdef SQLITE_DEBUG
70734	sqlite3PagerPagecount(pPager, &nPage);
70735	assert( rc!=SQLITE_OK \|\| nPage==1 );
70736	#endif
70737	}
70738	if( rc==SQLITE_OK ){
70739	rc = sqlite3BtreeCreateTable(pCsr->pBt, &iRoot, BTREE_BLOBKEY);
70740	}
70741	if( rc==SQLITE_OK ){
70742	assert( iRoot==2 );
70743	rc = sqlite3BtreeCursor(pCsr->pBt, iRoot, 1, pCsr->pKeyInfo, p);
70744	}
70745	}
70746
70747	pSorter->nBtree += sqlite3VarintLen(nKey) + nKey;
70748	}
70749	return rc;
70750	}
70751
70752	/*
70753	** Helper function for sqlite3VdbeSorterRewind().
70754	*/
70755	static int vdbeSorterInitMerge(
70756	sqlite3 db, / Database handle */
70757	VdbeCursor pCsr, / Cursor handle for this sorter */
70758	i64 pnByte / Sum of bytes in all opened PMAs */
70759	){
70760	VdbeSorter *pSorter = pCsr->pSorter;
70761	int rc = SQLITE_OK; /* Return code */
70762	int i; /* Used to iterator through aIter[] */
70763	i64 nByte = 0; /* Total bytes in all opened PMAs */
70764
70765	/* Initialize the iterators. */
70766	for(i=0; rc==SQLITE_OK && i<SORTER_MAX_MERGE_COUNT; i++){
70767	VdbeSorterIter *pIter = &pSorter->aIter[i];
70768	rc = vdbeSorterIterInit(db, pSorter, pSorter->iReadOff, pIter, &nByte);
70769	pSorter->iReadOff = pIter->iEof;
70770	assert( pSorter->iReadOff<=pSorter->iWriteOff \|\| rc!=SQLITE_OK );
70771	if( pSorter->iReadOff>=pSorter->iWriteOff ) break;
70772	}
70773
70774	/* Initialize the aTree[] array. */
70775	for(i=pSorter->nTree-1; rc==SQLITE_OK && i>0; i--){
70776	rc = vdbeSorterDoCompare(pCsr, i);
70777	}
70778
70779	*pnByte = nByte;
70780	return rc;
70781	}
70782
70783	/*
70784	** Once the sorter has been populated, this function is called to prepare
70785	** for iterating through its contents in sorted order.
70786	*/
70787	SQLITE_PRIVATE int sqlite3VdbeSorterRewind(sqlite3 db, VdbeCursor pCsr, int *pbEof){
70788	VdbeSorter *pSorter = pCsr->pSorter;
70789	int rc; /* Return code */
70790	sqlite3_file pTemp2 = 0; / Second temp file to use */
70791	i64 iWrite2 = 0; /* Write offset for pTemp2 */
70792	int nIter; /* Number of iterators used */
70793	int nByte; /* Bytes of space required for aIter/aTree */
70794	int N = 2; /* Power of 2 >= nIter */
70795
70796	assert( pSorter );
70797
70798	/* Write the current b-tree to a PMA. Close the b-tree cursor. */
70799	rc = vdbeSorterBtreeToPMA(db, pCsr);
70800	sqlite3BtreeCloseCursor(pCsr->pCursor);
70801	if( rc!=SQLITE_OK ) return rc;
70802	if( pSorter->nPMA==0 ){
70803	*pbEof = 1;
70804	return SQLITE_OK;
70805	}
70806
70807	/* Allocate space for aIter[] and aTree[]. */
70808	nIter = pSorter->nPMA;
70809	if( nIter>SORTER_MAX_MERGE_COUNT ) nIter = SORTER_MAX_MERGE_COUNT;
70810	assert( nIter>0 );
70811	while( N<nIter ) N += N;
70812	nByte = N * (sizeof(int) + sizeof(VdbeSorterIter));
70813	pSorter->aIter = (VdbeSorterIter *)sqlite3DbMallocZero(db, nByte);
70814	if( !pSorter->aIter ) return SQLITE_NOMEM;
70815	pSorter->aTree = (int *)&pSorter->aIter[N];
70816	pSorter->nTree = N;
70817
70818	do {
70819	int iNew; /* Index of new, merged, PMA */
70820
70821	for(iNew=0;
70822	rc==SQLITE_OK && iNew*SORTER_MAX_MERGE_COUNT<pSorter->nPMA;
70823	iNew++
70824	){
70825	i64 nWrite; /* Number of bytes in new PMA */
70826
70827	/* If there are SORTER_MAX_MERGE_COUNT or less PMAs in file pTemp1,
70828	** initialize an iterator for each of them and break out of the loop.
70829	** These iterators will be incrementally merged as the VDBE layer calls
70830	** sqlite3VdbeSorterNext().
70831	**
70832	** Otherwise, if pTemp1 contains more than SORTER_MAX_MERGE_COUNT PMAs,
70833	** initialize interators for SORTER_MAX_MERGE_COUNT of them. These PMAs
70834	** are merged into a single PMA that is written to file pTemp2.
70835	*/
70836	rc = vdbeSorterInitMerge(db, pCsr, &nWrite);
70837	assert( rc!=SQLITE_OK \|\| pSorter->aIter[ pSorter->aTree[1] ].pFile );
70838	if( rc!=SQLITE_OK \|\| pSorter->nPMA<=SORTER_MAX_MERGE_COUNT ){
70839	break;
70840	}
70841
70842	/* Open the second temp file, if it is not already open. */
70843	if( pTemp2==0 ){
70844	assert( iWrite2==0 );
70845	rc = vdbeSorterOpenTempFile(db, &pTemp2);
70846	}
70847
70848	if( rc==SQLITE_OK ){
70849	rc = vdbeSorterWriteVarint(pTemp2, nWrite, &iWrite2);
70850	}
70851
70852	if( rc==SQLITE_OK ){
70853	int bEof = 0;
70854	while( rc==SQLITE_OK && bEof==0 ){
70855	int nToWrite;
70856	VdbeSorterIter *pIter = &pSorter->aIter[ pSorter->aTree[1] ];
70857	assert( pIter->pFile );
70858	nToWrite = pIter->nKey + sqlite3VarintLen(pIter->nKey);
70859	rc = sqlite3OsWrite(pTemp2, pIter->aAlloc, nToWrite, iWrite2);
70860	iWrite2 += nToWrite;
70861	if( rc==SQLITE_OK ){
70862	rc = sqlite3VdbeSorterNext(db, pCsr, &bEof);
70863	}
70864	}
70865	}
70866	}
70867
70868	if( pSorter->nPMA<=SORTER_MAX_MERGE_COUNT ){
70869	break;
70870	}else{
70871	sqlite3_file *pTmp = pSorter->pTemp1;
70872	pSorter->nPMA = iNew;
70873	pSorter->pTemp1 = pTemp2;
70874	pTemp2 = pTmp;
70875	pSorter->iWriteOff = iWrite2;
70876	pSorter->iReadOff = 0;
70877	iWrite2 = 0;
70878	}
70879	}while( rc==SQLITE_OK );
70880
70881	if( pTemp2 ){
70882	sqlite3OsCloseFree(pTemp2);
70883	}
70884	*pbEof = (pSorter->aIter[pSorter->aTree[1]].pFile==0);
70885	return rc;
70886	}
70887
70888	/*
70889	** Advance to the next element in the sorter.
70890	*/
70891	SQLITE_PRIVATE int sqlite3VdbeSorterNext(sqlite3 db, VdbeCursor pCsr, int *pbEof){
70892	VdbeSorter *pSorter = pCsr->pSorter;
70893	int iPrev = pSorter->aTree[1]; /* Index of iterator to advance */
70894	int i; /* Index of aTree[] to recalculate */
70895	int rc; /* Return code */
70896
70897	rc = vdbeSorterIterNext(db, &pSorter->aIter[iPrev]);
70898	for(i=(pSorter->nTree+iPrev)/2; rc==SQLITE_OK && i>0; i=i/2){
70899	rc = vdbeSorterDoCompare(pCsr, i);
70900	}
70901
70902	*pbEof = (pSorter->aIter[pSorter->aTree[1]].pFile==0);
70903	return rc;
70904	}
70905
70906	/*
70907	** Copy the current sorter key into the memory cell pOut.
70908	*/
70909	SQLITE_PRIVATE int sqlite3VdbeSorterRowkey(VdbeCursor pCsr, Mem pOut){
70910	VdbeSorter *pSorter = pCsr->pSorter;
70911	VdbeSorterIter *pIter;
70912
70913	pIter = &pSorter->aIter[ pSorter->aTree[1] ];
70914
70915	/* Coverage testing note: As things are currently, this call will always
70916	** succeed. This is because the memory cell passed by the VDBE layer
70917	** happens to be the same one as was used to assemble the keys before they
70918	** were passed to the sorter - meaning it is always large enough for the
70919	** largest key. But this could change very easily, so we leave the call
70920	** to sqlite3VdbeMemGrow() in. */
70921	if( NEVER(sqlite3VdbeMemGrow(pOut, pIter->nKey, 0)) ){
70922	return SQLITE_NOMEM;
70923	}
70924	pOut->n = pIter->nKey;
70925	MemSetTypeFlag(pOut, MEM_Blob);
70926	memcpy(pOut->z, pIter->aKey, pIter->nKey);
70927
70928	return SQLITE_OK;
70929	}
70930
70931	#endif /* #ifndef SQLITE_OMIT_MERGE_SORT */
70932
70933	/************ End of vdbesort.c ******************************************/
70934	/************ Begin file journal.c ***************************************/
70935	/*
70936	** 2007 August 22
70937	**
70938	** The author disclaims copyright to this source code. In place of
	@@ -69839,10 +71445,12 @@
71445	**
71446	*************************************************************************
71447	** This file contains routines used for walking the parser tree for
71448	** an SQL statement.
71449	*/
71450	/* #include <stdlib.h> */
71451	/* #include <string.h> */
71452
71453
71454	/*
71455	** Walk an expression tree. Invoke the callback once for each node
71456	** of the expression, while decending. (In other words, the callback
	@@ -69977,10 +71585,12 @@
71585	**
71586	** This file contains routines used for walking the parser tree and
71587	** resolve all identifiers by associating them with a particular
71588	** table and column.
71589	*/
71590	/* #include <stdlib.h> */
71591	/* #include <string.h> */
71592
71593	/*
71594	** Turn the pExpr expression into an alias for the iCol-th column of the
71595	** result set in pEList.
71596	**
	@@ -70956,15 +72566,29 @@
72566	/* Recursively resolve names in all subqueries
72567	*/
72568	for(i=0; i<p->pSrc->nSrc; i++){
72569	struct SrcList_item *pItem = &p->pSrc->a[i];
72570	if( pItem->pSelect ){
72571	NameContext pNC; / Used to iterate name contexts */
72572	int nRef = 0; /* Refcount for pOuterNC and outer contexts */
72573	const char *zSavedContext = pParse->zAuthContext;
72574
72575	/* Count the total number of references to pOuterNC and all of its
72576	** parent contexts. After resolving references to expressions in
72577	** pItem->pSelect, check if this value has changed. If so, then
72578	** SELECT statement pItem->pSelect must be correlated. Set the
72579	** pItem->isCorrelated flag if this is the case. */
72580	for(pNC=pOuterNC; pNC; pNC=pNC->pNext) nRef += pNC->nRef;
72581
72582	if( pItem->zName ) pParse->zAuthContext = pItem->zName;
72583	sqlite3ResolveSelectNames(pParse, pItem->pSelect, pOuterNC);
72584	pParse->zAuthContext = zSavedContext;
72585	if( pParse->nErr \|\| db->mallocFailed ) return WRC_Abort;
72586
72587	for(pNC=pOuterNC; pNC; pNC=pNC->pNext) nRef -= pNC->nRef;
72588	assert( pItem->isCorrelated==0 && nRef<=0 );
72589	pItem->isCorrelated = (nRef!=0);
72590	}
72591	}
72592
72593	/* If there are no aggregate functions in the result-set, and no GROUP BY
72594	** expression, do not allow aggregates in any of the other expressions.
	@@ -72068,10 +73692,11 @@
73692	pNewItem->zName = sqlite3DbStrDup(db, pOldItem->zName);
73693	pNewItem->zAlias = sqlite3DbStrDup(db, pOldItem->zAlias);
73694	pNewItem->jointype = pOldItem->jointype;
73695	pNewItem->iCursor = pOldItem->iCursor;
73696	pNewItem->isPopulated = pOldItem->isPopulated;
73697	pNewItem->isCorrelated = pOldItem->isCorrelated;
73698	pNewItem->zIndex = sqlite3DbStrDup(db, pOldItem->zIndex);
73699	pNewItem->notIndexed = pOldItem->notIndexed;
73700	pNewItem->pIndex = pOldItem->pIndex;
73701	pTab = pNewItem->pTab = pOldItem->pTab;
73702	if( pTab ){
	@@ -78958,11 +80583,11 @@
80583	Table *p;
80584	int n;
80585	const char *z;
80586	Token sEnd;
80587	DbFixer sFix;
80588	Token *pName = 0;
80589	int iDb;
80590	sqlite3 *db = pParse->db;
80591
80592	if( pParse->nVar>0 ){
80593	sqlite3ErrorMsg(pParse, "parameters are not allowed in views");
	@@ -79263,10 +80888,33 @@
80888	iDestroyed = iLargest;
80889	}
80890	}
80891	#endif
80892	}
80893
80894	/*
80895	** Remove entries from the sqlite_stat1 and sqlite_stat2 tables
80896	** after a DROP INDEX or DROP TABLE command.
80897	*/
80898	static void sqlite3ClearStatTables(
80899	Parse pParse, / The parsing context */
80900	int iDb, /* The database number */
80901	const char zType, / "idx" or "tbl" */
80902	const char zName / Name of index or table */
80903	){
80904	static const char *azStatTab[] = { "sqlite_stat1", "sqlite_stat2" };
80905	int i;
80906	const char *zDbName = pParse->db->aDb[iDb].zName;
80907	for(i=0; i<ArraySize(azStatTab); i++){
80908	if( sqlite3FindTable(pParse->db, azStatTab[i], zDbName) ){
80909	sqlite3NestedParse(pParse,
80910	"DELETE FROM %Q.%s WHERE %s=%Q",
80911	zDbName, azStatTab[i], zType, zName
80912	);
80913	}
80914	}
80915	}
80916
80917	/*
80918	** This routine is called to do the work of a DROP TABLE statement.
80919	** pName is the name of the table to be dropped.
80920	*/
	@@ -79403,18 +81051,11 @@
81051	** database.
81052	*/
81053	sqlite3NestedParse(pParse,
81054	"DELETE FROM %Q.%s WHERE tbl_name=%Q and type!='trigger'",
81055	pDb->zName, SCHEMA_TABLE(iDb), pTab->zName);
81056	sqlite3ClearStatTables(pParse, iDb, "tbl", pTab->zName);







81057	if( !isView && !IsVirtual(pTab) ){
81058	destroyTable(pParse, pTab);
81059	}
81060
81061	/* Remove the table entry from SQLite's internal schema and modify
	@@ -79592,18 +81233,28 @@
81233	*/
81234	static void sqlite3RefillIndex(Parse pParse, Index pIndex, int memRootPage){
81235	Table pTab = pIndex->pTable; / The table that is indexed */
81236	int iTab = pParse->nTab++; /* Btree cursor used for pTab */
81237	int iIdx = pParse->nTab++; /* Btree cursor used for pIndex */
81238	int iSorter = iTab; /* Cursor opened by OpenSorter (if in use) */
81239	int addr1; /* Address of top of loop */
81240	int tnum; /* Root page of index */
81241	Vdbe v; / Generate code into this virtual machine */
81242	KeyInfo pKey; / KeyInfo for index */
81243	int regIdxKey; /* Registers containing the index key */
81244	int regRecord; /* Register holding assemblied index record */
81245	sqlite3 db = pParse->db; / The database connection */
81246	int iDb = sqlite3SchemaToIndex(db, pIndex->pSchema);
81247
81248	/* Set bUseSorter to use OP_OpenSorter, or clear it to insert directly
81249	** into the index. The sorter is used unless either OMIT_MERGE_SORT is
81250	** defined or the system is configured to store temp files in-memory. */
81251	#ifdef SQLITE_OMIT_MERGE_SORT
81252	static const int bUseSorter = 0;
81253	#else
81254	const int bUseSorter = !sqlite3TempInMemory(pParse->db);
81255	#endif
81256
81257	#ifndef SQLITE_OMIT_AUTHORIZATION
81258	if( sqlite3AuthCheck(pParse, SQLITE_REINDEX, pIndex->zName, 0,
81259	db->aDb[iDb].zName ) ){
81260	return;
	@@ -79625,14 +81276,33 @@
81276	sqlite3VdbeAddOp4(v, OP_OpenWrite, iIdx, tnum, iDb,
81277	(char *)pKey, P4_KEYINFO_HANDOFF);
81278	if( memRootPage>=0 ){
81279	sqlite3VdbeChangeP5(v, 1);
81280	}
81281
81282	/* Open the sorter cursor if we are to use one. */
81283	if( bUseSorter ){
81284	iSorter = pParse->nTab++;
81285	sqlite3VdbeAddOp4(v, OP_OpenSorter, iSorter, 0, 0, (char*)pKey, P4_KEYINFO);
81286	sqlite3VdbeChangeP5(v, BTREE_SORTER);
81287	}
81288
81289	/* Open the table. Loop through all rows of the table, inserting index
81290	** records into the sorter. */
81291	sqlite3OpenTable(pParse, iTab, iDb, pTab, OP_OpenRead);
81292	addr1 = sqlite3VdbeAddOp2(v, OP_Rewind, iTab, 0);
81293	regRecord = sqlite3GetTempReg(pParse);
81294	regIdxKey = sqlite3GenerateIndexKey(pParse, pIndex, iTab, regRecord, 1);
81295
81296	if( bUseSorter ){
81297	sqlite3VdbeAddOp2(v, OP_IdxInsert, iSorter, regRecord);
81298	sqlite3VdbeAddOp2(v, OP_Next, iTab, addr1+1);
81299	sqlite3VdbeJumpHere(v, addr1);
81300	addr1 = sqlite3VdbeAddOp2(v, OP_Sort, iSorter, 0);
81301	sqlite3VdbeAddOp2(v, OP_RowKey, iSorter, regRecord);
81302	}
81303
81304	if( pIndex->onError!=OE_None ){
81305	const int regRowid = regIdxKey + pIndex->nColumn;
81306	const int j2 = sqlite3VdbeCurrentAddr(v) + 2;
81307	void * const pRegKey = SQLITE_INT_TO_PTR(regIdxKey);
81308
	@@ -79647,17 +81317,19 @@
81317	*/
81318	sqlite3VdbeAddOp4(v, OP_IsUnique, iIdx, j2, regRowid, pRegKey, P4_INT32);
81319	sqlite3HaltConstraint(
81320	pParse, OE_Abort, "indexed columns are not unique", P4_STATIC);
81321	}
81322	sqlite3VdbeAddOp3(v, OP_IdxInsert, iIdx, regRecord, bUseSorter);
81323	sqlite3VdbeChangeP5(v, OPFLAG_USESEEKRESULT);
81324	sqlite3ReleaseTempReg(pParse, regRecord);
81325	sqlite3VdbeAddOp2(v, OP_Next, iSorter, addr1+1);
81326	sqlite3VdbeJumpHere(v, addr1);
81327
81328	sqlite3VdbeAddOp1(v, OP_Close, iTab);
81329	sqlite3VdbeAddOp1(v, OP_Close, iIdx);
81330	sqlite3VdbeAddOp1(v, OP_Close, iSorter);
81331	}
81332
81333	/*
81334	** Create a new index for an SQL table. pName1.pName2 is the name of the index
81335	** and pTblList is the name of the table that is to be indexed. Both will
	@@ -80075,11 +81747,11 @@
81747	if( pStart ){
81748	assert( pEnd!=0 );
81749	/* A named index with an explicit CREATE INDEX statement */
81750	zStmt = sqlite3MPrintf(db, "CREATE%s INDEX %.*s",
81751	onError==OE_None ? "" : " UNIQUE",
81752	(int)(pEnd->z - pName->z) + 1,
81753	pName->z);
81754	}else{
81755	/* An automatic index created by a PRIMARY KEY or UNIQUE constraint */
81756	/* zStmt = sqlite3MPrintf(""); */
81757	zStmt = 0;
	@@ -80233,19 +81905,13 @@
81905	v = sqlite3GetVdbe(pParse);
81906	if( v ){
81907	sqlite3BeginWriteOperation(pParse, 1, iDb);
81908	sqlite3NestedParse(pParse,
81909	"DELETE FROM %Q.%s WHERE name=%Q AND type='index'",
81910	db->aDb[iDb].zName, SCHEMA_TABLE(iDb), pIndex->zName
81911	);
81912	sqlite3ClearStatTables(pParse, iDb, "idx", pIndex->zName);






81913	sqlite3ChangeCookie(pParse, iDb);
81914	destroyRootPage(pParse, pIndex->tnum, iDb);
81915	sqlite3VdbeAddOp4(v, OP_DropIndex, iDb, 0, 0, pIndex->zName, 0);
81916	}
81917
	@@ -80613,12 +82279,13 @@
82279	** The operator is "natural cross join". The A and B operands are stored
82280	** in p->a[0] and p->a[1], respectively. The parser initially stores the
82281	** operator with A. This routine shifts that operator over to B.
82282	*/
82283	SQLITE_PRIVATE void sqlite3SrcListShiftJoinType(SrcList *p){
82284	if( p ){
82285	int i;
82286	assert( p->a \|\| p->nSrc==0 );
82287	for(i=p->nSrc-1; i>0; i--){
82288	p->a[i].jointype = p->a[i-1].jointype;
82289	}
82290	p->a[0].jointype = 0;
82291	}
	@@ -81870,11 +83537,13 @@
83537	int regRowid; /* Actual register containing rowids */
83538
83539	/* Collect rowids of every row to be deleted.
83540	*/
83541	sqlite3VdbeAddOp2(v, OP_Null, 0, iRowSet);
83542	pWInfo = sqlite3WhereBegin(
83543	pParse, pTabList, pWhere, 0, 0, WHERE_DUPLICATES_OK
83544	);
83545	if( pWInfo==0 ) goto delete_from_cleanup;
83546	regRowid = sqlite3ExprCodeGetColumn(pParse, pTab, -1, iCur, iRowid);
83547	sqlite3VdbeAddOp2(v, OP_RowSetAdd, iRowSet, regRowid);
83548	if( db->flags & SQLITE_CountRows ){
83549	sqlite3VdbeAddOp2(v, OP_AddImm, memCnt, 1);
	@@ -82135,12 +83804,18 @@
83804	sqlite3VdbeAddOp3(v, OP_Column, iCur, idx, regBase+j);
83805	sqlite3ColumnDefault(v, pTab, idx, -1);
83806	}
83807	}
83808	if( doMakeRec ){
83809	const char *zAff;
83810	if( pTab->pSelect \|\| (pParse->db->flags & SQLITE_IdxRealAsInt)!=0 ){
83811	zAff = 0;
83812	}else{
83813	zAff = sqlite3IndexAffinityStr(v, pIdx);
83814	}
83815	sqlite3VdbeAddOp3(v, OP_MakeRecord, regBase, nCol+1, regOut);
83816	sqlite3VdbeChangeP4(v, -1, zAff, P4_TRANSIENT);
83817	}
83818	sqlite3ReleaseTempRange(pParse, regBase, nCol+1);
83819	return regBase;
83820	}
83821
	@@ -82162,10 +83837,12 @@
83837	**
83838	** There is only one exported symbol in this file - the function
83839	** sqliteRegisterBuildinFunctions() found at the bottom of the file.
83840	** All other code has file scope.
83841	*/
83842	/* #include <stdlib.h> */
83843	/* #include <assert.h> */
83844
83845	/*
83846	** Return the collating function associated with a function.
83847	*/
83848	static CollSeq sqlite3GetFuncCollSeq(sqlite3_context context){
	@@ -84311,11 +85988,11 @@
85988
85989	/* Create VDBE to loop through the entries in pSrc that match the WHERE
85990	** clause. If the constraint is not deferred, throw an exception for
85991	** each row found. Otherwise, for deferred constraints, increment the
85992	** deferred constraint counter by nIncr for each row selected. */
85993	pWInfo = sqlite3WhereBegin(pParse, pSrc, pWhere, 0, 0, 0);
85994	if( nIncr>0 && pFKey->isDeferred==0 ){
85995	sqlite3ParseToplevel(pParse)->mayAbort = 1;
85996	}
85997	sqlite3VdbeAddOp2(v, OP_FkCounter, pFKey->isDeferred, nIncr);
85998	if( pWInfo ){
	@@ -84485,11 +86162,28 @@
86162	pTo = sqlite3FindTable(db, pFKey->zTo, zDb);
86163	}else{
86164	pTo = sqlite3LocateTable(pParse, 0, pFKey->zTo, zDb);
86165	}
86166	if( !pTo \|\| locateFkeyIndex(pParse, pTo, pFKey, &pIdx, &aiFree) ){
86167	assert( isIgnoreErrors==0 \|\| (regOld!=0 && regNew==0) );
86168	if( !isIgnoreErrors \|\| db->mallocFailed ) return;
86169	if( pTo==0 ){
86170	/* If isIgnoreErrors is true, then a table is being dropped. In this
86171	** case SQLite runs a "DELETE FROM xxx" on the table being dropped
86172	** before actually dropping it in order to check FK constraints.
86173	** If the parent table of an FK constraint on the current table is
86174	** missing, behave as if it is empty. i.e. decrement the relevant
86175	** FK counter for each row of the current table with non-NULL keys.
86176	*/
86177	Vdbe *v = sqlite3GetVdbe(pParse);
86178	int iJump = sqlite3VdbeCurrentAddr(v) + pFKey->nCol + 1;
86179	for(i=0; i<pFKey->nCol; i++){
86180	int iReg = pFKey->aCol[i].iFrom + regOld + 1;
86181	sqlite3VdbeAddOp2(v, OP_IsNull, iReg, iJump);
86182	}
86183	sqlite3VdbeAddOp2(v, OP_FkCounter, pFKey->isDeferred, -1);
86184	}
86185	continue;
86186	}
86187	assert( pFKey->nCol==1 \|\| (aiFree && pIdx) );
86188
86189	if( aiFree ){
	@@ -87178,10 +88872,13 @@
88872	int (strnicmp)(const char,const char*,int);
88873	int (unlock_notify)(sqlite3,void()(void,int),void);
88874	int (wal_autocheckpoint)(sqlite3,int);
88875	int (wal_checkpoint)(sqlite3,const char*);
88876	void (wal_hook)(sqlite3,int()(void,sqlite3,const char,int),void);
88877	int (blob_reopen)(sqlite3_blob,sqlite3_int64);
88878	int (vtab_config)(sqlite3,int op,...);
88879	int (vtab_on_conflict)(sqlite3);
88880	};
88881
88882	/*
88883	** The following macros redefine the API routines so that they are
88884	** redirected throught the global sqlite3_api structure.
	@@ -87378,19 +89075,23 @@
89075	#define sqlite3_strnicmp sqlite3_api->strnicmp
89076	#define sqlite3_unlock_notify sqlite3_api->unlock_notify
89077	#define sqlite3_wal_autocheckpoint sqlite3_api->wal_autocheckpoint
89078	#define sqlite3_wal_checkpoint sqlite3_api->wal_checkpoint
89079	#define sqlite3_wal_hook sqlite3_api->wal_hook
89080	#define sqlite3_blob_reopen sqlite3_api->blob_reopen
89081	#define sqlite3_vtab_config sqlite3_api->vtab_config
89082	#define sqlite3_vtab_on_conflict sqlite3_api->vtab_on_conflict
89083	#endif /* SQLITE_CORE */
89084
89085	#define SQLITE_EXTENSION_INIT1 const sqlite3_api_routines *sqlite3_api = 0;
89086	#define SQLITE_EXTENSION_INIT2(v) sqlite3_api = v;
89087
89088	#endif /* _SQLITE3EXT_H_ */
89089
89090	/************ End of sqlite3ext.h ****************************************/
89091	/************ Continuing where we left off in loadext.c ******************/
89092	/* #include <string.h> */
89093
89094	#ifndef SQLITE_OMIT_LOAD_EXTENSION
89095
89096	/*
89097	** Some API routines are omitted when various features are
	@@ -87452,10 +89153,12 @@
89153
89154	#ifdef SQLITE_OMIT_VIRTUALTABLE
89155	# define sqlite3_create_module 0
89156	# define sqlite3_create_module_v2 0
89157	# define sqlite3_declare_vtab 0
89158	# define sqlite3_vtab_config 0
89159	# define sqlite3_vtab_on_conflict 0
89160	#endif
89161
89162	#ifdef SQLITE_OMIT_SHARED_CACHE
89163	# define sqlite3_enable_shared_cache 0
89164	#endif
	@@ -87475,10 +89178,11 @@
89178	#define sqlite3_blob_bytes 0
89179	#define sqlite3_blob_close 0
89180	#define sqlite3_blob_open 0
89181	#define sqlite3_blob_read 0
89182	#define sqlite3_blob_write 0
89183	#define sqlite3_blob_reopen 0
89184	#endif
89185
89186	/*
89187	** The following structure contains pointers to all SQLite API routines.
89188	** A pointer to this structure is passed into extensions when they are
	@@ -87740,10 +89444,13 @@
89444	#else
89445	0,
89446	0,
89447	0,
89448	#endif
89449	sqlite3_blob_reopen,
89450	sqlite3_vtab_config,
89451	sqlite3_vtab_on_conflict,
89452	};
89453
89454	/*
89455	** Attempt to load an SQLite extension library contained in the file
89456	** zFile. The entry point is zProc. zProc may be 0 in which case a
	@@ -94129,10 +95836,11 @@
95836	Expr pHaving; / The HAVING clause. May be NULL */
95837	int isDistinct; /* True if the DISTINCT keyword is present */
95838	int distinct; /* Table to use for the distinct set */
95839	int rc = 1; /* Value to return from this function */
95840	int addrSortIndex; /* Address of an OP_OpenEphemeral instruction */
95841	int addrDistinctIndex; /* Address of an OP_OpenEphemeral instruction */
95842	AggInfo sAggInfo; /* Information used by aggregate queries */
95843	int iEnd; /* Address of the end of the query */
95844	sqlite3 db; / The database connection */
95845
95846	#ifndef SQLITE_OMIT_EXPLAIN
	@@ -94255,20 +95963,10 @@
95963	explainSetInteger(pParse->iSelectId, iRestoreSelectId);
95964	return rc;
95965	}
95966	#endif
95967










95968	/* If there is both a GROUP BY and an ORDER BY clause and they are
95969	** identical, then disable the ORDER BY clause since the GROUP BY
95970	** will cause elements to come out in the correct order. This is
95971	** an optimization - the correct answer should result regardless.
95972	** Use the SQLITE_GroupByOrder flag with SQLITE_TESTCTRL_OPTIMIZER
	@@ -94276,10 +95974,34 @@
95974	*/
95975	if( sqlite3ExprListCompare(p->pGroupBy, pOrderBy)==0
95976	&& (db->flags & SQLITE_GroupByOrder)==0 ){
95977	pOrderBy = 0;
95978	}
95979
95980	/* If the query is DISTINCT with an ORDER BY but is not an aggregate, and
95981	** if the select-list is the same as the ORDER BY list, then this query
95982	** can be rewritten as a GROUP BY. In other words, this:
95983	**
95984	** SELECT DISTINCT xyz FROM ... ORDER BY xyz
95985	**
95986	** is transformed to:
95987	**
95988	** SELECT xyz FROM ... GROUP BY xyz
95989	**
95990	** The second form is preferred as a single index (or temp-table) may be
95991	** used for both the ORDER BY and DISTINCT processing. As originally
95992	** written the query must use a temp-table for at least one of the ORDER
95993	** BY and DISTINCT, and an index or separate temp-table for the other.
95994	*/
95995	if( (p->selFlags & (SF_Distinct\|SF_Aggregate))==SF_Distinct
95996	&& sqlite3ExprListCompare(pOrderBy, p->pEList)==0
95997	){
95998	p->selFlags &= ~SF_Distinct;
95999	p->pGroupBy = sqlite3ExprListDup(db, p->pEList, 0);
96000	pGroupBy = p->pGroupBy;
96001	pOrderBy = 0;
96002	}
96003
96004	/* If there is an ORDER BY clause, then this sorting
96005	** index might end up being unused if the data can be
96006	** extracted in pre-sorted order. If that is the case, then the
96007	** OP_OpenEphemeral instruction will be changed to an OP_Noop once
	@@ -94312,26 +96034,25 @@
96034
96035	/* Open a virtual index to use for the distinct set.
96036	*/
96037	if( p->selFlags & SF_Distinct ){
96038	KeyInfo *pKeyInfo;

96039	distinct = pParse->nTab++;
96040	pKeyInfo = keyInfoFromExprList(pParse, p->pEList);
96041	addrDistinctIndex = sqlite3VdbeAddOp4(v, OP_OpenEphemeral, distinct, 0, 0,
96042	(char*)pKeyInfo, P4_KEYINFO_HANDOFF);
96043	sqlite3VdbeChangeP5(v, BTREE_UNORDERED);
96044	}else{
96045	distinct = addrDistinctIndex = -1;
96046	}
96047
96048	/* Aggregate and non-aggregate queries are handled differently */
96049	if( !isAgg && pGroupBy==0 ){
96050	ExprList *pDist = (isDistinct ? p->pEList : 0);
96051
96052	/* Begin the database scan. */
96053	pWInfo = sqlite3WhereBegin(pParse, pTabList, pWhere, &pOrderBy, pDist, 0);
96054	if( pWInfo==0 ) goto select_end;
96055	if( pWInfo->nRowOut < p->nSelectRow ) p->nSelectRow = pWInfo->nRowOut;
96056
96057	/* If sorting index that was created by a prior OP_OpenEphemeral
96058	** instruction ended up not being needed, then change the OP_OpenEphemeral
	@@ -94340,14 +96061,56 @@
96061	if( addrSortIndex>=0 && pOrderBy==0 ){
96062	sqlite3VdbeChangeToNoop(v, addrSortIndex, 1);
96063	p->addrOpenEphm[2] = -1;
96064	}
96065
96066	if( pWInfo->eDistinct ){
96067	VdbeOp pOp; / No longer required OpenEphemeral instr. */
96068
96069	assert( addrDistinctIndex>0 );
96070	pOp = sqlite3VdbeGetOp(v, addrDistinctIndex);
96071
96072	assert( isDistinct );
96073	assert( pWInfo->eDistinct==WHERE_DISTINCT_ORDERED
96074	\|\| pWInfo->eDistinct==WHERE_DISTINCT_UNIQUE
96075	);
96076	distinct = -1;
96077	if( pWInfo->eDistinct==WHERE_DISTINCT_ORDERED ){
96078	int iJump;
96079	int iExpr;
96080	int iFlag = ++pParse->nMem;
96081	int iBase = pParse->nMem+1;
96082	int iBase2 = iBase + pEList->nExpr;
96083	pParse->nMem += (pEList->nExpr*2);
96084
96085	/* Change the OP_OpenEphemeral coded earlier to an OP_Integer. The
96086	** OP_Integer initializes the "first row" flag. */
96087	pOp->opcode = OP_Integer;
96088	pOp->p1 = 1;
96089	pOp->p2 = iFlag;
96090
96091	sqlite3ExprCodeExprList(pParse, pEList, iBase, 1);
96092	iJump = sqlite3VdbeCurrentAddr(v) + 1 + pEList->nExpr + 1 + 1;
96093	sqlite3VdbeAddOp2(v, OP_If, iFlag, iJump-1);
96094	for(iExpr=0; iExpr<pEList->nExpr; iExpr++){
96095	CollSeq *pColl = sqlite3ExprCollSeq(pParse, pEList->a[iExpr].pExpr);
96096	sqlite3VdbeAddOp3(v, OP_Ne, iBase+iExpr, iJump, iBase2+iExpr);
96097	sqlite3VdbeChangeP4(v, -1, (const char *)pColl, P4_COLLSEQ);
96098	sqlite3VdbeChangeP5(v, SQLITE_NULLEQ);
96099	}
96100	sqlite3VdbeAddOp2(v, OP_Goto, 0, pWInfo->iContinue);
96101
96102	sqlite3VdbeAddOp2(v, OP_Integer, 0, iFlag);
96103	assert( sqlite3VdbeCurrentAddr(v)==iJump );
96104	sqlite3VdbeAddOp3(v, OP_Move, iBase, iBase2, pEList->nExpr);
96105	}else{
96106	pOp->opcode = OP_Noop;
96107	}
96108	}
96109
96110	/* Use the standard inner loop. */
96111	selectInnerLoop(pParse, p, pEList, 0, 0, pOrderBy, distinct, pDest,
96112	pWInfo->iContinue, pWInfo->iBreak);
96113
96114	/* End the database scan loop.
96115	*/
96116	sqlite3WhereEnd(pWInfo);
	@@ -94453,11 +96216,11 @@
96216	** This might involve two separate loops with an OP_Sort in between, or
96217	** it might be a single loop that uses an index to extract information
96218	** in the right order to begin with.
96219	*/
96220	sqlite3VdbeAddOp2(v, OP_Gosub, regReset, addrReset);
96221	pWInfo = sqlite3WhereBegin(pParse, pTabList, pWhere, &pGroupBy, 0, 0);
96222	if( pWInfo==0 ) goto select_end;
96223	if( pGroupBy==0 ){
96224	/* The optimizer is able to deliver rows in group by order so
96225	** we do not have to sort. The OP_OpenEphemeral table will be
96226	** cancelled later because we still need to use the pKeyInfo
	@@ -94715,11 +96478,11 @@
96478	/* This case runs if the aggregate has no GROUP BY clause. The
96479	** processing is much simpler since there is only a single row
96480	** of output.
96481	*/
96482	resetAccumulator(pParse, &sAggInfo);
96483	pWInfo = sqlite3WhereBegin(pParse, pTabList, pWhere, &pMinMax, 0, flag);
96484	if( pWInfo==0 ){
96485	sqlite3ExprListDelete(db, pDel);
96486	goto select_end;
96487	}
96488	updateAccumulator(pParse, &sAggInfo);
	@@ -94894,10 +96657,12 @@
96657	** interface routine of sqlite3_exec().
96658	**
96659	** These routines are in a separate files so that they will not be linked
96660	** if they are not used.
96661	*/
96662	/* #include <stdlib.h> */
96663	/* #include <string.h> */
96664
96665	#ifndef SQLITE_OMIT_GET_TABLE
96666
96667	/*
96668	** This structure is used to pass data from sqlite3_get_table() through
	@@ -95191,19 +96956,32 @@
96956	iDb = sqlite3TwoPartName(pParse, pName1, pName2, &pName);
96957	if( iDb<0 ){
96958	goto trigger_cleanup;
96959	}
96960	}
96961	if( !pTableName \|\| db->mallocFailed ){
96962	goto trigger_cleanup;
96963	}
96964
96965	/* A long-standing parser bug is that this syntax was allowed:
96966	**
96967	** CREATE TRIGGER attached.demo AFTER INSERT ON attached.tab ....
96968	** ^^^^^^^^
96969	**
96970	** To maintain backwards compatibility, ignore the database
96971	** name on pTableName if we are reparsing our of SQLITE_MASTER.
96972	*/
96973	if( db->init.busy && iDb!=1 ){
96974	sqlite3DbFree(db, pTableName->a[0].zDatabase);
96975	pTableName->a[0].zDatabase = 0;
96976	}
96977
96978	/* If the trigger name was unqualified, and the table is a temp table,
96979	** then set iDb to 1 to create the trigger in the temporary database.
96980	** If sqlite3SrcListLookup() returns 0, indicating the table does not
96981	** exist, the error is caught by the block below.
96982	*/



96983	pTab = sqlite3SrcListLookup(pParse, pTableName);
96984	if( db->init.busy==0 && pName2->n==0 && pTab
96985	&& pTab->pSchema==db->aDb[1].pSchema ){
96986	iDb = 1;
96987	}
	@@ -96497,11 +98275,13 @@
98275	}
98276
98277	/* Begin the database scan
98278	*/
98279	sqlite3VdbeAddOp2(v, OP_Null, 0, regOldRowid);
98280	pWInfo = sqlite3WhereBegin(
98281	pParse, pTabList, pWhere, 0, 0, WHERE_ONEPASS_DESIRED
98282	);
98283	if( pWInfo==0 ) goto update_cleanup;
98284	okOnePass = pWInfo->okOnePass;
98285
98286	/* Remember the rowid of every item to be updated.
98287	*/
	@@ -98523,10 +100303,11 @@
100303	#define WHERE_REVERSE 0x02000000 /* Scan in reverse order */
100304	#define WHERE_UNIQUE 0x04000000 /* Selects no more than one row */
100305	#define WHERE_VIRTUALTABLE 0x08000000 /* Use virtual-table processing */
100306	#define WHERE_MULTI_OR 0x10000000 /* OR using multiple indices */
100307	#define WHERE_TEMP_INDEX 0x20000000 /* Uses an ephemeral index */
100308	#define WHERE_DISTINCT 0x40000000 /* Correct order for DISTINCT */
100309
100310	/*
100311	** Initialize a preallocated WhereClause structure.
100312	*/
100313	static void whereClauseInit(
	@@ -99667,10 +101448,166 @@
101448	}
101449	}
101450	return 0;
101451	}
101452
101453	/*
101454	** This function searches the expression list passed as the second argument
101455	** for an expression of type TK_COLUMN that refers to the same column and
101456	** uses the same collation sequence as the iCol'th column of index pIdx.
101457	** Argument iBase is the cursor number used for the table that pIdx refers
101458	** to.
101459	**
101460	** If such an expression is found, its index in pList->a[] is returned. If
101461	** no expression is found, -1 is returned.
101462	*/
101463	static int findIndexCol(
101464	Parse pParse, / Parse context */
101465	ExprList pList, / Expression list to search */
101466	int iBase, /* Cursor for table associated with pIdx */
101467	Index pIdx, / Index to match column of */
101468	int iCol /* Column of index to match */
101469	){
101470	int i;
101471	const char *zColl = pIdx->azColl[iCol];
101472
101473	for(i=0; i<pList->nExpr; i++){
101474	Expr *p = pList->a[i].pExpr;
101475	if( p->op==TK_COLUMN
101476	&& p->iColumn==pIdx->aiColumn[iCol]
101477	&& p->iTable==iBase
101478	){
101479	CollSeq *pColl = sqlite3ExprCollSeq(pParse, p);
101480	if( ALWAYS(pColl) && 0==sqlite3StrICmp(pColl->zName, zColl) ){
101481	return i;
101482	}
101483	}
101484	}
101485
101486	return -1;
101487	}
101488
101489	/*
101490	** This routine determines if pIdx can be used to assist in processing a
101491	** DISTINCT qualifier. In other words, it tests whether or not using this
101492	** index for the outer loop guarantees that rows with equal values for
101493	** all expressions in the pDistinct list are delivered grouped together.
101494	**
101495	** For example, the query
101496	**
101497	** SELECT DISTINCT a, b, c FROM tbl WHERE a = ?
101498	**
101499	** can benefit from any index on columns "b" and "c".
101500	*/
101501	static int isDistinctIndex(
101502	Parse pParse, / Parsing context */
101503	WhereClause pWC, / The WHERE clause */
101504	Index pIdx, / The index being considered */
101505	int base, /* Cursor number for the table pIdx is on */
101506	ExprList pDistinct, / The DISTINCT expressions */
101507	int nEqCol /* Number of index columns with == */
101508	){
101509	Bitmask mask = 0; /* Mask of unaccounted for pDistinct exprs */
101510	int i; /* Iterator variable */
101511
101512	if( pIdx->zName==0 \|\| pDistinct==0 \|\| pDistinct->nExpr>=BMS ) return 0;
101513	testcase( pDistinct->nExpr==BMS-1 );
101514
101515	/* Loop through all the expressions in the distinct list. If any of them
101516	** are not simple column references, return early. Otherwise, test if the
101517	** WHERE clause contains a "col=X" clause. If it does, the expression
101518	** can be ignored. If it does not, and the column does not belong to the
101519	** same table as index pIdx, return early. Finally, if there is no
101520	** matching "col=X" expression and the column is on the same table as pIdx,
101521	** set the corresponding bit in variable mask.
101522	*/
101523	for(i=0; i<pDistinct->nExpr; i++){
101524	WhereTerm *pTerm;
101525	Expr *p = pDistinct->a[i].pExpr;
101526	if( p->op!=TK_COLUMN ) return 0;
101527	pTerm = findTerm(pWC, p->iTable, p->iColumn, ~(Bitmask)0, WO_EQ, 0);
101528	if( pTerm ){
101529	Expr *pX = pTerm->pExpr;
101530	CollSeq *p1 = sqlite3BinaryCompareCollSeq(pParse, pX->pLeft, pX->pRight);
101531	CollSeq *p2 = sqlite3ExprCollSeq(pParse, p);
101532	if( p1==p2 ) continue;
101533	}
101534	if( p->iTable!=base ) return 0;
101535	mask \|= (((Bitmask)1) << i);
101536	}
101537
101538	for(i=nEqCol; mask && i<pIdx->nColumn; i++){
101539	int iExpr = findIndexCol(pParse, pDistinct, base, pIdx, i);
101540	if( iExpr<0 ) break;
101541	mask &= ~(((Bitmask)1) << iExpr);
101542	}
101543
101544	return (mask==0);
101545	}
101546
101547
101548	/*
101549	** Return true if the DISTINCT expression-list passed as the third argument
101550	** is redundant. A DISTINCT list is redundant if the database contains a
101551	** UNIQUE index that guarantees that the result of the query will be distinct
101552	** anyway.
101553	*/
101554	static int isDistinctRedundant(
101555	Parse *pParse,
101556	SrcList *pTabList,
101557	WhereClause *pWC,
101558	ExprList *pDistinct
101559	){
101560	Table *pTab;
101561	Index *pIdx;
101562	int i;
101563	int iBase;
101564
101565	/* If there is more than one table or sub-select in the FROM clause of
101566	** this query, then it will not be possible to show that the DISTINCT
101567	** clause is redundant. */
101568	if( pTabList->nSrc!=1 ) return 0;
101569	iBase = pTabList->a[0].iCursor;
101570	pTab = pTabList->a[0].pTab;
101571
101572	/* If any of the expressions is an IPK column on table iBase, then return
101573	** true. Note: The (p->iTable==iBase) part of this test may be false if the
101574	** current SELECT is a correlated sub-query.
101575	*/
101576	for(i=0; i<pDistinct->nExpr; i++){
101577	Expr *p = pDistinct->a[i].pExpr;
101578	if( p->op==TK_COLUMN && p->iTable==iBase && p->iColumn<0 ) return 1;
101579	}
101580
101581	/* Loop through all indices on the table, checking each to see if it makes
101582	** the DISTINCT qualifier redundant. It does so if:
101583	**
101584	** 1. The index is itself UNIQUE, and
101585	**
101586	** 2. All of the columns in the index are either part of the pDistinct
101587	** list, or else the WHERE clause contains a term of the form "col=X",
101588	** where X is a constant value. The collation sequences of the
101589	** comparison and select-list expressions must match those of the index.
101590	*/
101591	for(pIdx=pTab->pIndex; pIdx; pIdx=pIdx->pNext){
101592	if( pIdx->onError==OE_None ) continue;
101593	for(i=0; i<pIdx->nColumn; i++){
101594	int iCol = pIdx->aiColumn[i];
101595	if( 0==findTerm(pWC, iBase, iCol, ~(Bitmask)0, WO_EQ, pIdx)
101596	&& 0>findIndexCol(pParse, pDistinct, iBase, pIdx, i)
101597	){
101598	break;
101599	}
101600	}
101601	if( i==pIdx->nColumn ){
101602	/* This index implies that the DISTINCT qualifier is redundant. */
101603	return 1;
101604	}
101605	}
101606
101607	return 0;
101608	}
101609
101610	/*
101611	** This routine decides if pIdx can be used to satisfy the ORDER BY
101612	** clause. If it can, it returns 1. If pIdx cannot satisfy the
101613	** ORDER BY clause, this routine returns 0.
	@@ -99703,11 +101640,14 @@
101640	int sortOrder = 0; /* XOR of index and ORDER BY sort direction */
101641	int nTerm; /* Number of ORDER BY terms */
101642	struct ExprList_item pTerm; / A term of the ORDER BY clause */
101643	sqlite3 *db = pParse->db;
101644
101645	if( !pOrderBy ) return 0;
101646	if( wsFlags & WHERE_COLUMN_IN ) return 0;
101647	if( pIdx->bUnordered ) return 0;
101648
101649	nTerm = pOrderBy->nExpr;
101650	assert( nTerm>0 );
101651
101652	/* Argument pIdx must either point to a 'real' named index structure,
101653	** or an index structure allocated on the stack by bestBtreeIndex() to
	@@ -100016,10 +101956,14 @@
101956	double costTempIdx; /* per-query cost of the transient index */
101957	WhereTerm pTerm; / A single term of the WHERE clause */
101958	WhereTerm pWCEnd; / End of pWC->a[] */
101959	Table pTable; / Table tht might be indexed */
101960
101961	if( pParse->nQueryLoop<=(double)1 ){
101962	/* There is no point in building an automatic index for a single scan */
101963	return;
101964	}
101965	if( (pParse->db->flags & SQLITE_AutoIndex)==0 ){
101966	/* Automatic indices are disabled at run-time */
101967	return;
101968	}
101969	if( (pCost->plan.wsFlags & WHERE_NOT_FULLSCAN)!=0 ){
	@@ -100027,10 +101971,14 @@
101971	return;
101972	}
101973	if( pSrc->notIndexed ){
101974	/* The NOT INDEXED clause appears in the SQL. */
101975	return;
101976	}
101977	if( pSrc->isCorrelated ){
101978	/* The source is a correlated sub-query. No point in indexing it. */
101979	return;
101980	}
101981
101982	assert( pParse->nQueryLoop >= (double)1 );
101983	pTable = pSrc->pTab;
101984	nTableRow = pTable->nRowEst;
	@@ -100244,10 +102192,11 @@
102192	if( pTerm->leftCursor != pSrc->iCursor ) continue;
102193	assert( (pTerm->eOperator&(pTerm->eOperator-1))==0 );
102194	testcase( pTerm->eOperator==WO_IN );
102195	testcase( pTerm->eOperator==WO_ISNULL );
102196	if( pTerm->eOperator & (WO_IN\|WO_ISNULL) ) continue;
102197	if( pTerm->wtFlags & TERM_VNULL ) continue;
102198	nTerm++;
102199	}
102200
102201	/* If the ORDER BY clause contains only columns in the current
102202	** virtual table then allocate space for the aOrderBy part of
	@@ -100294,10 +102243,11 @@
102243	if( pTerm->leftCursor != pSrc->iCursor ) continue;
102244	assert( (pTerm->eOperator&(pTerm->eOperator-1))==0 );
102245	testcase( pTerm->eOperator==WO_IN );
102246	testcase( pTerm->eOperator==WO_ISNULL );
102247	if( pTerm->eOperator & (WO_IN\|WO_ISNULL) ) continue;
102248	if( pTerm->wtFlags & TERM_VNULL ) continue;
102249	pIdxCons[j].iColumn = pTerm->u.leftColumn;
102250	pIdxCons[j].iTermOffset = i;
102251	pIdxCons[j].op = (u8)pTerm->eOperator;
102252	/* The direct assignment in the previous line is possible only because
102253	** the WO_ and SQLITE_INDEX_CONSTRAINT_ codes are identical. The
	@@ -100959,10 +102909,11 @@
102909	WhereClause pWC, / The WHERE clause */
102910	struct SrcList_item pSrc, / The FROM clause term to search */
102911	Bitmask notReady, /* Mask of cursors not available for indexing */
102912	Bitmask notValid, /* Cursors not available for any purpose */
102913	ExprList pOrderBy, / The ORDER BY clause */
102914	ExprList pDistinct, / The select-list if query is DISTINCT */
102915	WhereCost pCost / Lowest cost query plan */
102916	){
102917	int iCur = pSrc->iCursor; /* The cursor of the table to be accessed */
102918	Index pProbe; / An index we are evaluating */
102919	Index pIdx; / Copy of pProbe, or zero for IPK index */
	@@ -101099,11 +103050,12 @@
103050	int nEq; /* Number of == or IN terms matching index */
103051	int bInEst = 0; /* True if "x IN (SELECT...)" seen */
103052	int nInMul = 1; /* Number of distinct equalities to lookup */
103053	int estBound = 100; /* Estimated reduction in search space */
103054	int nBound = 0; /* Number of range constraints seen */
103055	int bSort = !!pOrderBy; /* True if external sort required */
103056	int bDist = !!pDistinct; /* True if index cannot help with DISTINCT */
103057	int bLookup = 0; /* True if not a covering index */
103058	WhereTerm pTerm; / A single term of the WHERE clause */
103059	#ifdef SQLITE_ENABLE_STAT2
103060	WhereTerm pFirstTerm = 0; / First term matching the index */
103061	#endif
	@@ -101163,21 +103115,24 @@
103115
103116	/* If there is an ORDER BY clause and the index being considered will
103117	** naturally scan rows in the required order, set the appropriate flags
103118	** in wsFlags. Otherwise, if there is an ORDER BY clause but the index
103119	** will scan rows in a different order, set the bSort variable. */
103120	if( isSortingIndex(
103121	pParse, pWC->pMaskSet, pProbe, iCur, pOrderBy, nEq, wsFlags, &rev)
103122	){
103123	bSort = 0;
103124	wsFlags \|= WHERE_ROWID_RANGE\|WHERE_COLUMN_RANGE\|WHERE_ORDERBY;
103125	wsFlags \|= (rev ? WHERE_REVERSE : 0);
103126	}
103127
103128	/* If there is a DISTINCT qualifier and this index will scan rows in
103129	** order of the DISTINCT expressions, clear bDist and set the appropriate
103130	** flags in wsFlags. */
103131	if( isDistinctIndex(pParse, pWC, pProbe, iCur, pDistinct, nEq) ){
103132	bDist = 0;
103133	wsFlags \|= WHERE_ROWID_RANGE\|WHERE_COLUMN_RANGE\|WHERE_DISTINCT;
103134	}
103135
103136	/* If currently calculating the cost of using an index (not the IPK
103137	** index), determine if all required column data may be obtained without
103138	** using the main table (i.e. if the index is a covering
	@@ -101208,16 +103163,17 @@
103163	nRow = aiRowEst[0]/2;
103164	nInMul = (int)(nRow / aiRowEst[nEq]);
103165	}
103166
103167	#ifdef SQLITE_ENABLE_STAT2
103168	/* If the constraint is of the form x=VALUE or x IN (E1,E2,...)
103169	** and we do not think that values of x are unique and if histogram
103170	** data is available for column x, then it might be possible
103171	** to get a better estimate on the number of rows based on
103172	** VALUE and how common that value is according to the histogram.
103173	*/
103174	if( nRow>(double)1 && nEq==1 && pFirstTerm!=0 && aiRowEst[1]>1 ){
103175	if( pFirstTerm->eOperator & (WO_EQ\|WO_ISNULL) ){
103176	testcase( pFirstTerm->eOperator==WO_EQ );
103177	testcase( pFirstTerm->eOperator==WO_ISNULL );
103178	whereEqualScanEst(pParse, pProbe, pFirstTerm->pExpr->pRight, &nRow);
103179	}else if( pFirstTerm->eOperator==WO_IN && bInEst==0 ){
	@@ -101290,10 +103246,13 @@
103246	** difference and select C of 3.0.
103247	*/
103248	if( bSort ){
103249	cost += nRowestLog(nRow)3;
103250	}
103251	if( bDist ){
103252	cost += nRowestLog(nRow)3;
103253	}
103254
103255	/** Cost of using this index has now been computed **/
103256
103257	/* If there are additional constraints on this table that cannot
103258	** be used with the current index, but which might lower the number
	@@ -101435,11 +103394,11 @@
103394	}
103395	sqlite3DbFree(pParse->db, p);
103396	}else
103397	#endif
103398	{
103399	bestBtreeIndex(pParse, pWC, pSrc, notReady, notValid, pOrderBy, 0, pCost);
103400	}
103401	}
103402
103403	/*
103404	** Disable a term in the WHERE clause. Except, do not disable the term
	@@ -102397,11 +104356,11 @@
104356	for(ii=0; ii<pOrWc->nTerm; ii++){
104357	WhereTerm *pOrTerm = &pOrWc->a[ii];
104358	if( pOrTerm->leftCursor==iCur \|\| pOrTerm->eOperator==WO_AND ){
104359	WhereInfo pSubWInfo; / Info for single OR-term scan */
104360	/* Loop through table entries that match term pOrTerm. */
104361	pSubWInfo = sqlite3WhereBegin(pParse, pOrTab, pOrTerm->pExpr, 0, 0,
104362	WHERE_OMIT_OPEN \| WHERE_OMIT_CLOSE \|
104363	WHERE_FORCE_TABLE \| WHERE_ONETABLE_ONLY);
104364	if( pSubWInfo ){
104365	explainOneScan(
104366	pParse, pOrTab, &pSubWInfo->a[0], iLevel, pLevel->iFrom, 0
	@@ -102638,10 +104597,11 @@
104597	SQLITE_PRIVATE WhereInfo *sqlite3WhereBegin(
104598	Parse pParse, / The parser context */
104599	SrcList pTabList, / A list of all tables to be scanned */
104600	Expr pWhere, / The WHERE clause */
104601	ExprList *ppOrderBy, / An ORDER BY clause, or NULL */
104602	ExprList pDistinct, / The select-list for DISTINCT queries - or NULL */
104603	u16 wctrlFlags /* One of the WHERE_* flags defined in sqliteInt.h */
104604	){
104605	int i; /* Loop counter */
104606	int nByteWInfo; /* Num. bytes allocated for WhereInfo struct */
104607	int nTabList; /* Number of elements in pTabList */
	@@ -102697,10 +104657,14 @@
104657	pWInfo->iBreak = sqlite3VdbeMakeLabel(v);
104658	pWInfo->pWC = pWC = (WhereClause )&((u8 )pWInfo)[nByteWInfo];
104659	pWInfo->wctrlFlags = wctrlFlags;
104660	pWInfo->savedNQueryLoop = pParse->nQueryLoop;
104661	pMaskSet = (WhereMaskSet*)&pWC[1];
104662
104663	/* Disable the DISTINCT optimization if SQLITE_DistinctOpt is set via
104664	** sqlite3_test_ctrl(SQLITE_TESTCTRL_OPTIMIZATIONS,...) */
104665	if( db->flags & SQLITE_DistinctOpt ) pDistinct = 0;
104666
104667	/* Split the WHERE clause into separate subexpressions where each
104668	** subexpression is separated by an AND operator.
104669	*/
104670	initMaskSet(pMaskSet);
	@@ -102764,10 +104728,19 @@
104728	*/
104729	exprAnalyzeAll(pTabList, pWC);
104730	if( db->mallocFailed ){
104731	goto whereBeginError;
104732	}
104733
104734	/* Check if the DISTINCT qualifier, if there is one, is redundant.
104735	** If it is, then set pDistinct to NULL and WhereInfo.eDistinct to
104736	** WHERE_DISTINCT_UNIQUE to tell the caller to ignore the DISTINCT.
104737	*/
104738	if( pDistinct && isDistinctRedundant(pParse, pTabList, pWC, pDistinct) ){
104739	pDistinct = 0;
104740	pWInfo->eDistinct = WHERE_DISTINCT_UNIQUE;
104741	}
104742
104743	/* Chose the best index to use for each table in the FROM clause.
104744	**
104745	** This loop fills in the following fields:
104746	**
	@@ -102848,10 +104821,11 @@
104821	Bitmask mask; /* Mask of tables not yet ready */
104822	for(j=iFrom, pTabItem=&pTabList->a[j]; j<nTabList; j++, pTabItem++){
104823	int doNotReorder; /* True if this table should not be reordered */
104824	WhereCost sCost; /* Cost information from best[Virtual]Index() */
104825	ExprList pOrderBy; / ORDER BY clause for index to optimize */
104826	ExprList pDist; / DISTINCT clause for index to optimize */
104827
104828	doNotReorder = (pTabItem->jointype & (JT_LEFT\|JT_CROSS))!=0;
104829	if( j!=iFrom && doNotReorder ) break;
104830	m = getMask(pMaskSet, pTabItem->iCursor);
104831	if( (m & notReady)==0 ){
	@@ -102858,10 +104832,11 @@
104832	if( j==iFrom ) iFrom++;
104833	continue;
104834	}
104835	mask = (isOptimal ? m : notReady);
104836	pOrderBy = ((i==0 && ppOrderBy )?*ppOrderBy:0);
104837	pDist = (i==0 ? pDistinct : 0);
104838	if( pTabItem->pIndex==0 ) nUnconstrained++;
104839
104840	WHERETRACE(("=== trying table %d with isOptimal=%d ===\n",
104841	j, isOptimal));
104842	assert( pTabItem->pTab );
	@@ -102872,11 +104847,11 @@
104847	&sCost, pp);
104848	}else
104849	#endif
104850	{
104851	bestBtreeIndex(pParse, pWC, pTabItem, mask, notReady, pOrderBy,
104852	pDist, &sCost);
104853	}
104854	assert( isOptimal \|\| (sCost.used&notReady)==0 );
104855
104856	/* If an INDEXED BY clause is present, then the plan must use that
104857	** index if it uses any index at all */
	@@ -102932,10 +104907,14 @@
104907	WHERETRACE(("*** Optimizer selects table %d for loop %d"
104908	" with cost=%g and nRow=%g\n",
104909	bestJ, pLevel-pWInfo->a, bestPlan.rCost, bestPlan.plan.nRow));
104910	if( (bestPlan.plan.wsFlags & WHERE_ORDERBY)!=0 ){
104911	*ppOrderBy = 0;
104912	}
104913	if( (bestPlan.plan.wsFlags & WHERE_DISTINCT)!=0 ){
104914	assert( pWInfo->eDistinct==0 );
104915	pWInfo->eDistinct = WHERE_DISTINCT_ORDERED;
104916	}
104917	andFlags &= bestPlan.plan.wsFlags;
104918	pLevel->plan = bestPlan.plan;
104919	testcase( bestPlan.plan.wsFlags & WHERE_INDEXED );
104920	testcase( bestPlan.plan.wsFlags & WHERE_TEMP_INDEX );
	@@ -103267,10 +105246,11 @@
105246	** LALR(1) grammar but which are always false in the
105247	** specific grammar used by SQLite.
105248	*/
105249	/* First off, code is included that follows the "include" declaration
105250	** in the input grammar file. */
105251	/* #include <stdio.h> */
105252
105253
105254	/*
105255	** Disable all error recovery processing in the parser push-down
105256	** automaton.
	@@ -104127,10 +106107,11 @@
106107	#endif
106108	};
106109	typedef struct yyParser yyParser;
106110
106111	#ifndef NDEBUG
106112	/* #include <stdio.h> */
106113	static FILE *yyTraceFILE = 0;
106114	static char *yyTracePrompt = 0;
106115	#endif /* NDEBUG */
106116
106117	#ifndef NDEBUG
	@@ -106702,10 +108683,11 @@
108683	**
108684	** This file contains C code that splits an SQL input string up into
108685	** individual tokens and sends those tokens one-by-one over to the
108686	** parser for analysis.
108687	*/
108688	/* #include <stdlib.h> */
108689
108690	/*
108691	** The charMap() macro maps alphabetic characters into their
108692	** lower-case ASCII equivalent. On ASCII machines, this is just
108693	** an upper-to-lower case map. On EBCDIC machines we also need
	@@ -108093,10 +110075,20 @@
110075	memcpy(&y, &x, 8);
110076	assert( sqlite3IsNaN(y) );
110077	}
110078	#endif
110079	#endif
110080
110081	/* Do extra initialization steps requested by the SQLITE_EXTRA_INIT
110082	** compile-time option.
110083	*/
110084	#ifdef SQLITE_EXTRA_INIT
110085	if( rc==SQLITE_OK && sqlite3GlobalConfig.isInit ){
110086	int SQLITE_EXTRA_INIT(void);
110087	rc = SQLITE_EXTRA_INIT();
110088	}
110089	#endif
110090
110091	return rc;
110092	}
110093
110094	/*
	@@ -110763,10 +112755,21 @@
112755	if( sz ) *ppNew = sqlite3ScratchMalloc(sz);
112756	sqlite3ScratchFree(pFree);
112757	break;
112758	}
112759
112760	/* sqlite3_test_control(SQLITE_TESTCTRL_LOCALTIME_FAULT, int onoff);
112761	**
112762	** If parameter onoff is non-zero, configure the wrappers so that all
112763	** subsequent calls to localtime() and variants fail. If onoff is zero,
112764	** undo this setting.
112765	*/
112766	case SQLITE_TESTCTRL_LOCALTIME_FAULT: {
112767	sqlite3GlobalConfig.bLocaltimeFault = va_arg(ap, int);
112768	break;
112769	}
112770
112771	}
112772	va_end(ap);
112773	#endif /* SQLITE_OMIT_BUILTIN_TEST */
112774	return rc;
112775	}
	@@ -111420,16 +113423,10 @@
113423	** TODO(shess) Provide a VACUUM type operation to clear out all
113424	** deletions and duplications. This would basically be a forced merge
113425	** into a single segment.
113426	*/
113427






113428	/************ Include fts3Int.h in the middle of fts3.c ******************/
113429	/************ Begin file fts3Int.h ***************************************/
113430	/*
113431	** 2009 Nov 12
113432	**
	@@ -111441,18 +113438,33 @@
113438	** May you share freely, never taking more than you give.
113439	**
113440	******************************************************************************
113441	**
113442	*/

113443	#ifndef _FTSINT_H
113444	#define _FTSINT_H
113445
113446	#if !defined(NDEBUG) && !defined(SQLITE_DEBUG)
113447	# define NDEBUG 1
113448	#endif
113449
113450	/*
113451	** FTS4 is really an extension for FTS3. It is enabled using the
113452	** SQLITE_ENABLE_FTS3 macro. But to avoid confusion we also all
113453	** the SQLITE_ENABLE_FTS4 macro to serve as an alisse for SQLITE_ENABLE_FTS3.
113454	*/
113455	#if defined(SQLITE_ENABLE_FTS4) && !defined(SQLITE_ENABLE_FTS3)
113456	# define SQLITE_ENABLE_FTS3
113457	#endif
113458
113459	#if !defined(SQLITE_CORE) \|\| defined(SQLITE_ENABLE_FTS3)
113460
113461	/* If not building as part of the core, include sqlite3ext.h. */
113462	#ifndef SQLITE_CORE
113463	SQLITE_API extern const sqlite3_api_routines *sqlite3_api;
113464	#endif
113465
113466	/************ Include fts3_tokenizer.h in the middle of fts3Int.h ********/
113467	/************ Begin file fts3_tokenizer.h ********************************/
113468	/*
113469	** 2006 July 10
113470	**
	@@ -111940,11 +113952,11 @@
113952	char aDoclist; / List of docids for full-text queries */
113953	int nDoclist; /* Size of buffer at aDoclist */
113954	u8 bDesc; /* True to sort in descending order */
113955	int eEvalmode; /* An FTS3_EVAL_XX constant */
113956	int nRowAvg; /* Average size of database rows, in pages */
113957	sqlite3_int64 nDoc; /* Documents in table */
113958
113959	int isMatchinfoNeeded; /* True when aMatchinfo[] needs filling in */
113960	u32 aMatchinfo; / Information about most recent match */
113961	int nMatchinfo; /* Number of elements in aMatchinfo[] */
113962	char zMatchinfo; / Matchinfo specification */
	@@ -111981,11 +113993,11 @@
113993
113994	sqlite3_int64 iDocid; /* Current docid (if pList!=0) */
113995	int bFreeList; /* True if pList should be sqlite3_free()d */
113996	char pList; / Pointer to position list following iDocid */
113997	int nList; /* Length of position list */
113998	};
113999
114000	/*
114001	** A "phrase" is a sequence of one or more tokens that must match in
114002	** sequence. A single token is the base case and the most common case.
114003	** For a sequence of tokens contained in double-quotes (i.e. "one two three")
	@@ -111997,19 +114009,19 @@
114009	int isPrefix; /* True if token ends with a "" character /
114010
114011	/* Variables above this point are populated when the expression is
114012	** parsed (by code in fts3_expr.c). Below this point the variables are
114013	** used when evaluating the expression. */

114014	Fts3DeferredToken pDeferred; / Deferred token object for this token */
114015	Fts3MultiSegReader pSegcsr; / Segment-reader for this token */
114016	};
114017
114018	struct Fts3Phrase {
114019	/* Cache of doclist for this phrase. */
114020	Fts3Doclist doclist;
114021	int bIncr; /* True if doclist is loaded incrementally */
114022	int iDoclistToken;
114023
114024	/* Variables below this point are populated by fts3_expr.c when parsing
114025	** a MATCH expression. Everything above is part of the evaluation phase.
114026	*/
114027	int nToken; /* Number of tokens in the phrase */
	@@ -112130,10 +114142,11 @@
114142	Fts3SegFilter pFilter; / Pointer to filter object */
114143	char aBuffer; / Buffer to merge doclists in */
114144	int nBuffer; /* Allocated size of aBuffer[] in bytes */
114145
114146	int iColFilter; /* If >=0, filter for this column */
114147	int bRestart;
114148
114149	/* Used by fts3.c only. */
114150	int nCost; /* Cost of running iterator */
114151	int bLookup; /* True if a lookup of a single entry. */
114152
	@@ -112180,43 +114193,49 @@
114193	#endif
114194
114195	/* fts3_aux.c */
114196	SQLITE_PRIVATE int sqlite3Fts3InitAux(sqlite3 *db);
114197








114198	SQLITE_PRIVATE void sqlite3Fts3EvalPhraseCleanup(Fts3Phrase *);
114199



114200	SQLITE_PRIVATE int sqlite3Fts3MsrIncrStart(
114201	Fts3Table, Fts3MultiSegReader, int, const char*, int);
114202	SQLITE_PRIVATE int sqlite3Fts3MsrIncrNext(
114203	Fts3Table , Fts3MultiSegReader , sqlite3_int64 , char , int );
114204	SQLITE_PRIVATE char sqlite3Fts3EvalPhrasePoslist(Fts3Cursor , Fts3Expr *, int iCol);
114205	SQLITE_PRIVATE int sqlite3Fts3MsrOvfl(Fts3Cursor , Fts3MultiSegReader , int *);
114206	SQLITE_PRIVATE int sqlite3Fts3MsrIncrRestart(Fts3MultiSegReader *pCsr);
114207
114208	SQLITE_PRIVATE int sqlite3Fts3DeferredTokenList(Fts3DeferredToken , char , int );
114209
114210	#endif /* !SQLITE_CORE \|\| SQLITE_ENABLE_FTS3 */
114211	#endif /* _FTSINT_H */
114212
114213	/************ End of fts3Int.h *******************************************/
114214	/************ Continuing where we left off in fts3.c *********************/
114215	#if !defined(SQLITE_CORE) \|\| defined(SQLITE_ENABLE_FTS3)
114216
114217	#if defined(SQLITE_ENABLE_FTS3) && !defined(SQLITE_CORE)
114218	# define SQLITE_CORE 1
114219	#endif
114220
114221	/* #include <assert.h> */
114222	/* #include <stdlib.h> */
114223	/* #include <stddef.h> */
114224	/* #include <stdio.h> */
114225	/* #include <string.h> */
114226	/* #include <stdarg.h> */
114227
114228	#ifndef SQLITE_CORE
114229	SQLITE_EXTENSION_INIT1
114230	#endif
114231
114232	static int fts3EvalNext(Fts3Cursor *pCsr);
114233	static int fts3EvalStart(Fts3Cursor *pCsr);
114234	static int fts3TermSegReaderCursor(
114235	Fts3Cursor , const char , int, int, Fts3MultiSegReader **);
114236
114237	/*
114238	** Write a 64-bit variable-length integer to memory starting at p[0].
114239	** The length of data written will be between 1 and FTS3_VARINT_MAX bytes.
114240	** The number of bytes written is returned.
114241	*/
	@@ -112545,10 +114564,13 @@
114564	rc = sqlite3_prepare(p->db, zSql, -1, &pStmt, 0);
114565	if( rc==SQLITE_OK ){
114566	sqlite3_step(pStmt);
114567	p->nPgsz = sqlite3_column_int(pStmt, 0);
114568	rc = sqlite3_finalize(pStmt);
114569	}else if( rc==SQLITE_AUTH ){
114570	p->nPgsz = 1024;
114571	rc = SQLITE_OK;
114572	}
114573	}
114574	assert( p->nPgsz>0 \|\| rc!=SQLITE_OK );
114575	sqlite3_free(zSql);
114576	*pRc = rc;
	@@ -112718,31 +114740,60 @@
114740	}
114741	sqlite3_free(zFree);
114742	return zRet;
114743	}
114744
114745	/*
114746	** This function interprets the string at (*pp) as a non-negative integer
114747	** value. It reads the integer and sets *pnOut to the value read, then
114748	** sets *pp to point to the byte immediately following the last byte of
114749	** the integer value.
114750	**
114751	** Only decimal digits ('0'..'9') may be part of an integer value.
114752	**
114753	** If *pp does not being with a decimal digit SQLITE_ERROR is returned and
114754	** the output value undefined. Otherwise SQLITE_OK is returned.
114755	**
114756	** This function is used when parsing the "prefix=" FTS4 parameter.
114757	*/
114758	static int fts3GobbleInt(const char *pp, int pnOut){
114759	const char p = pp; /* Iterator pointer */
114760	int nInt = 0; /* Output value */
114761
114762	for(p=*pp; p[0]>='0' && p[0]<='9'; p++){
114763	nInt = nInt * 10 + (p[0] - '0');
114764	}
114765	if( p==*pp ) return SQLITE_ERROR;
114766	*pnOut = nInt;
114767	*pp = p;
114768	return SQLITE_OK;
114769	}
114770
114771	/*
114772	** This function is called to allocate an array of Fts3Index structures
114773	** representing the indexes maintained by the current FTS table. FTS tables
114774	** always maintain the main "terms" index, but may also maintain one or
114775	** more "prefix" indexes, depending on the value of the "prefix=" parameter
114776	** (if any) specified as part of the CREATE VIRTUAL TABLE statement.
114777	**
114778	** Argument zParam is passed the value of the "prefix=" option if one was
114779	** specified, or NULL otherwise.
114780	**
114781	** If no error occurs, SQLITE_OK is returned and *apIndex set to point to
114782	** the allocated array. *pnIndex is set to the number of elements in the
114783	** array. If an error does occur, an SQLite error code is returned.
114784	**
114785	** Regardless of whether or not an error is returned, it is the responsibility
114786	** of the caller to call sqlite3_free() on the output array to free it.
114787	*/
114788	static int fts3PrefixParameter(
114789	const char zParam, / ABC in prefix=ABC parameter to parse */
114790	int pnIndex, / OUT: size of apIndex[] array /
114791	struct Fts3Index *apIndex / OUT: Array of indexes for this table */

114792	){
114793	struct Fts3Index aIndex; / Allocated array */
114794	int nIndex = 1; /* Number of entries in array */
114795
114796	if( zParam && zParam[0] ){
114797	const char *p;
114798	nIndex++;
114799	for(p=zParam; *p; p++){
	@@ -112749,11 +114800,11 @@
114800	if( *p==',' ) nIndex++;
114801	}
114802	}
114803
114804	aIndex = sqlite3_malloc(sizeof(struct Fts3Index) * nIndex);
114805	*apIndex = aIndex;
114806	*pnIndex = nIndex;
114807	if( !aIndex ){
114808	return SQLITE_NOMEM;
114809	}
114810
	@@ -112806,12 +114857,11 @@
114857	int isFts4 = (argv[0][3]=='4'); /* True for FTS4, false for FTS3 */
114858	const char *aCol; / Array of column names */
114859	sqlite3_tokenizer pTokenizer = 0; / Tokenizer for this table */
114860
114861	int nIndex; /* Size of aIndex[] array */
114862	struct Fts3Index aIndex = 0; / Array of indexes for this table */

114863
114864	/* The results of parsing supported FTS4 key=value options: */
114865	int bNoDocsize = 0; /* True to omit %_docsize table */
114866	int bDescIdx = 0; /* True to store descending indexes */
114867	char zPrefix = 0; / Prefix parameter value (or NULL) */
	@@ -112944,11 +114994,11 @@
114994	rc = sqlite3Fts3InitTokenizer(pHash, "simple", &pTokenizer, pzErr);
114995	if( rc!=SQLITE_OK ) goto fts3_init_out;
114996	}
114997	assert( pTokenizer );
114998
114999	rc = fts3PrefixParameter(zPrefix, &nIndex, &aIndex);
115000	if( rc==SQLITE_ERROR ){
115001	assert( zPrefix );
115002	*pzErr = sqlite3_mprintf("error parsing prefix parameter: %s", zPrefix);
115003	}
115004	if( rc!=SQLITE_OK ) goto fts3_init_out;
	@@ -112995,11 +115045,11 @@
115045	zCsr += nDb;
115046
115047	/* Fill in the azColumn array */
115048	for(iCol=0; iCol<nCol; iCol++){
115049	char *z;
115050	int n = 0;
115051	z = (char *)sqlite3Fts3NextToken(aCol[iCol], &n);
115052	memcpy(zCsr, z, n);
115053	zCsr[n] = '\0';
115054	sqlite3Fts3Dequote(zCsr);
115055	p->azColumn[iCol] = zCsr;
	@@ -113031,11 +115081,11 @@
115081	/* Declare the table schema to SQLite. */
115082	fts3DeclareVtab(&rc, p);
115083
115084	fts3_init_out:
115085	sqlite3_free(zPrefix);
115086	sqlite3_free(aIndex);
115087	sqlite3_free(zCompress);
115088	sqlite3_free(zUncompress);
115089	sqlite3_free((void *)aCol);
115090	if( rc!=SQLITE_OK ){
115091	if( p ){
	@@ -113622,12 +115672,10 @@
115672	*pp1 = p1 + 1;
115673	*pp2 = p2 + 1;
115674	}
115675
115676	/*


115677	** This function is used to merge two position lists into one. When it is
115678	** called, pp1 and pp2 must both point to position lists. A position-list is
115679	** the part of a doclist that follows each document id. For example, if a row
115680	** contains:
115681	**
	@@ -113643,10 +115691,12 @@
115691	** If isSaveLeft is 0, an entry is added to the output position list for
115692	** each position in *pp2 for which there exists one or more positions in
115693	** pp1 so that (pos(pp2)>pos(pp1) && pos(pp2)-pos(*pp1)<=nToken). i.e.
115694	** when the pp1 token appears before the pp2 token, but not more than nToken
115695	** slots before it.
115696	**
115697	** e.g. nToken==1 searches for adjacent positions.
115698	*/
115699	static int fts3PoslistPhraseMerge(
115700	char *pp, / IN/OUT: Preallocated output buffer */
115701	int nToken, /* Maximum difference in token positions */
115702	int isSaveLeft, /* Save the left position */
	@@ -113809,26 +115859,38 @@
115859
115860	return res;
115861	}
115862
115863	/*
115864	** An instance of this function is used to merge together the (potentially
115865	** large number of) doclists for each term that matches a prefix query.
115866	** See function fts3TermSelectMerge() for details.
115867	*/
115868	typedef struct TermSelect TermSelect;
115869	struct TermSelect {
115870	char aaOutput[16]; / Malloc'd output buffers */
115871	int anOutput[16]; /* Size each output buffer in bytes */

115872	};
115873
115874	/*
115875	** This function is used to read a single varint from a buffer. Parameter
115876	** pEnd points 1 byte past the end of the buffer. When this function is
115877	** called, if *pp points to pEnd or greater, then the end of the buffer
115878	** has been reached. In this case *pp is set to 0 and the function returns.
115879	**
115880	** If *pp does not point to or past pEnd, then a single varint is read
115881	** from pp. pp is then set to point 1 byte past the end of the read varint.
115882	**
115883	** If bDescIdx is false, the value read is added to *pVal before returning.
115884	** If it is true, the value read is subtracted from *pVal before this
115885	** function returns.
115886	*/
115887	static void fts3GetDeltaVarint3(
115888	char *pp, / IN/OUT: Point to read varint from */
115889	char pEnd, / End of buffer */
115890	int bDescIdx, /* True if docids are descending */
115891	sqlite3_int64 pVal / IN/OUT: Integer value */
115892	){
115893	if( *pp>=pEnd ){
115894	*pp = 0;
115895	}else{
115896	sqlite3_int64 iVal;
	@@ -113839,10 +115901,25 @@
115901	*pVal += iVal;
115902	}
115903	}
115904	}
115905
115906	/*
115907	** This function is used to write a single varint to a buffer. The varint
115908	** is written to pp. Before returning, pp is set to point 1 byte past the
115909	** end of the value written.
115910	**
115911	** If *pbFirst is zero when this function is called, the value written to
115912	** the buffer is that of parameter iVal.
115913	**
115914	** If *pbFirst is non-zero when this function is called, then the value
115915	** written is either (iVal-piPrev) (if bDescIdx is zero) or (piPrev-iVal)
115916	** (if bDescIdx is non-zero).
115917	**
115918	** Before returning, this function always sets pbFirst to 1 and piPrev
115919	** to the value of parameter iVal.
115920	*/
115921	static void fts3PutDeltaVarint3(
115922	char *pp, / IN/OUT: Output pointer */
115923	int bDescIdx, /* True for descending docids */
115924	sqlite3_int64 piPrev, / IN/OUT: Previous value written to list */
115925	int pbFirst, / IN/OUT: True after first int written */
	@@ -113859,14 +115936,38 @@
115936	pp += sqlite3Fts3PutVarint(pp, iWrite);
115937	*piPrev = iVal;
115938	*pbFirst = 1;
115939	}
115940

115941
115942	/*
115943	** This macro is used by various functions that merge doclists. The two
115944	** arguments are 64-bit docid values. If the value of the stack variable
115945	** bDescDoclist is 0 when this macro is invoked, then it returns (i1-i2).
115946	** Otherwise, (i2-i1).
115947	**
115948	** Using this makes it easier to write code that can merge doclists that are
115949	** sorted in either ascending or descending order.
115950	*/
115951	#define DOCID_CMP(i1, i2) ((bDescDoclist?-1:1) * (i1-i2))
115952
115953	/*
115954	** This function does an "OR" merge of two doclists (output contains all
115955	** positions contained in either argument doclist). If the docids in the
115956	** input doclists are sorted in ascending order, parameter bDescDoclist
115957	** should be false. If they are sorted in ascending order, it should be
115958	** passed a non-zero value.
115959	**
115960	** If no error occurs, *paOut is set to point at an sqlite3_malloc'd buffer
115961	** containing the output doclist and SQLITE_OK is returned. In this case
115962	** *pnOut is set to the number of bytes in the output doclist.
115963	**
115964	** If an error occurs, an SQLite error code is returned. The output values
115965	** are undefined in this case.
115966	*/
115967	static int fts3DoclistOrMerge(
115968	int bDescDoclist, /* True if arguments are desc */
115969	char a1, int n1, / First doclist */
115970	char a2, int n2, / Second doclist */
115971	char *paOut, int pnOut /* OUT: Malloc'd doclist */
115972	){
115973	sqlite3_int64 i1 = 0;
	@@ -113887,35 +115988,47 @@
115988
115989	p = aOut;
115990	fts3GetDeltaVarint3(&p1, pEnd1, 0, &i1);
115991	fts3GetDeltaVarint3(&p2, pEnd2, 0, &i2);
115992	while( p1 \|\| p2 ){
115993	sqlite3_int64 iDiff = DOCID_CMP(i1, i2);
115994
115995	if( p2 && p1 && iDiff==0 ){
115996	fts3PutDeltaVarint3(&p, bDescDoclist, &iPrev, &bFirstOut, i1);
115997	fts3PoslistMerge(&p, &p1, &p2);
115998	fts3GetDeltaVarint3(&p1, pEnd1, bDescDoclist, &i1);
115999	fts3GetDeltaVarint3(&p2, pEnd2, bDescDoclist, &i2);
116000	}else if( !p2 \|\| (p1 && iDiff<0) ){
116001	fts3PutDeltaVarint3(&p, bDescDoclist, &iPrev, &bFirstOut, i1);
116002	fts3PoslistCopy(&p, &p1);
116003	fts3GetDeltaVarint3(&p1, pEnd1, bDescDoclist, &i1);
116004	}else{
116005	fts3PutDeltaVarint3(&p, bDescDoclist, &iPrev, &bFirstOut, i2);
116006	fts3PoslistCopy(&p, &p2);
116007	fts3GetDeltaVarint3(&p2, pEnd2, bDescDoclist, &i2);
116008	}
116009	}
116010
116011	*paOut = aOut;
116012	*pnOut = (p-aOut);
116013	return SQLITE_OK;
116014	}
116015
116016	/*
116017	** This function does a "phrase" merge of two doclists. In a phrase merge,
116018	** the output contains a copy of each position from the right-hand input
116019	** doclist for which there is a position in the left-hand input doclist
116020	** exactly nDist tokens before it.
116021	**
116022	** If the docids in the input doclists are sorted in ascending order,
116023	** parameter bDescDoclist should be false. If they are sorted in ascending
116024	** order, it should be passed a non-zero value.
116025	**
116026	** The right-hand input doclist is overwritten by this function.
116027	*/
116028	static void fts3DoclistPhraseMerge(
116029	int bDescDoclist, /* True if arguments are desc */
116030	int nDist, /* Distance from left to right (1=adjacent) */
116031	char aLeft, int nLeft, / Left doclist */
116032	char aRight, int pnRight /* IN/OUT: Right/output doclist */
116033	){
116034	sqlite3_int64 i1 = 0;
	@@ -113934,30 +116047,30 @@
116047	p = aOut;
116048	fts3GetDeltaVarint3(&p1, pEnd1, 0, &i1);
116049	fts3GetDeltaVarint3(&p2, pEnd2, 0, &i2);
116050
116051	while( p1 && p2 ){
116052	sqlite3_int64 iDiff = DOCID_CMP(i1, i2);
116053	if( iDiff==0 ){
116054	char *pSave = p;
116055	sqlite3_int64 iPrevSave = iPrev;
116056	int bFirstOutSave = bFirstOut;
116057
116058	fts3PutDeltaVarint3(&p, bDescDoclist, &iPrev, &bFirstOut, i1);
116059	if( 0==fts3PoslistPhraseMerge(&p, nDist, 0, 1, &p1, &p2) ){
116060	p = pSave;
116061	iPrev = iPrevSave;
116062	bFirstOut = bFirstOutSave;
116063	}
116064	fts3GetDeltaVarint3(&p1, pEnd1, bDescDoclist, &i1);
116065	fts3GetDeltaVarint3(&p2, pEnd2, bDescDoclist, &i2);
116066	}else if( iDiff<0 ){
116067	fts3PoslistCopy(0, &p1);
116068	fts3GetDeltaVarint3(&p1, pEnd1, bDescDoclist, &i1);
116069	}else{
116070	fts3PoslistCopy(0, &p2);
116071	fts3GetDeltaVarint3(&p2, pEnd2, bDescDoclist, &i2);
116072	}
116073	}
116074
116075	*pnRight = p - aOut;
116076	}
	@@ -113970,11 +116083,11 @@
116083	**
116084	** If an OOM error occurs, return SQLITE_NOMEM. In this case it is
116085	** the responsibility of the caller to free any doclists left in the
116086	** TermSelect.aaOutput[] array.
116087	*/
116088	static int fts3TermSelectFinishMerge(Fts3Table p, TermSelect pTS){
116089	char *aOut = 0;
116090	int nOut = 0;
116091	int i;
116092
116093	/* Loop through the doclists in the aaOutput[] array. Merge them all
	@@ -114011,28 +116124,29 @@
116124	pTS->anOutput[0] = nOut;
116125	return SQLITE_OK;
116126	}
116127
116128	/*
116129	** Merge the doclist aDoclist/nDoclist into the TermSelect object passed
116130	** as the first argument. The merge is an "OR" merge (see function
116131	** fts3DoclistOrMerge() for details).
116132	**
116133	** This function is called with the doclist for each term that matches
116134	** a queried prefix. It merges all these doclists into one, the doclist
116135	** for the specified prefix. Since there can be a very large number of
116136	** doclists to merge, the merging is done pair-wise using the TermSelect
116137	** object.
116138	**
116139	** This function returns SQLITE_OK if the merge is successful, or an
116140	** SQLite error code (SQLITE_NOMEM) if an error occurs.
116141	*/
116142	static int fts3TermSelectMerge(
116143	Fts3Table p, / FTS table handle */
116144	TermSelect pTS, / TermSelect object to merge into */
116145	char aDoclist, / Pointer to doclist */
116146	int nDoclist /* Size of aDoclist in bytes */


116147	){






116148	if( pTS->aaOutput[0]==0 ){
116149	/* If this is the first term selected, copy the doclist to the output
116150	** buffer using memcpy(). */
116151	pTS->aaOutput[0] = sqlite3_malloc(nDoclist);
116152	pTS->anOutput[0] = nDoclist;
	@@ -114099,23 +116213,30 @@
116213	}
116214	pCsr->apSegment[pCsr->nSegment++] = pNew;
116215	return SQLITE_OK;
116216	}
116217
116218	/*
116219	** Add seg-reader objects to the Fts3MultiSegReader object passed as the
116220	** 8th argument.
116221	**
116222	** This function returns SQLITE_OK if successful, or an SQLite error code
116223	** otherwise.
116224	*/
116225	static int fts3SegReaderCursor(
116226	Fts3Table p, / FTS3 table handle */
116227	int iIndex, /* Index to search (from 0 to p->nIndex-1) */
116228	int iLevel, /* Level of segments to scan */
116229	const char zTerm, / Term to query for */
116230	int nTerm, /* Size of zTerm in bytes */
116231	int isPrefix, /* True for a prefix search */
116232	int isScan, /* True to scan from zTerm to EOF */
116233	Fts3MultiSegReader pCsr / Cursor object to populate */
116234	){
116235	int rc = SQLITE_OK; /* Error code */
116236	sqlite3_stmt pStmt = 0; / Statement to iterate through segments */
116237	int rc2; /* Result of sqlite3_reset() */
116238
116239	/* If iLevel is less than 0 and this is not a scan, include a seg-reader
116240	** for the pending-terms. If this is a scan, then this call must be being
116241	** made by an fts4aux module, not an FTS table. In this case calling
116242	** Fts3SegReaderPending might segfault, as the data structures used by
	@@ -114200,28 +116321,46 @@
116321	return fts3SegReaderCursor(
116322	p, iIndex, iLevel, zTerm, nTerm, isPrefix, isScan, pCsr
116323	);
116324	}
116325
116326	/*
116327	** In addition to its current configuration, have the Fts3MultiSegReader
116328	** passed as the 4th argument also scan the doclist for term zTerm/nTerm.
116329	**
116330	** SQLITE_OK is returned if no error occurs, otherwise an SQLite error code.
116331	*/
116332	static int fts3SegReaderCursorAddZero(
116333	Fts3Table p, / FTS virtual table handle */
116334	const char zTerm, / Term to scan doclist of */
116335	int nTerm, /* Number of bytes in zTerm */
116336	Fts3MultiSegReader pCsr / Fts3MultiSegReader to modify */
116337	){
116338	return fts3SegReaderCursor(p, 0, FTS3_SEGCURSOR_ALL, zTerm, nTerm, 0, 0,pCsr);
116339	}
116340
116341	/*
116342	** Open an Fts3MultiSegReader to scan the doclist for term zTerm/nTerm. Or,
116343	** if isPrefix is true, to scan the doclist for all terms for which
116344	** zTerm/nTerm is a prefix. If successful, return SQLITE_OK and write
116345	** a pointer to the new Fts3MultiSegReader to *ppSegcsr. Otherwise, return
116346	** an SQLite error code.
116347	**
116348	** It is the responsibility of the caller to free this object by eventually
116349	** passing it to fts3SegReaderCursorFree()
116350	**
116351	** SQLITE_OK is returned if no error occurs, otherwise an SQLite error code.
116352	** Output parameter *ppSegcsr is set to 0 if an error occurs.
116353	*/
116354	static int fts3TermSegReaderCursor(
116355	Fts3Cursor pCsr, / Virtual table cursor handle */
116356	const char zTerm, / Term to query for */
116357	int nTerm, /* Size of zTerm in bytes */
116358	int isPrefix, /* True for a prefix search */
116359	Fts3MultiSegReader *ppSegcsr / OUT: Allocated seg-reader cursor */
116360	){
116361	Fts3MultiSegReader pSegcsr; / Object to allocate and return */
116362	int rc = SQLITE_NOMEM; /* Return code */
116363
116364	pSegcsr = sqlite3_malloc(sizeof(Fts3MultiSegReader));
116365	if( pSegcsr ){
116366	int i;
	@@ -114261,62 +116400,53 @@
116400
116401	*ppSegcsr = pSegcsr;
116402	return rc;
116403	}
116404
116405	/*
116406	** Free an Fts3MultiSegReader allocated by fts3TermSegReaderCursor().
116407	*/
116408	static void fts3SegReaderCursorFree(Fts3MultiSegReader *pSegcsr){
116409	sqlite3Fts3SegReaderFinish(pSegcsr);
116410	sqlite3_free(pSegcsr);
116411	}
116412
116413	/*
116414	** This function retreives the doclist for the specified term (or term
116415	** prefix) from the database.







116416	*/
116417	static int fts3TermSelect(
116418	Fts3Table p, / Virtual table handle */
116419	Fts3PhraseToken pTok, / Token to query for */
116420	int iColumn, /* Column to query (or -ve for all columns) */

116421	int pnOut, / OUT: Size of buffer at ppOut /
116422	char *ppOut / OUT: Malloced result buffer */
116423	){
116424	int rc; /* Return code */
116425	Fts3MultiSegReader pSegcsr; / Seg-reader cursor for this term */
116426	TermSelect tsc; /* Object for pair-wise doclist merging */
116427	Fts3SegFilter filter; /* Segment term filter configuration */
116428
116429	pSegcsr = pTok->pSegcsr;
116430	memset(&tsc, 0, sizeof(TermSelect));

116431
116432	filter.flags = FTS3_SEGMENT_IGNORE_EMPTY \| FTS3_SEGMENT_REQUIRE_POS
116433	\| (pTok->isPrefix ? FTS3_SEGMENT_PREFIX : 0)

116434	\| (iColumn<p->nColumn ? FTS3_SEGMENT_COLUMN_FILTER : 0);
116435	filter.iCol = iColumn;
116436	filter.zTerm = pTok->z;
116437	filter.nTerm = pTok->n;
116438
116439	rc = sqlite3Fts3SegReaderStart(p, pSegcsr, &filter);
116440	while( SQLITE_OK==rc
116441	&& SQLITE_ROW==(rc = sqlite3Fts3SegReaderStep(p, pSegcsr))
116442	){
116443	rc = fts3TermSelectMerge(p, &tsc, pSegcsr->aDoclist, pSegcsr->nDoclist);


116444	}
116445
116446	if( rc==SQLITE_OK ){
116447	rc = fts3TermSelectFinishMerge(p, &tsc);
116448	}
116449	if( rc==SQLITE_OK ){
116450	*ppOut = tsc.aaOutput[0];
116451	*pnOut = tsc.anOutput[0];
116452	}else{
	@@ -114338,28 +116468,19 @@
116468	** If the isPoslist argument is true, then it is assumed that the doclist
116469	** contains a position-list following each docid. Otherwise, it is assumed
116470	** that the doclist is simply a list of docids stored as delta encoded
116471	** varints.
116472	*/
116473	static int fts3DoclistCountDocids(char *aList, int nList){
116474	int nDoc = 0; /* Return value */
116475	if( aList ){
116476	char aEnd = &aList[nList]; / Pointer to one byte after EOF */
116477	char p = aList; / Cursor */
116478	while( p<aEnd ){
116479	nDoc++;
116480	while( (p++)&0x80 ); / Skip docid varint */
116481	fts3PoslistCopy(0, &p); /* Skip over position list */









116482	}
116483	}
116484
116485	return nDoc;
116486	}
	@@ -114385,11 +116506,11 @@
116506	}else{
116507	pCsr->iPrevId = sqlite3_column_int64(pCsr->pStmt, 0);
116508	rc = SQLITE_OK;
116509	}
116510	}else{
116511	rc = fts3EvalNext((Fts3Cursor *)pCursor);
116512	}
116513	assert( ((Fts3Table *)pCsr->base.pVtab)->pSegments==0 );
116514	return rc;
116515	}
116516
	@@ -114462,11 +116583,11 @@
116583	}
116584
116585	rc = sqlite3Fts3ReadLock(p);
116586	if( rc!=SQLITE_OK ) return rc;
116587
116588	rc = fts3EvalStart(pCsr);
116589
116590	sqlite3Fts3SegmentsClose(p);
116591	if( rc!=SQLITE_OK ) return rc;
116592	pCsr->pNextId = pCsr->aDoclist;
116593	pCsr->iPrevId = 0;
	@@ -114581,12 +116702,12 @@
116702
116703	/*
116704	** Implementation of xBegin() method. This is a no-op.
116705	*/
116706	static int fts3BeginMethod(sqlite3_vtab *pVtab){

116707	TESTONLY( Fts3Table p = (Fts3Table)pVtab );
116708	UNUSED_PARAMETER(pVtab);
116709	assert( p->pSegments==0 );
116710	assert( p->nPendingData==0 );
116711	assert( p->inTransaction!=1 );
116712	TESTONLY( p->inTransaction = 1 );
116713	TESTONLY( p->mxSavepoint = -1; );
	@@ -114597,12 +116718,12 @@
116718	** Implementation of xCommit() method. This is a no-op. The contents of
116719	** the pending-terms hash-table have already been flushed into the database
116720	** by fts3SyncMethod().
116721	*/
116722	static int fts3CommitMethod(sqlite3_vtab *pVtab){

116723	TESTONLY( Fts3Table p = (Fts3Table)pVtab );
116724	UNUSED_PARAMETER(pVtab);
116725	assert( p->nPendingData==0 );
116726	assert( p->inTransaction!=0 );
116727	assert( p->pSegments==0 );
116728	TESTONLY( p->inTransaction = 0 );
116729	TESTONLY( p->mxSavepoint = -1; );
	@@ -114869,26 +116990,43 @@
116990	p->zDb, p->zName, zName
116991	);
116992	return rc;
116993	}
116994
116995	/*
116996	** The xSavepoint() method.
116997	**
116998	** Flush the contents of the pending-terms table to disk.
116999	*/
117000	static int fts3SavepointMethod(sqlite3_vtab *pVtab, int iSavepoint){
117001	UNUSED_PARAMETER(iSavepoint);
117002	assert( ((Fts3Table *)pVtab)->inTransaction );
117003	assert( ((Fts3Table *)pVtab)->mxSavepoint < iSavepoint );
117004	TESTONLY( ((Fts3Table *)pVtab)->mxSavepoint = iSavepoint );
117005	return fts3SyncMethod(pVtab);
117006	}
117007
117008	/*
117009	** The xRelease() method.
117010	**
117011	** This is a no-op.
117012	*/
117013	static int fts3ReleaseMethod(sqlite3_vtab *pVtab, int iSavepoint){
117014	TESTONLY( Fts3Table p = (Fts3Table)pVtab );
117015	UNUSED_PARAMETER(iSavepoint);
117016	UNUSED_PARAMETER(pVtab);
117017	assert( p->inTransaction );
117018	assert( p->mxSavepoint >= iSavepoint );
117019	TESTONLY( p->mxSavepoint = iSavepoint-1 );
117020	return SQLITE_OK;
117021	}
117022
117023	/*
117024	** The xRollbackTo() method.
117025	**
117026	** Discard the contents of the pending terms table.
117027	*/
117028	static int fts3RollbackToMethod(sqlite3_vtab *pVtab, int iSavepoint){
117029	Fts3Table p = (Fts3Table)pVtab;
117030	UNUSED_PARAMETER(iSavepoint);
117031	assert( p->inTransaction );
117032	assert( p->mxSavepoint >= iSavepoint );
	@@ -115033,22 +117171,10 @@
117171	sqlite3Fts3HashClear(pHash);
117172	sqlite3_free(pHash);
117173	}
117174	return rc;
117175	}












117176
117177	/*
117178	** Allocate an Fts3MultiSegReader for each token in the expression headed
117179	** by pExpr.
117180	**
	@@ -115062,99 +117188,151 @@
117188	** there exists prefix b-tree of the right length) then it may be traversed
117189	** and merged incrementally. Otherwise, it has to be merged into an in-memory
117190	** doclist and then traversed.
117191	*/
117192	static void fts3EvalAllocateReaders(
117193	Fts3Cursor pCsr, / FTS cursor handle */
117194	Fts3Expr pExpr, / Allocate readers for this expression */
117195	int pnToken, / OUT: Total number of tokens in phrase. */
117196	int pnOr, / OUT: Total number of OR nodes in expr. */
117197	int pRc / IN/OUT: Error code */
117198	){
117199	if( pExpr && SQLITE_OK==*pRc ){
117200	if( pExpr->eType==FTSQUERY_PHRASE ){
117201	int i;
117202	int nToken = pExpr->pPhrase->nToken;
117203	*pnToken += nToken;
117204	for(i=0; i<nToken; i++){
117205	Fts3PhraseToken *pToken = &pExpr->pPhrase->aToken[i];
117206	int rc = fts3TermSegReaderCursor(pCsr,
117207	pToken->z, pToken->n, pToken->isPrefix, &pToken->pSegcsr
117208	);
117209	if( rc!=SQLITE_OK ){
117210	*pRc = rc;
117211	return;
117212	}
117213	}
117214	assert( pExpr->pPhrase->iDoclistToken==0 );
117215	pExpr->pPhrase->iDoclistToken = -1;
117216	}else{
117217	*pnOr += (pExpr->eType==FTSQUERY_OR);
117218	fts3EvalAllocateReaders(pCsr, pExpr->pLeft, pnToken, pnOr, pRc);
117219	fts3EvalAllocateReaders(pCsr, pExpr->pRight, pnToken, pnOr, pRc);
117220	}
117221	}
117222	}
117223
117224	/*
117225	** Arguments pList/nList contain the doclist for token iToken of phrase p.
117226	** It is merged into the main doclist stored in p->doclist.aAll/nAll.
117227	**
117228	** This function assumes that pList points to a buffer allocated using
117229	** sqlite3_malloc(). This function takes responsibility for eventually
117230	** freeing the buffer.
117231	*/
117232	static void fts3EvalPhraseMergeToken(
117233	Fts3Table pTab, / FTS Table pointer */
117234	Fts3Phrase p, / Phrase to merge pList/nList into */
117235	int iToken, /* Token pList/nList corresponds to */
117236	char pList, / Pointer to doclist */
117237	int nList /* Number of bytes in pList */
117238	){
117239	assert( iToken!=p->iDoclistToken );
117240
117241	if( pList==0 ){
117242	sqlite3_free(p->doclist.aAll);
117243	p->doclist.aAll = 0;
117244	p->doclist.nAll = 0;
117245	}
117246
117247	else if( p->iDoclistToken<0 ){
117248	p->doclist.aAll = pList;
117249	p->doclist.nAll = nList;
117250	}
117251
117252	else if( p->doclist.aAll==0 ){
117253	sqlite3_free(pList);
117254	}
117255
117256	else {
117257	char *pLeft;
117258	char *pRight;
117259	int nLeft;
117260	int nRight;
117261	int nDiff;
117262
117263	if( p->iDoclistToken<iToken ){
117264	pLeft = p->doclist.aAll;
117265	nLeft = p->doclist.nAll;
117266	pRight = pList;
117267	nRight = nList;
117268	nDiff = iToken - p->iDoclistToken;
117269	}else{
117270	pRight = p->doclist.aAll;
117271	nRight = p->doclist.nAll;
117272	pLeft = pList;
117273	nLeft = nList;
117274	nDiff = p->iDoclistToken - iToken;
117275	}
117276
117277	fts3DoclistPhraseMerge(pTab->bDescIdx, nDiff, pLeft, nLeft, pRight,&nRight);
117278	sqlite3_free(pLeft);
117279	p->doclist.aAll = pRight;
117280	p->doclist.nAll = nRight;
117281	}
117282
117283	if( iToken>p->iDoclistToken ) p->iDoclistToken = iToken;
117284	}
117285
117286	/*
117287	** Load the doclist for phrase p into p->doclist.aAll/nAll. The loaded doclist
117288	** does not take deferred tokens into account.
117289	**
117290	** SQLITE_OK is returned if no error occurs, otherwise an SQLite error code.
117291	*/
117292	static int fts3EvalPhraseLoad(
117293	Fts3Cursor pCsr, / FTS Cursor handle */
117294	Fts3Phrase p / Phrase object */
117295	){
117296	Fts3Table pTab = (Fts3Table )pCsr->base.pVtab;
117297	int iToken;
117298	int rc = SQLITE_OK;
117299




117300	for(iToken=0; rc==SQLITE_OK && iToken<p->nToken; iToken++){
117301	Fts3PhraseToken *pToken = &p->aToken[iToken];
117302	assert( pToken->pDeferred==0 \|\| pToken->pSegcsr==0 );
117303
117304	if( pToken->pSegcsr ){
117305	int nThis = 0;
117306	char *pThis = 0;
117307	rc = fts3TermSelect(pTab, pToken, p->iColumn, &nThis, &pThis);
117308	if( rc==SQLITE_OK ){
117309	fts3EvalPhraseMergeToken(pTab, p, iToken, pThis, nThis);
117310	}
117311	}
117312	assert( pToken->pSegcsr==0 );
117313	}
117314
117315	return rc;
117316	}
117317
117318	/*
117319	** This function is called on each phrase after the position lists for
117320	** any deferred tokens have been loaded into memory. It updates the phrases
117321	** current position list to include only those positions that are really
117322	** instances of the phrase (after considering deferred tokens). If this
117323	** means that the phrase does not appear in the current row, doclist.pList
117324	** and doclist.nList are both zeroed.
117325	**
117326	** SQLITE_OK is returned if no error occurs, otherwise an SQLite error code.
117327	*/












117328	static int fts3EvalDeferredPhrase(Fts3Cursor pCsr, Fts3Phrase pPhrase){
117329	int iToken; /* Used to iterate through phrase tokens */
117330	int rc = SQLITE_OK; /* Return code */
117331	char aPoslist = 0; / Position list for deferred tokens */
117332	int nPoslist = 0; /* Number of bytes in aPoslist */
117333	int iPrev = -1; /* Token number of previous deferred token */


117334
117335	assert( pPhrase->doclist.bFreeList==0 );
117336
117337	for(iToken=0; rc==SQLITE_OK && iToken<pPhrase->nToken; iToken++){
117338	Fts3PhraseToken *pToken = &pPhrase->aToken[iToken];
	@@ -115192,16 +117370,15 @@
117370	pPhrase->doclist.nList = 0;
117371	return SQLITE_OK;
117372	}
117373	}
117374	iPrev = iToken;


117375	}
117376	}
117377
117378	if( iPrev>=0 ){
117379	int nMaxUndeferred = pPhrase->iDoclistToken;
117380	if( nMaxUndeferred<0 ){
117381	pPhrase->doclist.pList = aPoslist;
117382	pPhrase->doclist.nList = nPoslist;
117383	pPhrase->doclist.iDocid = pCsr->iPrevId;
117384	pPhrase->doclist.bFreeList = 1;
	@@ -115246,19 +117423,27 @@
117423	/*
117424	** This function is called for each Fts3Phrase in a full-text query
117425	** expression to initialize the mechanism for returning rows. Once this
117426	** function has been called successfully on an Fts3Phrase, it may be
117427	** used with fts3EvalPhraseNext() to iterate through the matching docids.
117428	**
117429	** If parameter bOptOk is true, then the phrase may (or may not) use the
117430	** incremental loading strategy. Otherwise, the entire doclist is loaded into
117431	** memory within this call.
117432	**
117433	** SQLITE_OK is returned if no error occurs, otherwise an SQLite error code.
117434	*/
117435	static int fts3EvalPhraseStart(Fts3Cursor pCsr, int bOptOk, Fts3Phrase p){
117436	int rc; /* Error code */
117437	Fts3PhraseToken *pFirst = &p->aToken[0];
117438	Fts3Table pTab = (Fts3Table )pCsr->base.pVtab;
117439
117440	if( pCsr->bDesc==pTab->bDescIdx
117441	&& bOptOk==1
117442	&& p->nToken==1
117443	&& pFirst->pSegcsr
117444	&& pFirst->pSegcsr->bLookup
117445	){
117446	/* Use the incremental approach. */
117447	int iCol = (p->iColumn >= pTab->nColumn ? -1 : p->iColumn);
117448	rc = sqlite3Fts3MsrIncrStart(
117449	pTab, pFirst->pSegcsr, iCol, pFirst->z, pFirst->n);
	@@ -115274,11 +117459,17 @@
117459	return rc;
117460	}
117461
117462	/*
117463	** This function is used to iterate backwards (from the end to start)
117464	** through doclists. It is used by this module to iterate through phrase
117465	** doclists in reverse and by the fts3_write.c module to iterate through
117466	** pending-terms lists when writing to databases with "order=desc".
117467	**
117468	** The doclist may be sorted in ascending (parameter bDescIdx==0) or
117469	** descending (parameter bDescIdx==1) order of docid. Regardless, this
117470	** function iterates from the end of the doclist to the beginning.
117471	*/
117472	SQLITE_PRIVATE void sqlite3Fts3DoclistPrev(
117473	int bDescIdx, /* True if the doclist is desc */
117474	char aDoclist, / Pointer to entire doclist */
117475	int nDoclist, /* Length of aDoclist in bytes */
	@@ -115339,13 +117530,13 @@
117530	** If there is no "next" entry and no error occurs, then *pbEof is set to
117531	** 1 before returning. Otherwise, if no error occurs and the iterator is
117532	** successfully advanced, *pbEof is set to 0.
117533	*/
117534	static int fts3EvalPhraseNext(
117535	Fts3Cursor pCsr, / FTS Cursor handle */
117536	Fts3Phrase p, / Phrase object to advance to next docid */
117537	u8 pbEof / OUT: Set to 1 if EOF */
117538	){
117539	int rc = SQLITE_OK;
117540	Fts3Doclist *pDL = &p->doclist;
117541	Fts3Table pTab = (Fts3Table )pCsr->base.pVtab;
117542
	@@ -115387,30 +117578,46 @@
117578	fts3PoslistCopy(0, &pIter);
117579	pDL->nList = (pIter - pDL->pList);
117580
117581	/* pIter now points just past the 0x00 that terminates the position-
117582	** list for document pDL->iDocid. However, if this position-list was
117583	** edited in place by fts3EvalNearTrim(), then pIter may not actually
117584	** point to the start of the next docid value. The following line deals
117585	** with this case by advancing pIter past the zero-padding added by
117586	** fts3EvalNearTrim(). */
117587	while( pIter<pEnd && *pIter==0 ) pIter++;
117588
117589	pDL->pNextDocid = pIter;
117590	assert( pIter>=&pDL->aAll[pDL->nAll] \|\| *pIter );
117591	*pbEof = 0;
117592	}
117593	}
117594
117595	return rc;
117596	}
117597
117598	/*
117599	**
117600	** If *pRc is not SQLITE_OK when this function is called, it is a no-op.
117601	** Otherwise, fts3EvalPhraseStart() is called on all phrases within the
117602	** expression. Also the Fts3Expr.bDeferred variable is set to true for any
117603	** expressions for which all descendent tokens are deferred.
117604	**
117605	** If parameter bOptOk is zero, then it is guaranteed that the
117606	** Fts3Phrase.doclist.aAll/nAll variables contain the entire doclist for
117607	** each phrase in the expression (subject to deferred token processing).
117608	** Or, if bOptOk is non-zero, then one or more tokens within the expression
117609	** may be loaded incrementally, meaning doclist.aAll/nAll is not available.
117610	**
117611	** If an error occurs within this function, *pRc is set to an SQLite error
117612	** code before returning.
117613	*/
117614	static void fts3EvalStartReaders(
117615	Fts3Cursor pCsr, / FTS Cursor handle */
117616	Fts3Expr pExpr, / Expression to initialize phrases in */
117617	int bOptOk, /* True to enable incremental loading */
117618	int pRc / IN/OUT: Error code */
117619	){
117620	if( pExpr && SQLITE_OK==*pRc ){
117621	if( pExpr->eType==FTSQUERY_PHRASE ){
117622	int i;
117623	int nToken = pExpr->pPhrase->nToken;
	@@ -115425,33 +117632,55 @@
117632	pExpr->bDeferred = (pExpr->pLeft->bDeferred && pExpr->pRight->bDeferred);
117633	}
117634	}
117635	}
117636
117637	/*
117638	** An array of the following structures is assembled as part of the process
117639	** of selecting tokens to defer before the query starts executing (as part
117640	** of the xFilter() method). There is one element in the array for each
117641	** token in the FTS expression.
117642	**
117643	** Tokens are divided into AND/NEAR clusters. All tokens in a cluster belong
117644	** to phrases that are connected only by AND and NEAR operators (not OR or
117645	** NOT). When determining tokens to defer, each AND/NEAR cluster is considered
117646	** separately. The root of a tokens AND/NEAR cluster is stored in
117647	** Fts3TokenAndCost.pRoot.
117648	*/
117649	typedef struct Fts3TokenAndCost Fts3TokenAndCost;
117650	struct Fts3TokenAndCost {
117651	Fts3Phrase pPhrase; / The phrase the token belongs to */
117652	int iToken; /* Position of token in phrase */
117653	Fts3PhraseToken pToken; / The token itself */
117654	Fts3Expr pRoot; / Root of NEAR/AND cluster */
117655	int nOvfl; /* Number of overflow pages to load doclist */
117656	int iCol; /* The column the token must match */
117657	};
117658
117659	/*
117660	** This function is used to populate an allocated Fts3TokenAndCost array.
117661	**
117662	** If *pRc is not SQLITE_OK when this function is called, it is a no-op.
117663	** Otherwise, if an error occurs during execution, *pRc is set to an
117664	** SQLite error code.
117665	*/
117666	static void fts3EvalTokenCosts(
117667	Fts3Cursor pCsr, / FTS Cursor handle */
117668	Fts3Expr pRoot, / Root of current AND/NEAR cluster */
117669	Fts3Expr pExpr, / Expression to consider */
117670	Fts3TokenAndCost *ppTC, / Write new entries to (ppTC)++ */
117671	Fts3Expr **ppOr, / Write new OR root to (ppOr)++ */
117672	int pRc / IN/OUT: Error code */
117673	){
117674	if( *pRc==SQLITE_OK && pExpr ){
117675	if( pExpr->eType==FTSQUERY_PHRASE ){
117676	Fts3Phrase *pPhrase = pExpr->pPhrase;
117677	int i;
117678	for(i=0; *pRc==SQLITE_OK && i<pPhrase->nToken; i++){
117679	Fts3TokenAndCost pTC = (ppTC)++;
117680	pTC->pPhrase = pPhrase;
117681	pTC->iToken = i;
117682	pTC->pRoot = pRoot;
117683	pTC->pToken = &pPhrase->aToken[i];
117684	pTC->iCol = pPhrase->iColumn;
117685	*pRc = sqlite3Fts3MsrOvfl(pCsr, pTC->pToken->pSegcsr, &pTC->nOvfl);
117686	}
	@@ -115470,23 +117699,34 @@
117699	fts3EvalTokenCosts(pCsr, pRoot, pExpr->pRight, ppTC, ppOr, pRc);
117700	}
117701	}
117702	}
117703
117704	/*
117705	** Determine the average document (row) size in pages. If successful,
117706	** write this value to *pnPage and return SQLITE_OK. Otherwise, return
117707	** an SQLite error code.
117708	**
117709	** The average document size in pages is calculated by first calculating
117710	** determining the average size in bytes, B. If B is less than the amount
117711	** of data that will fit on a single leaf page of an intkey table in
117712	** this database, then the average docsize is 1. Otherwise, it is 1 plus
117713	** the number of overflow pages consumed by a record B bytes in size.
117714	*/
117715	static int fts3EvalAverageDocsize(Fts3Cursor pCsr, int pnPage){
117716	if( pCsr->nRowAvg==0 ){
117717	/* The average document size, which is required to calculate the cost
117718	** of each doclist, has not yet been determined. Read the required
117719	** data from the %_stat table to calculate it.
117720	**
117721	** Entry 0 of the %_stat table is a blob containing (nCol+1) FTS3
117722	** varints, where nCol is the number of columns in the FTS3 table.
117723	** The first varint is the number of documents currently stored in
117724	** the table. The following nCol varints contain the total amount of
117725	** data stored in all rows of each column of the table, from left
117726	** to right.
117727	*/
117728	int rc;
117729	Fts3Table p = (Fts3Table)pCsr->base.pVtab;
117730	sqlite3_stmt *pStmt;
117731	sqlite3_int64 nDoc = 0;
117732	sqlite3_int64 nByte = 0;
	@@ -115517,113 +117757,151 @@
117757
117758	*pnPage = pCsr->nRowAvg;
117759	return SQLITE_OK;
117760	}
117761
117762	/*
117763	** This function is called to select the tokens (if any) that will be
117764	** deferred. The array aTC[] has already been populated when this is
117765	** called.
117766	**
117767	** This function is called once for each AND/NEAR cluster in the
117768	** expression. Each invocation determines which tokens to defer within
117769	** the cluster with root node pRoot. See comments above the definition
117770	** of struct Fts3TokenAndCost for more details.
117771	**
117772	** If no error occurs, SQLITE_OK is returned and sqlite3Fts3DeferToken()
117773	** called on each token to defer. Otherwise, an SQLite error code is
117774	** returned.
117775	*/
117776	static int fts3EvalSelectDeferred(
117777	Fts3Cursor pCsr, / FTS Cursor handle */
117778	Fts3Expr pRoot, / Consider tokens with this root node */
117779	Fts3TokenAndCost aTC, / Array of expression tokens and costs */
117780	int nTC /* Number of entries in aTC[] */
117781	){



117782	Fts3Table pTab = (Fts3Table )pCsr->base.pVtab;
117783	int nDocSize = 0; /* Number of pages per doc loaded */
117784	int rc = SQLITE_OK; /* Return code */
117785	int ii; /* Iterator variable for various purposes */
117786	int nOvfl = 0; /* Total overflow pages used by doclists */
117787	int nToken = 0; /* Total number of tokens in cluster */
117788
117789	int nMinEst = 0; /* The minimum count for any phrase so far. */
117790	int nLoad4 = 1; /* (Phrases that will be loaded)^4. */
117791
117792	/* Count the tokens in this AND/NEAR cluster. If none of the doclists
117793	** associated with the tokens spill onto overflow pages, or if there is
117794	** only 1 token, exit early. No tokens to defer in this case. */
117795	for(ii=0; ii<nTC; ii++){
117796	if( aTC[ii].pRoot==pRoot ){
117797	nOvfl += aTC[ii].nOvfl;
117798	nToken++;
117799	}
117800	}
117801	if( nOvfl==0 \|\| nToken<2 ) return SQLITE_OK;
117802
117803	/* Obtain the average docsize (in pages). */
117804	rc = fts3EvalAverageDocsize(pCsr, &nDocSize);
117805	assert( rc!=SQLITE_OK \|\| nDocSize>0 );
117806
117807
117808	/* Iterate through all tokens in this AND/NEAR cluster, in ascending order
117809	** of the number of overflow pages that will be loaded by the pager layer
117810	** to retrieve the entire doclist for the token from the full-text index.
117811	** Load the doclists for tokens that are either:
117812	**
117813	** a. The cheapest token in the entire query (i.e. the one visited by the
117814	** first iteration of this loop), or
117815	**
117816	** b. Part of a multi-token phrase.
117817	**
117818	** After each token doclist is loaded, merge it with the others from the
117819	** same phrase and count the number of documents that the merged doclist
117820	** contains. Set variable "nMinEst" to the smallest number of documents in
117821	** any phrase doclist for which 1 or more token doclists have been loaded.
117822	** Let nOther be the number of other phrases for which it is certain that
117823	** one or more tokens will not be deferred.
117824	**
117825	** Then, for each token, defer it if loading the doclist would result in
117826	** loading N or more overflow pages into memory, where N is computed as:
117827	**
117828	** (nMinEst + 4^nOther - 1) / (4^nOther)
117829	*/
117830	for(ii=0; ii<nToken && rc==SQLITE_OK; ii++){
117831	int iTC; /* Used to iterate through aTC[] array. */
117832	Fts3TokenAndCost pTC = 0; / Set to cheapest remaining token. */
117833
117834	/* Set pTC to point to the cheapest remaining token. */
117835	for(iTC=0; iTC<nTC; iTC++){
117836	if( aTC[iTC].pToken && aTC[iTC].pRoot==pRoot
117837	&& (!pTC \|\| aTC[iTC].nOvfl<pTC->nOvfl)
117838	){
117839	pTC = &aTC[iTC];
117840	}
117841	}
117842	assert( pTC );
117843
117844	if( ii && pTC->nOvfl>=((nMinEst+(nLoad4/4)-1)/(nLoad4/4))*nDocSize ){
117845	/* The number of overflow pages to load for this (and therefore all
117846	** subsequent) tokens is greater than the estimated number of pages
117847	** that will be loaded if all subsequent tokens are deferred.
117848	*/
117849	Fts3PhraseToken *pToken = pTC->pToken;
117850	rc = sqlite3Fts3DeferToken(pCsr, pToken, pTC->iCol);
117851	fts3SegReaderCursorFree(pToken->pSegcsr);
117852	pToken->pSegcsr = 0;
117853	}else{
117854	nLoad4 = nLoad4*4;
117855	if( ii==0 \|\| pTC->pPhrase->nToken>1 ){
117856	/* Either this is the cheapest token in the entire query, or it is
117857	** part of a multi-token phrase. Either way, the entire doclist will
117858	** (eventually) be loaded into memory. It may as well be now. */
117859	Fts3PhraseToken *pToken = pTC->pToken;
117860	int nList = 0;
117861	char *pList = 0;
117862	rc = fts3TermSelect(pTab, pToken, pTC->iCol, &nList, &pList);
117863	assert( rc==SQLITE_OK \|\| pList==0 );
117864	if( rc==SQLITE_OK ){
117865	int nCount;
117866	fts3EvalPhraseMergeToken(pTab, pTC->pPhrase, pTC->iToken,pList,nList);
117867	nCount = fts3DoclistCountDocids(
117868	pTC->pPhrase->doclist.aAll, pTC->pPhrase->doclist.nAll

117869	);
117870	if( ii==0 \|\| nCount<nMinEst ) nMinEst = nCount;
117871	}
117872	}








117873	}
117874	pTC->pToken = 0;
117875	}
117876
117877	return rc;
117878	}
117879
117880	/*
117881	** This function is called from within the xFilter method. It initializes
117882	** the full-text query currently stored in pCsr->pExpr. To iterate through
117883	** the results of a query, the caller does:
117884	**
117885	** fts3EvalStart(pCsr);
117886	** while( 1 ){
117887	** fts3EvalNext(pCsr);
117888	** if( pCsr->bEof ) break;
117889	** ... return row pCsr->iPrevId to the caller ...
117890	** }
117891	*/
117892	static int fts3EvalStart(Fts3Cursor *pCsr){
117893	Fts3Table pTab = (Fts3Table )pCsr->base.pVtab;
117894	int rc = SQLITE_OK;
117895	int nToken = 0;
117896	int nOr = 0;
117897
117898	/* Allocate a MultiSegReader for each token in the expression. */
117899	fts3EvalAllocateReaders(pCsr, pCsr->pExpr, &nToken, &nOr, &rc);
117900
117901	/* Determine which, if any, tokens in the expression should be deferred. */
117902	if( rc==SQLITE_OK && nToken>1 && pTab->bHasStat ){



















117903	Fts3TokenAndCost *aTC;
117904	Fts3Expr **apOr;
117905	aTC = (Fts3TokenAndCost *)sqlite3_malloc(
117906	sizeof(Fts3TokenAndCost) * nToken
117907	+ sizeof(Fts3Expr ) nOr * 2
	@@ -115635,11 +117913,11 @@
117913	}else{
117914	int ii;
117915	Fts3TokenAndCost *pTC = aTC;
117916	Fts3Expr **ppOr = apOr;
117917
117918	fts3EvalTokenCosts(pCsr, 0, pCsr->pExpr, &pTC, &ppOr, &rc);
117919	nToken = pTC-aTC;
117920	nOr = ppOr-apOr;
117921
117922	if( rc==SQLITE_OK ){
117923	rc = fts3EvalSelectDeferred(pCsr, 0, aTC, nToken);
	@@ -115650,25 +117928,50 @@
117928
117929	sqlite3_free(aTC);
117930	}
117931	}
117932
117933	fts3EvalStartReaders(pCsr, pCsr->pExpr, 1, &rc);
117934	return rc;
117935	}
117936
117937	/*
117938	** Invalidate the current position list for phrase pPhrase.
117939	*/
117940	static void fts3EvalInvalidatePoslist(Fts3Phrase *pPhrase){
117941	if( pPhrase->doclist.bFreeList ){
117942	sqlite3_free(pPhrase->doclist.pList);
117943	}
117944	pPhrase->doclist.pList = 0;
117945	pPhrase->doclist.nList = 0;
117946	pPhrase->doclist.bFreeList = 0;
117947	}
117948
117949	/*
117950	** This function is called to edit the position list associated with
117951	** the phrase object passed as the fifth argument according to a NEAR
117952	** condition. For example:
117953	**
117954	** abc NEAR/5 "def ghi"
117955	**
117956	** Parameter nNear is passed the NEAR distance of the expression (5 in
117957	** the example above). When this function is called, *paPoslist points to
117958	** the position list, and *pnToken is the number of phrase tokens in, the
117959	** phrase on the other side of the NEAR operator to pPhrase. For example,
117960	** if pPhrase refers to the "def ghi" phrase, then *paPoslist points to
117961	** the position list associated with phrase "abc".
117962	**
117963	** All positions in the pPhrase position list that are not sufficiently
117964	** close to a position in the *paPoslist position list are removed. If this
117965	** leaves 0 positions, zero is returned. Otherwise, non-zero.
117966	**
117967	** Before returning, *paPoslist is set to point to the position lsit
117968	** associated with pPhrase. And *pnToken is set to the number of tokens in
117969	** pPhrase.
117970	*/
117971	static int fts3EvalNearTrim(
117972	int nNear, /* NEAR distance. As in "NEAR/nNear". */
117973	char aTmp, / Temporary space to use */
117974	char *paPoslist, / IN/OUT: Position list */
117975	int pnToken, / IN/OUT: Tokens in phrase of paPoslist /
117976	Fts3Phrase pPhrase / The phrase object to trim the doclist of */
117977	){
	@@ -115696,10 +117999,176 @@
117999	}
118000
118001	return res;
118002	}
118003
118004	/*
118005	** This function is a no-op if *pRc is other than SQLITE_OK when it is called.
118006	** Otherwise, it advances the expression passed as the second argument to
118007	** point to the next matching row in the database. Expressions iterate through
118008	** matching rows in docid order. Ascending order if Fts3Cursor.bDesc is zero,
118009	** or descending if it is non-zero.
118010	**
118011	** If an error occurs, *pRc is set to an SQLite error code. Otherwise, if
118012	** successful, the following variables in pExpr are set:
118013	**
118014	** Fts3Expr.bEof (non-zero if EOF - there is no next row)
118015	** Fts3Expr.iDocid (valid if bEof==0. The docid of the next row)
118016	**
118017	** If the expression is of type FTSQUERY_PHRASE, and the expression is not
118018	** at EOF, then the following variables are populated with the position list
118019	** for the phrase for the visited row:
118020	**
118021	** FTs3Expr.pPhrase->doclist.nList (length of pList in bytes)
118022	** FTs3Expr.pPhrase->doclist.pList (pointer to position list)
118023	**
118024	** It says above that this function advances the expression to the next
118025	** matching row. This is usually true, but there are the following exceptions:
118026	**
118027	** 1. Deferred tokens are not taken into account. If a phrase consists
118028	** entirely of deferred tokens, it is assumed to match every row in
118029	** the db. In this case the position-list is not populated at all.
118030	**
118031	** Or, if a phrase contains one or more deferred tokens and one or
118032	** more non-deferred tokens, then the expression is advanced to the
118033	** next possible match, considering only non-deferred tokens. In other
118034	** words, if the phrase is "A B C", and "B" is deferred, the expression
118035	** is advanced to the next row that contains an instance of "A * C",
118036	** where "*" may match any single token. The position list in this case
118037	** is populated as for "A * C" before returning.
118038	**
118039	** 2. NEAR is treated as AND. If the expression is "x NEAR y", it is
118040	** advanced to point to the next row that matches "x AND y".
118041	**
118042	** See fts3EvalTestDeferredAndNear() for details on testing if a row is
118043	** really a match, taking into account deferred tokens and NEAR operators.
118044	*/
118045	static void fts3EvalNextRow(
118046	Fts3Cursor pCsr, / FTS Cursor handle */
118047	Fts3Expr pExpr, / Expr. to advance to next matching row */
118048	int pRc / IN/OUT: Error code */
118049	){
118050	if( *pRc==SQLITE_OK ){
118051	int bDescDoclist = pCsr->bDesc; /* Used by DOCID_CMP() macro */
118052	assert( pExpr->bEof==0 );
118053	pExpr->bStart = 1;
118054
118055	switch( pExpr->eType ){
118056	case FTSQUERY_NEAR:
118057	case FTSQUERY_AND: {
118058	Fts3Expr *pLeft = pExpr->pLeft;
118059	Fts3Expr *pRight = pExpr->pRight;
118060	assert( !pLeft->bDeferred \|\| !pRight->bDeferred );
118061
118062	if( pLeft->bDeferred ){
118063	/* LHS is entirely deferred. So we assume it matches every row.
118064	** Advance the RHS iterator to find the next row visited. */
118065	fts3EvalNextRow(pCsr, pRight, pRc);
118066	pExpr->iDocid = pRight->iDocid;
118067	pExpr->bEof = pRight->bEof;
118068	}else if( pRight->bDeferred ){
118069	/* RHS is entirely deferred. So we assume it matches every row.
118070	** Advance the LHS iterator to find the next row visited. */
118071	fts3EvalNextRow(pCsr, pLeft, pRc);
118072	pExpr->iDocid = pLeft->iDocid;
118073	pExpr->bEof = pLeft->bEof;
118074	}else{
118075	/* Neither the RHS or LHS are deferred. */
118076	fts3EvalNextRow(pCsr, pLeft, pRc);
118077	fts3EvalNextRow(pCsr, pRight, pRc);
118078	while( !pLeft->bEof && !pRight->bEof && *pRc==SQLITE_OK ){
118079	sqlite3_int64 iDiff = DOCID_CMP(pLeft->iDocid, pRight->iDocid);
118080	if( iDiff==0 ) break;
118081	if( iDiff<0 ){
118082	fts3EvalNextRow(pCsr, pLeft, pRc);
118083	}else{
118084	fts3EvalNextRow(pCsr, pRight, pRc);
118085	}
118086	}
118087	pExpr->iDocid = pLeft->iDocid;
118088	pExpr->bEof = (pLeft->bEof \|\| pRight->bEof);
118089	}
118090	break;
118091	}
118092
118093	case FTSQUERY_OR: {
118094	Fts3Expr *pLeft = pExpr->pLeft;
118095	Fts3Expr *pRight = pExpr->pRight;
118096	sqlite3_int64 iCmp = DOCID_CMP(pLeft->iDocid, pRight->iDocid);
118097
118098	assert( pLeft->bStart \|\| pLeft->iDocid==pRight->iDocid );
118099	assert( pRight->bStart \|\| pLeft->iDocid==pRight->iDocid );
118100
118101	if( pRight->bEof \|\| (pLeft->bEof==0 && iCmp<0) ){
118102	fts3EvalNextRow(pCsr, pLeft, pRc);
118103	}else if( pLeft->bEof \|\| (pRight->bEof==0 && iCmp>0) ){
118104	fts3EvalNextRow(pCsr, pRight, pRc);
118105	}else{
118106	fts3EvalNextRow(pCsr, pLeft, pRc);
118107	fts3EvalNextRow(pCsr, pRight, pRc);
118108	}
118109
118110	pExpr->bEof = (pLeft->bEof && pRight->bEof);
118111	iCmp = DOCID_CMP(pLeft->iDocid, pRight->iDocid);
118112	if( pRight->bEof \|\| (pLeft->bEof==0 && iCmp<0) ){
118113	pExpr->iDocid = pLeft->iDocid;
118114	}else{
118115	pExpr->iDocid = pRight->iDocid;
118116	}
118117
118118	break;
118119	}
118120
118121	case FTSQUERY_NOT: {
118122	Fts3Expr *pLeft = pExpr->pLeft;
118123	Fts3Expr *pRight = pExpr->pRight;
118124
118125	if( pRight->bStart==0 ){
118126	fts3EvalNextRow(pCsr, pRight, pRc);
118127	assert( *pRc!=SQLITE_OK \|\| pRight->bStart );
118128	}
118129
118130	fts3EvalNextRow(pCsr, pLeft, pRc);
118131	if( pLeft->bEof==0 ){
118132	while( !*pRc
118133	&& !pRight->bEof
118134	&& DOCID_CMP(pLeft->iDocid, pRight->iDocid)>0
118135	){
118136	fts3EvalNextRow(pCsr, pRight, pRc);
118137	}
118138	}
118139	pExpr->iDocid = pLeft->iDocid;
118140	pExpr->bEof = pLeft->bEof;
118141	break;
118142	}
118143
118144	default: {
118145	Fts3Phrase *pPhrase = pExpr->pPhrase;
118146	fts3EvalInvalidatePoslist(pPhrase);
118147	*pRc = fts3EvalPhraseNext(pCsr, pPhrase, &pExpr->bEof);
118148	pExpr->iDocid = pPhrase->doclist.iDocid;
118149	break;
118150	}
118151	}
118152	}
118153	}
118154
118155	/*
118156	** If *pRc is not SQLITE_OK, or if pExpr is not the root node of a NEAR
118157	** cluster, then this function returns 1 immediately.
118158	**
118159	** Otherwise, it checks if the current row really does match the NEAR
118160	** expression, using the data currently stored in the position lists
118161	** (Fts3Expr->pPhrase.doclist.pList/nList) for each phrase in the expression.
118162	**
118163	** If the current row is a match, the position list associated with each
118164	** phrase in the NEAR expression is edited in place to contain only those
118165	** phrase instances sufficiently close to their peers to satisfy all NEAR
118166	** constraints. In this case it returns 1. If the NEAR expression does not
118167	** match the current row, 0 is returned. The position lists may or may not
118168	** be edited if 0 is returned.
118169	*/
118170	static int fts3EvalNearTest(Fts3Expr pExpr, int pRc){
118171	int res = 1;
118172
118173	/* The following block runs if pExpr is the root of a NEAR query.
118174	** For example, the query:
	@@ -115717,11 +118186,11 @@
118186	** \| \|
118187	** "w" "x"
118188	**
118189	** The right-hand child of a NEAR node is always a phrase. The
118190	** left-hand child may be either a phrase or a NEAR node. There are
118191	** no exceptions to this - it's the way the parser in fts3_expr.c works.
118192	*/
118193	if( *pRc==SQLITE_OK
118194	&& pExpr->eType==FTSQUERY_NEAR
118195	&& pExpr->bEof==0
118196	&& (pExpr->pParent==0 \|\| pExpr->pParent->eType!=FTSQUERY_NEAR)
	@@ -115744,21 +118213,21 @@
118213	int nToken = p->pPhrase->nToken;
118214
118215	for(p=p->pParent;res && p && p->eType==FTSQUERY_NEAR; p=p->pParent){
118216	Fts3Phrase *pPhrase = p->pRight->pPhrase;
118217	int nNear = p->nNear;
118218	res = fts3EvalNearTrim(nNear, aTmp, &aPoslist, &nToken, pPhrase);
118219	}
118220
118221	aPoslist = pExpr->pRight->pPhrase->doclist.pList;
118222	nToken = pExpr->pRight->pPhrase->nToken;
118223	for(p=pExpr->pLeft; p && res; p=p->pLeft){
118224	int nNear = p->pParent->nNear;
118225	Fts3Phrase *pPhrase = (
118226	p->eType==FTSQUERY_NEAR ? p->pRight->pPhrase : p->pPhrase
118227	);
118228	res = fts3EvalNearTrim(nNear, aTmp, &aPoslist, &nToken, pPhrase);
118229	}
118230	}
118231
118232	sqlite3_free(aTmp);
118233	}
	@@ -115765,132 +118234,33 @@
118234
118235	return res;
118236	}
118237
118238	/*
118239	** This function is a helper function for fts3EvalTestDeferredAndNear().
118240	** Assuming no error occurs or has occurred, It returns non-zero if the
118241	** expression passed as the second argument matches the row that pCsr
118242	** currently points to, or zero if it does not.
118243	**
118244	** If *pRc is not SQLITE_OK when this function is called, it is a no-op.
118245	** If an error occurs during execution of this function, *pRc is set to
118246	** the appropriate SQLite error code. In this case the returned value is
118247	** undefined.
118248	*/
118249	static int fts3EvalTestExpr(
118250	Fts3Cursor pCsr, / FTS cursor handle */
118251	Fts3Expr pExpr, / Expr to test. May or may not be root. */
118252	int pRc / IN/OUT: Error code */
118253	){
118254	int bHit = 1; /* Return value */







































































































118255	if( *pRc==SQLITE_OK ){
118256	switch( pExpr->eType ){
118257	case FTSQUERY_NEAR:
118258	case FTSQUERY_AND:
118259	bHit = (
118260	fts3EvalTestExpr(pCsr, pExpr->pLeft, pRc)
118261	&& fts3EvalTestExpr(pCsr, pExpr->pRight, pRc)
118262	&& fts3EvalNearTest(pExpr, pRc)
118263	);
118264
118265	/* If the NEAR expression does not match any rows, zero the doclist for
118266	** all phrases involved in the NEAR. This is because the snippet(),
	@@ -115912,31 +118282,31 @@
118282	&& (pExpr->pParent==0 \|\| pExpr->pParent->eType!=FTSQUERY_NEAR)
118283	){
118284	Fts3Expr *p;
118285	for(p=pExpr; p->pPhrase==0; p=p->pLeft){
118286	if( p->pRight->iDocid==pCsr->iPrevId ){
118287	fts3EvalInvalidatePoslist(p->pRight->pPhrase);
118288	}
118289	}
118290	if( p->iDocid==pCsr->iPrevId ){
118291	fts3EvalInvalidatePoslist(p->pPhrase);
118292	}
118293	}
118294
118295	break;
118296
118297	case FTSQUERY_OR: {
118298	int bHit1 = fts3EvalTestExpr(pCsr, pExpr->pLeft, pRc);
118299	int bHit2 = fts3EvalTestExpr(pCsr, pExpr->pRight, pRc);
118300	bHit = bHit1 \|\| bHit2;
118301	break;
118302	}
118303
118304	case FTSQUERY_NOT:
118305	bHit = (
118306	fts3EvalTestExpr(pCsr, pExpr->pLeft, pRc)
118307	&& !fts3EvalTestExpr(pCsr, pExpr->pRight, pRc)
118308	);
118309	break;
118310
118311	default: {
118312	if( pCsr->pDeferred
	@@ -115943,11 +118313,11 @@
118313	&& (pExpr->iDocid==pCsr->iPrevId \|\| pExpr->bDeferred)
118314	){
118315	Fts3Phrase *pPhrase = pExpr->pPhrase;
118316	assert( pExpr->bDeferred \|\| pPhrase->doclist.bFreeList==0 );
118317	if( pExpr->bDeferred ){
118318	fts3EvalInvalidatePoslist(pPhrase);
118319	}
118320	*pRc = fts3EvalDeferredPhrase(pCsr, pPhrase);
118321	bHit = (pPhrase->doclist.pList!=0);
118322	pExpr->iDocid = pCsr->iPrevId;
118323	}else{
	@@ -115959,31 +118329,53 @@
118329	}
118330	return bHit;
118331	}
118332
118333	/*
118334	** This function is called as the second part of each xNext operation when
118335	** iterating through the results of a full-text query. At this point the
118336	** cursor points to a row that matches the query expression, with the
118337	** following caveats:
118338	**
118339	** * Up until this point, "NEAR" operators in the expression have been
118340	** treated as "AND".
118341	**
118342	** * Deferred tokens have not yet been considered.
118343	**
118344	** If *pRc is not SQLITE_OK when this function is called, it immediately
118345	** returns 0. Otherwise, it tests whether or not after considering NEAR
118346	** operators and deferred tokens the current row is still a match for the
118347	** expression. It returns 1 if both of the following are true:
118348	**
118349	** 1. *pRc is SQLITE_OK when this function returns, and
118350	**
118351	** 2. After scanning the current FTS table row for the deferred tokens,
118352	** it is determined that the row does not match the query.
118353	**
118354	** Or, if no error occurs and it seems the current row does match the FTS
118355	** query, return 0.
118356	*/
118357	static int fts3EvalTestDeferredAndNear(Fts3Cursor pCsr, int pRc){
118358	int rc = *pRc;
118359	int bMiss = 0;
118360	if( rc==SQLITE_OK ){
118361
118362	/* If there are one or more deferred tokens, load the current row into
118363	** memory and scan it to determine the position list for each deferred
118364	** token. Then, see if this row is really a match, considering deferred
118365	** tokens and NEAR operators (neither of which were taken into account
118366	** earlier, by fts3EvalNextRow()).
118367	*/
118368	if( pCsr->pDeferred ){
118369	rc = fts3CursorSeek(0, pCsr);
118370	if( rc==SQLITE_OK ){
118371	rc = sqlite3Fts3CacheDeferredDoclists(pCsr);
118372	}
118373	}
118374	bMiss = (0==fts3EvalTestExpr(pCsr, pCsr->pExpr, &rc));
118375
118376	/* Free the position-lists accumulated for each deferred token above. */
118377	sqlite3Fts3FreeDeferredDoclists(pCsr);
118378	*pRc = rc;
118379	}
118380	return (rc==SQLITE_OK && bMiss);
118381	}
	@@ -115990,11 +118382,11 @@
118382
118383	/*
118384	** Advance to the next document that matches the FTS expression in
118385	** Fts3Cursor.pExpr.
118386	*/
118387	static int fts3EvalNext(Fts3Cursor *pCsr){
118388	int rc = SQLITE_OK; /* Return Code */
118389	Fts3Expr *pExpr = pCsr->pExpr;
118390	assert( pCsr->isEof==0 );
118391	if( pExpr==0 ){
118392	pCsr->isEof = 1;
	@@ -116002,23 +118394,23 @@
118394	do {
118395	if( pCsr->isRequireSeek==0 ){
118396	sqlite3_reset(pCsr->pStmt);
118397	}
118398	assert( sqlite3_data_count(pCsr->pStmt)==0 );
118399	fts3EvalNextRow(pCsr, pExpr, &rc);
118400	pCsr->isEof = pExpr->bEof;
118401	pCsr->isRequireSeek = 1;
118402	pCsr->isMatchinfoNeeded = 1;
118403	pCsr->iPrevId = pExpr->iDocid;
118404	}while( pCsr->isEof==0 && fts3EvalTestDeferredAndNear(pCsr, &rc) );
118405	}
118406	return rc;
118407	}
118408
118409	/*
118410	** Restart interation for expression pExpr so that the next call to
118411	** fts3EvalNext() visits the first row. Do not allow incremental
118412	** loading or merging of phrase doclists for this iteration.
118413	**
118414	** If *pRc is other than SQLITE_OK when this function is called, it is
118415	** a no-op. If an error occurs within this function, *pRc is set to an
118416	** SQLite error code before returning.
	@@ -116030,19 +118422,20 @@
118422	){
118423	if( pExpr && *pRc==SQLITE_OK ){
118424	Fts3Phrase *pPhrase = pExpr->pPhrase;
118425
118426	if( pPhrase ){
118427	fts3EvalInvalidatePoslist(pPhrase);
118428	if( pPhrase->bIncr ){
118429	assert( pPhrase->nToken==1 );
118430	assert( pPhrase->aToken[0].pSegcsr );
118431	sqlite3Fts3MsrIncrRestart(pPhrase->aToken[0].pSegcsr);
118432	*pRc = fts3EvalPhraseStart(pCsr, 0, pPhrase);


118433	}
118434
118435	pPhrase->doclist.pNextDocid = 0;
118436	pPhrase->doclist.iDocid = 0;
118437	}
118438
118439	pExpr->iDocid = 0;
118440	pExpr->bEof = 0;
118441	pExpr->bStart = 0;
	@@ -116145,18 +118538,18 @@
118538	/* Ensure the %_content statement is reset. */
118539	if( pCsr->isRequireSeek==0 ) sqlite3_reset(pCsr->pStmt);
118540	assert( sqlite3_data_count(pCsr->pStmt)==0 );
118541
118542	/* Advance to the next document */
118543	fts3EvalNextRow(pCsr, pRoot, &rc);
118544	pCsr->isEof = pRoot->bEof;
118545	pCsr->isRequireSeek = 1;
118546	pCsr->isMatchinfoNeeded = 1;
118547	pCsr->iPrevId = pRoot->iDocid;
118548	}while( pCsr->isEof==0
118549	&& pRoot->eType==FTSQUERY_NEAR
118550	&& fts3EvalTestDeferredAndNear(pCsr, &rc)
118551	);
118552
118553	if( rc==SQLITE_OK && pCsr->isEof==0 ){
118554	fts3EvalUpdateCounts(pRoot);
118555	}
	@@ -116174,14 +118567,14 @@
118567	**
118568	** do {...} while( pRoot->iDocid<iDocid && rc==SQLITE_OK );
118569	*/
118570	fts3EvalRestart(pCsr, pRoot, &rc);
118571	do {
118572	fts3EvalNextRow(pCsr, pRoot, &rc);
118573	assert( pRoot->bEof==0 );
118574	}while( pRoot->iDocid!=iDocid && rc==SQLITE_OK );
118575	fts3EvalTestDeferredAndNear(pCsr, &rc);
118576	}
118577	}
118578	return rc;
118579	}
118580
	@@ -116225,12 +118618,12 @@
118618	int iCol;
118619
118620	if( pExpr->bDeferred && pExpr->pParent->eType!=FTSQUERY_NEAR ){
118621	assert( pCsr->nDoc>0 );
118622	for(iCol=0; iCol<pTab->nColumn; iCol++){
118623	aiOut[iCol*3 + 1] = (u32)pCsr->nDoc;
118624	aiOut[iCol*3 + 2] = (u32)pCsr->nDoc;
118625	}
118626	}else{
118627	rc = fts3EvalGatherStats(pCsr, pExpr);
118628	if( rc==SQLITE_OK ){
118629	assert( pExpr->aMI );
	@@ -116308,18 +118701,32 @@
118701	*/
118702	SQLITE_PRIVATE void sqlite3Fts3EvalPhraseCleanup(Fts3Phrase *pPhrase){
118703	if( pPhrase ){
118704	int i;
118705	sqlite3_free(pPhrase->doclist.aAll);
118706	fts3EvalInvalidatePoslist(pPhrase);
118707	memset(&pPhrase->doclist, 0, sizeof(Fts3Doclist));
118708	for(i=0; i<pPhrase->nToken; i++){
118709	fts3SegReaderCursorFree(pPhrase->aToken[i].pSegcsr);
118710	pPhrase->aToken[i].pSegcsr = 0;
118711	}
118712	}
118713	}
118714
118715	#if !SQLITE_CORE
118716	/*
118717	** Initialize API pointer table, if required.
118718	*/
118719	SQLITE_API int sqlite3_extension_init(
118720	sqlite3 *db,
118721	char **pzErrMsg,
118722	const sqlite3_api_routines *pApi
118723	){
118724	SQLITE_EXTENSION_INIT2(pApi)
118725	return sqlite3Fts3Init(db);
118726	}
118727	#endif
118728
118729	#endif
118730
118731	/************ End of fts3.c **********************************************/
118732	/************ Begin file fts3_aux.c **************************************/
	@@ -116334,13 +118741,14 @@
118741	** May you share freely, never taking more than you give.
118742	**
118743	******************************************************************************
118744	**
118745	*/

118746	#if !defined(SQLITE_CORE) \|\| defined(SQLITE_ENABLE_FTS3)
118747
118748	/* #include <string.h> */
118749	/* #include <assert.h> */
118750
118751	typedef struct Fts3auxTable Fts3auxTable;
118752	typedef struct Fts3auxCursor Fts3auxCursor;
118753
118754	struct Fts3auxTable {
	@@ -116875,10 +119283,12 @@
119283	/*
119284	** Default span for NEAR operators.
119285	*/
119286	#define SQLITE_FTS3_DEFAULT_NEAR_PARAM 10
119287
119288	/* #include <string.h> */
119289	/* #include <assert.h> */
119290
119291	/*
119292	** isNot:
119293	** This variable is used by function getNextNode(). When getNextNode() is
119294	** called, it sets ParseContext.isNot to true if the 'next node' is a
	@@ -117576,10 +119986,11 @@
119986	** Everything after this point is just test code.
119987	*/
119988
119989	#ifdef SQLITE_TEST
119990
119991	/* #include <stdio.h> */
119992
119993	/*
119994	** Function to query the hash-table of tokenizers (see README.tokenizers).
119995	*/
119996	static int queryTestTokenizer(
	@@ -117786,10 +120197,13 @@
120197	** * The FTS3 module is being built into the core of
120198	** SQLite (in which case SQLITE_ENABLE_FTS3 is defined).
120199	*/
120200	#if !defined(SQLITE_CORE) \|\| defined(SQLITE_ENABLE_FTS3)
120201
120202	/* #include <assert.h> */
120203	/* #include <stdlib.h> */
120204	/* #include <string.h> */
120205
120206
120207	/*
120208	** Malloc and Free functions
120209	*/
	@@ -118166,11 +120580,14 @@
120580	** * The FTS3 module is being built into the core of
120581	** SQLite (in which case SQLITE_ENABLE_FTS3 is defined).
120582	*/
120583	#if !defined(SQLITE_CORE) \|\| defined(SQLITE_ENABLE_FTS3)
120584
120585	/* #include <assert.h> */
120586	/* #include <stdlib.h> */
120587	/* #include <stdio.h> */
120588	/* #include <string.h> */
120589
120590
120591	/*
120592	** Class derived from sqlite3_tokenizer
120593	*/
	@@ -118810,14 +121227,12 @@
121227	** * The FTS3 module is being built into the core of
121228	** SQLite (in which case SQLITE_ENABLE_FTS3 is defined).
121229	*/
121230	#if !defined(SQLITE_CORE) \|\| defined(SQLITE_ENABLE_FTS3)
121231
121232	/* #include <assert.h> */
121233	/* #include <string.h> */


121234
121235	/*
121236	** Implementation of the SQL scalar function for accessing the underlying
121237	** hash table. This function may be called as follows:
121238	**
	@@ -118937,11 +121352,11 @@
121352	sqlite3_tokenizer *ppTok, / OUT: Tokenizer (if applicable) */
121353	char *pzErr / OUT: Set to malloced error message */
121354	){
121355	int rc;
121356	char z = (char )zArg;
121357	int n = 0;
121358	char *zCopy;
121359	char zEnd; / Pointer to nul-term of zCopy */
121360	sqlite3_tokenizer_module *m;
121361
121362	zCopy = sqlite3_mprintf("%s", zArg);
	@@ -118989,10 +121404,12 @@
121404	}
121405
121406
121407	#ifdef SQLITE_TEST
121408
121409	/* #include <tcl.h> */
121410	/* #include <string.h> */
121411
121412	/*
121413	** Implementation of a special SQL scalar function for testing tokenizers
121414	** designed to be used in concert with the Tcl testing framework. This
121415	** function must be called with two arguments:
	@@ -119300,11 +121717,14 @@
121717	** * The FTS3 module is being built into the core of
121718	** SQLite (in which case SQLITE_ENABLE_FTS3 is defined).
121719	*/
121720	#if !defined(SQLITE_CORE) \|\| defined(SQLITE_ENABLE_FTS3)
121721
121722	/* #include <assert.h> */
121723	/* #include <stdlib.h> */
121724	/* #include <stdio.h> */
121725	/* #include <string.h> */
121726
121727
121728	typedef struct simple_tokenizer {
121729	sqlite3_tokenizer base;
121730	char delim[128]; /* flag ASCII delimiters */
	@@ -119526,10 +121946,13 @@
121946	** code in fts3.c.
121947	*/
121948
121949	#if !defined(SQLITE_CORE) \|\| defined(SQLITE_ENABLE_FTS3)
121950
121951	/* #include <string.h> */
121952	/* #include <assert.h> */
121953	/* #include <stdlib.h> */
121954
121955	/*
121956	** When full-text index nodes are loaded from disk, the buffer that they
121957	** are loaded into has the following number of bytes of padding at the end
121958	** of it. i.e. if a full-text index node is 900 bytes in size, then a buffer
	@@ -120750,10 +123173,11 @@
123173	pReader->pOffsetList = 0;
123174	}else{
123175	pReader->pOffsetList = p;
123176	}
123177	}else{
123178	char *pEnd = &pReader->aDoclist[pReader->nDoclist];
123179
123180	/* Pointer p currently points at the first byte of an offset list. The
123181	** following block advances it to point one byte past the end of
123182	** the same offset list. */
123183	while( 1 ){
	@@ -120778,17 +123202,19 @@
123202	*/
123203	if( ppOffsetList ){
123204	*ppOffsetList = pReader->pOffsetList;
123205	*pnOffsetList = (int)(p - pReader->pOffsetList - 1);
123206	}
123207
123208	while( p<pEnd && *p==0 ) p++;
123209
123210	/* If there are no more entries in the doclist, set pOffsetList to
123211	** NULL. Otherwise, set Fts3SegReader.iDocid to the next docid and
123212	** Fts3SegReader.pOffsetList to point to the next offset list before
123213	** returning.
123214	*/
123215	if( p>=pEnd ){
123216	pReader->pOffsetList = 0;
123217	}else{
123218	rc = fts3SegReaderRequire(pReader, p, FTS3_VARINT_MAX);
123219	if( rc==SQLITE_OK ){
123220	sqlite3_int64 iDelta;
	@@ -120823,11 +123249,11 @@
123249	for(ii=0; rc==SQLITE_OK && ii<pMsr->nSegment; ii++){
123250	Fts3SegReader *pReader = pMsr->apSegment[ii];
123251	if( !fts3SegReaderIsPending(pReader)
123252	&& !fts3SegReaderIsRootOnly(pReader)
123253	){
123254	sqlite3_int64 jj;
123255	for(jj=pReader->iStartBlock; jj<=pReader->iLeafEndBlock; jj++){
123256	int nBlob;
123257	rc = sqlite3Fts3ReadBlock(p, jj, 0, &nBlob, 0);
123258	if( rc!=SQLITE_OK ) break;
123259	if( (nBlob+35)>pgsz ){
	@@ -121768,55 +124194,31 @@
124194
124195	*ppList = pList;
124196	*pnList = nList;
124197	}
124198
124199	/*
124200	** Cache data in the Fts3MultiSegReader.aBuffer[] buffer (overwriting any
124201	** existing data). Grow the buffer if required.
124202	**
124203	** If successful, return SQLITE_OK. Otherwise, if an OOM error is encountered
124204	** trying to resize the buffer, return SQLITE_NOMEM.
124205	*/
124206	static int fts3MsrBufferData(
124207	Fts3MultiSegReader pMsr, / Multi-segment-reader handle */
124208	char *pList,
124209	int nList
124210	){
124211	if( nList>pMsr->nBuffer ){
124212	char *pNew;
124213	pMsr->nBuffer = nList*2;
124214	pNew = (char *)sqlite3_realloc(pMsr->aBuffer, pMsr->nBuffer);
124215	if( !pNew ) return SQLITE_NOMEM;
124216	pMsr->aBuffer = pNew;
124217	}
124218
124219	memcpy(pMsr->aBuffer, pList, nList);





























124220	return SQLITE_OK;
124221	}
124222
124223	SQLITE_PRIVATE int sqlite3Fts3MsrIncrNext(
124224	Fts3Table p, / Virtual table handle */
	@@ -121866,52 +124268,138 @@
124268	if( pMsr->iColFilter>=0 ){
124269	fts3ColumnFilter(pMsr->iColFilter, &pList, &nList);
124270	}
124271
124272	if( nList>0 ){
124273	if( fts3SegReaderIsPending(apSegment[0]) ){
124274	rc = fts3MsrBufferData(pMsr, pList, nList+1);
124275	if( rc!=SQLITE_OK ) return rc;
124276	*paPoslist = pMsr->aBuffer;
124277	assert( (pMsr->aBuffer[nList] & 0xFE)==0x00 );
124278	}else{
124279	*paPoslist = pList;
124280	}
124281	*piDocid = iDocid;

124282	*pnPoslist = nList;
124283	break;
124284	}
124285	}

124286	}
124287
124288	return SQLITE_OK;
124289	}
124290
124291	static int fts3SegReaderStart(
124292	Fts3Table p, / Virtual table handle */
124293	Fts3MultiSegReader pCsr, / Cursor object */
124294	const char zTerm, / Term searched for (or NULL) */
124295	int nTerm /* Length of zTerm in bytes */
124296	){
124297	int i;
124298	int nSeg = pCsr->nSegment;
124299
124300	/* If the Fts3SegFilter defines a specific term (or term prefix) to search
124301	** for, then advance each segment iterator until it points to a term of
124302	** equal or greater value than the specified term. This prevents many
124303	** unnecessary merge/sort operations for the case where single segment
124304	** b-tree leaf nodes contain more than one term.
124305	*/
124306	for(i=0; pCsr->bRestart==0 && i<pCsr->nSegment; i++){
124307	Fts3SegReader *pSeg = pCsr->apSegment[i];
124308	do {
124309	int rc = fts3SegReaderNext(p, pSeg, 0);
124310	if( rc!=SQLITE_OK ) return rc;
124311	}while( zTerm && fts3SegReaderTermCmp(pSeg, zTerm, nTerm)<0 );
124312	}
124313	fts3SegReaderSort(pCsr->apSegment, nSeg, nSeg, fts3SegReaderCmp);
124314
124315	return SQLITE_OK;
124316	}
124317
124318	SQLITE_PRIVATE int sqlite3Fts3SegReaderStart(
124319	Fts3Table p, / Virtual table handle */
124320	Fts3MultiSegReader pCsr, / Cursor object */
124321	Fts3SegFilter pFilter / Restrictions on range of iteration */
124322	){



124323	pCsr->pFilter = pFilter;
124324	return fts3SegReaderStart(p, pCsr, pFilter->zTerm, pFilter->nTerm);
124325	}
124326
124327	SQLITE_PRIVATE int sqlite3Fts3MsrIncrStart(
124328	Fts3Table p, / Virtual table handle */
124329	Fts3MultiSegReader pCsr, / Cursor object */
124330	int iCol, /* Column to match on. */
124331	const char zTerm, / Term to iterate through a doclist for */
124332	int nTerm /* Number of bytes in zTerm */
124333	){
124334	int i;
124335	int rc;
124336	int nSegment = pCsr->nSegment;
124337	int (xCmp)(Fts3SegReader , Fts3SegReader *) = (
124338	p->bDescIdx ? fts3SegReaderDoclistCmpRev : fts3SegReaderDoclistCmp
124339	);
124340
124341	assert( pCsr->pFilter==0 );
124342	assert( zTerm && nTerm>0 );
124343
124344	/* Advance each segment iterator until it points to the term zTerm/nTerm. */
124345	rc = fts3SegReaderStart(p, pCsr, zTerm, nTerm);
124346	if( rc!=SQLITE_OK ) return rc;
124347
124348	/* Determine how many of the segments actually point to zTerm/nTerm. */
124349	for(i=0; i<nSegment; i++){
124350	Fts3SegReader *pSeg = pCsr->apSegment[i];
124351	if( !pSeg->aNode \|\| fts3SegReaderTermCmp(pSeg, zTerm, nTerm) ){
124352	break;
124353	}

124354	}
124355	pCsr->nAdvance = i;
124356
124357	/* Advance each of the segments to point to the first docid. */
124358	for(i=0; i<pCsr->nAdvance; i++){
124359	rc = fts3SegReaderFirstDocid(p, pCsr->apSegment[i]);
124360	if( rc!=SQLITE_OK ) return rc;
124361	}
124362	fts3SegReaderSort(pCsr->apSegment, i, i, xCmp);
124363
124364	assert( iCol<0 \|\| iCol<p->nColumn );
124365	pCsr->iColFilter = iCol;
124366
124367	return SQLITE_OK;
124368	}
124369
124370	/*
124371	** This function is called on a MultiSegReader that has been started using
124372	** sqlite3Fts3MsrIncrStart(). One or more calls to MsrIncrNext() may also
124373	** have been made. Calling this function puts the MultiSegReader in such
124374	** a state that if the next two calls are:
124375	**
124376	** sqlite3Fts3SegReaderStart()
124377	** sqlite3Fts3SegReaderStep()
124378	**
124379	** then the entire doclist for the term is available in
124380	** MultiSegReader.aDoclist/nDoclist.
124381	*/
124382	SQLITE_PRIVATE int sqlite3Fts3MsrIncrRestart(Fts3MultiSegReader *pCsr){
124383	int i; /* Used to iterate through segment-readers */
124384
124385	assert( pCsr->zTerm==0 );
124386	assert( pCsr->nTerm==0 );
124387	assert( pCsr->aDoclist==0 );
124388	assert( pCsr->nDoclist==0 );
124389
124390	pCsr->nAdvance = 0;
124391	pCsr->bRestart = 1;
124392	for(i=0; i<pCsr->nSegment; i++){
124393	pCsr->apSegment[i]->pOffsetList = 0;
124394	pCsr->apSegment[i]->nOffsetList = 0;
124395	pCsr->apSegment[i]->iDocid = 0;
124396	}
124397
124398	return SQLITE_OK;
124399	}
124400
124401
124402	SQLITE_PRIVATE int sqlite3Fts3SegReaderStep(
124403	Fts3Table p, / Virtual table handle */
124404	Fts3MultiSegReader pCsr / Cursor object */
124405	){
	@@ -121981,13 +124469,18 @@
124469	assert( isIgnoreEmpty \|\| (isRequirePos && !isColFilter) );
124470	if( nMerge==1
124471	&& !isIgnoreEmpty
124472	&& (p->bDescIdx==0 \|\| fts3SegReaderIsPending(apSegment[0])==0)
124473	){

124474	pCsr->nDoclist = apSegment[0]->nDoclist;
124475	if( fts3SegReaderIsPending(apSegment[0]) ){
124476	rc = fts3MsrBufferData(pCsr, apSegment[0]->aDoclist, pCsr->nDoclist);
124477	pCsr->aDoclist = pCsr->aBuffer;
124478	}else{
124479	pCsr->aDoclist = apSegment[0]->aDoclist;
124480	}
124481	if( rc==SQLITE_OK ) rc = SQLITE_ROW;
124482	}else{
124483	int nDoclist = 0; /* Size of doclist */
124484	sqlite3_int64 iPrev = 0; /* Previous docid stored in doclist */
124485
124486	/* The current term of the first nMerge entries in the array
	@@ -122717,10 +125210,12 @@
125210	******************************************************************************
125211	*/
125212
125213	#if !defined(SQLITE_CORE) \|\| defined(SQLITE_ENABLE_FTS3)
125214
125215	/* #include <string.h> */
125216	/* #include <assert.h> */
125217
125218	/*
125219	** Characters that may appear in the second argument to matchinfo().
125220	*/
125221	#define FTS3_MATCHINFO_NPHRASE 'p' /* 1 value */
	@@ -123725,11 +126220,11 @@
126220	if( bGlobal ) pInfo->aMatchinfo[0] = pInfo->nCol;
126221	break;
126222
126223	case FTS3_MATCHINFO_NDOC:
126224	if( bGlobal ){
126225	sqlite3_int64 nDoc = 0;
126226	rc = fts3MatchinfoSelectDoctotal(pTab, &pSelect, &nDoc, 0);
126227	pInfo->aMatchinfo[0] = (u32)nDoc;
126228	}
126229	break;
126230
	@@ -124304,10 +126799,12 @@
126799	#ifndef SQLITE_CORE
126800	SQLITE_EXTENSION_INIT1
126801	#else
126802	#endif
126803
126804	/* #include <string.h> */
126805	/* #include <assert.h> */
126806
126807	#ifndef SQLITE_AMALGAMATION
126808	#include "sqlite3rtree.h"
126809	typedef sqlite3_int64 i64;
126810	typedef unsigned char u8;
	@@ -125620,11 +128117,11 @@
128117	*/
128118	static float cellArea(Rtree pRtree, RtreeCell p){
128119	float area = 1.0;
128120	int ii;
128121	for(ii=0; ii<(pRtree->nDim*2); ii+=2){
128122	area = (float)(area * (DCOORD(p->aCoord[ii+1]) - DCOORD(p->aCoord[ii])));
128123	}
128124	return area;
128125	}
128126
128127	/*
	@@ -125633,11 +128130,11 @@
128130	*/
128131	static float cellMargin(Rtree pRtree, RtreeCell p){
128132	float margin = 0.0;
128133	int ii;
128134	for(ii=0; ii<(pRtree->nDim*2); ii+=2){
128135	margin += (float)(DCOORD(p->aCoord[ii+1]) - DCOORD(p->aCoord[ii]));
128136	}
128137	return margin;
128138	}
128139
128140	/*
	@@ -125718,11 +128215,11 @@
128215
128216	if( x2<x1 ){
128217	o = 0.0;
128218	break;
128219	}else{
128220	o = o * (float)(x2-x1);
128221	}
128222	}
128223	overlap += o;
128224	}
128225	}
	@@ -125737,16 +128234,16 @@
128234	RtreeCell *pInsert,
128235	RtreeCell *aCell,
128236	int nCell,
128237	int iExclude
128238	){
128239	double before;
128240	double after;
128241	before = cellOverlap(pRtree, p, aCell, nCell, iExclude);
128242	cellUnion(pRtree, p, pInsert);
128243	after = cellOverlap(pRtree, p, aCell, nCell, iExclude);
128244	return (float)(after-before);
128245	}
128246	#endif
128247
128248
128249	/*
	@@ -125764,15 +128261,15 @@
128261	RtreeNode *pNode;
128262	rc = nodeAcquire(pRtree, 1, 0, &pNode);
128263
128264	for(ii=0; rc==SQLITE_OK && ii<(pRtree->iDepth-iHeight); ii++){
128265	int iCell;
128266	sqlite3_int64 iBest = 0;
128267
128268	float fMinGrowth = 0.0;
128269	float fMinArea = 0.0;
128270	float fMinOverlap = 0.0;
128271
128272	int nCell = NCELL(pNode);
128273	RtreeCell cell;
128274	RtreeNode *pChild;
128275
	@@ -126198,13 +128695,13 @@
128695	){
128696	int **aaSorted;
128697	int *aSpare;
128698	int ii;
128699
128700	int iBestDim = 0;
128701	int iBestSplit = 0;
128702	float fBestMargin = 0.0;
128703
128704	int nByte = (pRtree->nDim+1)(sizeof(int)+nCell*sizeof(int));
128705
128706	aaSorted = (int **)sqlite3_malloc(nByte);
128707	if( !aaSorted ){
	@@ -126222,13 +128719,13 @@
128719	SortByDimension(pRtree, aaSorted[ii], nCell, ii, aCell, aSpare);
128720	}
128721
128722	for(ii=0; ii<pRtree->nDim; ii++){
128723	float margin = 0.0;
128724	float fBestOverlap = 0.0;
128725	float fBestArea = 0.0;
128726	int iBestLeft = 0;
128727	int nLeft;
128728
128729	for(
128730	nLeft=RTREE_MINCELLS(pRtree);
128731	nLeft<=(nCell-RTREE_MINCELLS(pRtree));
	@@ -126539,11 +129036,11 @@
129036	static int deleteCell(Rtree , RtreeNode , int, int);
129037
129038	static int removeNode(Rtree pRtree, RtreeNode pNode, int iHeight){
129039	int rc;
129040	int rc2;
129041	RtreeNode *pParent = 0;
129042	int iCell;
129043
129044	assert( pNode->nRef==1 );
129045
129046	/* Remove the entry in the parent cell. */
	@@ -126687,23 +129184,23 @@
129184	}else{
129185	nodeGetCell(pRtree, pNode, ii, &aCell[ii]);
129186	}
129187	aOrder[ii] = ii;
129188	for(iDim=0; iDim<pRtree->nDim; iDim++){
129189	aCenterCoord[iDim] += (float)DCOORD(aCell[ii].aCoord[iDim*2]);
129190	aCenterCoord[iDim] += (float)DCOORD(aCell[ii].aCoord[iDim*2+1]);
129191	}
129192	}
129193	for(iDim=0; iDim<pRtree->nDim; iDim++){
129194	aCenterCoord[iDim] = (float)(aCenterCoord[iDim]/((float)nCell*2.0));
129195	}
129196
129197	for(ii=0; ii<nCell; ii++){
129198	aDistance[ii] = 0.0;
129199	for(iDim=0; iDim<pRtree->nDim; iDim++){
129200	float coord = (float)(DCOORD(aCell[ii].aCoord[iDim*2+1]) -
129201	DCOORD(aCell[ii].aCoord[iDim*2]));
129202	aDistance[ii] += (coord-aCenterCoord[iDim])*(coord-aCenterCoord[iDim]);
129203	}
129204	}
129205
129206	SortByDistance(aOrder, nCell, aDistance, aSpare);
	@@ -126798,14 +129295,14 @@
129295	nodeGetCell(pRtree, pNode, ii, &cell);
129296
129297	/* Find a node to store this cell in. pNode->iNode currently contains
129298	** the height of the sub-tree headed by the cell.
129299	*/
129300	rc = ChooseLeaf(pRtree, &cell, (int)pNode->iNode, &pInsert);
129301	if( rc==SQLITE_OK ){
129302	int rc2;
129303	rc = rtreeInsertCell(pRtree, pInsert, &cell, (int)pNode->iNode);
129304	rc2 = nodeRelease(pRtree, pInsert);
129305	if( rc==SQLITE_OK ){
129306	rc = rc2;
129307	}
129308	}
	@@ -127190,11 +129687,11 @@
129687	int isCreate /* True for xCreate, false for xConnect */
129688	){
129689	int rc;
129690	char *zSql;
129691	if( isCreate ){
129692	int iPageSize = 0;
129693	zSql = sqlite3_mprintf("PRAGMA %Q.page_size", pRtree->zDb);
129694	rc = getIntFromStmt(db, zSql, &iPageSize);
129695	if( rc==SQLITE_OK ){
129696	pRtree->iNodeSize = iPageSize-64;
129697	if( (4+pRtree->nBytesPerCell*RTREE_MAXCELLS)<pRtree->iNodeSize ){
	@@ -127518,10 +130015,11 @@
130015	#include <unicode/utypes.h>
130016	#include <unicode/uregex.h>
130017	#include <unicode/ustring.h>
130018	#include <unicode/ucol.h>
130019
130020	/* #include <assert.h> */
130021
130022	#ifndef SQLITE_CORE
130023	SQLITE_EXTENSION_INIT1
130024	#else
130025	#endif
	@@ -127993,19 +130491,20 @@
130491	** May you find forgiveness for yourself and forgive others.
130492	** May you share freely, never taking more than you give.
130493	**
130494	*************************************************************************
130495	** This file implements a tokenizer for fts3 based on the ICU library.


130496	*/

130497	#if !defined(SQLITE_CORE) \|\| defined(SQLITE_ENABLE_FTS3)
130498	#ifdef SQLITE_ENABLE_ICU
130499
130500	/* #include <assert.h> */
130501	/* #include <string.h> */
130502
130503	#include <unicode/ubrk.h>
130504	/* #include <unicode/ucol.h> */
130505	/* #include <unicode/ustring.h> */
130506	#include <unicode/utf16.h>
130507
130508	typedef struct IcuTokenizer IcuTokenizer;
130509	typedef struct IcuCursor IcuCursor;
130510
130511

M src/sqlite3.h

+50 -23

		--- src/sqlite3.h
		+++ src/sqlite3.h
		@@ -105,13 +105,13 @@
105	105	**
106	106	** See also: [sqlite3_libversion()],
107	107	** [sqlite3_libversion_number()], [sqlite3_sourceid()],
108	108	** [sqlite_version()] and [sqlite_source_id()].
109	109	*/
110		-#define SQLITE_VERSION "3.7.7"
111		-#define SQLITE_VERSION_NUMBER 3007007
112		-#define SQLITE_SOURCE_ID "2011-06-15 13:11:06 f9750870ee04935f338e4d808900fee5a8b2b389"
	110	+#define SQLITE_VERSION "3.7.8"
	111	+#define SQLITE_VERSION_NUMBER 3007008
	112	+#define SQLITE_SOURCE_ID "2011-08-29 11:56:14 639cc85a911454bffdcccb33f2976c683953ae64"
113	113
114	114	/*
115	115	** CAPI3REF: Run-Time Library Version Numbers
116	116	** KEYWORDS: sqlite3_version, sqlite3_sourceid
117	117	**
		@@ -308,11 +308,11 @@
308	308	** semicolon-separate SQL statements passed into its 2nd argument,
309	309	** in the context of the [database connection] passed in as its 1st
310	310	** argument. ^If the callback function of the 3rd argument to
311	311	** sqlite3_exec() is not NULL, then it is invoked for each result row
312	312	** coming out of the evaluated SQL statements. ^The 4th argument to
313		-** to sqlite3_exec() is relayed through to the 1st argument of each
	313	+** sqlite3_exec() is relayed through to the 1st argument of each
314	314	** callback invocation. ^If the callback pointer to sqlite3_exec()
315	315	** is NULL, then no callback is ever invoked and result rows are
316	316	** ignored.
317	317	**
318	318	** ^If an error occurs while evaluating the SQL statements passed into
		@@ -739,20 +739,52 @@
739	739	** when [PRAGMA synchronous \| PRAGMA synchronous=OFF] is set, but most
740	740	** VFSes do not need this signal and should silently ignore this opcode.
741	741	** Applications should not call [sqlite3_file_control()] with this
742	742	** opcode as doing so may disrupt the operation of the specialized VFSes
743	743	** that do require it.
	744	+**
	745	+** ^The [SQLITE_FCNTL_WIN32_AV_RETRY] opcode is used to configure automatic
	746	+** retry counts and intervals for certain disk I/O operations for the
	747	+** windows [VFS] in order to work to provide robustness against
	748	+** anti-virus programs. By default, the windows VFS will retry file read,
	749	+** file write, and file delete opertions up to 10 times, with a delay
	750	+** of 25 milliseconds before the first retry and with the delay increasing
	751	+** by an additional 25 milliseconds with each subsequent retry. This
	752	+** opcode allows those to values (10 retries and 25 milliseconds of delay)
	753	+** to be adjusted. The values are changed for all database connections
	754	+** within the same process. The argument is a pointer to an array of two
	755	+** integers where the first integer i the new retry count and the second
	756	+** integer is the delay. If either integer is negative, then the setting
	757	+** is not changed but instead the prior value of that setting is written
	758	+** into the array entry, allowing the current retry settings to be
	759	+** interrogated. The zDbName parameter is ignored.
	760	+**
	761	+** ^The [SQLITE_FCNTL_PERSIST_WAL] opcode is used to set or query the
	762	+** persistent [WAL \| Write AHead Log] setting. By default, the auxiliary
	763	+** write ahead log and shared memory files used for transaction control
	764	+** are automatically deleted when the latest connection to the database
	765	+** closes. Setting persistent WAL mode causes those files to persist after
	766	+** close. Persisting the files is useful when other processes that do not
	767	+** have write permission on the directory containing the database file want
	768	+** to read the database file, as the WAL and shared memory files must exist
	769	+** in order for the database to be readable. The fourth parameter to
	770	+** [sqlite3_file_control()] for this opcode should be a pointer to an integer.
	771	+** That integer is 0 to disable persistent WAL mode or 1 to enable persistent
	772	+** WAL mode. If the integer is -1, then it is overwritten with the current
	773	+** WAL persistence setting.
	774	+**
744	775	*/
745	776	#define SQLITE_FCNTL_LOCKSTATE 1
746	777	#define SQLITE_GET_LOCKPROXYFILE 2
747	778	#define SQLITE_SET_LOCKPROXYFILE 3
748	779	#define SQLITE_LAST_ERRNO 4
749	780	#define SQLITE_FCNTL_SIZE_HINT 5
750	781	#define SQLITE_FCNTL_CHUNK_SIZE 6
751	782	#define SQLITE_FCNTL_FILE_POINTER 7
752	783	#define SQLITE_FCNTL_SYNC_OMITTED 8
753		-
	784	+#define SQLITE_FCNTL_WIN32_AV_RETRY 9
	785	+#define SQLITE_FCNTL_PERSIST_WAL 10
754	786
755	787	/*
756	788	** CAPI3REF: Mutex Handle
757	789	**
758	790	** The mutex module within SQLite defines [sqlite3_mutex] to be an
		@@ -900,11 +932,11 @@
900	932	** The xSleep() method causes the calling thread to sleep for at
901	933	** least the number of microseconds given. ^The xCurrentTime()
902	934	** method returns a Julian Day Number for the current date and time as
903	935	** a floating point value.
904	936	** ^The xCurrentTimeInt64() method returns, as an integer, the Julian
905		-** Day Number multipled by 86400000 (the number of milliseconds in
	937	+** Day Number multiplied by 86400000 (the number of milliseconds in
906	938	** a 24-hour day).
907	939	** ^SQLite will use the xCurrentTimeInt64() method to get the current
908	940	** date and time if that method is available (if iVersion is 2 or
909	941	** greater and the function pointer is not NULL) and will fall back
910	942	** to xCurrentTime() if xCurrentTimeInt64() is unavailable.
		@@ -1176,20 +1208,14 @@
1176	1208	** also used during testing of SQLite in order to specify an alternative
1177	1209	** memory allocator that simulates memory out-of-memory conditions in
1178	1210	** order to verify that SQLite recovers gracefully from such
1179	1211	** conditions.
1180	1212	**
1181		-** The xMalloc and xFree methods must work like the
1182		-** malloc() and free() functions from the standard C library.
1183		-** The xRealloc method must work like realloc() from the standard C library
1184		-** with the exception that if the second argument to xRealloc is zero,
1185		-** xRealloc must be a no-op - it must not perform any allocation or
1186		-** deallocation. ^SQLite guarantees that the second argument to
	1213	+** The xMalloc, xRealloc, and xFree methods must work like the
	1214	+** malloc(), realloc() and free() functions from the standard C library.
	1215	+** ^SQLite guarantees that the second argument to
1187	1216	** xRealloc is always a value returned by a prior call to xRoundup.
1188		-** And so in cases where xRoundup always returns a positive number,
1189		-** xRealloc can perform exactly as the standard library realloc() and
1190		-** still be in compliance with this specification.
1191	1217	**
1192	1218	** xSize should return the allocated size of a memory allocation
1193	1219	** previously obtained from xMalloc or xRealloc. The allocated size
1194	1220	** is always at least as big as the requested size but may be larger.
1195	1221	**
		@@ -1338,11 +1364,11 @@
1338	1364	** scratch memory beyond what is provided by this configuration option, then
1339	1365	** [sqlite3_malloc()] will be used to obtain the memory needed.</dd>
1340	1366	**
1341	1367	** [[SQLITE_CONFIG_PAGECACHE]] <dt>SQLITE_CONFIG_PAGECACHE</dt>
1342	1368	** <dd> ^This option specifies a static memory buffer that SQLite can use for
1343		-** the database page cache with the default page cache implemenation.
	1369	+** the database page cache with the default page cache implementation.
1344	1370	** This configuration should not be used if an application-define page
1345	1371	** cache implementation is loaded using the SQLITE_CONFIG_PCACHE option.
1346	1372	** There are three arguments to this option: A pointer to 8-byte aligned
1347	1373	** memory, the size of each page buffer (sz), and the number of pages (N).
1348	1374	** The sz argument should be the size of the largest database page
		@@ -2436,16 +2462,16 @@
2436	2462	** [[URI filenames in sqlite3_open()]] <h3>URI Filenames</h3>
2437	2463	**
2438	2464	** ^If [URI filename] interpretation is enabled, and the filename argument
2439	2465	** begins with "file:", then the filename is interpreted as a URI. ^URI
2440	2466	** filename interpretation is enabled if the [SQLITE_OPEN_URI] flag is
2441		-** is set in the fourth argument to sqlite3_open_v2(), or if it has
	2467	+** set in the fourth argument to sqlite3_open_v2(), or if it has
2442	2468	** been enabled globally using the [SQLITE_CONFIG_URI] option with the
2443	2469	** [sqlite3_config()] method or by the [SQLITE_USE_URI] compile-time option.
2444	2470	** As of SQLite version 3.7.7, URI filename interpretation is turned off
2445	2471	** by default, but future releases of SQLite might enable URI filename
2446		-** intepretation by default. See "[URI filenames]" for additional
	2472	+** interpretation by default. See "[URI filenames]" for additional
2447	2473	** information.
2448	2474	**
2449	2475	** URI filenames are parsed according to RFC 3986. ^If the URI contains an
2450	2476	** authority, then it must be either an empty string or the string
2451	2477	** "localhost". ^If the authority is not an empty string or "localhost", an
		@@ -3260,11 +3286,11 @@
3260	3286	** [extended result codes] might be returned as well.
3261	3287	**
3262	3288	** ^[SQLITE_BUSY] means that the database engine was unable to acquire the
3263	3289	** database locks it needs to do its job. ^If the statement is a [COMMIT]
3264	3290	** or occurs outside of an explicit transaction, then you can retry the
3265		-** statement. If the statement is not a [COMMIT] and occurs within a
	3291	+** statement. If the statement is not a [COMMIT] and occurs within an
3266	3292	** explicit transaction then you should rollback the transaction before
3267	3293	** continuing.
3268	3294	**
3269	3295	** ^[SQLITE_DONE] means that the statement has finished executing
3270	3296	** successfully. sqlite3_step() should not be called again on this virtual
		@@ -3539,11 +3565,11 @@
3539	3565
3540	3566	/*
3541	3567	** CAPI3REF: Destroy A Prepared Statement Object
3542	3568	**
3543	3569	** ^The sqlite3_finalize() function is called to delete a [prepared statement].
3544		-** ^If the most recent evaluation of the statement encountered no errors or
	3570	+** ^If the most recent evaluation of the statement encountered no errors
3545	3571	** or if the statement is never been evaluated, then sqlite3_finalize() returns
3546	3572	** SQLITE_OK. ^If the most recent evaluation of statement S failed, then
3547	3573	** sqlite3_finalize(S) returns the appropriate [error code] or
3548	3574	** [extended error code].
3549	3575	**
		@@ -5453,11 +5479,11 @@
5453	5479	** versions of these routines, it should at least provide stubs that always
5454	5480	** return true so that one does not get spurious assertion failures.
5455	5481	**
5456	5482	** ^If the argument to sqlite3_mutex_held() is a NULL pointer then
5457	5483	** the routine should return 1. This seems counter-intuitive since
5458		-** clearly the mutex cannot be held if it does not exist. But the
	5484	+** clearly the mutex cannot be held if it does not exist. But
5459	5485	** the reason the mutex does not exist is because the build is not
5460	5486	** using mutexes. And we do not want the assert() containing the
5461	5487	** call to sqlite3_mutex_held() to fail, so a non-zero return is
5462	5488	** the appropriate thing to do. ^The sqlite3_mutex_notheld()
5463	5489	** interface should also return 1 when given a NULL pointer.
		@@ -5576,11 +5602,12 @@
5576	5602	#define SQLITE_TESTCTRL_RESERVE 14
5577	5603	#define SQLITE_TESTCTRL_OPTIMIZATIONS 15
5578	5604	#define SQLITE_TESTCTRL_ISKEYWORD 16
5579	5605	#define SQLITE_TESTCTRL_PGHDRSZ 17
5580	5606	#define SQLITE_TESTCTRL_SCRATCHMALLOC 18
5581		-#define SQLITE_TESTCTRL_LAST 18
	5607	+#define SQLITE_TESTCTRL_LOCALTIME_FAULT 19
	5608	+#define SQLITE_TESTCTRL_LAST 19
5582	5609
5583	5610	/*
5584	5611	** CAPI3REF: SQLite Runtime Status
5585	5612	**
5586	5613	** ^This interface is used to retrieve runtime status information
		@@ -5962,11 +5989,11 @@
5962	5989	** [[the xFetch() page cache methods]]
5963	5990	** The xFetch() method locates a page in the cache and returns a pointer to
5964	5991	** the page, or a NULL pointer.
5965	5992	** A "page", in this context, means a buffer of szPage bytes aligned at an
5966	5993	** 8-byte boundary. The page to be fetched is determined by the key. ^The
5967		-** mimimum key value is 1. After it has been retrieved using xFetch, the page
	5994	+** minimum key value is 1. After it has been retrieved using xFetch, the page
5968	5995	** is considered to be "pinned".
5969	5996	**
5970	5997	** If the requested page is already in the page cache, then the page cache
5971	5998	** implementation must return a pointer to the page buffer with its content
5972	5999	** intact. If the requested page is not already in the cache, then the
5973	6000

	--- src/sqlite3.h
	+++ src/sqlite3.h
	@@ -105,13 +105,13 @@
105	**
106	** See also: [sqlite3_libversion()],
107	** [sqlite3_libversion_number()], [sqlite3_sourceid()],
108	** [sqlite_version()] and [sqlite_source_id()].
109	*/
110	#define SQLITE_VERSION "3.7.7"
111	#define SQLITE_VERSION_NUMBER 3007007
112	#define SQLITE_SOURCE_ID "2011-06-15 13:11:06 f9750870ee04935f338e4d808900fee5a8b2b389"
113
114	/*
115	** CAPI3REF: Run-Time Library Version Numbers
116	** KEYWORDS: sqlite3_version, sqlite3_sourceid
117	**
	@@ -308,11 +308,11 @@
308	** semicolon-separate SQL statements passed into its 2nd argument,
309	** in the context of the [database connection] passed in as its 1st
310	** argument. ^If the callback function of the 3rd argument to
311	** sqlite3_exec() is not NULL, then it is invoked for each result row
312	** coming out of the evaluated SQL statements. ^The 4th argument to
313	** to sqlite3_exec() is relayed through to the 1st argument of each
314	** callback invocation. ^If the callback pointer to sqlite3_exec()
315	** is NULL, then no callback is ever invoked and result rows are
316	** ignored.
317	**
318	** ^If an error occurs while evaluating the SQL statements passed into
	@@ -739,20 +739,52 @@
739	** when [PRAGMA synchronous \| PRAGMA synchronous=OFF] is set, but most
740	** VFSes do not need this signal and should silently ignore this opcode.
741	** Applications should not call [sqlite3_file_control()] with this
742	** opcode as doing so may disrupt the operation of the specialized VFSes
743	** that do require it.































744	*/
745	#define SQLITE_FCNTL_LOCKSTATE 1
746	#define SQLITE_GET_LOCKPROXYFILE 2
747	#define SQLITE_SET_LOCKPROXYFILE 3
748	#define SQLITE_LAST_ERRNO 4
749	#define SQLITE_FCNTL_SIZE_HINT 5
750	#define SQLITE_FCNTL_CHUNK_SIZE 6
751	#define SQLITE_FCNTL_FILE_POINTER 7
752	#define SQLITE_FCNTL_SYNC_OMITTED 8
753

754
755	/*
756	** CAPI3REF: Mutex Handle
757	**
758	** The mutex module within SQLite defines [sqlite3_mutex] to be an
	@@ -900,11 +932,11 @@
900	** The xSleep() method causes the calling thread to sleep for at
901	** least the number of microseconds given. ^The xCurrentTime()
902	** method returns a Julian Day Number for the current date and time as
903	** a floating point value.
904	** ^The xCurrentTimeInt64() method returns, as an integer, the Julian
905	** Day Number multipled by 86400000 (the number of milliseconds in
906	** a 24-hour day).
907	** ^SQLite will use the xCurrentTimeInt64() method to get the current
908	** date and time if that method is available (if iVersion is 2 or
909	** greater and the function pointer is not NULL) and will fall back
910	** to xCurrentTime() if xCurrentTimeInt64() is unavailable.
	@@ -1176,20 +1208,14 @@
1176	** also used during testing of SQLite in order to specify an alternative
1177	** memory allocator that simulates memory out-of-memory conditions in
1178	** order to verify that SQLite recovers gracefully from such
1179	** conditions.
1180	**
1181	** The xMalloc and xFree methods must work like the
1182	** malloc() and free() functions from the standard C library.
1183	** The xRealloc method must work like realloc() from the standard C library
1184	** with the exception that if the second argument to xRealloc is zero,
1185	** xRealloc must be a no-op - it must not perform any allocation or
1186	** deallocation. ^SQLite guarantees that the second argument to
1187	** xRealloc is always a value returned by a prior call to xRoundup.
1188	** And so in cases where xRoundup always returns a positive number,
1189	** xRealloc can perform exactly as the standard library realloc() and
1190	** still be in compliance with this specification.
1191	**
1192	** xSize should return the allocated size of a memory allocation
1193	** previously obtained from xMalloc or xRealloc. The allocated size
1194	** is always at least as big as the requested size but may be larger.
1195	**
	@@ -1338,11 +1364,11 @@
1338	** scratch memory beyond what is provided by this configuration option, then
1339	** [sqlite3_malloc()] will be used to obtain the memory needed.</dd>
1340	**
1341	** [[SQLITE_CONFIG_PAGECACHE]] <dt>SQLITE_CONFIG_PAGECACHE</dt>
1342	** <dd> ^This option specifies a static memory buffer that SQLite can use for
1343	** the database page cache with the default page cache implemenation.
1344	** This configuration should not be used if an application-define page
1345	** cache implementation is loaded using the SQLITE_CONFIG_PCACHE option.
1346	** There are three arguments to this option: A pointer to 8-byte aligned
1347	** memory, the size of each page buffer (sz), and the number of pages (N).
1348	** The sz argument should be the size of the largest database page
	@@ -2436,16 +2462,16 @@
2436	** [[URI filenames in sqlite3_open()]] <h3>URI Filenames</h3>
2437	**
2438	** ^If [URI filename] interpretation is enabled, and the filename argument
2439	** begins with "file:", then the filename is interpreted as a URI. ^URI
2440	** filename interpretation is enabled if the [SQLITE_OPEN_URI] flag is
2441	** is set in the fourth argument to sqlite3_open_v2(), or if it has
2442	** been enabled globally using the [SQLITE_CONFIG_URI] option with the
2443	** [sqlite3_config()] method or by the [SQLITE_USE_URI] compile-time option.
2444	** As of SQLite version 3.7.7, URI filename interpretation is turned off
2445	** by default, but future releases of SQLite might enable URI filename
2446	** intepretation by default. See "[URI filenames]" for additional
2447	** information.
2448	**
2449	** URI filenames are parsed according to RFC 3986. ^If the URI contains an
2450	** authority, then it must be either an empty string or the string
2451	** "localhost". ^If the authority is not an empty string or "localhost", an
	@@ -3260,11 +3286,11 @@
3260	** [extended result codes] might be returned as well.
3261	**
3262	** ^[SQLITE_BUSY] means that the database engine was unable to acquire the
3263	** database locks it needs to do its job. ^If the statement is a [COMMIT]
3264	** or occurs outside of an explicit transaction, then you can retry the
3265	** statement. If the statement is not a [COMMIT] and occurs within a
3266	** explicit transaction then you should rollback the transaction before
3267	** continuing.
3268	**
3269	** ^[SQLITE_DONE] means that the statement has finished executing
3270	** successfully. sqlite3_step() should not be called again on this virtual
	@@ -3539,11 +3565,11 @@
3539
3540	/*
3541	** CAPI3REF: Destroy A Prepared Statement Object
3542	**
3543	** ^The sqlite3_finalize() function is called to delete a [prepared statement].
3544	** ^If the most recent evaluation of the statement encountered no errors or
3545	** or if the statement is never been evaluated, then sqlite3_finalize() returns
3546	** SQLITE_OK. ^If the most recent evaluation of statement S failed, then
3547	** sqlite3_finalize(S) returns the appropriate [error code] or
3548	** [extended error code].
3549	**
	@@ -5453,11 +5479,11 @@
5453	** versions of these routines, it should at least provide stubs that always
5454	** return true so that one does not get spurious assertion failures.
5455	**
5456	** ^If the argument to sqlite3_mutex_held() is a NULL pointer then
5457	** the routine should return 1. This seems counter-intuitive since
5458	** clearly the mutex cannot be held if it does not exist. But the
5459	** the reason the mutex does not exist is because the build is not
5460	** using mutexes. And we do not want the assert() containing the
5461	** call to sqlite3_mutex_held() to fail, so a non-zero return is
5462	** the appropriate thing to do. ^The sqlite3_mutex_notheld()
5463	** interface should also return 1 when given a NULL pointer.
	@@ -5576,11 +5602,12 @@
5576	#define SQLITE_TESTCTRL_RESERVE 14
5577	#define SQLITE_TESTCTRL_OPTIMIZATIONS 15
5578	#define SQLITE_TESTCTRL_ISKEYWORD 16
5579	#define SQLITE_TESTCTRL_PGHDRSZ 17
5580	#define SQLITE_TESTCTRL_SCRATCHMALLOC 18
5581	#define SQLITE_TESTCTRL_LAST 18

5582
5583	/*
5584	** CAPI3REF: SQLite Runtime Status
5585	**
5586	** ^This interface is used to retrieve runtime status information
	@@ -5962,11 +5989,11 @@
5962	** [[the xFetch() page cache methods]]
5963	** The xFetch() method locates a page in the cache and returns a pointer to
5964	** the page, or a NULL pointer.
5965	** A "page", in this context, means a buffer of szPage bytes aligned at an
5966	** 8-byte boundary. The page to be fetched is determined by the key. ^The
5967	** mimimum key value is 1. After it has been retrieved using xFetch, the page
5968	** is considered to be "pinned".
5969	**
5970	** If the requested page is already in the page cache, then the page cache
5971	** implementation must return a pointer to the page buffer with its content
5972	** intact. If the requested page is not already in the cache, then the
5973

	--- src/sqlite3.h
	+++ src/sqlite3.h
	@@ -105,13 +105,13 @@
105	**
106	** See also: [sqlite3_libversion()],
107	** [sqlite3_libversion_number()], [sqlite3_sourceid()],
108	** [sqlite_version()] and [sqlite_source_id()].
109	*/
110	#define SQLITE_VERSION "3.7.8"
111	#define SQLITE_VERSION_NUMBER 3007008
112	#define SQLITE_SOURCE_ID "2011-08-29 11:56:14 639cc85a911454bffdcccb33f2976c683953ae64"
113
114	/*
115	** CAPI3REF: Run-Time Library Version Numbers
116	** KEYWORDS: sqlite3_version, sqlite3_sourceid
117	**
	@@ -308,11 +308,11 @@
308	** semicolon-separate SQL statements passed into its 2nd argument,
309	** in the context of the [database connection] passed in as its 1st
310	** argument. ^If the callback function of the 3rd argument to
311	** sqlite3_exec() is not NULL, then it is invoked for each result row
312	** coming out of the evaluated SQL statements. ^The 4th argument to
313	** sqlite3_exec() is relayed through to the 1st argument of each
314	** callback invocation. ^If the callback pointer to sqlite3_exec()
315	** is NULL, then no callback is ever invoked and result rows are
316	** ignored.
317	**
318	** ^If an error occurs while evaluating the SQL statements passed into
	@@ -739,20 +739,52 @@
739	** when [PRAGMA synchronous \| PRAGMA synchronous=OFF] is set, but most
740	** VFSes do not need this signal and should silently ignore this opcode.
741	** Applications should not call [sqlite3_file_control()] with this
742	** opcode as doing so may disrupt the operation of the specialized VFSes
743	** that do require it.
744	**
745	** ^The [SQLITE_FCNTL_WIN32_AV_RETRY] opcode is used to configure automatic
746	** retry counts and intervals for certain disk I/O operations for the
747	** windows [VFS] in order to work to provide robustness against
748	** anti-virus programs. By default, the windows VFS will retry file read,
749	** file write, and file delete opertions up to 10 times, with a delay
750	** of 25 milliseconds before the first retry and with the delay increasing
751	** by an additional 25 milliseconds with each subsequent retry. This
752	** opcode allows those to values (10 retries and 25 milliseconds of delay)
753	** to be adjusted. The values are changed for all database connections
754	** within the same process. The argument is a pointer to an array of two
755	** integers where the first integer i the new retry count and the second
756	** integer is the delay. If either integer is negative, then the setting
757	** is not changed but instead the prior value of that setting is written
758	** into the array entry, allowing the current retry settings to be
759	** interrogated. The zDbName parameter is ignored.
760	**
761	** ^The [SQLITE_FCNTL_PERSIST_WAL] opcode is used to set or query the
762	** persistent [WAL \| Write AHead Log] setting. By default, the auxiliary
763	** write ahead log and shared memory files used for transaction control
764	** are automatically deleted when the latest connection to the database
765	** closes. Setting persistent WAL mode causes those files to persist after
766	** close. Persisting the files is useful when other processes that do not
767	** have write permission on the directory containing the database file want
768	** to read the database file, as the WAL and shared memory files must exist
769	** in order for the database to be readable. The fourth parameter to
770	** [sqlite3_file_control()] for this opcode should be a pointer to an integer.
771	** That integer is 0 to disable persistent WAL mode or 1 to enable persistent
772	** WAL mode. If the integer is -1, then it is overwritten with the current
773	** WAL persistence setting.
774	**
775	*/
776	#define SQLITE_FCNTL_LOCKSTATE 1
777	#define SQLITE_GET_LOCKPROXYFILE 2
778	#define SQLITE_SET_LOCKPROXYFILE 3
779	#define SQLITE_LAST_ERRNO 4
780	#define SQLITE_FCNTL_SIZE_HINT 5
781	#define SQLITE_FCNTL_CHUNK_SIZE 6
782	#define SQLITE_FCNTL_FILE_POINTER 7
783	#define SQLITE_FCNTL_SYNC_OMITTED 8
784	#define SQLITE_FCNTL_WIN32_AV_RETRY 9
785	#define SQLITE_FCNTL_PERSIST_WAL 10
786
787	/*
788	** CAPI3REF: Mutex Handle
789	**
790	** The mutex module within SQLite defines [sqlite3_mutex] to be an
	@@ -900,11 +932,11 @@
932	** The xSleep() method causes the calling thread to sleep for at
933	** least the number of microseconds given. ^The xCurrentTime()
934	** method returns a Julian Day Number for the current date and time as
935	** a floating point value.
936	** ^The xCurrentTimeInt64() method returns, as an integer, the Julian
937	** Day Number multiplied by 86400000 (the number of milliseconds in
938	** a 24-hour day).
939	** ^SQLite will use the xCurrentTimeInt64() method to get the current
940	** date and time if that method is available (if iVersion is 2 or
941	** greater and the function pointer is not NULL) and will fall back
942	** to xCurrentTime() if xCurrentTimeInt64() is unavailable.
	@@ -1176,20 +1208,14 @@
1208	** also used during testing of SQLite in order to specify an alternative
1209	** memory allocator that simulates memory out-of-memory conditions in
1210	** order to verify that SQLite recovers gracefully from such
1211	** conditions.
1212	**
1213	** The xMalloc, xRealloc, and xFree methods must work like the
1214	** malloc(), realloc() and free() functions from the standard C library.
1215	** ^SQLite guarantees that the second argument to



1216	** xRealloc is always a value returned by a prior call to xRoundup.



1217	**
1218	** xSize should return the allocated size of a memory allocation
1219	** previously obtained from xMalloc or xRealloc. The allocated size
1220	** is always at least as big as the requested size but may be larger.
1221	**
	@@ -1338,11 +1364,11 @@
1364	** scratch memory beyond what is provided by this configuration option, then
1365	** [sqlite3_malloc()] will be used to obtain the memory needed.</dd>
1366	**
1367	** [[SQLITE_CONFIG_PAGECACHE]] <dt>SQLITE_CONFIG_PAGECACHE</dt>
1368	** <dd> ^This option specifies a static memory buffer that SQLite can use for
1369	** the database page cache with the default page cache implementation.
1370	** This configuration should not be used if an application-define page
1371	** cache implementation is loaded using the SQLITE_CONFIG_PCACHE option.
1372	** There are three arguments to this option: A pointer to 8-byte aligned
1373	** memory, the size of each page buffer (sz), and the number of pages (N).
1374	** The sz argument should be the size of the largest database page
	@@ -2436,16 +2462,16 @@
2462	** [[URI filenames in sqlite3_open()]] <h3>URI Filenames</h3>
2463	**
2464	** ^If [URI filename] interpretation is enabled, and the filename argument
2465	** begins with "file:", then the filename is interpreted as a URI. ^URI
2466	** filename interpretation is enabled if the [SQLITE_OPEN_URI] flag is
2467	** set in the fourth argument to sqlite3_open_v2(), or if it has
2468	** been enabled globally using the [SQLITE_CONFIG_URI] option with the
2469	** [sqlite3_config()] method or by the [SQLITE_USE_URI] compile-time option.
2470	** As of SQLite version 3.7.7, URI filename interpretation is turned off
2471	** by default, but future releases of SQLite might enable URI filename
2472	** interpretation by default. See "[URI filenames]" for additional
2473	** information.
2474	**
2475	** URI filenames are parsed according to RFC 3986. ^If the URI contains an
2476	** authority, then it must be either an empty string or the string
2477	** "localhost". ^If the authority is not an empty string or "localhost", an
	@@ -3260,11 +3286,11 @@
3286	** [extended result codes] might be returned as well.
3287	**
3288	** ^[SQLITE_BUSY] means that the database engine was unable to acquire the
3289	** database locks it needs to do its job. ^If the statement is a [COMMIT]
3290	** or occurs outside of an explicit transaction, then you can retry the
3291	** statement. If the statement is not a [COMMIT] and occurs within an
3292	** explicit transaction then you should rollback the transaction before
3293	** continuing.
3294	**
3295	** ^[SQLITE_DONE] means that the statement has finished executing
3296	** successfully. sqlite3_step() should not be called again on this virtual
	@@ -3539,11 +3565,11 @@
3565
3566	/*
3567	** CAPI3REF: Destroy A Prepared Statement Object
3568	**
3569	** ^The sqlite3_finalize() function is called to delete a [prepared statement].
3570	** ^If the most recent evaluation of the statement encountered no errors
3571	** or if the statement is never been evaluated, then sqlite3_finalize() returns
3572	** SQLITE_OK. ^If the most recent evaluation of statement S failed, then
3573	** sqlite3_finalize(S) returns the appropriate [error code] or
3574	** [extended error code].
3575	**
	@@ -5453,11 +5479,11 @@
5479	** versions of these routines, it should at least provide stubs that always
5480	** return true so that one does not get spurious assertion failures.
5481	**
5482	** ^If the argument to sqlite3_mutex_held() is a NULL pointer then
5483	** the routine should return 1. This seems counter-intuitive since
5484	** clearly the mutex cannot be held if it does not exist. But
5485	** the reason the mutex does not exist is because the build is not
5486	** using mutexes. And we do not want the assert() containing the
5487	** call to sqlite3_mutex_held() to fail, so a non-zero return is
5488	** the appropriate thing to do. ^The sqlite3_mutex_notheld()
5489	** interface should also return 1 when given a NULL pointer.
	@@ -5576,11 +5602,12 @@
5602	#define SQLITE_TESTCTRL_RESERVE 14
5603	#define SQLITE_TESTCTRL_OPTIMIZATIONS 15
5604	#define SQLITE_TESTCTRL_ISKEYWORD 16
5605	#define SQLITE_TESTCTRL_PGHDRSZ 17
5606	#define SQLITE_TESTCTRL_SCRATCHMALLOC 18
5607	#define SQLITE_TESTCTRL_LOCALTIME_FAULT 19
5608	#define SQLITE_TESTCTRL_LAST 19
5609
5610	/*
5611	** CAPI3REF: SQLite Runtime Status
5612	**
5613	** ^This interface is used to retrieve runtime status information
	@@ -5962,11 +5989,11 @@
5989	** [[the xFetch() page cache methods]]
5990	** The xFetch() method locates a page in the cache and returns a pointer to
5991	** the page, or a NULL pointer.
5992	** A "page", in this context, means a buffer of szPage bytes aligned at an
5993	** 8-byte boundary. The page to be fetched is determined by the key. ^The
5994	** minimum key value is 1. After it has been retrieved using xFetch, the page
5995	** is considered to be "pinned".
5996	**
5997	** If the requested page is already in the page cache, then the page cache
5998	** implementation must return a pointer to the page buffer with its content
5999	** intact. If the requested page is not already in the cache, then the
6000

M src/stash.c

+4 -2

		--- src/stash.c
		+++ src/stash.c
		@@ -100,10 +100,11 @@
100	100	if( rid==0 ){
101	101	/* A new file */
102	102	blob_read_from_file(&content, zPath);
103	103	db_bind_blob(&ins, ":content", &content);
104	104	}else if( deleted ){
	105	+ blob_zero(&content);
105	106	db_bind_null(&ins, ":content");
106	107	}else{
107	108	/* A modified file */
108	109	Blob orig;
109	110	Blob disk;
		@@ -274,11 +275,11 @@
274	275	}
275	276	db_finalize(&q);
276	277	}
277	278
278	279	/*
279		-** Drop the indicates stash
	280	+** Drop the indicated stash
280	281	*/
281	282	static void stash_drop(int stashid){
282	283	db_multi_exec(
283	284	"DELETE FROM stash WHERE stashid=%d;"
284	285	"DELETE FROM stashfile WHERE stashid=%d;",
		@@ -344,10 +345,11 @@
344	345	** Update to the baseline checkout for STASHID then apply the
345	346	** changes of STASHID. Keep STASHID so that it can be reused
346	347	** This command is undoable.
347	348	**
348	349	** fossil stash drop ?STASHID? ?--all?
	350	+** fossil stash rm ?STASHID? ?--all?
349	351	**
350	352	** Forget everything about STASHID. Forget the whole stash if the
351	353	** --all flag is used. Individual drops are undoable but --all is not.
352	354	**
353	355	** fossil stash snapshot ?-m COMMENT? ?FILES...?
		@@ -427,11 +429,11 @@
427	429	}
428	430	}
429	431	db_finalize(&q);
430	432	if( n==0 ) fossil_print("empty stash\n");
431	433	}else
432		- if( memcmp(zCmd, "drop", nCmd)==0 ){
	434	+ if( memcmp(zCmd, "drop", nCmd)==0 \|\| memcmp(zCmd, "rm", nCmd)==0 ){
433	435	int allFlag = find_option("all", 0, 0)!=0;
434	436	if( g.argc>4 ) usage("stash apply STASHID");
435	437	if( allFlag ){
436	438	db_multi_exec("DELETE FROM stash; DELETE FROM stashfile;");
437	439	}else{
438	440

	--- src/stash.c
	+++ src/stash.c
	@@ -100,10 +100,11 @@
100	if( rid==0 ){
101	/* A new file */
102	blob_read_from_file(&content, zPath);
103	db_bind_blob(&ins, ":content", &content);
104	}else if( deleted ){

105	db_bind_null(&ins, ":content");
106	}else{
107	/* A modified file */
108	Blob orig;
109	Blob disk;
	@@ -274,11 +275,11 @@
274	}
275	db_finalize(&q);
276	}
277
278	/*
279	** Drop the indicates stash
280	*/
281	static void stash_drop(int stashid){
282	db_multi_exec(
283	"DELETE FROM stash WHERE stashid=%d;"
284	"DELETE FROM stashfile WHERE stashid=%d;",
	@@ -344,10 +345,11 @@
344	** Update to the baseline checkout for STASHID then apply the
345	** changes of STASHID. Keep STASHID so that it can be reused
346	** This command is undoable.
347	**
348	** fossil stash drop ?STASHID? ?--all?

349	**
350	** Forget everything about STASHID. Forget the whole stash if the
351	** --all flag is used. Individual drops are undoable but --all is not.
352	**
353	** fossil stash snapshot ?-m COMMENT? ?FILES...?
	@@ -427,11 +429,11 @@
427	}
428	}
429	db_finalize(&q);
430	if( n==0 ) fossil_print("empty stash\n");
431	}else
432	if( memcmp(zCmd, "drop", nCmd)==0 ){
433	int allFlag = find_option("all", 0, 0)!=0;
434	if( g.argc>4 ) usage("stash apply STASHID");
435	if( allFlag ){
436	db_multi_exec("DELETE FROM stash; DELETE FROM stashfile;");
437	}else{
438

	--- src/stash.c
	+++ src/stash.c
	@@ -100,10 +100,11 @@
100	if( rid==0 ){
101	/* A new file */
102	blob_read_from_file(&content, zPath);
103	db_bind_blob(&ins, ":content", &content);
104	}else if( deleted ){
105	blob_zero(&content);
106	db_bind_null(&ins, ":content");
107	}else{
108	/* A modified file */
109	Blob orig;
110	Blob disk;
	@@ -274,11 +275,11 @@
275	}
276	db_finalize(&q);
277	}
278
279	/*
280	** Drop the indicated stash
281	*/
282	static void stash_drop(int stashid){
283	db_multi_exec(
284	"DELETE FROM stash WHERE stashid=%d;"
285	"DELETE FROM stashfile WHERE stashid=%d;",
	@@ -344,10 +345,11 @@
345	** Update to the baseline checkout for STASHID then apply the
346	** changes of STASHID. Keep STASHID so that it can be reused
347	** This command is undoable.
348	**
349	** fossil stash drop ?STASHID? ?--all?
350	** fossil stash rm ?STASHID? ?--all?
351	**
352	** Forget everything about STASHID. Forget the whole stash if the
353	** --all flag is used. Individual drops are undoable but --all is not.
354	**
355	** fossil stash snapshot ?-m COMMENT? ?FILES...?
	@@ -427,11 +429,11 @@
429	}
430	}
431	db_finalize(&q);
432	if( n==0 ) fossil_print("empty stash\n");
433	}else
434	if( memcmp(zCmd, "drop", nCmd)==0 \|\| memcmp(zCmd, "rm", nCmd)==0 ){
435	int allFlag = find_option("all", 0, 0)!=0;
436	if( g.argc>4 ) usage("stash apply STASHID");
437	if( allFlag ){
438	db_multi_exec("DELETE FROM stash; DELETE FROM stashfile;");
439	}else{
440

M src/stat.c

+2 -1

		--- src/stat.c
		+++ src/stat.c
		@@ -104,11 +104,12 @@
104	104	@ <tr><th>Server ID:</th><td>
105	105	@ %h(db_get("server-code",""))
106	106	@ </td></tr>
107	107
108	108	@ <tr><th>Fossil Version:</th><td>
109		- @ %h(MANIFEST_DATE) %h(MANIFEST_VERSION) (%h(COMPILER_NAME))
	109	+ @ %h(RELEASE_VERSION) %h(MANIFEST_DATE) %h(MANIFEST_VERSION)
	110	+ @ (%h(COMPILER_NAME))
110	111	@ </td></tr>
111	112	@ <tr><th>SQLite Version:</th><td>
112	113	sqlite3_snprintf(sizeof(zBuf), zBuf, "%.19s [%.10s] (%s)",
113	114	SQLITE_SOURCE_ID, &SQLITE_SOURCE_ID[20], SQLITE_VERSION);
114	115	zDb = db_name("repository");
115	116

	--- src/stat.c
	+++ src/stat.c
	@@ -104,11 +104,12 @@
104	@ <tr><th>Server ID:</th><td>
105	@ %h(db_get("server-code",""))
106	@ </td></tr>
107
108	@ <tr><th>Fossil Version:</th><td>
109	@ %h(MANIFEST_DATE) %h(MANIFEST_VERSION) (%h(COMPILER_NAME))

110	@ </td></tr>
111	@ <tr><th>SQLite Version:</th><td>
112	sqlite3_snprintf(sizeof(zBuf), zBuf, "%.19s [%.10s] (%s)",
113	SQLITE_SOURCE_ID, &SQLITE_SOURCE_ID[20], SQLITE_VERSION);
114	zDb = db_name("repository");
115

	--- src/stat.c
	+++ src/stat.c
	@@ -104,11 +104,12 @@
104	@ <tr><th>Server ID:</th><td>
105	@ %h(db_get("server-code",""))
106	@ </td></tr>
107
108	@ <tr><th>Fossil Version:</th><td>
109	@ %h(RELEASE_VERSION) %h(MANIFEST_DATE) %h(MANIFEST_VERSION)
110	@ (%h(COMPILER_NAME))
111	@ </td></tr>
112	@ <tr><th>SQLite Version:</th><td>
113	sqlite3_snprintf(sizeof(zBuf), zBuf, "%.19s [%.10s] (%s)",
114	SQLITE_SOURCE_ID, &SQLITE_SOURCE_ID[20], SQLITE_VERSION);
115	zDb = db_name("repository");
116

M src/style.c

+2 -1

		--- src/style.c
		+++ src/style.c
		@@ -98,10 +98,11 @@
98	98	Th_Store("title", zTitle);
99	99	Th_Store("baseurl", g.zBaseURL);
100	100	Th_Store("home", g.zTop);
101	101	Th_Store("index_page", db_get("index-page","/home"));
102	102	Th_Store("current_page", g.zPath);
	103	+ Th_Store("release_version", RELEASE_VERSION);
103	104	Th_Store("manifest_version", MANIFEST_VERSION);
104	105	Th_Store("manifest_date", MANIFEST_DATE);
105	106	Th_Store("compiler_name", COMPILER_NAME);
106	107	if( g.zLogin ){
107	108	Th_Store("login", g.zLogin);
		@@ -245,11 +246,11 @@
245	246	/*
246	247	** The default page footer
247	248	*/
248	249	const char zDefaultFooter[] =
249	250	@ <div class="footer">
250		-@ Fossil version $manifest_version $manifest_date
	251	+@ Fossil version $release_version $manifest_version $manifest_date
251	252	@ </div>
252	253	@ </body></html>
253	254	;
254	255
255	256	/*
256	257

	--- src/style.c
	+++ src/style.c
	@@ -98,10 +98,11 @@
98	Th_Store("title", zTitle);
99	Th_Store("baseurl", g.zBaseURL);
100	Th_Store("home", g.zTop);
101	Th_Store("index_page", db_get("index-page","/home"));
102	Th_Store("current_page", g.zPath);

103	Th_Store("manifest_version", MANIFEST_VERSION);
104	Th_Store("manifest_date", MANIFEST_DATE);
105	Th_Store("compiler_name", COMPILER_NAME);
106	if( g.zLogin ){
107	Th_Store("login", g.zLogin);
	@@ -245,11 +246,11 @@
245	/*
246	** The default page footer
247	*/
248	const char zDefaultFooter[] =
249	@ <div class="footer">
250	@ Fossil version $manifest_version $manifest_date
251	@ </div>
252	@ </body></html>
253	;
254
255	/*
256

	--- src/style.c
	+++ src/style.c
	@@ -98,10 +98,11 @@
98	Th_Store("title", zTitle);
99	Th_Store("baseurl", g.zBaseURL);
100	Th_Store("home", g.zTop);
101	Th_Store("index_page", db_get("index-page","/home"));
102	Th_Store("current_page", g.zPath);
103	Th_Store("release_version", RELEASE_VERSION);
104	Th_Store("manifest_version", MANIFEST_VERSION);
105	Th_Store("manifest_date", MANIFEST_DATE);
106	Th_Store("compiler_name", COMPILER_NAME);
107	if( g.zLogin ){
108	Th_Store("login", g.zLogin);
	@@ -245,11 +246,11 @@
246	/*
247	** The default page footer
248	*/
249	const char zDefaultFooter[] =
250	@ <div class="footer">
251	@ Fossil version $release_version $manifest_version $manifest_date
252	@ </div>
253	@ </body></html>
254	;
255
256	/*
257

M src/tag.c

+1 -1

		--- src/tag.c
		+++ src/tag.c
		@@ -458,11 +458,11 @@
458	458	" WHERE tagtype>0 AND tagid=%d"
459	459	")"
460	460	" ORDER BY event.mtime DESC",
461	461	timeline_query_for_tty(), zType, tagid
462	462	);
463		- print_timeline(&q, 2000);
	463	+ print_timeline(&q, 2000, 0);
464	464	db_finalize(&q);
465	465	}
466	466	}
467	467	}else
468	468
469	469

	--- src/tag.c
	+++ src/tag.c
	@@ -458,11 +458,11 @@
458	" WHERE tagtype>0 AND tagid=%d"
459	")"
460	" ORDER BY event.mtime DESC",
461	timeline_query_for_tty(), zType, tagid
462	);
463	print_timeline(&q, 2000);
464	db_finalize(&q);
465	}
466	}
467	}else
468
469

	--- src/tag.c
	+++ src/tag.c
	@@ -458,11 +458,11 @@
458	" WHERE tagtype>0 AND tagid=%d"
459	")"
460	" ORDER BY event.mtime DESC",
461	timeline_query_for_tty(), zType, tagid
462	);
463	print_timeline(&q, 2000, 0);
464	db_finalize(&q);
465	}
466	}
467	}else
468
469

M src/tar.c

+280 -29

		--- src/tar.c
		+++ src/tar.c
		@@ -27,31 +27,252 @@
27	27	*/
28	28	static struct tarball_t {
29	29	unsigned char aHdr; / Space for building headers */
30	30	char zSpaces; / Spaces for padding */
31	31	char zPrevDir; / Name of directory for previous entry */
	32	+ int nPrevDirAlloc; /* size of zPrevDir */
	33	+ Blob pax; /* PAX data */
32	34	} tball;
	35	+
	36	+
	37	+/*
	38	+** field lengths of 'ustar' name and prefix fields.
	39	+*/
	40	+#define USTAR_NAME_LEN 100
	41	+#define USTAR_PREFIX_LEN 155
	42	+
33	43
34	44	/*
35	45	** Begin the process of generating a tarball.
36	46	**
37	47	** Initialize the GZIP compressor and the table of directory names.
38	48	*/
39	49	static void tar_begin(void){
40	50	assert( tball.aHdr==0 );
41		- tball.aHdr = fossil_malloc(512+512+256);
42		- memset(tball.aHdr, 0, 512+512+256);
	51	+ tball.aHdr = fossil_malloc(512+512);
	52	+ memset(tball.aHdr, 0, 512+512);
43	53	tball.zSpaces = (char*)&tball.aHdr[512];
44		- tball.zPrevDir = (char*)&tball.zSpaces[512];
	54	+ /* zPrevDir init */
	55	+ tball.zPrevDir = NULL;
	56	+ tball.nPrevDirAlloc = 0;
	57	+ /* scratch buffer init */
	58	+ blob_zero(&tball.pax);
	59	+
45	60	memcpy(&tball.aHdr[108], "0000000", 8); /* Owner ID */
46	61	memcpy(&tball.aHdr[116], "0000000", 8); /* Group ID */
47		- memcpy(&tball.aHdr[257], "ustar ", 7); /* Format */
	62	+ memcpy(&tball.aHdr[257], "ustar\00000", 8); /* POSIX.1 format */
	63	+ memcpy(&tball.aHdr[265], "nobody", 7); /* Owner name */
	64	+ memcpy(&tball.aHdr[297], "nobody", 7); /* Group name */
48	65	gzip_begin();
49	66	db_multi_exec(
50	67	"CREATE TEMP TABLE dir(name UNIQUE);"
51	68	);
52	69	}
	70	+
	71	+
	72	+/*
	73	+** verify that lla characters in 'zName' are in the
	74	+** ISO646 (=ASCII) character set.
	75	+*/
	76	+static int is_iso646_name(
	77	+ const char zName, / file path */
	78	+ int nName /* path length */
	79	+){
	80	+ int i;
	81	+ for(i = 0; i < nName; i++){
	82	+ unsigned char c = (unsigned char)zName[i];
	83	+ if( c>0x7e ) return 0;
	84	+ }
	85	+ return 1;
	86	+}
	87	+
	88	+
	89	+/*
	90	+** copy string pSrc into pDst, truncating or padding with 0 if necessary
	91	+*/
	92	+static void padded_copy(
	93	+ char *pDest,
	94	+ int nDest,
	95	+ const char *pSrc,
	96	+ int nSrc
	97	+){
	98	+ if(nSrc >= nDest){
	99	+ memcpy(pDest, pSrc, nDest);
	100	+ }else{
	101	+ memcpy(pDest, pSrc, nSrc);
	102	+ memset(&pDest[nSrc], 0, nDest - nSrc);
	103	+ }
	104	+}
	105	+
	106	+
	107	+
	108	+/******************************************************************************
	109	+**
	110	+** The 'tar' format has evolved over time. Initially the name was stored
	111	+** in a 100 byte null-terminated field 'name'. File path names were
	112	+** limited to 99 bytes.
	113	+**
	114	+** The Posix.1 'ustar' format added a 155 byte field 'prefix', allowing
	115	+** for up to 255 characters to be stored. The full file path is formed by
	116	+** concatenating the field 'prefix', a slash, and the field 'name'. This
	117	+** gives some measure of compatibility with programs that only understand
	118	+** the oldest format.
	119	+**
	120	+** The latest Posix extension is called the 'pax Interchange Format'.
	121	+** It removes all the limitations of the previous two formats by allowing
	122	+** the storage of arbitrary-length attributes in a separate object that looks
	123	+** like a file to programs that do not understand this extension. So the
	124	+** contents of the 'name' and 'prefix' fields should contain values that allow
	125	+** versions of tar that do not understand this extension to still do
	126	+** something useful.
	127	+**
	128	+******************************************************************************/
	129	+
	130	+/*
	131	+** The position we use to split a file path into the 'name' and 'prefix'
	132	+** fields needs to meet the following criteria:
	133	+**
	134	+** - not at the beginning or end of the string
	135	+** - the position must contain a slash
	136	+** - no more than 100 characters follow the slash
	137	+** - no more than 155 characters precede it
	138	+**
	139	+** The routine 'find_split_pos' finds a split position. It will meet the
	140	+** criteria of listed above if such a position exists. If no such
	141	+** position exists it generates one that useful for generating the
	142	+** values used for backward compatibility.
	143	+*/
	144	+static int find_split_pos(
	145	+ const char zName, / file path */
	146	+ int nName /* path length */
	147	+){
	148	+ int i, split = 0;
	149	+ /* only search if the string needs splitting */
	150	+ if(nName > USTAR_NAME_LEN){
	151	+ for(i = 1; i+1 < nName; i++)
	152	+ if(zName[i] == '/'){
	153	+ split = i+1;
	154	+ /* if the split position is within USTAR_NAME_LEN bytes from
	155	+ * the end we can quit */
	156	+ if(nName - split <= USTAR_NAME_LEN) break;
	157	+ }
	158	+ }
	159	+ return split;
	160	+}
	161	+
	162	+
	163	+/*
	164	+** attempt to split the file name path to meet 'ustar' header
	165	+** criteria.
	166	+*/
	167	+static int tar_split_path(
	168	+ const char zName, / path */
	169	+ int nName, /* path length */
	170	+ char pName, / name field */
	171	+ char pPrefix / prefix field */
	172	+){
	173	+ int split = find_split_pos(zName, nName);
	174	+ /* check whether both pieces fit */
	175	+ if(nName - split > USTAR_NAME_LEN \|\| split > USTAR_PREFIX_LEN+1){
	176	+ return 0; /* no */
	177	+ }
	178	+
	179	+ /* extract name */
	180	+ padded_copy(pName, USTAR_NAME_LEN, &zName[split], nName - split);
	181	+
	182	+ /* extract prefix */
	183	+ padded_copy(pPrefix, USTAR_PREFIX_LEN, zName, (split > 0 ? split - 1 : 0));
	184	+
	185	+ return 1; /* success */
	186	+}
	187	+
	188	+
	189	+/*
	190	+** When using an extension header we still need to put something
	191	+** reasonable in the name and prefix fields. This is probably as
	192	+** good as it gets.
	193	+*/
	194	+static void approximate_split_path(
	195	+ const char zName, / path */
	196	+ int nName, /* path length */
	197	+ char pName, / name field */
	198	+ char pPrefix, / prefix field */
	199	+ int bHeader /* is this a 'x' type tar header? */
	200	+){
	201	+ int split;
	202	+
	203	+ /* if this is a Pax Interchange header prepend "PaxHeader/"
	204	+ ** so we can tell files apart from metadata */
	205	+ if( bHeader ){
	206	+ blob_reset(&tball.pax);
	207	+ blob_appendf(&tball.pax, "PaxHeader/%.s", nName, nName, zName);
	208	+ zName = blob_buffer(&tball.pax);
	209	+ nName = blob_size(&tball.pax);
	210	+ }
	211	+
	212	+ /* find the split position */
	213	+ split = find_split_pos(zName, nName);
	214	+
	215	+ /* extract a name, truncate if needed */
	216	+ padded_copy(pName, USTAR_NAME_LEN, &zName[split], nName - split);
	217	+
	218	+ /* extract a prefix field, truncate when needed */
	219	+ padded_copy(pPrefix, USTAR_PREFIX_LEN, zName, (split > 0 ? split-1 : 0));
	220	+}
	221	+
	222	+
	223	+/*
	224	+** add a Pax Interchange header to the scratch buffer
	225	+**
	226	+** format: <length> <key>=<value>\n
	227	+** the tricky part is that each header contains its own
	228	+** size in decimal, counting that length.
	229	+*/
	230	+static void add_pax_header(
	231	+ const char *zField,
	232	+ const char *zValue,
	233	+ int nValue
	234	+){
	235	+ /* calculate length without length field */
	236	+ int blen = strlen(zField) + nValue + 3;
	237	+ /* calculate the length of the length field */
	238	+ int next10 = 1;
	239	+ int n;
	240	+ for(n = blen; n > 0; ){
	241	+ blen++; next10 *= 10;
	242	+ n /= 10;
	243	+ }
	244	+ /* adding the length extended the length field? */
	245	+ if(blen > next10){
	246	+ blen++;
	247	+ }
	248	+ /* build the string */
	249	+ blob_appendf(&tball.pax, "%d %s=%.s\n", blen, zField, nValue, nValue, zValue);
	250	+ /* this _must_ be right */
	251	+ if(blob_size(&tball.pax) != blen){
	252	+ fossil_fatal("internal error: PAX tar header has bad length");
	253	+ }
	254	+}
	255	+
	256	+
	257	+/*
	258	+** set the header type, calculate the checksum and output
	259	+** the header
	260	+*/
	261	+static void cksum_and_write_header(
	262	+ char cType
	263	+){
	264	+ unsigned int cksum = 0;
	265	+ int i;
	266	+ memset(&tball.aHdr[148], ' ', 8);
	267	+ tball.aHdr[156] = cType;
	268	+ for(i=0; i<512; i++) cksum += tball.aHdr[i];
	269	+ sqlite3_snprintf(8, (char*)&tball.aHdr[148], "%07o", cksum);
	270	+ tball.aHdr[155] = 0;
	271	+ gzip_step((char*)tball.aHdr, 512);
	272	+}
	273	+
53	274
54	275	/*
55	276	** Build a header for a file or directory and write that header
56	277	** into the growing tarball.
57	278	*/
		@@ -59,33 +280,52 @@
59	280	const char zName, / Name of the object */
60	281	int nName, /* Number of characters in zName */
61	282	int iMode, /* Mode. 0644 or 0755 */
62	283	unsigned int mTime, /* File modification time */
63	284	int iSize, /* Size of the object in bytes */
64		- int iType /* Type of object. 0==file. 5==directory */
	285	+ char cType /* Type of object. '0'==file. '5'==directory */
65	286	){
66		- unsigned int cksum = 0;
67		- int i;
68		- if( nName>100 ){
69		- memcpy(&tball.aHdr[345], zName, nName-100);
70		- memcpy(tball.aHdr, &zName[nName-100], 100);
71		- memset(&tball.aHdr[245+nName], 0, 267-nName);
72		- }else{
73		- memcpy(tball.aHdr, zName, nName);
74		- memset(&tball.aHdr[nName], 0, 100-nName);
75		- memset(&tball.aHdr[345], 0, 167);
76		- }
	287	+ /* set mode and modification time */
77	288	sqlite3_snprintf(8, (char*)&tball.aHdr[100], "%07o", iMode);
78		- sqlite3_snprintf(12, (char*)&tball.aHdr[124], "%011o", iSize);
79	289	sqlite3_snprintf(12, (char*)&tball.aHdr[136], "%011o", mTime);
80		- memset(&tball.aHdr[148], ' ', 8);
81		- tball.aHdr[156] = iType + '0';
82		- for(i=0; i<512; i++) cksum += tball.aHdr[i];
83		- sqlite3_snprintf(7, (char*)&tball.aHdr[148], "%06o", cksum);
84		- tball.aHdr[154] = 0;
85		- gzip_step((char*)tball.aHdr, 512);
	290	+
	291	+ /* see if we need to output a Pax Interchange Header */
	292	+ if( !is_iso646_name(zName, nName)
	293	+ \|\| !tar_split_path(zName, nName, (char)tball.aHdr, (char)&tball.aHdr[345])
	294	+ ){
	295	+ int lastPage;
	296	+ /* add a file name for interoperability with older programs */
	297	+ approximate_split_path(zName, nName, (char*)tball.aHdr,
	298	+ (char*)&tball.aHdr[345], 1);
	299	+
	300	+ /* generate the Pax Interchange path header */
	301	+ blob_reset(&tball.pax);
	302	+ add_pax_header("path", zName, nName);
	303	+
	304	+ /* set the header length, and write the header */
	305	+ sqlite3_snprintf(12, (char*)&tball.aHdr[124], "%011o",
	306	+ blob_size(&tball.pax));
	307	+ cksum_and_write_header('x');
	308	+
	309	+ /* write the Pax Interchange data */
	310	+ gzip_step(blob_buffer(&tball.pax), blob_size(&tball.pax));
	311	+ lastPage = blob_size(&tball.pax) % 512;
	312	+ if( lastPage!=0 ){
	313	+ gzip_step(tball.zSpaces, 512 - lastPage);
	314	+ }
	315	+
	316	+ /* generate an approximate path for the regular header */
	317	+ approximate_split_path(zName, nName, (char*)tball.aHdr,
	318	+ (char*)&tball.aHdr[345], 0);
	319	+ }
	320	+ /* set the size */
	321	+ sqlite3_snprintf(12, (char*)&tball.aHdr[124], "%011o", iSize);
	322	+
	323	+ /* write the regular header */
	324	+ cksum_and_write_header(cType);
86	325	}
	326	+
87	327
88	328	/*
89	329	** Recursively add an directory entry for the given file if those
90	330	** directories have not previously been seen.
91	331	*/
		@@ -95,18 +335,27 @@
95	335	unsigned int mTime /* Modification time */
96	336	){
97	337	int i;
98	338	for(i=nName-1; i>0 && zName[i]!='/'; i--){}
99	339	if( i<=0 ) return;
100		- if( tball.zPrevDir[i]==0 && memcmp(tball.zPrevDir, zName, i)==0 ) return;
	340	+ if( i < tball.nPrevDirAlloc && tball.zPrevDir[i]==0 &&
	341	+ memcmp(tball.zPrevDir, zName, i)==0 ) return;
101	342	db_multi_exec("INSERT OR IGNORE INTO dir VALUES('%#q')", i, zName);
102	343	if( sqlite3_changes(g.db)==0 ) return;
103	344	tar_add_directory_of(zName, i-1, mTime);
104		- tar_add_header(zName, i, 0755, mTime, 0, 5);
	345	+ tar_add_header(zName, i, 0755, mTime, 0, '5');
	346	+ if( i >= tball.nPrevDirAlloc ){
	347	+ int nsize = tball.nPrevDirAlloc * 2;
	348	+ if(i+1 > nsize)
	349	+ nsize = i+1;
	350	+ tball.zPrevDir = fossil_realloc(tball.zPrevDir, nsize);
	351	+ tball.nPrevDirAlloc = nsize;
	352	+ }
105	353	memcpy(tball.zPrevDir, zName, i);
106	354	tball.zPrevDir[i] = 0;
107	355	}
	356	+
108	357
109	358	/*
110	359	** Add a single file to the growing tarball.
111	360	*/
112	361	static void tar_add_file(
		@@ -117,15 +366,13 @@
117	366	){
118	367	int nName = strlen(zName);
119	368	int n = blob_size(pContent);
120	369	int lastPage;
121	370
122		- if( nName>=250 ){
123		- fossil_fatal("name too long for ustar format: \"%s\"", zName);
124		- }
	371	+ /* length check moved to tar_split_path */
125	372	tar_add_directory_of(zName, nName, mTime);
126		- tar_add_header(zName, nName, isExe ? 0755 : 0644, mTime, n, 0);
	373	+ tar_add_header(zName, nName, isExe ? 0755 : 0644, mTime, n, '0');
127	374	if( n ){
128	375	gzip_step(blob_buffer(pContent), n);
129	376	lastPage = n % 512;
130	377	if( lastPage!=0 ){
131	378	gzip_step(tball.zSpaces, 512 - lastPage);
		@@ -142,10 +389,14 @@
142	389	gzip_step(tball.zSpaces, 512);
143	390	gzip_step(tball.zSpaces, 512);
144	391	gzip_finish(pOut);
145	392	fossil_free(tball.aHdr);
146	393	tball.aHdr = 0;
	394	+ fossil_free(tball.zPrevDir);
	395	+ tball.zPrevDir = NULL;
	396	+ tball.nPrevDirAlloc = 0;
	397	+ blob_reset(&tball.pax);
147	398	}
148	399
149	400
150	401	/*
151	402	** COMMAND: test-tarball
152	403

	--- src/tar.c
	+++ src/tar.c
	@@ -27,31 +27,252 @@
27	*/
28	static struct tarball_t {
29	unsigned char aHdr; / Space for building headers */
30	char zSpaces; / Spaces for padding */
31	char zPrevDir; / Name of directory for previous entry */


32	} tball;








33
34	/*
35	** Begin the process of generating a tarball.
36	**
37	** Initialize the GZIP compressor and the table of directory names.
38	*/
39	static void tar_begin(void){
40	assert( tball.aHdr==0 );
41	tball.aHdr = fossil_malloc(512+512+256);
42	memset(tball.aHdr, 0, 512+512+256);
43	tball.zSpaces = (char*)&tball.aHdr[512];
44	tball.zPrevDir = (char*)&tball.zSpaces[512];





45	memcpy(&tball.aHdr[108], "0000000", 8); /* Owner ID */
46	memcpy(&tball.aHdr[116], "0000000", 8); /* Group ID */
47	memcpy(&tball.aHdr[257], "ustar ", 7); /* Format */


48	gzip_begin();
49	db_multi_exec(
50	"CREATE TEMP TABLE dir(name UNIQUE);"
51	);
52	}












































































































































































































53
54	/*
55	** Build a header for a file or directory and write that header
56	** into the growing tarball.
57	*/
	@@ -59,33 +280,52 @@
59	const char zName, / Name of the object */
60	int nName, /* Number of characters in zName */
61	int iMode, /* Mode. 0644 or 0755 */
62	unsigned int mTime, /* File modification time */
63	int iSize, /* Size of the object in bytes */
64	int iType /* Type of object. 0==file. 5==directory */
65	){
66	unsigned int cksum = 0;
67	int i;
68	if( nName>100 ){
69	memcpy(&tball.aHdr[345], zName, nName-100);
70	memcpy(tball.aHdr, &zName[nName-100], 100);
71	memset(&tball.aHdr[245+nName], 0, 267-nName);
72	}else{
73	memcpy(tball.aHdr, zName, nName);
74	memset(&tball.aHdr[nName], 0, 100-nName);
75	memset(&tball.aHdr[345], 0, 167);
76	}
77	sqlite3_snprintf(8, (char*)&tball.aHdr[100], "%07o", iMode);
78	sqlite3_snprintf(12, (char*)&tball.aHdr[124], "%011o", iSize);
79	sqlite3_snprintf(12, (char*)&tball.aHdr[136], "%011o", mTime);
80	memset(&tball.aHdr[148], ' ', 8);
81	tball.aHdr[156] = iType + '0';
82	for(i=0; i<512; i++) cksum += tball.aHdr[i];
83	sqlite3_snprintf(7, (char*)&tball.aHdr[148], "%06o", cksum);
84	tball.aHdr[154] = 0;
85	gzip_step((char*)tball.aHdr, 512);





























86	}

87
88	/*
89	** Recursively add an directory entry for the given file if those
90	** directories have not previously been seen.
91	*/
	@@ -95,18 +335,27 @@
95	unsigned int mTime /* Modification time */
96	){
97	int i;
98	for(i=nName-1; i>0 && zName[i]!='/'; i--){}
99	if( i<=0 ) return;
100	if( tball.zPrevDir[i]==0 && memcmp(tball.zPrevDir, zName, i)==0 ) return;

101	db_multi_exec("INSERT OR IGNORE INTO dir VALUES('%#q')", i, zName);
102	if( sqlite3_changes(g.db)==0 ) return;
103	tar_add_directory_of(zName, i-1, mTime);
104	tar_add_header(zName, i, 0755, mTime, 0, 5);







105	memcpy(tball.zPrevDir, zName, i);
106	tball.zPrevDir[i] = 0;
107	}

108
109	/*
110	** Add a single file to the growing tarball.
111	*/
112	static void tar_add_file(
	@@ -117,15 +366,13 @@
117	){
118	int nName = strlen(zName);
119	int n = blob_size(pContent);
120	int lastPage;
121
122	if( nName>=250 ){
123	fossil_fatal("name too long for ustar format: \"%s\"", zName);
124	}
125	tar_add_directory_of(zName, nName, mTime);
126	tar_add_header(zName, nName, isExe ? 0755 : 0644, mTime, n, 0);
127	if( n ){
128	gzip_step(blob_buffer(pContent), n);
129	lastPage = n % 512;
130	if( lastPage!=0 ){
131	gzip_step(tball.zSpaces, 512 - lastPage);
	@@ -142,10 +389,14 @@
142	gzip_step(tball.zSpaces, 512);
143	gzip_step(tball.zSpaces, 512);
144	gzip_finish(pOut);
145	fossil_free(tball.aHdr);
146	tball.aHdr = 0;




147	}
148
149
150	/*
151	** COMMAND: test-tarball
152

	--- src/tar.c
	+++ src/tar.c
	@@ -27,31 +27,252 @@
27	*/
28	static struct tarball_t {
29	unsigned char aHdr; / Space for building headers */
30	char zSpaces; / Spaces for padding */
31	char zPrevDir; / Name of directory for previous entry */
32	int nPrevDirAlloc; /* size of zPrevDir */
33	Blob pax; /* PAX data */
34	} tball;
35
36
37	/*
38	** field lengths of 'ustar' name and prefix fields.
39	*/
40	#define USTAR_NAME_LEN 100
41	#define USTAR_PREFIX_LEN 155
42
43
44	/*
45	** Begin the process of generating a tarball.
46	**
47	** Initialize the GZIP compressor and the table of directory names.
48	*/
49	static void tar_begin(void){
50	assert( tball.aHdr==0 );
51	tball.aHdr = fossil_malloc(512+512);
52	memset(tball.aHdr, 0, 512+512);
53	tball.zSpaces = (char*)&tball.aHdr[512];
54	/* zPrevDir init */
55	tball.zPrevDir = NULL;
56	tball.nPrevDirAlloc = 0;
57	/* scratch buffer init */
58	blob_zero(&tball.pax);
59
60	memcpy(&tball.aHdr[108], "0000000", 8); /* Owner ID */
61	memcpy(&tball.aHdr[116], "0000000", 8); /* Group ID */
62	memcpy(&tball.aHdr[257], "ustar\00000", 8); /* POSIX.1 format */
63	memcpy(&tball.aHdr[265], "nobody", 7); /* Owner name */
64	memcpy(&tball.aHdr[297], "nobody", 7); /* Group name */
65	gzip_begin();
66	db_multi_exec(
67	"CREATE TEMP TABLE dir(name UNIQUE);"
68	);
69	}
70
71
72	/*
73	** verify that lla characters in 'zName' are in the
74	** ISO646 (=ASCII) character set.
75	*/
76	static int is_iso646_name(
77	const char zName, / file path */
78	int nName /* path length */
79	){
80	int i;
81	for(i = 0; i < nName; i++){
82	unsigned char c = (unsigned char)zName[i];
83	if( c>0x7e ) return 0;
84	}
85	return 1;
86	}
87
88
89	/*
90	** copy string pSrc into pDst, truncating or padding with 0 if necessary
91	*/
92	static void padded_copy(
93	char *pDest,
94	int nDest,
95	const char *pSrc,
96	int nSrc
97	){
98	if(nSrc >= nDest){
99	memcpy(pDest, pSrc, nDest);
100	}else{
101	memcpy(pDest, pSrc, nSrc);
102	memset(&pDest[nSrc], 0, nDest - nSrc);
103	}
104	}
105
106
107
108	/******************************************************************************
109	**
110	** The 'tar' format has evolved over time. Initially the name was stored
111	** in a 100 byte null-terminated field 'name'. File path names were
112	** limited to 99 bytes.
113	**
114	** The Posix.1 'ustar' format added a 155 byte field 'prefix', allowing
115	** for up to 255 characters to be stored. The full file path is formed by
116	** concatenating the field 'prefix', a slash, and the field 'name'. This
117	** gives some measure of compatibility with programs that only understand
118	** the oldest format.
119	**
120	** The latest Posix extension is called the 'pax Interchange Format'.
121	** It removes all the limitations of the previous two formats by allowing
122	** the storage of arbitrary-length attributes in a separate object that looks
123	** like a file to programs that do not understand this extension. So the
124	** contents of the 'name' and 'prefix' fields should contain values that allow
125	** versions of tar that do not understand this extension to still do
126	** something useful.
127	**
128	******************************************************************************/
129
130	/*
131	** The position we use to split a file path into the 'name' and 'prefix'
132	** fields needs to meet the following criteria:
133	**
134	** - not at the beginning or end of the string
135	** - the position must contain a slash
136	** - no more than 100 characters follow the slash
137	** - no more than 155 characters precede it
138	**
139	** The routine 'find_split_pos' finds a split position. It will meet the
140	** criteria of listed above if such a position exists. If no such
141	** position exists it generates one that useful for generating the
142	** values used for backward compatibility.
143	*/
144	static int find_split_pos(
145	const char zName, / file path */
146	int nName /* path length */
147	){
148	int i, split = 0;
149	/* only search if the string needs splitting */
150	if(nName > USTAR_NAME_LEN){
151	for(i = 1; i+1 < nName; i++)
152	if(zName[i] == '/'){
153	split = i+1;
154	/* if the split position is within USTAR_NAME_LEN bytes from
155	* the end we can quit */
156	if(nName - split <= USTAR_NAME_LEN) break;
157	}
158	}
159	return split;
160	}
161
162
163	/*
164	** attempt to split the file name path to meet 'ustar' header
165	** criteria.
166	*/
167	static int tar_split_path(
168	const char zName, / path */
169	int nName, /* path length */
170	char pName, / name field */
171	char pPrefix / prefix field */
172	){
173	int split = find_split_pos(zName, nName);
174	/* check whether both pieces fit */
175	if(nName - split > USTAR_NAME_LEN \|\| split > USTAR_PREFIX_LEN+1){
176	return 0; /* no */
177	}
178
179	/* extract name */
180	padded_copy(pName, USTAR_NAME_LEN, &zName[split], nName - split);
181
182	/* extract prefix */
183	padded_copy(pPrefix, USTAR_PREFIX_LEN, zName, (split > 0 ? split - 1 : 0));
184
185	return 1; /* success */
186	}
187
188
189	/*
190	** When using an extension header we still need to put something
191	** reasonable in the name and prefix fields. This is probably as
192	** good as it gets.
193	*/
194	static void approximate_split_path(
195	const char zName, / path */
196	int nName, /* path length */
197	char pName, / name field */
198	char pPrefix, / prefix field */
199	int bHeader /* is this a 'x' type tar header? */
200	){
201	int split;
202
203	/* if this is a Pax Interchange header prepend "PaxHeader/"
204	** so we can tell files apart from metadata */
205	if( bHeader ){
206	blob_reset(&tball.pax);
207	blob_appendf(&tball.pax, "PaxHeader/%.s", nName, nName, zName);
208	zName = blob_buffer(&tball.pax);
209	nName = blob_size(&tball.pax);
210	}
211
212	/* find the split position */
213	split = find_split_pos(zName, nName);
214
215	/* extract a name, truncate if needed */
216	padded_copy(pName, USTAR_NAME_LEN, &zName[split], nName - split);
217
218	/* extract a prefix field, truncate when needed */
219	padded_copy(pPrefix, USTAR_PREFIX_LEN, zName, (split > 0 ? split-1 : 0));
220	}
221
222
223	/*
224	** add a Pax Interchange header to the scratch buffer
225	**
226	** format: <length> <key>=<value>\n
227	** the tricky part is that each header contains its own
228	** size in decimal, counting that length.
229	*/
230	static void add_pax_header(
231	const char *zField,
232	const char *zValue,
233	int nValue
234	){
235	/* calculate length without length field */
236	int blen = strlen(zField) + nValue + 3;
237	/* calculate the length of the length field */
238	int next10 = 1;
239	int n;
240	for(n = blen; n > 0; ){
241	blen++; next10 *= 10;
242	n /= 10;
243	}
244	/* adding the length extended the length field? */
245	if(blen > next10){
246	blen++;
247	}
248	/* build the string */
249	blob_appendf(&tball.pax, "%d %s=%.s\n", blen, zField, nValue, nValue, zValue);
250	/* this _must_ be right */
251	if(blob_size(&tball.pax) != blen){
252	fossil_fatal("internal error: PAX tar header has bad length");
253	}
254	}
255
256
257	/*
258	** set the header type, calculate the checksum and output
259	** the header
260	*/
261	static void cksum_and_write_header(
262	char cType
263	){
264	unsigned int cksum = 0;
265	int i;
266	memset(&tball.aHdr[148], ' ', 8);
267	tball.aHdr[156] = cType;
268	for(i=0; i<512; i++) cksum += tball.aHdr[i];
269	sqlite3_snprintf(8, (char*)&tball.aHdr[148], "%07o", cksum);
270	tball.aHdr[155] = 0;
271	gzip_step((char*)tball.aHdr, 512);
272	}
273
274
275	/*
276	** Build a header for a file or directory and write that header
277	** into the growing tarball.
278	*/
	@@ -59,33 +280,52 @@
280	const char zName, / Name of the object */
281	int nName, /* Number of characters in zName */
282	int iMode, /* Mode. 0644 or 0755 */
283	unsigned int mTime, /* File modification time */
284	int iSize, /* Size of the object in bytes */
285	char cType /* Type of object. '0'==file. '5'==directory */
286	){
287	/* set mode and modification time */










288	sqlite3_snprintf(8, (char*)&tball.aHdr[100], "%07o", iMode);

289	sqlite3_snprintf(12, (char*)&tball.aHdr[136], "%011o", mTime);
290
291	/* see if we need to output a Pax Interchange Header */
292	if( !is_iso646_name(zName, nName)
293	\|\| !tar_split_path(zName, nName, (char)tball.aHdr, (char)&tball.aHdr[345])
294	){
295	int lastPage;
296	/* add a file name for interoperability with older programs */
297	approximate_split_path(zName, nName, (char*)tball.aHdr,
298	(char*)&tball.aHdr[345], 1);
299
300	/* generate the Pax Interchange path header */
301	blob_reset(&tball.pax);
302	add_pax_header("path", zName, nName);
303
304	/* set the header length, and write the header */
305	sqlite3_snprintf(12, (char*)&tball.aHdr[124], "%011o",
306	blob_size(&tball.pax));
307	cksum_and_write_header('x');
308
309	/* write the Pax Interchange data */
310	gzip_step(blob_buffer(&tball.pax), blob_size(&tball.pax));
311	lastPage = blob_size(&tball.pax) % 512;
312	if( lastPage!=0 ){
313	gzip_step(tball.zSpaces, 512 - lastPage);
314	}
315
316	/* generate an approximate path for the regular header */
317	approximate_split_path(zName, nName, (char*)tball.aHdr,
318	(char*)&tball.aHdr[345], 0);
319	}
320	/* set the size */
321	sqlite3_snprintf(12, (char*)&tball.aHdr[124], "%011o", iSize);
322
323	/* write the regular header */
324	cksum_and_write_header(cType);
325	}
326
327
328	/*
329	** Recursively add an directory entry for the given file if those
330	** directories have not previously been seen.
331	*/
	@@ -95,18 +335,27 @@
335	unsigned int mTime /* Modification time */
336	){
337	int i;
338	for(i=nName-1; i>0 && zName[i]!='/'; i--){}
339	if( i<=0 ) return;
340	if( i < tball.nPrevDirAlloc && tball.zPrevDir[i]==0 &&
341	memcmp(tball.zPrevDir, zName, i)==0 ) return;
342	db_multi_exec("INSERT OR IGNORE INTO dir VALUES('%#q')", i, zName);
343	if( sqlite3_changes(g.db)==0 ) return;
344	tar_add_directory_of(zName, i-1, mTime);
345	tar_add_header(zName, i, 0755, mTime, 0, '5');
346	if( i >= tball.nPrevDirAlloc ){
347	int nsize = tball.nPrevDirAlloc * 2;
348	if(i+1 > nsize)
349	nsize = i+1;
350	tball.zPrevDir = fossil_realloc(tball.zPrevDir, nsize);
351	tball.nPrevDirAlloc = nsize;
352	}
353	memcpy(tball.zPrevDir, zName, i);
354	tball.zPrevDir[i] = 0;
355	}
356
357
358	/*
359	** Add a single file to the growing tarball.
360	*/
361	static void tar_add_file(
	@@ -117,15 +366,13 @@
366	){
367	int nName = strlen(zName);
368	int n = blob_size(pContent);
369	int lastPage;
370
371	/* length check moved to tar_split_path */


372	tar_add_directory_of(zName, nName, mTime);
373	tar_add_header(zName, nName, isExe ? 0755 : 0644, mTime, n, '0');
374	if( n ){
375	gzip_step(blob_buffer(pContent), n);
376	lastPage = n % 512;
377	if( lastPage!=0 ){
378	gzip_step(tball.zSpaces, 512 - lastPage);
	@@ -142,10 +389,14 @@
389	gzip_step(tball.zSpaces, 512);
390	gzip_step(tball.zSpaces, 512);
391	gzip_finish(pOut);
392	fossil_free(tball.aHdr);
393	tball.aHdr = 0;
394	fossil_free(tball.zPrevDir);
395	tball.zPrevDir = NULL;
396	tball.nPrevDirAlloc = 0;
397	blob_reset(&tball.pax);
398	}
399
400
401	/*
402	** COMMAND: test-tarball
403

M src/timeline.c

+136 -20

		--- src/timeline.c
		+++ src/timeline.c
		@@ -107,11 +107,70 @@
107	107	#define TIMELINE_LEAFONLY 0x0002 /* Show "Leaf", but not "Merge", "Fork" etc */
108	108	#define TIMELINE_BRIEF 0x0004 /* Combine adjacent elements of same object */
109	109	#define TIMELINE_GRAPH 0x0008 /* Compute a graph */
110	110	#define TIMELINE_DISJOINT 0x0010 /* Elements are not contiguous */
111	111	#define TIMELINE_FCHANGES 0x0020 /* Detail file changes */
	112	+#define TIMELINE_BRCOLOR 0x0040 /* Background color by branch name */
	113	+#define TIMELINE_UCOLOR 0x0080 /* Background color by user */
112	114	#endif
	115	+
	116	+/*
	117	+** Hash a string and use the hash to determine a background color.
	118	+*/
	119	+char hash_color(const char z){
	120	+ int i; /* Loop counter */
	121	+ unsigned int h = 0; /* Hash on the branch name */
	122	+ int r, g, b; /* Values for red, green, and blue */
	123	+ int h1, h2, h3, h4; /* Elements of the hash value */
	124	+ int mx, mn; /* Components of HSV */
	125	+ static char zColor[10]; /* The resulting color */
	126	+ static int ix[2] = {0,0}; /* Color chooser parameters */
	127	+
	128	+ if( ix[0]==0 ){
	129	+ if( db_get_boolean("white-foreground", 0) ){
	130	+ ix[0] = 140;
	131	+ ix[1] = 40;
	132	+ }else{
	133	+ ix[0] = 216;
	134	+ ix[1] = 16;
	135	+ }
	136	+ }
	137	+ for(i=0; z[i]; i++ ){
	138	+ h = (h<<11) ^ (h<<1) ^ (h>>3) ^ z[i];
	139	+ }
	140	+ h1 = h % 6; h /= 6;
	141	+ h3 = h % 30; h /= 30;
	142	+ h4 = h % 40; h /= 40;
	143	+ mx = ix[0] - h3;
	144	+ mn = mx - h4 - ix[1];
	145	+ h2 = (h%(mx - mn)) + mn;
	146	+ switch( h1 ){
	147	+ case 0: r = mx; g = h2, b = mn; break;
	148	+ case 1: r = h2; g = mx, b = mn; break;
	149	+ case 2: r = mn; g = mx, b = h2; break;
	150	+ case 3: r = mn; g = h2, b = mx; break;
	151	+ case 4: r = h2; g = mn, b = mx; break;
	152	+ default: r = mx; g = mn, b = h2; break;
	153	+ }
	154	+ sqlite3_snprintf(8, zColor, "#%02x%02x%02x", r,g,b);
	155	+ return zColor;
	156	+}
	157	+
	158	+/*
	159	+** COMMAND: test-hash-color
	160	+**
	161	+** Usage: %fossil test-hash-color TAG ...
	162	+**
	163	+** Print out the color names associated with each tag. Used for
	164	+** testing the hash_color() function.
	165	+*/
	166	+void test_hash_color(void){
	167	+ int i;
	168	+ for(i=2; i<g.argc; i++){
	169	+ fossil_print("%20s: %s\n", g.argv[i], hash_color(g.argv[i]));
	170	+ }
	171	+}
113	172
114	173	/*
115	174	** Output a timeline in the web format given a query. The query
116	175	** should return these columns:
117	176	**
		@@ -142,10 +201,11 @@
142	201	char zPrevDate[20];
143	202	GraphContext *pGraph = 0;
144	203	int prevWasDivider = 0; /* True if previous output row was <hr> */
145	204	int fchngQueryInit = 0; /* True if fchngQuery is initialized */
146	205	Stmt fchngQuery; /* Query for file changes on check-ins */
	206	+ static Stmt qbranch;
147	207
148	208	zPrevDate[0] = 0;
149	209	mxWikiLen = db_get_int("timeline-max-comment", 0);
150	210	if( db_get_boolean("timeline-block-markup", 0) ){
151	211	wikiFlags = WIKI_INLINE;
		@@ -157,10 +217,14 @@
157	217	/* style is not moved to css, because this is
158	218	** a technical div for the timeline graph
159	219	*/
160	220	@ <div id="canvas" style="position:relative;width:1px;height:1px;"></div>
161	221	}
	222	+ db_static_prepare(&qbranch,
	223	+ "SELECT value FROM tagxref WHERE tagid=%d AND tagtype>0 AND rid=:rid",
	224	+ TAG_BRANCH
	225	+ );
162	226
163	227	@ <table id="timelineTable" class="timelineTable">
164	228	blob_zero(&comment);
165	229	while( db_step(pQuery)==SQLITE_ROW ){
166	230	int rid = db_column_int(pQuery, 0);
		@@ -170,10 +234,11 @@
170	234	const char *zDate = db_column_text(pQuery, 2);
171	235	const char *zType = db_column_text(pQuery, 7);
172	236	const char *zUser = db_column_text(pQuery, 4);
173	237	const char *zTagList = db_column_text(pQuery, 8);
174	238	int tagid = db_column_int(pQuery, 9);
	239	+ const char zBr = 0; / Branch */
175	240	int commentColumn = 3; /* Column containing comment text */
176	241	char zTime[8];
177	242	if( tagid ){
178	243	if( tagid==prevTagid ){
179	244	if( tmFlags & TIMELINE_BRIEF ){
		@@ -208,37 +273,44 @@
208	273	memcpy(zTime, &zDate[11], 5);
209	274	zTime[5] = 0;
210	275	@ <tr>
211	276	@ <td class="timelineTime">%s(zTime)</td>
212	277	@ <td class="timelineGraph">
213		- if( pGraph && zType[0]=='c' ){
	278	+ if( tmFlags & TIMELINE_UCOLOR ) zBgClr = zUser ? hash_color(zUser) : 0;
	279	+ if( zType[0]=='c'
	280	+ && (pGraph \|\| zBgClr==0 \|\| (tmFlags & TIMELINE_BRCOLOR)!=0)
	281	+ ){
	282	+ db_reset(&qbranch);
	283	+ db_bind_int(&qbranch, ":rid", rid);
	284	+ if( db_step(&qbranch)==SQLITE_ROW ){
	285	+ zBr = db_column_text(&qbranch, 0);
	286	+ }else{
	287	+ zBr = "trunk";
	288	+ }
	289	+ if( zBgClr==0 \|\| (tmFlags & TIMELINE_BRCOLOR)!=0 ){
	290	+ if( zBr==0 \|\| strcmp(zBr,"trunk")==0 ){
	291	+ zBgClr = 0;
	292	+ }else{
	293	+ zBgClr = hash_color(zBr);
	294	+ }
	295	+ }
	296	+ }
	297	+ if( zType[0]=='c' && (pGraph \|\| (tmFlags & TIMELINE_BRCOLOR)!=0) ){
214	298	int nParent = 0;
215	299	int aParent[32];
216		- const char *zBr;
217	300	int gidx;
218	301	static Stmt qparent;
219		- static Stmt qbranch;
220	302	db_static_prepare(&qparent,
221	303	"SELECT pid FROM plink"
222	304	" WHERE cid=:rid AND pid NOT IN phantom"
223	305	" ORDER BY isprim DESC /sort/"
224	306	);
225		- db_static_prepare(&qbranch,
226		- "SELECT value FROM tagxref WHERE tagid=%d AND tagtype>0 AND rid=:rid",
227		- TAG_BRANCH
228		- );
229	307	db_bind_int(&qparent, ":rid", rid);
230	308	while( db_step(&qparent)==SQLITE_ROW && nParent<32 ){
231	309	aParent[nParent++] = db_column_int(&qparent, 0);
232	310	}
233	311	db_reset(&qparent);
234		- db_bind_int(&qbranch, ":rid", rid);
235		- if( db_step(&qbranch)==SQLITE_ROW ){
236		- zBr = db_column_text(&qbranch, 0);
237		- }else{
238		- zBr = "trunk";
239		- }
240	312	gidx = graph_add_row(pGraph, rid, nParent, aParent, zBr, zBgClr, isLeaf);
241	313	db_reset(&qbranch);
242	314	@ <div id="m%d(gidx)"></div>
243	315	}
244	316	@</td>
		@@ -397,11 +469,11 @@
397	469	/*
398	470	** Generate all of the necessary javascript to generate a timeline
399	471	** graph.
400	472	*/
401	473	void timeline_output_graph_javascript(GraphContext *pGraph, int omitDescenders){
402		- if( pGraph && pGraph->nErr==0 ){
	474	+ if( pGraph && pGraph->nErr==0 && pGraph->nRow>0 ){
403	475	GraphRow *pRow;
404	476	int i;
405	477	char cSep;
406	478	@ <script type="text/JavaScript">
407	479	@ /* <![CDATA[ */
		@@ -768,10 +840,12 @@
768	840	** fc Show details of files changed
769	841	** f=RID Show family (immediate parents and children) of RID
770	842	** from=RID Path from...
771	843	** to=RID ... to this
772	844	** nomerge ... avoid merge links on the path
	845	+** brbg Background color from branch name
	846	+** ubg Background color from user
773	847	**
774	848	** p= and d= can appear individually or together. If either p= or d=
775	849	** appear, then u=, y=, a=, and b= are ignored.
776	850	**
777	851	** If a= and b= appear, only a= is used. If neither appear, the most
		@@ -826,10 +900,12 @@
826	900	tmFlags = TIMELINE_GRAPH;
827	901	}
828	902	if( P("ng")!=0 \|\| zSearch!=0 ){
829	903	tmFlags &= ~TIMELINE_GRAPH;
830	904	}
	905	+ if( P("brbg")!=0 ) tmFlags \|= TIMELINE_BRCOLOR;
	906	+ if( P("ubg")!=0 ) tmFlags \|= TIMELINE_UCOLOR;
831	907
832	908	style_header("Timeline");
833	909	login_anonymous_available();
834	910	timeline_temp_table();
835	911	blob_zero(&sql);
		@@ -1020,11 +1096,12 @@
1020	1096	}else if( zType[0]=='e' ){
1021	1097	zEType = "event";
1022	1098	}
1023	1099	}
1024	1100	if( zUser ){
1025		- blob_appendf(&sql, " AND event.user=%Q", zUser);
	1101	+ blob_appendf(&sql, " AND (event.user=%Q OR event.euser=%Q)",
	1102	+ zUser, zUser);
1026	1103	url_add_parameter(&url, "u", zUser);
1027	1104	zThisUser = zUser;
1028	1105	}
1029	1106	if ( zSearch ){
1030	1107	blob_appendf(&sql,
		@@ -1178,14 +1255,16 @@
1178	1255	** 2. Date/Time
1179	1256	** 3. Comment string and user
1180	1257	** 4. Number of non-merge children
1181	1258	** 5. Number of parents
1182	1259	*/
1183		-void print_timeline(Stmt *q, int mxLine){
	1260	+void print_timeline(Stmt *q, int mxLine, int showfiles){
1184	1261	int nLine = 0;
1185	1262	char zPrevDate[20];
1186	1263	const char *zCurrentUuid=0;
	1264	+ int fchngQueryInit = 0; /* True if fchngQuery is initialized */
	1265	+ Stmt fchngQuery; /* Query for file changes on check-ins */
1187	1266	zPrevDate[0] = 0;
1188	1267
1189	1268	if( g.localOpen ){
1190	1269	int rid = db_lget_int("checkout", 0);
1191	1270	zCurrentUuid = db_text(0, "SELECT uuid FROM blob WHERE rid=%d", rid);
		@@ -1231,11 +1310,42 @@
1231	1310	n += strlen(zPrefix);
1232	1311	}
1233	1312	zFree = sqlite3_mprintf("[%.10s] %s%s", zUuid, zPrefix, zCom);
1234	1313	nLine += comment_print(zFree, 9, 79);
1235	1314	sqlite3_free(zFree);
	1315	+
	1316	+ if(showfiles){
	1317	+ if( !fchngQueryInit ){
	1318	+ db_prepare(&fchngQuery,
	1319	+ "SELECT (pid==0) AS isnew,"
	1320	+ " (fid==0) AS isdel,"
	1321	+ " (SELECT name FROM filename WHERE fnid=mlink.fnid) AS name,"
	1322	+ " (SELECT uuid FROM blob WHERE rid=fid),"
	1323	+ " (SELECT uuid FROM blob WHERE rid=pid)"
	1324	+ " FROM mlink"
	1325	+ " WHERE mid=:mid AND pid!=fid"
	1326	+ " ORDER BY 3 /sort/"
	1327	+ );
	1328	+ fchngQueryInit = 1;
	1329	+ }
	1330	+ db_bind_int(&fchngQuery, ":mid", rid);
	1331	+ while( db_step(&fchngQuery)==SQLITE_ROW ){
	1332	+ const char *zFilename = db_column_text(&fchngQuery, 2);
	1333	+ int isNew = db_column_int(&fchngQuery, 0);
	1334	+ int isDel = db_column_int(&fchngQuery, 1);
	1335	+ if( isNew ){
	1336	+ fossil_print(" ADDED %s\n",zFilename);
	1337	+ }else if( isDel ){
	1338	+ fossil_print(" DELETED %s\n",zFilename);
	1339	+ }else{
	1340	+ fossil_print(" EDITED %s\n", zFilename);
	1341	+ }
	1342	+ }
	1343	+ db_reset(&fchngQuery);
	1344	+ }
1236	1345	}
	1346	+ if( fchngQueryInit ) db_finalize(&fchngQuery);
1237	1347	}
1238	1348
1239	1349	/*
1240	1350	** Return a pointer to a static string that forms the basis for
1241	1351	** a timeline query for display on a TTY.
		@@ -1276,11 +1386,11 @@
1276	1386	}
1277	1387
1278	1388	/*
1279	1389	** COMMAND: timeline
1280	1390	**
1281		-** Usage: %fossil timeline ?WHEN? ?BASELINE\|DATETIME? ?-n N? ?-t TYPE?
	1391	+** Usage: %fossil timeline ?WHEN? ?BASELINE\|DATETIME? ?-n N? ?-t TYPE? ?-showfiles?
1282	1392	**
1283	1393	** Print a summary of activity going backwards in date and time
1284	1394	** specified or from the current date and time if no arguments
1285	1395	** are given. Show as many as N (default 20) check-ins. The
1286	1396	** WHEN argument can be any unique abbreviation of one of these
		@@ -1300,10 +1410,14 @@
1300	1410	** page. For example:
1301	1411	**
1302	1412	** w = wiki commits only
1303	1413	** ci = file commits only
1304	1414	** t = tickets only
	1415	+**
	1416	+** The optional showfiles argument if specified prints the list of
	1417	+** files changed in a checkin after the checkin comment
	1418	+**
1305	1419	*/
1306	1420	void timeline_cmd(void){
1307	1421	Stmt q;
1308	1422	int n, k;
1309	1423	const char *zCount;
		@@ -1312,10 +1426,12 @@
1312	1426	char *zDate;
1313	1427	Blob sql;
1314	1428	int objid = 0;
1315	1429	Blob uuid;
1316	1430	int mode = 0 ; /* 0:none 1: before 2:after 3:children 4:parents */
	1431	+ int showfilesFlag = 0 ;
	1432	+ showfilesFlag = find_option("showfiles","f", 0)!=0;
1317	1433	db_find_and_open_repository(0, 0);
1318	1434	zCount = find_option("count","n",1);
1319	1435	zType = find_option("type","t",1);
1320	1436	if( zCount ){
1321	1437	n = atoi(zCount);
		@@ -1338,13 +1454,13 @@
1338	1454	mode = 4;
1339	1455	}else if(!zType && !zCount){
1340	1456	usage("?WHEN? ?BASELINE\|DATETIME? ?-n\|--count N? ?-t TYPE?");
1341	1457	}
1342	1458	if( '-' != *g.argv[3] ){
1343		- zOrigin = g.argv[3];
	1459	+ zOrigin = g.argv[3];
1344	1460	}else{
1345		- zOrigin = "now";
	1461	+ zOrigin = "now";
1346	1462	}
1347	1463	}else if( g.argc==3 ){
1348	1464	zOrigin = g.argv[2];
1349	1465	}else{
1350	1466	zOrigin = "now";
		@@ -1398,11 +1514,11 @@
1398	1514	}
1399	1515
1400	1516	blob_appendf(&sql, " ORDER BY event.mtime DESC");
1401	1517	db_prepare(&q, blob_str(&sql));
1402	1518	blob_reset(&sql);
1403		- print_timeline(&q, n);
	1519	+ print_timeline(&q, n, showfilesFlag);
1404	1520	db_finalize(&q);
1405	1521	}
1406	1522
1407	1523	/*
1408	1524	** This is a version of the "localtime()" function from the standard
1409	1525

	--- src/timeline.c
	+++ src/timeline.c
	@@ -107,11 +107,70 @@
107	#define TIMELINE_LEAFONLY 0x0002 /* Show "Leaf", but not "Merge", "Fork" etc */
108	#define TIMELINE_BRIEF 0x0004 /* Combine adjacent elements of same object */
109	#define TIMELINE_GRAPH 0x0008 /* Compute a graph */
110	#define TIMELINE_DISJOINT 0x0010 /* Elements are not contiguous */
111	#define TIMELINE_FCHANGES 0x0020 /* Detail file changes */


112	#endif

























































113
114	/*
115	** Output a timeline in the web format given a query. The query
116	** should return these columns:
117	**
	@@ -142,10 +201,11 @@
142	char zPrevDate[20];
143	GraphContext *pGraph = 0;
144	int prevWasDivider = 0; /* True if previous output row was <hr> */
145	int fchngQueryInit = 0; /* True if fchngQuery is initialized */
146	Stmt fchngQuery; /* Query for file changes on check-ins */

147
148	zPrevDate[0] = 0;
149	mxWikiLen = db_get_int("timeline-max-comment", 0);
150	if( db_get_boolean("timeline-block-markup", 0) ){
151	wikiFlags = WIKI_INLINE;
	@@ -157,10 +217,14 @@
157	/* style is not moved to css, because this is
158	** a technical div for the timeline graph
159	*/
160	@ <div id="canvas" style="position:relative;width:1px;height:1px;"></div>
161	}




162
163	@ <table id="timelineTable" class="timelineTable">
164	blob_zero(&comment);
165	while( db_step(pQuery)==SQLITE_ROW ){
166	int rid = db_column_int(pQuery, 0);
	@@ -170,10 +234,11 @@
170	const char *zDate = db_column_text(pQuery, 2);
171	const char *zType = db_column_text(pQuery, 7);
172	const char *zUser = db_column_text(pQuery, 4);
173	const char *zTagList = db_column_text(pQuery, 8);
174	int tagid = db_column_int(pQuery, 9);

175	int commentColumn = 3; /* Column containing comment text */
176	char zTime[8];
177	if( tagid ){
178	if( tagid==prevTagid ){
179	if( tmFlags & TIMELINE_BRIEF ){
	@@ -208,37 +273,44 @@
208	memcpy(zTime, &zDate[11], 5);
209	zTime[5] = 0;
210	@ <tr>
211	@ <td class="timelineTime">%s(zTime)</td>
212	@ <td class="timelineGraph">
213	if( pGraph && zType[0]=='c' ){



















214	int nParent = 0;
215	int aParent[32];
216	const char *zBr;
217	int gidx;
218	static Stmt qparent;
219	static Stmt qbranch;
220	db_static_prepare(&qparent,
221	"SELECT pid FROM plink"
222	" WHERE cid=:rid AND pid NOT IN phantom"
223	" ORDER BY isprim DESC /sort/"
224	);
225	db_static_prepare(&qbranch,
226	"SELECT value FROM tagxref WHERE tagid=%d AND tagtype>0 AND rid=:rid",
227	TAG_BRANCH
228	);
229	db_bind_int(&qparent, ":rid", rid);
230	while( db_step(&qparent)==SQLITE_ROW && nParent<32 ){
231	aParent[nParent++] = db_column_int(&qparent, 0);
232	}
233	db_reset(&qparent);
234	db_bind_int(&qbranch, ":rid", rid);
235	if( db_step(&qbranch)==SQLITE_ROW ){
236	zBr = db_column_text(&qbranch, 0);
237	}else{
238	zBr = "trunk";
239	}
240	gidx = graph_add_row(pGraph, rid, nParent, aParent, zBr, zBgClr, isLeaf);
241	db_reset(&qbranch);
242	@ <div id="m%d(gidx)"></div>
243	}
244	@</td>
	@@ -397,11 +469,11 @@
397	/*
398	** Generate all of the necessary javascript to generate a timeline
399	** graph.
400	*/
401	void timeline_output_graph_javascript(GraphContext *pGraph, int omitDescenders){
402	if( pGraph && pGraph->nErr==0 ){
403	GraphRow *pRow;
404	int i;
405	char cSep;
406	@ <script type="text/JavaScript">
407	@ /* <![CDATA[ */
	@@ -768,10 +840,12 @@
768	** fc Show details of files changed
769	** f=RID Show family (immediate parents and children) of RID
770	** from=RID Path from...
771	** to=RID ... to this
772	** nomerge ... avoid merge links on the path


773	**
774	** p= and d= can appear individually or together. If either p= or d=
775	** appear, then u=, y=, a=, and b= are ignored.
776	**
777	** If a= and b= appear, only a= is used. If neither appear, the most
	@@ -826,10 +900,12 @@
826	tmFlags = TIMELINE_GRAPH;
827	}
828	if( P("ng")!=0 \|\| zSearch!=0 ){
829	tmFlags &= ~TIMELINE_GRAPH;
830	}


831
832	style_header("Timeline");
833	login_anonymous_available();
834	timeline_temp_table();
835	blob_zero(&sql);
	@@ -1020,11 +1096,12 @@
1020	}else if( zType[0]=='e' ){
1021	zEType = "event";
1022	}
1023	}
1024	if( zUser ){
1025	blob_appendf(&sql, " AND event.user=%Q", zUser);

1026	url_add_parameter(&url, "u", zUser);
1027	zThisUser = zUser;
1028	}
1029	if ( zSearch ){
1030	blob_appendf(&sql,
	@@ -1178,14 +1255,16 @@
1178	** 2. Date/Time
1179	** 3. Comment string and user
1180	** 4. Number of non-merge children
1181	** 5. Number of parents
1182	*/
1183	void print_timeline(Stmt *q, int mxLine){
1184	int nLine = 0;
1185	char zPrevDate[20];
1186	const char *zCurrentUuid=0;


1187	zPrevDate[0] = 0;
1188
1189	if( g.localOpen ){
1190	int rid = db_lget_int("checkout", 0);
1191	zCurrentUuid = db_text(0, "SELECT uuid FROM blob WHERE rid=%d", rid);
	@@ -1231,11 +1310,42 @@
1231	n += strlen(zPrefix);
1232	}
1233	zFree = sqlite3_mprintf("[%.10s] %s%s", zUuid, zPrefix, zCom);
1234	nLine += comment_print(zFree, 9, 79);
1235	sqlite3_free(zFree);






























1236	}

1237	}
1238
1239	/*
1240	** Return a pointer to a static string that forms the basis for
1241	** a timeline query for display on a TTY.
	@@ -1276,11 +1386,11 @@
1276	}
1277
1278	/*
1279	** COMMAND: timeline
1280	**
1281	** Usage: %fossil timeline ?WHEN? ?BASELINE\|DATETIME? ?-n N? ?-t TYPE?
1282	**
1283	** Print a summary of activity going backwards in date and time
1284	** specified or from the current date and time if no arguments
1285	** are given. Show as many as N (default 20) check-ins. The
1286	** WHEN argument can be any unique abbreviation of one of these
	@@ -1300,10 +1410,14 @@
1300	** page. For example:
1301	**
1302	** w = wiki commits only
1303	** ci = file commits only
1304	** t = tickets only




1305	*/
1306	void timeline_cmd(void){
1307	Stmt q;
1308	int n, k;
1309	const char *zCount;
	@@ -1312,10 +1426,12 @@
1312	char *zDate;
1313	Blob sql;
1314	int objid = 0;
1315	Blob uuid;
1316	int mode = 0 ; /* 0:none 1: before 2:after 3:children 4:parents */


1317	db_find_and_open_repository(0, 0);
1318	zCount = find_option("count","n",1);
1319	zType = find_option("type","t",1);
1320	if( zCount ){
1321	n = atoi(zCount);
	@@ -1338,13 +1454,13 @@
1338	mode = 4;
1339	}else if(!zType && !zCount){
1340	usage("?WHEN? ?BASELINE\|DATETIME? ?-n\|--count N? ?-t TYPE?");
1341	}
1342	if( '-' != *g.argv[3] ){
1343	zOrigin = g.argv[3];
1344	}else{
1345	zOrigin = "now";
1346	}
1347	}else if( g.argc==3 ){
1348	zOrigin = g.argv[2];
1349	}else{
1350	zOrigin = "now";
	@@ -1398,11 +1514,11 @@
1398	}
1399
1400	blob_appendf(&sql, " ORDER BY event.mtime DESC");
1401	db_prepare(&q, blob_str(&sql));
1402	blob_reset(&sql);
1403	print_timeline(&q, n);
1404	db_finalize(&q);
1405	}
1406
1407	/*
1408	** This is a version of the "localtime()" function from the standard
1409

	--- src/timeline.c
	+++ src/timeline.c
	@@ -107,11 +107,70 @@
107	#define TIMELINE_LEAFONLY 0x0002 /* Show "Leaf", but not "Merge", "Fork" etc */
108	#define TIMELINE_BRIEF 0x0004 /* Combine adjacent elements of same object */
109	#define TIMELINE_GRAPH 0x0008 /* Compute a graph */
110	#define TIMELINE_DISJOINT 0x0010 /* Elements are not contiguous */
111	#define TIMELINE_FCHANGES 0x0020 /* Detail file changes */
112	#define TIMELINE_BRCOLOR 0x0040 /* Background color by branch name */
113	#define TIMELINE_UCOLOR 0x0080 /* Background color by user */
114	#endif
115
116	/*
117	** Hash a string and use the hash to determine a background color.
118	*/
119	char hash_color(const char z){
120	int i; /* Loop counter */
121	unsigned int h = 0; /* Hash on the branch name */
122	int r, g, b; /* Values for red, green, and blue */
123	int h1, h2, h3, h4; /* Elements of the hash value */
124	int mx, mn; /* Components of HSV */
125	static char zColor[10]; /* The resulting color */
126	static int ix[2] = {0,0}; /* Color chooser parameters */
127
128	if( ix[0]==0 ){
129	if( db_get_boolean("white-foreground", 0) ){
130	ix[0] = 140;
131	ix[1] = 40;
132	}else{
133	ix[0] = 216;
134	ix[1] = 16;
135	}
136	}
137	for(i=0; z[i]; i++ ){
138	h = (h<<11) ^ (h<<1) ^ (h>>3) ^ z[i];
139	}
140	h1 = h % 6; h /= 6;
141	h3 = h % 30; h /= 30;
142	h4 = h % 40; h /= 40;
143	mx = ix[0] - h3;
144	mn = mx - h4 - ix[1];
145	h2 = (h%(mx - mn)) + mn;
146	switch( h1 ){
147	case 0: r = mx; g = h2, b = mn; break;
148	case 1: r = h2; g = mx, b = mn; break;
149	case 2: r = mn; g = mx, b = h2; break;
150	case 3: r = mn; g = h2, b = mx; break;
151	case 4: r = h2; g = mn, b = mx; break;
152	default: r = mx; g = mn, b = h2; break;
153	}
154	sqlite3_snprintf(8, zColor, "#%02x%02x%02x", r,g,b);
155	return zColor;
156	}
157
158	/*
159	** COMMAND: test-hash-color
160	**
161	** Usage: %fossil test-hash-color TAG ...
162	**
163	** Print out the color names associated with each tag. Used for
164	** testing the hash_color() function.
165	*/
166	void test_hash_color(void){
167	int i;
168	for(i=2; i<g.argc; i++){
169	fossil_print("%20s: %s\n", g.argv[i], hash_color(g.argv[i]));
170	}
171	}
172
173	/*
174	** Output a timeline in the web format given a query. The query
175	** should return these columns:
176	**
	@@ -142,10 +201,11 @@
201	char zPrevDate[20];
202	GraphContext *pGraph = 0;
203	int prevWasDivider = 0; /* True if previous output row was <hr> */
204	int fchngQueryInit = 0; /* True if fchngQuery is initialized */
205	Stmt fchngQuery; /* Query for file changes on check-ins */
206	static Stmt qbranch;
207
208	zPrevDate[0] = 0;
209	mxWikiLen = db_get_int("timeline-max-comment", 0);
210	if( db_get_boolean("timeline-block-markup", 0) ){
211	wikiFlags = WIKI_INLINE;
	@@ -157,10 +217,14 @@
217	/* style is not moved to css, because this is
218	** a technical div for the timeline graph
219	*/
220	@ <div id="canvas" style="position:relative;width:1px;height:1px;"></div>
221	}
222	db_static_prepare(&qbranch,
223	"SELECT value FROM tagxref WHERE tagid=%d AND tagtype>0 AND rid=:rid",
224	TAG_BRANCH
225	);
226
227	@ <table id="timelineTable" class="timelineTable">
228	blob_zero(&comment);
229	while( db_step(pQuery)==SQLITE_ROW ){
230	int rid = db_column_int(pQuery, 0);
	@@ -170,10 +234,11 @@
234	const char *zDate = db_column_text(pQuery, 2);
235	const char *zType = db_column_text(pQuery, 7);
236	const char *zUser = db_column_text(pQuery, 4);
237	const char *zTagList = db_column_text(pQuery, 8);
238	int tagid = db_column_int(pQuery, 9);
239	const char zBr = 0; / Branch */
240	int commentColumn = 3; /* Column containing comment text */
241	char zTime[8];
242	if( tagid ){
243	if( tagid==prevTagid ){
244	if( tmFlags & TIMELINE_BRIEF ){
	@@ -208,37 +273,44 @@
273	memcpy(zTime, &zDate[11], 5);
274	zTime[5] = 0;
275	@ <tr>
276	@ <td class="timelineTime">%s(zTime)</td>
277	@ <td class="timelineGraph">
278	if( tmFlags & TIMELINE_UCOLOR ) zBgClr = zUser ? hash_color(zUser) : 0;
279	if( zType[0]=='c'
280	&& (pGraph \|\| zBgClr==0 \|\| (tmFlags & TIMELINE_BRCOLOR)!=0)
281	){
282	db_reset(&qbranch);
283	db_bind_int(&qbranch, ":rid", rid);
284	if( db_step(&qbranch)==SQLITE_ROW ){
285	zBr = db_column_text(&qbranch, 0);
286	}else{
287	zBr = "trunk";
288	}
289	if( zBgClr==0 \|\| (tmFlags & TIMELINE_BRCOLOR)!=0 ){
290	if( zBr==0 \|\| strcmp(zBr,"trunk")==0 ){
291	zBgClr = 0;
292	}else{
293	zBgClr = hash_color(zBr);
294	}
295	}
296	}
297	if( zType[0]=='c' && (pGraph \|\| (tmFlags & TIMELINE_BRCOLOR)!=0) ){
298	int nParent = 0;
299	int aParent[32];

300	int gidx;
301	static Stmt qparent;

302	db_static_prepare(&qparent,
303	"SELECT pid FROM plink"
304	" WHERE cid=:rid AND pid NOT IN phantom"
305	" ORDER BY isprim DESC /sort/"
306	);




307	db_bind_int(&qparent, ":rid", rid);
308	while( db_step(&qparent)==SQLITE_ROW && nParent<32 ){
309	aParent[nParent++] = db_column_int(&qparent, 0);
310	}
311	db_reset(&qparent);






312	gidx = graph_add_row(pGraph, rid, nParent, aParent, zBr, zBgClr, isLeaf);
313	db_reset(&qbranch);
314	@ <div id="m%d(gidx)"></div>
315	}
316	@</td>
	@@ -397,11 +469,11 @@
469	/*
470	** Generate all of the necessary javascript to generate a timeline
471	** graph.
472	*/
473	void timeline_output_graph_javascript(GraphContext *pGraph, int omitDescenders){
474	if( pGraph && pGraph->nErr==0 && pGraph->nRow>0 ){
475	GraphRow *pRow;
476	int i;
477	char cSep;
478	@ <script type="text/JavaScript">
479	@ /* <![CDATA[ */
	@@ -768,10 +840,12 @@
840	** fc Show details of files changed
841	** f=RID Show family (immediate parents and children) of RID
842	** from=RID Path from...
843	** to=RID ... to this
844	** nomerge ... avoid merge links on the path
845	** brbg Background color from branch name
846	** ubg Background color from user
847	**
848	** p= and d= can appear individually or together. If either p= or d=
849	** appear, then u=, y=, a=, and b= are ignored.
850	**
851	** If a= and b= appear, only a= is used. If neither appear, the most
	@@ -826,10 +900,12 @@
900	tmFlags = TIMELINE_GRAPH;
901	}
902	if( P("ng")!=0 \|\| zSearch!=0 ){
903	tmFlags &= ~TIMELINE_GRAPH;
904	}
905	if( P("brbg")!=0 ) tmFlags \|= TIMELINE_BRCOLOR;
906	if( P("ubg")!=0 ) tmFlags \|= TIMELINE_UCOLOR;
907
908	style_header("Timeline");
909	login_anonymous_available();
910	timeline_temp_table();
911	blob_zero(&sql);
	@@ -1020,11 +1096,12 @@
1096	}else if( zType[0]=='e' ){
1097	zEType = "event";
1098	}
1099	}
1100	if( zUser ){
1101	blob_appendf(&sql, " AND (event.user=%Q OR event.euser=%Q)",
1102	zUser, zUser);
1103	url_add_parameter(&url, "u", zUser);
1104	zThisUser = zUser;
1105	}
1106	if ( zSearch ){
1107	blob_appendf(&sql,
	@@ -1178,14 +1255,16 @@
1255	** 2. Date/Time
1256	** 3. Comment string and user
1257	** 4. Number of non-merge children
1258	** 5. Number of parents
1259	*/
1260	void print_timeline(Stmt *q, int mxLine, int showfiles){
1261	int nLine = 0;
1262	char zPrevDate[20];
1263	const char *zCurrentUuid=0;
1264	int fchngQueryInit = 0; /* True if fchngQuery is initialized */
1265	Stmt fchngQuery; /* Query for file changes on check-ins */
1266	zPrevDate[0] = 0;
1267
1268	if( g.localOpen ){
1269	int rid = db_lget_int("checkout", 0);
1270	zCurrentUuid = db_text(0, "SELECT uuid FROM blob WHERE rid=%d", rid);
	@@ -1231,11 +1310,42 @@
1310	n += strlen(zPrefix);
1311	}
1312	zFree = sqlite3_mprintf("[%.10s] %s%s", zUuid, zPrefix, zCom);
1313	nLine += comment_print(zFree, 9, 79);
1314	sqlite3_free(zFree);
1315
1316	if(showfiles){
1317	if( !fchngQueryInit ){
1318	db_prepare(&fchngQuery,
1319	"SELECT (pid==0) AS isnew,"
1320	" (fid==0) AS isdel,"
1321	" (SELECT name FROM filename WHERE fnid=mlink.fnid) AS name,"
1322	" (SELECT uuid FROM blob WHERE rid=fid),"
1323	" (SELECT uuid FROM blob WHERE rid=pid)"
1324	" FROM mlink"
1325	" WHERE mid=:mid AND pid!=fid"
1326	" ORDER BY 3 /sort/"
1327	);
1328	fchngQueryInit = 1;
1329	}
1330	db_bind_int(&fchngQuery, ":mid", rid);
1331	while( db_step(&fchngQuery)==SQLITE_ROW ){
1332	const char *zFilename = db_column_text(&fchngQuery, 2);
1333	int isNew = db_column_int(&fchngQuery, 0);
1334	int isDel = db_column_int(&fchngQuery, 1);
1335	if( isNew ){
1336	fossil_print(" ADDED %s\n",zFilename);
1337	}else if( isDel ){
1338	fossil_print(" DELETED %s\n",zFilename);
1339	}else{
1340	fossil_print(" EDITED %s\n", zFilename);
1341	}
1342	}
1343	db_reset(&fchngQuery);
1344	}
1345	}
1346	if( fchngQueryInit ) db_finalize(&fchngQuery);
1347	}
1348
1349	/*
1350	** Return a pointer to a static string that forms the basis for
1351	** a timeline query for display on a TTY.
	@@ -1276,11 +1386,11 @@
1386	}
1387
1388	/*
1389	** COMMAND: timeline
1390	**
1391	** Usage: %fossil timeline ?WHEN? ?BASELINE\|DATETIME? ?-n N? ?-t TYPE? ?-showfiles?
1392	**
1393	** Print a summary of activity going backwards in date and time
1394	** specified or from the current date and time if no arguments
1395	** are given. Show as many as N (default 20) check-ins. The
1396	** WHEN argument can be any unique abbreviation of one of these
	@@ -1300,10 +1410,14 @@
1410	** page. For example:
1411	**
1412	** w = wiki commits only
1413	** ci = file commits only
1414	** t = tickets only
1415	**
1416	** The optional showfiles argument if specified prints the list of
1417	** files changed in a checkin after the checkin comment
1418	**
1419	*/
1420	void timeline_cmd(void){
1421	Stmt q;
1422	int n, k;
1423	const char *zCount;
	@@ -1312,10 +1426,12 @@
1426	char *zDate;
1427	Blob sql;
1428	int objid = 0;
1429	Blob uuid;
1430	int mode = 0 ; /* 0:none 1: before 2:after 3:children 4:parents */
1431	int showfilesFlag = 0 ;
1432	showfilesFlag = find_option("showfiles","f", 0)!=0;
1433	db_find_and_open_repository(0, 0);
1434	zCount = find_option("count","n",1);
1435	zType = find_option("type","t",1);
1436	if( zCount ){
1437	n = atoi(zCount);
	@@ -1338,13 +1454,13 @@
1454	mode = 4;
1455	}else if(!zType && !zCount){
1456	usage("?WHEN? ?BASELINE\|DATETIME? ?-n\|--count N? ?-t TYPE?");
1457	}
1458	if( '-' != *g.argv[3] ){
1459	zOrigin = g.argv[3];
1460	}else{
1461	zOrigin = "now";
1462	}
1463	}else if( g.argc==3 ){
1464	zOrigin = g.argv[2];
1465	}else{
1466	zOrigin = "now";
	@@ -1398,11 +1514,11 @@
1514	}
1515
1516	blob_appendf(&sql, " ORDER BY event.mtime DESC");
1517	db_prepare(&q, blob_str(&sql));
1518	blob_reset(&sql);
1519	print_timeline(&q, n, showfilesFlag);
1520	db_finalize(&q);
1521	}
1522
1523	/*
1524	** This is a version of the "localtime()" function from the standard
1525

M src/tkt.c

+1 -1

		--- src/tkt.c
		+++ src/tkt.c
		@@ -731,11 +731,11 @@
731	731	zUuid = PD("name","");
732	732	zTitle = mprintf("History Of Ticket %h", zUuid);
733	733	style_submenu_element("Status", "Status",
734	734	"%s/info/%s", g.zTop, zUuid);
735	735	style_submenu_element("Check-ins", "Check-ins",
736		- "%s/tkttimeline?name=%s?y=ci", g.zTop, zUuid);
	736	+ "%s/tkttimeline?name=%s&y=ci", g.zTop, zUuid);
737	737	style_submenu_element("Timeline", "Timeline",
738	738	"%s/tkttimeline?name=%s", g.zTop, zUuid);
739	739	style_header(zTitle);
740	740	free(zTitle);
741	741
742	742

	--- src/tkt.c
	+++ src/tkt.c
	@@ -731,11 +731,11 @@
731	zUuid = PD("name","");
732	zTitle = mprintf("History Of Ticket %h", zUuid);
733	style_submenu_element("Status", "Status",
734	"%s/info/%s", g.zTop, zUuid);
735	style_submenu_element("Check-ins", "Check-ins",
736	"%s/tkttimeline?name=%s?y=ci", g.zTop, zUuid);
737	style_submenu_element("Timeline", "Timeline",
738	"%s/tkttimeline?name=%s", g.zTop, zUuid);
739	style_header(zTitle);
740	free(zTitle);
741
742

	--- src/tkt.c
	+++ src/tkt.c
	@@ -731,11 +731,11 @@
731	zUuid = PD("name","");
732	zTitle = mprintf("History Of Ticket %h", zUuid);
733	style_submenu_element("Status", "Status",
734	"%s/info/%s", g.zTop, zUuid);
735	style_submenu_element("Check-ins", "Check-ins",
736	"%s/tkttimeline?name=%s&y=ci", g.zTop, zUuid);
737	style_submenu_element("Timeline", "Timeline",
738	"%s/tkttimeline?name=%s", g.zTop, zUuid);
739	style_header(zTitle);
740	free(zTitle);
741
742

M src/update.c

+58 -2

		--- src/update.c
		+++ src/update.c
		@@ -115,10 +115,13 @@
115	115	}
116	116	if( !nochangeFlag && db_exists("SELECT 1 FROM vmerge") ){
117	117	fossil_fatal("cannot update an uncommitted merge");
118	118	}
119	119	if( !nochangeFlag && !internalUpdate ) autosync(AUTOSYNC_PULL);
	120	+
	121	+ /* Create any empty directories now, as well as after the update, so changes in settings are reflected now */
	122	+ ensure_empty_dirs_created();
120	123
121	124	if( internalUpdate ){
122	125	tid = internalUpdate;
123	126	}else if( g.argc>=3 ){
124	127	if( fossil_strcmp(g.argv[2], "current")==0 ){
		@@ -167,11 +170,11 @@
167	170	"%s "
168	171	" AND event.objid IN leaves"
169	172	" ORDER BY event.mtime DESC",
170	173	timeline_query_for_tty()
171	174	);
172		- print_timeline(&q, 100);
	175	+ print_timeline(&q, 100, 0);
173	176	db_finalize(&q);
174	177	fossil_fatal("Multiple descendants");
175	178	}
176	179	}
177	180	tid = db_int(0, "SELECT rid FROM leaves, event"
		@@ -182,11 +185,10 @@
182	185
183	186	if( tid==0 ){
184	187	fossil_panic("Internal Error: unable to find a version to update to.");
185	188	}
186	189
187		- if( tid==vid && !verboseFlag ) return; /* Nothing to update */
188	190	db_begin_transaction();
189	191	vfile_check_signature(vid, 1, 0);
190	192	if( !nochangeFlag && !internalUpdate ) undo_begin();
191	193	load_vfile_from_rid(tid);
192	194
		@@ -441,10 +443,11 @@
441	443	** Clean up the mid and pid VFILE entries. Then commit the changes.
442	444	*/
443	445	if( nochangeFlag ){
444	446	db_end_transaction(1); /* With --nochange, rollback changes */
445	447	}else{
	448	+ ensure_empty_dirs_created();
446	449	if( g.argc<=3 ){
447	450	/* All files updated. Shift the current checkout to the target. */
448	451	db_multi_exec("DELETE FROM vfile WHERE vid!=%d", tid);
449	452	checkout_set_all_exe(tid);
450	453	manifest_to_disk(tid);
		@@ -456,10 +459,63 @@
456	459	}
457	460	if( !internalUpdate ) undo_finish();
458	461	db_end_transaction(0);
459	462	}
460	463	}
	464	+
	465	+/*
	466	+** Make sure empty directories are created
	467	+*/
	468	+void ensure_empty_dirs_created(void){
	469	+ /* Make empty directories? */
	470	+ char *zEmptyDirs = db_get("empty-dirs", 0);
	471	+ if( zEmptyDirs!=0 ){
	472	+ char *bc;
	473	+ Blob dirName;
	474	+ Blob dirsList;
	475	+
	476	+ blob_zero(&dirsList);
	477	+ blob_init(&dirsList, zEmptyDirs, strlen(zEmptyDirs));
	478	+ /* Replace commas by spaces */
	479	+ bc = blob_str(&dirsList);
	480	+ while( (*bc)!='\0' ){
	481	+ if( (bc)==',' ) { bc = ' '; }
	482	+ ++bc;
	483	+ }
	484	+ /* Make directories */
	485	+ blob_zero(&dirName);
	486	+ while( blob_token(&dirsList, &dirName) ){
	487	+ const char *zDir = blob_str(&dirName);
	488	+ /* Make full pathname of the directory */
	489	+ Blob path;
	490	+ const char *zPath;
	491	+
	492	+ blob_zero(&path);
	493	+ blob_appendf(&path, "%s/%s", g.zLocalRoot, zDir);
	494	+ zPath = blob_str(&path);
	495	+ /* Handle various cases of existence of the directory */
	496	+ switch( file_isdir(zPath) ){
	497	+ case 0: { /* doesn't exist */
	498	+ if( file_mkdir(zPath, 0)!=0 ) {
	499	+ fossil_warning("couldn't create directory %s as "
	500	+ "required by empty-dirs setting", zDir);
	501	+ }
	502	+ break;
	503	+ }
	504	+ case 1: { /* exists, and is a directory */
	505	+ /* do nothing - required directory exists already */
	506	+ break;
	507	+ }
	508	+ case 2: { /* exists, but isn't a directory */
	509	+ fossil_warning("file %s found, but a directory is required "
	510	+ "by empty-dirs setting", zDir);
	511	+ }
	512	+ }
	513	+ blob_reset(&path);
	514	+ }
	515	+ }
	516	+}
461	517
462	518
463	519	/*
464	520	** Get the contents of a file within the checking "revision". If
465	521	** revision==NULL then get the file content for the current checkout.
466	522

	--- src/update.c
	+++ src/update.c
	@@ -115,10 +115,13 @@
115	}
116	if( !nochangeFlag && db_exists("SELECT 1 FROM vmerge") ){
117	fossil_fatal("cannot update an uncommitted merge");
118	}
119	if( !nochangeFlag && !internalUpdate ) autosync(AUTOSYNC_PULL);



120
121	if( internalUpdate ){
122	tid = internalUpdate;
123	}else if( g.argc>=3 ){
124	if( fossil_strcmp(g.argv[2], "current")==0 ){
	@@ -167,11 +170,11 @@
167	"%s "
168	" AND event.objid IN leaves"
169	" ORDER BY event.mtime DESC",
170	timeline_query_for_tty()
171	);
172	print_timeline(&q, 100);
173	db_finalize(&q);
174	fossil_fatal("Multiple descendants");
175	}
176	}
177	tid = db_int(0, "SELECT rid FROM leaves, event"
	@@ -182,11 +185,10 @@
182
183	if( tid==0 ){
184	fossil_panic("Internal Error: unable to find a version to update to.");
185	}
186
187	if( tid==vid && !verboseFlag ) return; /* Nothing to update */
188	db_begin_transaction();
189	vfile_check_signature(vid, 1, 0);
190	if( !nochangeFlag && !internalUpdate ) undo_begin();
191	load_vfile_from_rid(tid);
192
	@@ -441,10 +443,11 @@
441	** Clean up the mid and pid VFILE entries. Then commit the changes.
442	*/
443	if( nochangeFlag ){
444	db_end_transaction(1); /* With --nochange, rollback changes */
445	}else{

446	if( g.argc<=3 ){
447	/* All files updated. Shift the current checkout to the target. */
448	db_multi_exec("DELETE FROM vfile WHERE vid!=%d", tid);
449	checkout_set_all_exe(tid);
450	manifest_to_disk(tid);
	@@ -456,10 +459,63 @@
456	}
457	if( !internalUpdate ) undo_finish();
458	db_end_transaction(0);
459	}
460	}





















































461
462
463	/*
464	** Get the contents of a file within the checking "revision". If
465	** revision==NULL then get the file content for the current checkout.
466

	--- src/update.c
	+++ src/update.c
	@@ -115,10 +115,13 @@
115	}
116	if( !nochangeFlag && db_exists("SELECT 1 FROM vmerge") ){
117	fossil_fatal("cannot update an uncommitted merge");
118	}
119	if( !nochangeFlag && !internalUpdate ) autosync(AUTOSYNC_PULL);
120
121	/* Create any empty directories now, as well as after the update, so changes in settings are reflected now */
122	ensure_empty_dirs_created();
123
124	if( internalUpdate ){
125	tid = internalUpdate;
126	}else if( g.argc>=3 ){
127	if( fossil_strcmp(g.argv[2], "current")==0 ){
	@@ -167,11 +170,11 @@
170	"%s "
171	" AND event.objid IN leaves"
172	" ORDER BY event.mtime DESC",
173	timeline_query_for_tty()
174	);
175	print_timeline(&q, 100, 0);
176	db_finalize(&q);
177	fossil_fatal("Multiple descendants");
178	}
179	}
180	tid = db_int(0, "SELECT rid FROM leaves, event"
	@@ -182,11 +185,10 @@
185
186	if( tid==0 ){
187	fossil_panic("Internal Error: unable to find a version to update to.");
188	}
189

190	db_begin_transaction();
191	vfile_check_signature(vid, 1, 0);
192	if( !nochangeFlag && !internalUpdate ) undo_begin();
193	load_vfile_from_rid(tid);
194
	@@ -441,10 +443,11 @@
443	** Clean up the mid and pid VFILE entries. Then commit the changes.
444	*/
445	if( nochangeFlag ){
446	db_end_transaction(1); /* With --nochange, rollback changes */
447	}else{
448	ensure_empty_dirs_created();
449	if( g.argc<=3 ){
450	/* All files updated. Shift the current checkout to the target. */
451	db_multi_exec("DELETE FROM vfile WHERE vid!=%d", tid);
452	checkout_set_all_exe(tid);
453	manifest_to_disk(tid);
	@@ -456,10 +459,63 @@
459	}
460	if( !internalUpdate ) undo_finish();
461	db_end_transaction(0);
462	}
463	}
464
465	/*
466	** Make sure empty directories are created
467	*/
468	void ensure_empty_dirs_created(void){
469	/* Make empty directories? */
470	char *zEmptyDirs = db_get("empty-dirs", 0);
471	if( zEmptyDirs!=0 ){
472	char *bc;
473	Blob dirName;
474	Blob dirsList;
475
476	blob_zero(&dirsList);
477	blob_init(&dirsList, zEmptyDirs, strlen(zEmptyDirs));
478	/* Replace commas by spaces */
479	bc = blob_str(&dirsList);
480	while( (*bc)!='\0' ){
481	if( (bc)==',' ) { bc = ' '; }
482	++bc;
483	}
484	/* Make directories */
485	blob_zero(&dirName);
486	while( blob_token(&dirsList, &dirName) ){
487	const char *zDir = blob_str(&dirName);
488	/* Make full pathname of the directory */
489	Blob path;
490	const char *zPath;
491
492	blob_zero(&path);
493	blob_appendf(&path, "%s/%s", g.zLocalRoot, zDir);
494	zPath = blob_str(&path);
495	/* Handle various cases of existence of the directory */
496	switch( file_isdir(zPath) ){
497	case 0: { /* doesn't exist */
498	if( file_mkdir(zPath, 0)!=0 ) {
499	fossil_warning("couldn't create directory %s as "
500	"required by empty-dirs setting", zDir);
501	}
502	break;
503	}
504	case 1: { /* exists, and is a directory */
505	/* do nothing - required directory exists already */
506	break;
507	}
508	case 2: { /* exists, but isn't a directory */
509	fossil_warning("file %s found, but a directory is required "
510	"by empty-dirs setting", zDir);
511	}
512	}
513	blob_reset(&path);
514	}
515	}
516	}
517
518
519	/*
520	** Get the contents of a file within the checking "revision". If
521	** revision==NULL then get the file content for the current checkout.
522

M src/wiki.c

+8 -20

		--- src/wiki.c
		+++ src/wiki.c
		@@ -182,12 +182,11 @@
182	182	" ORDER BY mtime DESC", zTag
183	183	);
184	184	free(zTag);
185	185	pWiki = manifest_get(rid, CFTYPE_WIKI);
186	186	if( pWiki ){
187		- while( fossil_isspace(pWiki->zWiki[0]) ) pWiki->zWiki++;
188		- if( pWiki->zWiki[0] ) zBody = pWiki->zWiki;
	187	+ zBody = pWiki->zWiki;
189	188	}
190	189	}
191	190	if( !g.isHome ){
192	191	if( (rid && g.okWrWiki) \|\| (!rid && g.okNewWiki) ){
193	192	style_submenu_element("Edit", "Edit Wiki Page", "%s/wikiedit?name=%T",
		@@ -891,10 +890,11 @@
891	890	char const zPageName; / Name of the wiki page to export */
892	891	char const zFile; / Name of the output file (0=stdout) */
893	892	int rid; /* Artifact ID of the wiki page */
894	893	int i; /* Loop counter */
895	894	char zBody = 0; / Wiki page content */
	895	+ Blob body; /* Wiki page content */
896	896	Manifest pWiki = 0; / Parsed wiki page content */
897	897
898	898	if( (g.argc!=4) && (g.argc!=5) ){
899	899	usage("export PAGENAME ?FILE?");
900	900	}
		@@ -909,28 +909,16 @@
909	909	}
910	910	if( zBody==0 ){
911	911	fossil_fatal("wiki page [%s] not found",zPageName);
912	912	}
913	913	for(i=strlen(zBody); i>0 && fossil_isspace(zBody[i-1]); i--){}
914		- zFile = (g.argc==4) ? 0 : g.argv[4];
915		- if( zFile ){
916		- FILE * zF;
917		- short doClose = 0;
918		- if( (1 == strlen(zFile)) && ('-'==zFile[0]) ){
919		- zF = stdout;
920		- }else{
921		- zF = fossil_fopen( zFile, "w" );
922		- doClose = zF ? 1 : 0;
923		- }
924		- if( ! zF ){
925		- fossil_fatal("wiki export could not open output file for writing.");
926		- }
927		- fprintf(zF,"%.*s\n", i, zBody);
928		- if( doClose ) fclose(zF);
929		- }else{
930		- fossil_print("%.*s\n", i, zBody);
931		- }
	914	+ zBody[i] = 0;
	915	+ zFile = (g.argc==4) ? "-" : g.argv[4];
	916	+ blob_init(&body, zBody, -1);
	917	+ blob_append(&body, "\n", 1);
	918	+ blob_write_to_file(&body, zFile);
	919	+ blob_reset(&body);
932	920	manifest_destroy(pWiki);
933	921	return;
934	922	}else
935	923	if( strncmp(g.argv[2],"commit",n)==0
936	924	\|\| strncmp(g.argv[2],"create",n)==0 ){
937	925

	--- src/wiki.c
	+++ src/wiki.c
	@@ -182,12 +182,11 @@
182	" ORDER BY mtime DESC", zTag
183	);
184	free(zTag);
185	pWiki = manifest_get(rid, CFTYPE_WIKI);
186	if( pWiki ){
187	while( fossil_isspace(pWiki->zWiki[0]) ) pWiki->zWiki++;
188	if( pWiki->zWiki[0] ) zBody = pWiki->zWiki;
189	}
190	}
191	if( !g.isHome ){
192	if( (rid && g.okWrWiki) \|\| (!rid && g.okNewWiki) ){
193	style_submenu_element("Edit", "Edit Wiki Page", "%s/wikiedit?name=%T",
	@@ -891,10 +890,11 @@
891	char const zPageName; / Name of the wiki page to export */
892	char const zFile; / Name of the output file (0=stdout) */
893	int rid; /* Artifact ID of the wiki page */
894	int i; /* Loop counter */
895	char zBody = 0; / Wiki page content */

896	Manifest pWiki = 0; / Parsed wiki page content */
897
898	if( (g.argc!=4) && (g.argc!=5) ){
899	usage("export PAGENAME ?FILE?");
900	}
	@@ -909,28 +909,16 @@
909	}
910	if( zBody==0 ){
911	fossil_fatal("wiki page [%s] not found",zPageName);
912	}
913	for(i=strlen(zBody); i>0 && fossil_isspace(zBody[i-1]); i--){}
914	zFile = (g.argc==4) ? 0 : g.argv[4];
915	if( zFile ){
916	FILE * zF;
917	short doClose = 0;
918	if( (1 == strlen(zFile)) && ('-'==zFile[0]) ){
919	zF = stdout;
920	}else{
921	zF = fossil_fopen( zFile, "w" );
922	doClose = zF ? 1 : 0;
923	}
924	if( ! zF ){
925	fossil_fatal("wiki export could not open output file for writing.");
926	}
927	fprintf(zF,"%.*s\n", i, zBody);
928	if( doClose ) fclose(zF);
929	}else{
930	fossil_print("%.*s\n", i, zBody);
931	}
932	manifest_destroy(pWiki);
933	return;
934	}else
935	if( strncmp(g.argv[2],"commit",n)==0
936	\|\| strncmp(g.argv[2],"create",n)==0 ){
937

	--- src/wiki.c
	+++ src/wiki.c
	@@ -182,12 +182,11 @@
182	" ORDER BY mtime DESC", zTag
183	);
184	free(zTag);
185	pWiki = manifest_get(rid, CFTYPE_WIKI);
186	if( pWiki ){
187	zBody = pWiki->zWiki;

188	}
189	}
190	if( !g.isHome ){
191	if( (rid && g.okWrWiki) \|\| (!rid && g.okNewWiki) ){
192	style_submenu_element("Edit", "Edit Wiki Page", "%s/wikiedit?name=%T",
	@@ -891,10 +890,11 @@
890	char const zPageName; / Name of the wiki page to export */
891	char const zFile; / Name of the output file (0=stdout) */
892	int rid; /* Artifact ID of the wiki page */
893	int i; /* Loop counter */
894	char zBody = 0; / Wiki page content */
895	Blob body; /* Wiki page content */
896	Manifest pWiki = 0; / Parsed wiki page content */
897
898	if( (g.argc!=4) && (g.argc!=5) ){
899	usage("export PAGENAME ?FILE?");
900	}
	@@ -909,28 +909,16 @@
909	}
910	if( zBody==0 ){
911	fossil_fatal("wiki page [%s] not found",zPageName);
912	}
913	for(i=strlen(zBody); i>0 && fossil_isspace(zBody[i-1]); i--){}
914	zBody[i] = 0;
915	zFile = (g.argc==4) ? "-" : g.argv[4];
916	blob_init(&body, zBody, -1);
917	blob_append(&body, "\n", 1);
918	blob_write_to_file(&body, zFile);
919	blob_reset(&body);












920	manifest_destroy(pWiki);
921	return;
922	}else
923	if( strncmp(g.argv[2],"commit",n)==0
924	\|\| strncmp(g.argv[2],"create",n)==0 ){
925

M src/winhttp.c

+644 -9

		--- src/winhttp.c
		+++ src/winhttp.c
		@@ -14,17 +14,18 @@
14	14	** http://www.hwaci.com/drh/
15	15	**
16	16	*******************************************************************************
17	17	**
18	18	** This file implements a very simple (and low-performance) HTTP server
19		-** for windows.
	19	+** for windows. It also implements a Windows Service which allows the HTTP
	20	+** server to be run without any user logged on.
20	21	*/
21	22	#include "config.h"
22	23	#ifdef _WIN32
23	24	/* This code is for win32 only */
24		-#include "winhttp.h"
25	25	#include <windows.h>
	26	+#include "winhttp.h"
26	27
27	28	/*
28	29	** The HttpRequest structure holds information about each incoming
29	30	** HTTP request.
30	31	*/
		@@ -60,11 +61,11 @@
60	61	}
61	62
62	63	/*
63	64	** Process a single incoming HTTP request.
64	65	*/
65		-void win32_process_one_http_request(void *pAppData){
	66	+static void win32_process_one_http_request(void *pAppData){
66	67	HttpRequest p = (HttpRequest)pAppData;
67	68	FILE in = 0, out = 0;
68	69	int amt, got;
69	70	int wanted = 0;
70	71	char *z;
		@@ -145,10 +146,11 @@
145	146	SOCKET s = INVALID_SOCKET;
146	147	SOCKADDR_IN addr;
147	148	int idCnt = 0;
148	149	int iPort = mnPort;
149	150	Blob options;
	151	+ char zTmpPath[MAX_PATH];
150	152
151	153	if( zStopper ) file_delete(zStopper);
152	154	blob_zero(&options);
153	155	if( zNotFound ){
154	156	blob_appendf(&options, " --notfound %s", zNotFound);
		@@ -189,31 +191,46 @@
189	191	}else{
190	192	fossil_fatal("unable to open listening socket on any"
191	193	" port in the range %d..%d", mnPort, mxPort);
192	194	}
193	195	}
194		- zTempPrefix = mprintf("fossil_server_P%d_", iPort);
	196	+ if( !GetTempPath(sizeof(zTmpPath), zTmpPath) ){
	197	+ fossil_fatal("unable to get path to the temporary directory.");
	198	+ }
	199	+ zTempPrefix = mprintf("%sfossil_server_P%d_", zTmpPath, iPort);
195	200	fossil_print("Listening for HTTP requests on TCP port %d\n", iPort);
196	201	if( zBrowser ){
197	202	zBrowser = mprintf(zBrowser, iPort);
198	203	fossil_print("Launch webbrowser: %s\n", zBrowser);
199	204	fossil_system(zBrowser);
200	205	}
201	206	fossil_print("Type Ctrl-C to stop the HTTP server\n");
	207	+ /* Set the service status to running and pass the listener socket to the
	208	+ ** service handling procedures. */
	209	+ win32_http_service_running(s);
202	210	for(;;){
203	211	SOCKET client;
204	212	SOCKADDR_IN client_addr;
205	213	HttpRequest *p;
206	214	int len = sizeof(client_addr);
	215	+ int wsaError;
207	216
208	217	client = accept(s, (struct sockaddr*)&client_addr, &len);
209		- if( zStopper && file_size(zStopper)>=0 ){
210		- break;
211		- }
212	218	if( client==INVALID_SOCKET ){
213		- closesocket(s);
214		- fossil_fatal("error from accept()");
	219	+ /* If the service control handler has closed the listener socket,
	220	+ ** cleanup and return, otherwise report a fatal error. */
	221	+ wsaError = WSAGetLastError();
	222	+ if( (wsaError==WSAEINTR) \|\| (wsaError==WSAENOTSOCK) ){
	223	+ WSACleanup();
	224	+ return;
	225	+ }else{
	226	+ closesocket(s);
	227	+ WSACleanup();
	228	+ fossil_fatal("error from accept()");
	229	+ }
	230	+ }else if( zStopper && file_size(zStopper)>=0 ){
	231	+ break;
215	232	}
216	233	p = fossil_malloc( sizeof(*p) );
217	234	p->id = ++idCnt;
218	235	p->s = client;
219	236	p->addr = client_addr;
		@@ -221,7 +238,625 @@
221	238	_beginthread(win32_process_one_http_request, 0, (void*)p);
222	239	}
223	240	closesocket(s);
224	241	WSACleanup();
225	242	}
	243	+
	244	+/*
	245	+** The HttpService structure is used to pass information to the service main
	246	+** function and to the service control handler function.
	247	+*/
	248	+typedef struct HttpService HttpService;
	249	+struct HttpService {
	250	+ int port; /* Port on which the http server should run */
	251	+ const char zNotFound; / The --notfound option, or NULL */
	252	+ int flags; /* One or more HTTP_SERVER_ flags */
	253	+ int isRunningAsService; /* Are we running as a service ? */
	254	+ const char zServiceName; / Name of the service */
	255	+ SOCKET s; /* Socket on which the http server listens */
	256	+};
	257	+
	258	+/*
	259	+** Variables used for running as windows service.
	260	+*/
	261	+static HttpService hsData = {8080, NULL, 0, 0, NULL, INVALID_SOCKET};
	262	+static SERVICE_STATUS ssStatus;
	263	+static SERVICE_STATUS_HANDLE sshStatusHandle;
	264	+
	265	+/*
	266	+** Get message string of the last system error. Return a pointer to the
	267	+** message string. Call fossil_mbcs_free() to deallocate any memory used
	268	+** to store the message string when done.
	269	+*/
	270	+static char *win32_get_last_errmsg(void){
	271	+ DWORD nMsg;
	272	+ LPTSTR tmp = NULL;
	273	+ char *zMsg = NULL;
	274	+
	275	+ nMsg = FormatMessage(
	276	+ FORMAT_MESSAGE_ALLOCATE_BUFFER \|
	277	+ FORMAT_MESSAGE_FROM_SYSTEM \|
	278	+ FORMAT_MESSAGE_IGNORE_INSERTS,
	279	+ NULL,
	280	+ GetLastError(),
	281	+ MAKELANGID(LANG_ENGLISH, SUBLANG_ENGLISH_US),
	282	+ (LPTSTR) &tmp,
	283	+ 0,
	284	+ NULL
	285	+ );
	286	+ if( nMsg ){
	287	+ zMsg = fossil_mbcs_to_utf8(tmp);
	288	+ }else{
	289	+ fossil_fatal("unable to get system error message.");
	290	+ }
	291	+ if( tmp ){
	292	+ LocalFree((HLOCAL) tmp);
	293	+ }
	294	+ return zMsg;
	295	+}
	296	+
	297	+/*
	298	+** Report the current status of the service to the service control manager.
	299	+** Make sure that during service startup no control codes are accepted.
	300	+*/
	301	+static void win32_report_service_status(
	302	+ DWORD dwCurrentState, /* The current state of the service */
	303	+ DWORD dwWin32ExitCode, /* The error code to report */
	304	+ DWORD dwWaitHint /* The estimated time for a pending operation */
	305	+){
	306	+ if( dwCurrentState==SERVICE_START_PENDING) {
	307	+ ssStatus.dwControlsAccepted = 0;
	308	+ }else{
	309	+ ssStatus.dwControlsAccepted = SERVICE_ACCEPT_STOP;
	310	+ }
	311	+ ssStatus.dwCurrentState = dwCurrentState;
	312	+ ssStatus.dwWin32ExitCode = dwWin32ExitCode;
	313	+ ssStatus.dwWaitHint = dwWaitHint;
	314	+
	315	+ if( (dwCurrentState==SERVICE_RUNNING) \|\|
	316	+ (dwCurrentState==SERVICE_STOPPED) ){
	317	+ ssStatus.dwCheckPoint = 0;
	318	+ }else{
	319	+ ssStatus.dwCheckPoint++;
	320	+ }
	321	+ SetServiceStatus(sshStatusHandle, &ssStatus);
	322	+ return ;
	323	+}
	324	+
	325	+/*
	326	+** Handle control codes sent from the service control manager.
	327	+** The control dispatcher in the main thread of the service process invokes
	328	+** this function whenever it receives a control request from the service
	329	+** control manager.
	330	+*/
	331	+static void WINAPI win32_http_service_ctrl(
	332	+ DWORD dwCtrlCode
	333	+){
	334	+ switch( dwCtrlCode ){
	335	+ case SERVICE_CONTROL_STOP: {
	336	+ win32_report_service_status(SERVICE_STOP_PENDING, NO_ERROR, 0);
	337	+ if( hsData.s != INVALID_SOCKET ){
	338	+ closesocket(hsData.s);
	339	+ }
	340	+ win32_report_service_status(ssStatus.dwCurrentState, NO_ERROR, 0);
	341	+ break;
	342	+ }
	343	+ default: {
	344	+ break;
	345	+ }
	346	+ }
	347	+ return;
	348	+}
	349	+
	350	+/*
	351	+** This is the main entry point for the service.
	352	+** When the service control manager receives a request to start the service,
	353	+** it starts the service process (if it is not already running). The main
	354	+** thread of the service process calls the StartServiceCtrlDispatcher
	355	+** function with a pointer to an array of SERVICE_TABLE_ENTRY structures.
	356	+** Then the service control manager sends a start request to the service
	357	+** control dispatcher for this service process. The service control dispatcher
	358	+** creates a new thread to execute the ServiceMain function (this function)
	359	+** of the service being started.
	360	+*/
	361	+static void WINAPI win32_http_service_main(
	362	+ DWORD argc, /* Number of arguments in argv */
	363	+ LPTSTR argv / Arguments passed */
	364	+){
	365	+
	366	+ /* Update the service information. */
	367	+ hsData.isRunningAsService = 1;
	368	+ if( argc>0 ){
	369	+ hsData.zServiceName = argv[0];
	370	+ }
	371	+
	372	+ /* Register the service control handler function */
	373	+ sshStatusHandle = RegisterServiceCtrlHandler("", win32_http_service_ctrl);
	374	+ if( !sshStatusHandle ){
	375	+ win32_report_service_status(SERVICE_STOPPED, NO_ERROR, 0);
	376	+ return;
	377	+ }
	378	+
	379	+ /* Set service specific data and report that the service is starting. */
	380	+ ssStatus.dwServiceType = SERVICE_WIN32_OWN_PROCESS;
	381	+ ssStatus.dwServiceSpecificExitCode = 0;
	382	+ win32_report_service_status(SERVICE_START_PENDING, NO_ERROR, 3000);
	383	+
	384	+ /* Execute the http server */
	385	+ win32_http_server(hsData.port, hsData.port,
	386	+ NULL, NULL, hsData.zNotFound, hsData.flags);
	387	+
	388	+ /* Service has stopped now. */
	389	+ win32_report_service_status(SERVICE_STOPPED, NO_ERROR, 0);
	390	+ return;
	391	+}
	392	+
	393	+/*
	394	+** When running as service, update the HttpService structure with the
	395	+** listener socket and update the service status. This procedure must be
	396	+** called from the http server when he is ready to accept connections.
	397	+*/
	398	+LOCAL void win32_http_service_running(SOCKET s){
	399	+ if( hsData.isRunningAsService ){
	400	+ hsData.s = s;
	401	+ win32_report_service_status(SERVICE_RUNNING, NO_ERROR, 0);
	402	+ }
	403	+}
	404	+
	405	+/*
	406	+** Try to start the http server as a windows service. If we are running in
	407	+** a interactive console session, this routine fails and returns a non zero
	408	+** integer value. When running as service, this routine does not return until
	409	+** the service is stopped. In this case, the return value is zero.
	410	+*/
	411	+int win32_http_service(
	412	+ int nPort, /* TCP port number */
	413	+ const char zNotFound, / The --notfound option, or NULL */
	414	+ int flags /* One or more HTTP_SERVER_ flags */
	415	+){
	416	+ /* Define the service table. */
	417	+ SERVICE_TABLE_ENTRY ServiceTable[] =
	418	+ {{"", (LPSERVICE_MAIN_FUNCTION)win32_http_service_main}, {NULL, NULL}};
	419	+
	420	+ /* Initialize the HttpService structure. */
	421	+ hsData.port = nPort;
	422	+ hsData.zNotFound = zNotFound;
	423	+ hsData.flags = flags;
	424	+
	425	+ /* Try to start the control dispatcher thread for the service. */
	426	+ if( !StartServiceCtrlDispatcher(ServiceTable) ){
	427	+ if( GetLastError()==ERROR_FAILED_SERVICE_CONTROLLER_CONNECT ){
	428	+ return 1;
	429	+ }else{
	430	+ fossil_fatal("error from StartServiceCtrlDispatcher()");
	431	+ }
	432	+ }
	433	+ return 0;
	434	+}
	435	+
	436	+/*
	437	+** COMMAND: winsrv
	438	+** Usage: fossil winsrv METHOD ?SERVICE-NAME? ?OPTIONS?
	439	+**
	440	+** Where METHOD is one of: create delete show start stop.
	441	+**
	442	+** The winsrv command manages Fossil as a Windows service. This allows
	443	+** (for example) Fossil to be running in the background when no user
	444	+** is logged in.
	445	+**
	446	+** In the following description of the methods, "Fossil-DSCM" will be
	447	+** used as the default SERVICE-NAME:
	448	+**
	449	+** fossil winsrv create ?SERVICE-NAME? ?OPTIONS?
	450	+**
	451	+** Creates a service. Available options include:
	452	+**
	453	+** -D\|--display DISPLAY-NAME
	454	+**
	455	+** Sets the display name of the service. This name is shown
	456	+** by graphical interface programs. By default, the display name
	457	+** equals to the service name.
	458	+**
	459	+** -S\|--start TYPE
	460	+**
	461	+** Sets the start type of the service. TYPE can be "manual",
	462	+** which means you need to start the service yourself with the
	463	+** 'fossil winsrv start' command or with the "net start" command
	464	+** from the operating system. If TYPE is set to "auto", the service
	465	+** will be started automatically by the system during startup.
	466	+**
	467	+** -U\|--username USERNAME
	468	+**
	469	+** Specifies the user account which will be used to run the
	470	+** service. The account needs the "Logon as a service" right
	471	+** enabled in its profile. Specify local accounts as follows:
	472	+** ".\\USERNAME". By default, the "LocalSystem" account will be
	473	+** used.
	474	+**
	475	+** -W\|--password PASSWORD
	476	+**
	477	+** Password for the user account.
	478	+**
	479	+** The following options are more or less the same as for the "server"
	480	+** command and influence the behaviour of the http server:
	481	+**
	482	+** -p\|--port TCPPORT
	483	+**
	484	+** Specifies the TCP port (default port is 8080) on which the
	485	+** server should listen.
	486	+**
	487	+** -R\|--repository REPOSITORY
	488	+**
	489	+** Specifies the name of the repository to be served.
	490	+** The repository option may be omitted if the working directory
	491	+** is within an open checkout.
	492	+** The REPOSITORY can be a directory (aka folder) that contains
	493	+** one or more respositories with names ending in ".fossil".
	494	+** In that case, the first element of the URL is used to select
	495	+** among the various repositories.
	496	+**
	497	+** --notfound URL
	498	+**
	499	+** If REPOSITORY is a directory that contains one or more
	500	+** respositories with names of the form "*.fossil" then the
	501	+** first element of the URL pathname selects among the various
	502	+** repositories. If the pathname does not select a valid
	503	+** repository and the --notfound option is available,
	504	+** then the server redirects (HTTP code 302) to the URL of
	505	+** --notfound.
	506	+**
	507	+** --localauth
	508	+**
	509	+** Enables automatic login if the --localauth option is present
	510	+** and the "localauth" setting is off and the connection is from
	511	+** localhost.
	512	+**
	513	+**
	514	+** fossil winsrv delete ?SERVICE-NAME?
	515	+**
	516	+** Deletes a service. If the service is currently running, it will be
	517	+** stopped first and then deleted.
	518	+**
	519	+**
	520	+** fossil winsrv show ?SERVICE-NAME?
	521	+**
	522	+** Shows how the service is configured and its current state.
	523	+**
	524	+**
	525	+** fossil winsrv start ?SERVICE-NAME?
	526	+**
	527	+** Start the service.
	528	+**
	529	+**
	530	+** fossil winsrv stop ?SERVICE-NAME?
	531	+**
	532	+** Stop the service.
	533	+**
	534	+**
	535	+** NOTE: This command is available on Windows operating systems only and
	536	+** requires administrative rights on the machine executed.
	537	+**
	538	+*/
	539	+void cmd_win32_service(void){
	540	+ int n;
	541	+ const char *zMethod;
	542	+ const char zSvcName = "Fossil-DSCM"; / Default service name */
	543	+
	544	+ if( g.argc<3 ){
	545	+ usage("create\|delete\|show\|start\|stop ...");
	546	+ }
	547	+ zMethod = g.argv[2];
	548	+ n = strlen(zMethod);
	549	+ if( g.argc==4 ){
	550	+ zSvcName = g.argv[3];
	551	+ }
	552	+
	553	+ if( strncmp(zMethod, "create", n)==0 ){
	554	+ SC_HANDLE hScm;
	555	+ SC_HANDLE hSvc;
	556	+ SERVICE_DESCRIPTION
	557	+ svcDescr = {"Fossil - Distributed Software Configuration Management"};
	558	+ char *zErrFmt = "unable to create service '%s': %s";
	559	+ DWORD dwStartType = SERVICE_DEMAND_START;
	560	+ const char *zDisplay;
	561	+ const char *zStart;
	562	+ const char *zUsername;
	563	+ const char *zPassword;
	564	+ const char *zPort;
	565	+ const char *zNotFound;
	566	+ const char *zLocalAuth;
	567	+ const char *zRepository;
	568	+ Blob binPath;
	569	+
	570	+ /* Process service creation specific options. */
	571	+ zDisplay = find_option("display", "D", 1);
	572	+ if( !zDisplay ){
	573	+ zDisplay = zSvcName;
	574	+ }
	575	+ zStart = find_option("start", "S", 1);
	576	+ if( zStart ){
	577	+ if( strncmp(zStart, "auto", strlen(zStart))==0 ){
	578	+ dwStartType = SERVICE_AUTO_START;
	579	+ }else if( strncmp(zStart, "manual", strlen(zStart))==0 ){
	580	+ dwStartType = SERVICE_DEMAND_START;
	581	+ }else{
	582	+ fossil_fatal(zErrFmt, zSvcName,
	583	+ "specify 'auto' or 'manual' for the '-S\|--start' option");
	584	+ }
	585	+ }
	586	+ zUsername = find_option("username", "U", 1);
	587	+ zPassword = find_option("password", "W", 1);
	588	+ /* Process options for Fossil running as server. */
	589	+ zPort = find_option("port", "P", 1);
	590	+ if( zPort && (atoi(zPort)<=0) ){
	591	+ fossil_fatal(zErrFmt, zSvcName,
	592	+ "port number must be in the range 1 - 65535.");
	593	+ }
	594	+ zNotFound = find_option("notfound", 0, 1);
	595	+ zLocalAuth = find_option("localauth", 0, 0);
	596	+ zRepository = find_option("repository", "R", 1);
	597	+ if( !zRepository ){
	598	+ db_must_be_within_tree();
	599	+ }else if( file_isdir(zRepository)==1 ){
	600	+ g.zRepositoryName = mprintf("%s", zRepository);
	601	+ file_simplify_name(g.zRepositoryName, -1);
	602	+ }else{
	603	+ db_open_repository(zRepository);
	604	+ }
	605	+ db_close(0);
	606	+ verify_all_options();
	607	+ if( g.argc>4 ) fossil_fatal("to much arguments for create method.");
	608	+ /* Build the fully-qualified path to the service binary file. */
	609	+ blob_zero(&binPath);
	610	+ blob_appendf(&binPath, "\"%s\" server", fossil_nameofexe());
	611	+ if( zPort ) blob_appendf(&binPath, " --port %s", zPort);
	612	+ if( zNotFound ) blob_appendf(&binPath, " --notfound \"%s\"", zNotFound);
	613	+ if( zLocalAuth ) blob_append(&binPath, " --localauth", -1);
	614	+ blob_appendf(&binPath, " \"%s\"", g.zRepositoryName);
	615	+ /* Create the service. */
	616	+ hScm = OpenSCManager(NULL, NULL, SC_MANAGER_ALL_ACCESS);
	617	+ if( !hScm ) fossil_fatal(zErrFmt, zSvcName, win32_get_last_errmsg());
	618	+ hSvc = CreateService(
	619	+ hScm, /* Handle to the SCM */
	620	+ fossil_utf8_to_mbcs(zSvcName), /* Name of the service */
	621	+ fossil_utf8_to_mbcs(zDisplay), /* Display name */
	622	+ SERVICE_ALL_ACCESS, /* Desired access */
	623	+ SERVICE_WIN32_OWN_PROCESS, /* Service type */
	624	+ dwStartType, /* Start type */
	625	+ SERVICE_ERROR_NORMAL, /* Error control */
	626	+ fossil_utf8_to_mbcs(blob_str(&binPath)), /* Binary path */
	627	+ NULL, /* Load ordering group */
	628	+ NULL, /* Tag value */
	629	+ NULL, /* Service dependencies */
	630	+ fossil_utf8_to_mbcs(zUsername), /* Service account */
	631	+ fossil_utf8_to_mbcs(zPassword) /* Account password */
	632	+ );
	633	+ if( !hSvc ) fossil_fatal(zErrFmt, zSvcName, win32_get_last_errmsg());
	634	+ /* Set the service description. */
	635	+ ChangeServiceConfig2(hSvc, SERVICE_CONFIG_DESCRIPTION, &svcDescr);
	636	+ fossil_print("Service '%s' successfully created.\n", zSvcName);
	637	+ CloseServiceHandle(hSvc);
	638	+ CloseServiceHandle(hScm);
	639	+ }else
	640	+ if( strncmp(zMethod, "delete", n)==0 ){
	641	+ SC_HANDLE hScm;
	642	+ SC_HANDLE hSvc;
	643	+ SERVICE_STATUS sstat;
	644	+ char *zErrFmt = "unable to delete service '%s': %s";
	645	+
	646	+ verify_all_options();
	647	+ if( g.argc>4 ) fossil_fatal("to much arguments for delete method.");
	648	+ hScm = OpenSCManager(NULL, NULL, SC_MANAGER_ALL_ACCESS);
	649	+ if( !hScm ) fossil_fatal(zErrFmt, zSvcName, win32_get_last_errmsg());
	650	+ hSvc = OpenService(hScm, fossil_utf8_to_mbcs(zSvcName), SERVICE_ALL_ACCESS);
	651	+ if( !hSvc ) fossil_fatal(zErrFmt, zSvcName, win32_get_last_errmsg());
	652	+ QueryServiceStatus(hSvc, &sstat);
	653	+ if( sstat.dwCurrentState!=SERVICE_STOPPED ){
	654	+ fossil_print("Stopping service '%s'", zSvcName);
	655	+ if( sstat.dwCurrentState!=SERVICE_STOP_PENDING ){
	656	+ if( !ControlService(hSvc, SERVICE_CONTROL_STOP, &sstat) ){
	657	+ fossil_fatal(zErrFmt, zSvcName, win32_get_last_errmsg());
	658	+ }
	659	+ }
	660	+ while( sstat.dwCurrentState!=SERVICE_STOPPED ){
	661	+ Sleep(100);
	662	+ fossil_print(".");
	663	+ QueryServiceStatus(hSvc, &sstat);
	664	+ }
	665	+ fossil_print("\nService '%s' stopped.\n", zSvcName);
	666	+ }
	667	+ if( !DeleteService(hSvc) ){
	668	+ if( GetLastError()==ERROR_SERVICE_MARKED_FOR_DELETE ){
	669	+ fossil_warning("Service '%s' already marked for delete.\n", zSvcName);
	670	+ }else{
	671	+ fossil_fatal(zErrFmt, zSvcName, win32_get_last_errmsg());
	672	+ }
	673	+ }else{
	674	+ fossil_print("Service '%s' successfully deleted.\n", zSvcName);
	675	+ }
	676	+ CloseServiceHandle(hSvc);
	677	+ CloseServiceHandle(hScm);
	678	+ }else
	679	+ if( strncmp(zMethod, "show", n)==0 ){
	680	+ SC_HANDLE hScm;
	681	+ SC_HANDLE hSvc;
	682	+ SERVICE_STATUS sstat;
	683	+ LPQUERY_SERVICE_CONFIG pSvcConfig;
	684	+ LPSERVICE_DESCRIPTION pSvcDescr;
	685	+ BOOL bStatus;
	686	+ DWORD nRequired;
	687	+ char *zErrFmt = "unable to show service '%s': %s";
	688	+ static const char *zSvcTypes[] = {
	689	+ "Driver service",
	690	+ "File system driver service",
	691	+ "Service runs in its own process",
	692	+ "Service shares a process with other services",
	693	+ "Service can interact with the desktop"
	694	+ };
	695	+ const char *zSvcType = "";
	696	+ static char *zSvcStartTypes[] = {
	697	+ "Started by the system loader",
	698	+ "Started by the IoInitSystem function",
	699	+ "Started automatically by the service control manager",
	700	+ "Started manually",
	701	+ "Service cannot be started"
	702	+ };
	703	+ const char *zSvcStartType = "";
	704	+ static const char *zSvcStates[] = {
	705	+ "Stopped", "Starting", "Stopping", "Running",
	706	+ "Continue pending", "Pause pending", "Paused"
	707	+ };
	708	+ const char *zSvcState = "";
	709	+
	710	+ verify_all_options();
	711	+ if( g.argc>4 ) fossil_fatal("to much arguments for show method.");
	712	+ hScm = OpenSCManager(NULL, NULL, GENERIC_READ);
	713	+ if( !hScm ) fossil_fatal(zErrFmt, zSvcName, win32_get_last_errmsg());
	714	+ hSvc = OpenService(hScm, fossil_utf8_to_mbcs(zSvcName), GENERIC_READ);
	715	+ if( !hSvc ) fossil_fatal(zErrFmt, zSvcName, win32_get_last_errmsg());
	716	+ /* Get the service configuration */
	717	+ bStatus = QueryServiceConfig(hSvc, NULL, 0, &nRequired);
	718	+ if( !bStatus && GetLastError()!=ERROR_INSUFFICIENT_BUFFER ){
	719	+ fossil_fatal(zErrFmt, zSvcName, win32_get_last_errmsg());
	720	+ }
	721	+ pSvcConfig = fossil_malloc(nRequired);
	722	+ bStatus = QueryServiceConfig(hSvc, pSvcConfig, nRequired, &nRequired);
	723	+ if( !bStatus ) fossil_fatal(zErrFmt, zSvcName, win32_get_last_errmsg());
	724	+ /* Translate the service type */
	725	+ switch( pSvcConfig->dwServiceType ){
	726	+ case SERVICE_KERNEL_DRIVER: zSvcType = zSvcTypes[0]; break;
	727	+ case SERVICE_FILE_SYSTEM_DRIVER: zSvcType = zSvcTypes[1]; break;
	728	+ case SERVICE_WIN32_OWN_PROCESS: zSvcType = zSvcTypes[2]; break;
	729	+ case SERVICE_WIN32_SHARE_PROCESS: zSvcType = zSvcTypes[3]; break;
	730	+ case SERVICE_INTERACTIVE_PROCESS: zSvcType = zSvcTypes[4]; break;
	731	+ }
	732	+ /* Translate the service start type */
	733	+ switch( pSvcConfig->dwStartType ){
	734	+ case SERVICE_BOOT_START: zSvcStartType = zSvcStartTypes[0]; break;
	735	+ case SERVICE_SYSTEM_START: zSvcStartType = zSvcStartTypes[1]; break;
	736	+ case SERVICE_AUTO_START: zSvcStartType = zSvcStartTypes[2]; break;
	737	+ case SERVICE_DEMAND_START: zSvcStartType = zSvcStartTypes[3]; break;
	738	+ case SERVICE_DISABLED: zSvcStartType = zSvcStartTypes[4]; break;
	739	+ }
	740	+ /* Get the service description. */
	741	+ bStatus = QueryServiceConfig2(hSvc, SERVICE_CONFIG_DESCRIPTION,
	742	+ NULL, 0, &nRequired);
	743	+ if( !bStatus && GetLastError()!=ERROR_INSUFFICIENT_BUFFER ){
	744	+ fossil_fatal(zErrFmt, zSvcName, win32_get_last_errmsg());
	745	+ }
	746	+ pSvcDescr = fossil_malloc(nRequired);
	747	+ bStatus = QueryServiceConfig2(hSvc, SERVICE_CONFIG_DESCRIPTION,
	748	+ (LPBYTE)pSvcDescr, nRequired, &nRequired);
	749	+ if( !bStatus ) fossil_fatal(zErrFmt, zSvcName, win32_get_last_errmsg());
	750	+ /* Retrieves the current status of the specified service. */
	751	+ bStatus = QueryServiceStatus(hSvc, &sstat);
	752	+ if( !bStatus ) fossil_fatal(zErrFmt, zSvcName, win32_get_last_errmsg());
	753	+ /* Translate the current state. */
	754	+ switch( sstat.dwCurrentState ){
	755	+ case SERVICE_STOPPED: zSvcState = zSvcStates[0]; break;
	756	+ case SERVICE_START_PENDING: zSvcState = zSvcStates[1]; break;
	757	+ case SERVICE_STOP_PENDING: zSvcState = zSvcStates[2]; break;
	758	+ case SERVICE_RUNNING: zSvcState = zSvcStates[3]; break;
	759	+ case SERVICE_CONTINUE_PENDING: zSvcState = zSvcStates[4]; break;
	760	+ case SERVICE_PAUSE_PENDING: zSvcState = zSvcStates[5]; break;
	761	+ case SERVICE_PAUSED: zSvcState = zSvcStates[6]; break;
	762	+ }
	763	+ /* Print service information to terminal */
	764	+ fossil_print("Service name .......: %s\n", zSvcName);
	765	+ fossil_print("Display name .......: %s\n",
	766	+ fossil_mbcs_to_utf8(pSvcConfig->lpDisplayName));
	767	+ fossil_print("Service description : %s\n",
	768	+ fossil_mbcs_to_utf8(pSvcDescr->lpDescription));
	769	+ fossil_print("Service type .......: %s.\n", zSvcType);
	770	+ fossil_print("Service start type .: %s.\n", zSvcStartType);
	771	+ fossil_print("Binary path name ...: %s\n",
	772	+ fossil_mbcs_to_utf8(pSvcConfig->lpBinaryPathName));
	773	+ fossil_print("Service username ...: %s\n",
	774	+ fossil_mbcs_to_utf8(pSvcConfig->lpServiceStartName));
	775	+ fossil_print("Current state ......: %s.\n", zSvcState);
	776	+ /* Cleanup */
	777	+ fossil_free(pSvcConfig);
	778	+ fossil_free(pSvcDescr);
	779	+ CloseServiceHandle(hSvc);
	780	+ CloseServiceHandle(hScm);
	781	+ }else
	782	+ if( strncmp(zMethod, "start", n)==0 ){
	783	+ SC_HANDLE hScm;
	784	+ SC_HANDLE hSvc;
	785	+ SERVICE_STATUS sstat;
	786	+ char *zErrFmt = "unable to start service '%s': %s";
	787	+
	788	+ verify_all_options();
	789	+ if( g.argc>4 ) fossil_fatal("to much arguments for start method.");
	790	+ hScm = OpenSCManager(NULL, NULL, SC_MANAGER_ALL_ACCESS);
	791	+ if( !hScm ) fossil_fatal(zErrFmt, zSvcName, win32_get_last_errmsg());
	792	+ hSvc = OpenService(hScm, fossil_utf8_to_mbcs(zSvcName), SERVICE_ALL_ACCESS);
	793	+ if( !hSvc ) fossil_fatal(zErrFmt, zSvcName, win32_get_last_errmsg());
	794	+ QueryServiceStatus(hSvc, &sstat);
	795	+ if( sstat.dwCurrentState!=SERVICE_RUNNING ){
	796	+ fossil_print("Starting service '%s'", zSvcName);
	797	+ if( sstat.dwCurrentState!=SERVICE_START_PENDING ){
	798	+ if( !StartService(hSvc, 0, NULL) ){
	799	+ fossil_fatal(zErrFmt, zSvcName, win32_get_last_errmsg());
	800	+ }
	801	+ }
	802	+ while( sstat.dwCurrentState!=SERVICE_RUNNING ){
	803	+ Sleep(100);
	804	+ fossil_print(".");
	805	+ QueryServiceStatus(hSvc, &sstat);
	806	+ }
	807	+ fossil_print("\nService '%s' started.\n", zSvcName);
	808	+ }else{
	809	+ fossil_print("Service '%s' is already started.\n", zSvcName);
	810	+ }
	811	+ CloseServiceHandle(hSvc);
	812	+ CloseServiceHandle(hScm);
	813	+ }else
	814	+ if( strncmp(zMethod, "stop", n)==0 ){
	815	+ SC_HANDLE hScm;
	816	+ SC_HANDLE hSvc;
	817	+ SERVICE_STATUS sstat;
	818	+ char *zErrFmt = "unable to stop service '%s': %s";
	819	+
	820	+ verify_all_options();
	821	+ if( g.argc>4 ) fossil_fatal("to much arguments for stop method.");
	822	+ hScm = OpenSCManager(NULL, NULL, SC_MANAGER_ALL_ACCESS);
	823	+ if( !hScm ) fossil_fatal(zErrFmt, zSvcName, win32_get_last_errmsg());
	824	+ hSvc = OpenService(hScm, fossil_utf8_to_mbcs(zSvcName), SERVICE_ALL_ACCESS);
	825	+ if( !hSvc ) fossil_fatal(zErrFmt, zSvcName, win32_get_last_errmsg());
	826	+ QueryServiceStatus(hSvc, &sstat);
	827	+ if( sstat.dwCurrentState!=SERVICE_STOPPED ){
	828	+ fossil_print("Stopping service '%s'", zSvcName);
	829	+ if( sstat.dwCurrentState!=SERVICE_STOP_PENDING ){
	830	+ if( !ControlService(hSvc, SERVICE_CONTROL_STOP, &sstat) ){
	831	+ fossil_fatal(zErrFmt, zSvcName, win32_get_last_errmsg());
	832	+ }
	833	+ }
	834	+ while( sstat.dwCurrentState!=SERVICE_STOPPED ){
	835	+ Sleep(100);
	836	+ fossil_print(".");
	837	+ QueryServiceStatus(hSvc, &sstat);
	838	+ }
	839	+ fossil_print("\nService '%s' stopped.\n", zSvcName);
	840	+ }else{
	841	+ fossil_print("Service '%s' is already stopped.\n", zSvcName);
	842	+ }
	843	+ CloseServiceHandle(hSvc);
	844	+ CloseServiceHandle(hScm);
	845	+ }else
	846	+ {
	847	+ fossil_fatal("METHOD should be one of:"
	848	+ " create delete show start stop");
	849	+ }
	850	+ return;
	851	+}
	852	+
	853	+#else /* _WIN32 -- This code is for win32 only */
	854	+#include "winhttp.h"
	855	+
	856	+void cmd_win32_service(void){
	857	+ fossil_fatal("The winsrv command is platform specific "
	858	+ "and not available on this platform.");
	859	+ return;
	860	+}
226	861
227	862	#endif /* _WIN32 -- This code is for win32 only */
228	863

	--- src/winhttp.c
	+++ src/winhttp.c
	@@ -14,17 +14,18 @@
14	** http://www.hwaci.com/drh/
15	**
16	*******************************************************************************
17	**
18	** This file implements a very simple (and low-performance) HTTP server
19	** for windows.

20	*/
21	#include "config.h"
22	#ifdef _WIN32
23	/* This code is for win32 only */
24	#include "winhttp.h"
25	#include <windows.h>

26
27	/*
28	** The HttpRequest structure holds information about each incoming
29	** HTTP request.
30	*/
	@@ -60,11 +61,11 @@
60	}
61
62	/*
63	** Process a single incoming HTTP request.
64	*/
65	void win32_process_one_http_request(void *pAppData){
66	HttpRequest p = (HttpRequest)pAppData;
67	FILE in = 0, out = 0;
68	int amt, got;
69	int wanted = 0;
70	char *z;
	@@ -145,10 +146,11 @@
145	SOCKET s = INVALID_SOCKET;
146	SOCKADDR_IN addr;
147	int idCnt = 0;
148	int iPort = mnPort;
149	Blob options;

150
151	if( zStopper ) file_delete(zStopper);
152	blob_zero(&options);
153	if( zNotFound ){
154	blob_appendf(&options, " --notfound %s", zNotFound);
	@@ -189,31 +191,46 @@
189	}else{
190	fossil_fatal("unable to open listening socket on any"
191	" port in the range %d..%d", mnPort, mxPort);
192	}
193	}
194	zTempPrefix = mprintf("fossil_server_P%d_", iPort);



195	fossil_print("Listening for HTTP requests on TCP port %d\n", iPort);
196	if( zBrowser ){
197	zBrowser = mprintf(zBrowser, iPort);
198	fossil_print("Launch webbrowser: %s\n", zBrowser);
199	fossil_system(zBrowser);
200	}
201	fossil_print("Type Ctrl-C to stop the HTTP server\n");



202	for(;;){
203	SOCKET client;
204	SOCKADDR_IN client_addr;
205	HttpRequest *p;
206	int len = sizeof(client_addr);

207
208	client = accept(s, (struct sockaddr*)&client_addr, &len);
209	if( zStopper && file_size(zStopper)>=0 ){
210	break;
211	}
212	if( client==INVALID_SOCKET ){
213	closesocket(s);
214	fossil_fatal("error from accept()");











215	}
216	p = fossil_malloc( sizeof(*p) );
217	p->id = ++idCnt;
218	p->s = client;
219	p->addr = client_addr;
	@@ -221,7 +238,625 @@
221	_beginthread(win32_process_one_http_request, 0, (void*)p);
222	}
223	closesocket(s);
224	WSACleanup();
225	}










































































































































































































































































































































































































































































































































































































































226
227	#endif /* _WIN32 -- This code is for win32 only */
228

	--- src/winhttp.c
	+++ src/winhttp.c
	@@ -14,17 +14,18 @@
14	** http://www.hwaci.com/drh/
15	**
16	*******************************************************************************
17	**
18	** This file implements a very simple (and low-performance) HTTP server
19	** for windows. It also implements a Windows Service which allows the HTTP
20	** server to be run without any user logged on.
21	*/
22	#include "config.h"
23	#ifdef _WIN32
24	/* This code is for win32 only */

25	#include <windows.h>
26	#include "winhttp.h"
27
28	/*
29	** The HttpRequest structure holds information about each incoming
30	** HTTP request.
31	*/
	@@ -60,11 +61,11 @@
61	}
62
63	/*
64	** Process a single incoming HTTP request.
65	*/
66	static void win32_process_one_http_request(void *pAppData){
67	HttpRequest p = (HttpRequest)pAppData;
68	FILE in = 0, out = 0;
69	int amt, got;
70	int wanted = 0;
71	char *z;
	@@ -145,10 +146,11 @@
146	SOCKET s = INVALID_SOCKET;
147	SOCKADDR_IN addr;
148	int idCnt = 0;
149	int iPort = mnPort;
150	Blob options;
151	char zTmpPath[MAX_PATH];
152
153	if( zStopper ) file_delete(zStopper);
154	blob_zero(&options);
155	if( zNotFound ){
156	blob_appendf(&options, " --notfound %s", zNotFound);
	@@ -189,31 +191,46 @@
191	}else{
192	fossil_fatal("unable to open listening socket on any"
193	" port in the range %d..%d", mnPort, mxPort);
194	}
195	}
196	if( !GetTempPath(sizeof(zTmpPath), zTmpPath) ){
197	fossil_fatal("unable to get path to the temporary directory.");
198	}
199	zTempPrefix = mprintf("%sfossil_server_P%d_", zTmpPath, iPort);
200	fossil_print("Listening for HTTP requests on TCP port %d\n", iPort);
201	if( zBrowser ){
202	zBrowser = mprintf(zBrowser, iPort);
203	fossil_print("Launch webbrowser: %s\n", zBrowser);
204	fossil_system(zBrowser);
205	}
206	fossil_print("Type Ctrl-C to stop the HTTP server\n");
207	/* Set the service status to running and pass the listener socket to the
208	** service handling procedures. */
209	win32_http_service_running(s);
210	for(;;){
211	SOCKET client;
212	SOCKADDR_IN client_addr;
213	HttpRequest *p;
214	int len = sizeof(client_addr);
215	int wsaError;
216
217	client = accept(s, (struct sockaddr*)&client_addr, &len);



218	if( client==INVALID_SOCKET ){
219	/* If the service control handler has closed the listener socket,
220	** cleanup and return, otherwise report a fatal error. */
221	wsaError = WSAGetLastError();
222	if( (wsaError==WSAEINTR) \|\| (wsaError==WSAENOTSOCK) ){
223	WSACleanup();
224	return;
225	}else{
226	closesocket(s);
227	WSACleanup();
228	fossil_fatal("error from accept()");
229	}
230	}else if( zStopper && file_size(zStopper)>=0 ){
231	break;
232	}
233	p = fossil_malloc( sizeof(*p) );
234	p->id = ++idCnt;
235	p->s = client;
236	p->addr = client_addr;
	@@ -221,7 +238,625 @@
238	_beginthread(win32_process_one_http_request, 0, (void*)p);
239	}
240	closesocket(s);
241	WSACleanup();
242	}
243
244	/*
245	** The HttpService structure is used to pass information to the service main
246	** function and to the service control handler function.
247	*/
248	typedef struct HttpService HttpService;
249	struct HttpService {
250	int port; /* Port on which the http server should run */
251	const char zNotFound; / The --notfound option, or NULL */
252	int flags; /* One or more HTTP_SERVER_ flags */
253	int isRunningAsService; /* Are we running as a service ? */
254	const char zServiceName; / Name of the service */
255	SOCKET s; /* Socket on which the http server listens */
256	};
257
258	/*
259	** Variables used for running as windows service.
260	*/
261	static HttpService hsData = {8080, NULL, 0, 0, NULL, INVALID_SOCKET};
262	static SERVICE_STATUS ssStatus;
263	static SERVICE_STATUS_HANDLE sshStatusHandle;
264
265	/*
266	** Get message string of the last system error. Return a pointer to the
267	** message string. Call fossil_mbcs_free() to deallocate any memory used
268	** to store the message string when done.
269	*/
270	static char *win32_get_last_errmsg(void){
271	DWORD nMsg;
272	LPTSTR tmp = NULL;
273	char *zMsg = NULL;
274
275	nMsg = FormatMessage(
276	FORMAT_MESSAGE_ALLOCATE_BUFFER \|
277	FORMAT_MESSAGE_FROM_SYSTEM \|
278	FORMAT_MESSAGE_IGNORE_INSERTS,
279	NULL,
280	GetLastError(),
281	MAKELANGID(LANG_ENGLISH, SUBLANG_ENGLISH_US),
282	(LPTSTR) &tmp,
283	0,
284	NULL
285	);
286	if( nMsg ){
287	zMsg = fossil_mbcs_to_utf8(tmp);
288	}else{
289	fossil_fatal("unable to get system error message.");
290	}
291	if( tmp ){
292	LocalFree((HLOCAL) tmp);
293	}
294	return zMsg;
295	}
296
297	/*
298	** Report the current status of the service to the service control manager.
299	** Make sure that during service startup no control codes are accepted.
300	*/
301	static void win32_report_service_status(
302	DWORD dwCurrentState, /* The current state of the service */
303	DWORD dwWin32ExitCode, /* The error code to report */
304	DWORD dwWaitHint /* The estimated time for a pending operation */
305	){
306	if( dwCurrentState==SERVICE_START_PENDING) {
307	ssStatus.dwControlsAccepted = 0;
308	}else{
309	ssStatus.dwControlsAccepted = SERVICE_ACCEPT_STOP;
310	}
311	ssStatus.dwCurrentState = dwCurrentState;
312	ssStatus.dwWin32ExitCode = dwWin32ExitCode;
313	ssStatus.dwWaitHint = dwWaitHint;
314
315	if( (dwCurrentState==SERVICE_RUNNING) \|\|
316	(dwCurrentState==SERVICE_STOPPED) ){
317	ssStatus.dwCheckPoint = 0;
318	}else{
319	ssStatus.dwCheckPoint++;
320	}
321	SetServiceStatus(sshStatusHandle, &ssStatus);
322	return ;
323	}
324
325	/*
326	** Handle control codes sent from the service control manager.
327	** The control dispatcher in the main thread of the service process invokes
328	** this function whenever it receives a control request from the service
329	** control manager.
330	*/
331	static void WINAPI win32_http_service_ctrl(
332	DWORD dwCtrlCode
333	){
334	switch( dwCtrlCode ){
335	case SERVICE_CONTROL_STOP: {
336	win32_report_service_status(SERVICE_STOP_PENDING, NO_ERROR, 0);
337	if( hsData.s != INVALID_SOCKET ){
338	closesocket(hsData.s);
339	}
340	win32_report_service_status(ssStatus.dwCurrentState, NO_ERROR, 0);
341	break;
342	}
343	default: {
344	break;
345	}
346	}
347	return;
348	}
349
350	/*
351	** This is the main entry point for the service.
352	** When the service control manager receives a request to start the service,
353	** it starts the service process (if it is not already running). The main
354	** thread of the service process calls the StartServiceCtrlDispatcher
355	** function with a pointer to an array of SERVICE_TABLE_ENTRY structures.
356	** Then the service control manager sends a start request to the service
357	** control dispatcher for this service process. The service control dispatcher
358	** creates a new thread to execute the ServiceMain function (this function)
359	** of the service being started.
360	*/
361	static void WINAPI win32_http_service_main(
362	DWORD argc, /* Number of arguments in argv */
363	LPTSTR argv / Arguments passed */
364	){
365
366	/* Update the service information. */
367	hsData.isRunningAsService = 1;
368	if( argc>0 ){
369	hsData.zServiceName = argv[0];
370	}
371
372	/* Register the service control handler function */
373	sshStatusHandle = RegisterServiceCtrlHandler("", win32_http_service_ctrl);
374	if( !sshStatusHandle ){
375	win32_report_service_status(SERVICE_STOPPED, NO_ERROR, 0);
376	return;
377	}
378
379	/* Set service specific data and report that the service is starting. */
380	ssStatus.dwServiceType = SERVICE_WIN32_OWN_PROCESS;
381	ssStatus.dwServiceSpecificExitCode = 0;
382	win32_report_service_status(SERVICE_START_PENDING, NO_ERROR, 3000);
383
384	/* Execute the http server */
385	win32_http_server(hsData.port, hsData.port,
386	NULL, NULL, hsData.zNotFound, hsData.flags);
387
388	/* Service has stopped now. */
389	win32_report_service_status(SERVICE_STOPPED, NO_ERROR, 0);
390	return;
391	}
392
393	/*
394	** When running as service, update the HttpService structure with the
395	** listener socket and update the service status. This procedure must be
396	** called from the http server when he is ready to accept connections.
397	*/
398	LOCAL void win32_http_service_running(SOCKET s){
399	if( hsData.isRunningAsService ){
400	hsData.s = s;
401	win32_report_service_status(SERVICE_RUNNING, NO_ERROR, 0);
402	}
403	}
404
405	/*
406	** Try to start the http server as a windows service. If we are running in
407	** a interactive console session, this routine fails and returns a non zero
408	** integer value. When running as service, this routine does not return until
409	** the service is stopped. In this case, the return value is zero.
410	*/
411	int win32_http_service(
412	int nPort, /* TCP port number */
413	const char zNotFound, / The --notfound option, or NULL */
414	int flags /* One or more HTTP_SERVER_ flags */
415	){
416	/* Define the service table. */
417	SERVICE_TABLE_ENTRY ServiceTable[] =
418	{{"", (LPSERVICE_MAIN_FUNCTION)win32_http_service_main}, {NULL, NULL}};
419
420	/* Initialize the HttpService structure. */
421	hsData.port = nPort;
422	hsData.zNotFound = zNotFound;
423	hsData.flags = flags;
424
425	/* Try to start the control dispatcher thread for the service. */
426	if( !StartServiceCtrlDispatcher(ServiceTable) ){
427	if( GetLastError()==ERROR_FAILED_SERVICE_CONTROLLER_CONNECT ){
428	return 1;
429	}else{
430	fossil_fatal("error from StartServiceCtrlDispatcher()");
431	}
432	}
433	return 0;
434	}
435
436	/*
437	** COMMAND: winsrv
438	** Usage: fossil winsrv METHOD ?SERVICE-NAME? ?OPTIONS?
439	**
440	** Where METHOD is one of: create delete show start stop.
441	**
442	** The winsrv command manages Fossil as a Windows service. This allows
443	** (for example) Fossil to be running in the background when no user
444	** is logged in.
445	**
446	** In the following description of the methods, "Fossil-DSCM" will be
447	** used as the default SERVICE-NAME:
448	**
449	** fossil winsrv create ?SERVICE-NAME? ?OPTIONS?
450	**
451	** Creates a service. Available options include:
452	**
453	** -D\|--display DISPLAY-NAME
454	**
455	** Sets the display name of the service. This name is shown
456	** by graphical interface programs. By default, the display name
457	** equals to the service name.
458	**
459	** -S\|--start TYPE
460	**
461	** Sets the start type of the service. TYPE can be "manual",
462	** which means you need to start the service yourself with the
463	** 'fossil winsrv start' command or with the "net start" command
464	** from the operating system. If TYPE is set to "auto", the service
465	** will be started automatically by the system during startup.
466	**
467	** -U\|--username USERNAME
468	**
469	** Specifies the user account which will be used to run the
470	** service. The account needs the "Logon as a service" right
471	** enabled in its profile. Specify local accounts as follows:
472	** ".\\USERNAME". By default, the "LocalSystem" account will be
473	** used.
474	**
475	** -W\|--password PASSWORD
476	**
477	** Password for the user account.
478	**
479	** The following options are more or less the same as for the "server"
480	** command and influence the behaviour of the http server:
481	**
482	** -p\|--port TCPPORT
483	**
484	** Specifies the TCP port (default port is 8080) on which the
485	** server should listen.
486	**
487	** -R\|--repository REPOSITORY
488	**
489	** Specifies the name of the repository to be served.
490	** The repository option may be omitted if the working directory
491	** is within an open checkout.
492	** The REPOSITORY can be a directory (aka folder) that contains
493	** one or more respositories with names ending in ".fossil".
494	** In that case, the first element of the URL is used to select
495	** among the various repositories.
496	**
497	** --notfound URL
498	**
499	** If REPOSITORY is a directory that contains one or more
500	** respositories with names of the form "*.fossil" then the
501	** first element of the URL pathname selects among the various
502	** repositories. If the pathname does not select a valid
503	** repository and the --notfound option is available,
504	** then the server redirects (HTTP code 302) to the URL of
505	** --notfound.
506	**
507	** --localauth
508	**
509	** Enables automatic login if the --localauth option is present
510	** and the "localauth" setting is off and the connection is from
511	** localhost.
512	**
513	**
514	** fossil winsrv delete ?SERVICE-NAME?
515	**
516	** Deletes a service. If the service is currently running, it will be
517	** stopped first and then deleted.
518	**
519	**
520	** fossil winsrv show ?SERVICE-NAME?
521	**
522	** Shows how the service is configured and its current state.
523	**
524	**
525	** fossil winsrv start ?SERVICE-NAME?
526	**
527	** Start the service.
528	**
529	**
530	** fossil winsrv stop ?SERVICE-NAME?
531	**
532	** Stop the service.
533	**
534	**
535	** NOTE: This command is available on Windows operating systems only and
536	** requires administrative rights on the machine executed.
537	**
538	*/
539	void cmd_win32_service(void){
540	int n;
541	const char *zMethod;
542	const char zSvcName = "Fossil-DSCM"; / Default service name */
543
544	if( g.argc<3 ){
545	usage("create\|delete\|show\|start\|stop ...");
546	}
547	zMethod = g.argv[2];
548	n = strlen(zMethod);
549	if( g.argc==4 ){
550	zSvcName = g.argv[3];
551	}
552
553	if( strncmp(zMethod, "create", n)==0 ){
554	SC_HANDLE hScm;
555	SC_HANDLE hSvc;
556	SERVICE_DESCRIPTION
557	svcDescr = {"Fossil - Distributed Software Configuration Management"};
558	char *zErrFmt = "unable to create service '%s': %s";
559	DWORD dwStartType = SERVICE_DEMAND_START;
560	const char *zDisplay;
561	const char *zStart;
562	const char *zUsername;
563	const char *zPassword;
564	const char *zPort;
565	const char *zNotFound;
566	const char *zLocalAuth;
567	const char *zRepository;
568	Blob binPath;
569
570	/* Process service creation specific options. */
571	zDisplay = find_option("display", "D", 1);
572	if( !zDisplay ){
573	zDisplay = zSvcName;
574	}
575	zStart = find_option("start", "S", 1);
576	if( zStart ){
577	if( strncmp(zStart, "auto", strlen(zStart))==0 ){
578	dwStartType = SERVICE_AUTO_START;
579	}else if( strncmp(zStart, "manual", strlen(zStart))==0 ){
580	dwStartType = SERVICE_DEMAND_START;
581	}else{
582	fossil_fatal(zErrFmt, zSvcName,
583	"specify 'auto' or 'manual' for the '-S\|--start' option");
584	}
585	}
586	zUsername = find_option("username", "U", 1);
587	zPassword = find_option("password", "W", 1);
588	/* Process options for Fossil running as server. */
589	zPort = find_option("port", "P", 1);
590	if( zPort && (atoi(zPort)<=0) ){
591	fossil_fatal(zErrFmt, zSvcName,
592	"port number must be in the range 1 - 65535.");
593	}
594	zNotFound = find_option("notfound", 0, 1);
595	zLocalAuth = find_option("localauth", 0, 0);
596	zRepository = find_option("repository", "R", 1);
597	if( !zRepository ){
598	db_must_be_within_tree();
599	}else if( file_isdir(zRepository)==1 ){
600	g.zRepositoryName = mprintf("%s", zRepository);
601	file_simplify_name(g.zRepositoryName, -1);
602	}else{
603	db_open_repository(zRepository);
604	}
605	db_close(0);
606	verify_all_options();
607	if( g.argc>4 ) fossil_fatal("to much arguments for create method.");
608	/* Build the fully-qualified path to the service binary file. */
609	blob_zero(&binPath);
610	blob_appendf(&binPath, "\"%s\" server", fossil_nameofexe());
611	if( zPort ) blob_appendf(&binPath, " --port %s", zPort);
612	if( zNotFound ) blob_appendf(&binPath, " --notfound \"%s\"", zNotFound);
613	if( zLocalAuth ) blob_append(&binPath, " --localauth", -1);
614	blob_appendf(&binPath, " \"%s\"", g.zRepositoryName);
615	/* Create the service. */
616	hScm = OpenSCManager(NULL, NULL, SC_MANAGER_ALL_ACCESS);
617	if( !hScm ) fossil_fatal(zErrFmt, zSvcName, win32_get_last_errmsg());
618	hSvc = CreateService(
619	hScm, /* Handle to the SCM */
620	fossil_utf8_to_mbcs(zSvcName), /* Name of the service */
621	fossil_utf8_to_mbcs(zDisplay), /* Display name */
622	SERVICE_ALL_ACCESS, /* Desired access */
623	SERVICE_WIN32_OWN_PROCESS, /* Service type */
624	dwStartType, /* Start type */
625	SERVICE_ERROR_NORMAL, /* Error control */
626	fossil_utf8_to_mbcs(blob_str(&binPath)), /* Binary path */
627	NULL, /* Load ordering group */
628	NULL, /* Tag value */
629	NULL, /* Service dependencies */
630	fossil_utf8_to_mbcs(zUsername), /* Service account */
631	fossil_utf8_to_mbcs(zPassword) /* Account password */
632	);
633	if( !hSvc ) fossil_fatal(zErrFmt, zSvcName, win32_get_last_errmsg());
634	/* Set the service description. */
635	ChangeServiceConfig2(hSvc, SERVICE_CONFIG_DESCRIPTION, &svcDescr);
636	fossil_print("Service '%s' successfully created.\n", zSvcName);
637	CloseServiceHandle(hSvc);
638	CloseServiceHandle(hScm);
639	}else
640	if( strncmp(zMethod, "delete", n)==0 ){
641	SC_HANDLE hScm;
642	SC_HANDLE hSvc;
643	SERVICE_STATUS sstat;
644	char *zErrFmt = "unable to delete service '%s': %s";
645
646	verify_all_options();
647	if( g.argc>4 ) fossil_fatal("to much arguments for delete method.");
648	hScm = OpenSCManager(NULL, NULL, SC_MANAGER_ALL_ACCESS);
649	if( !hScm ) fossil_fatal(zErrFmt, zSvcName, win32_get_last_errmsg());
650	hSvc = OpenService(hScm, fossil_utf8_to_mbcs(zSvcName), SERVICE_ALL_ACCESS);
651	if( !hSvc ) fossil_fatal(zErrFmt, zSvcName, win32_get_last_errmsg());
652	QueryServiceStatus(hSvc, &sstat);
653	if( sstat.dwCurrentState!=SERVICE_STOPPED ){
654	fossil_print("Stopping service '%s'", zSvcName);
655	if( sstat.dwCurrentState!=SERVICE_STOP_PENDING ){
656	if( !ControlService(hSvc, SERVICE_CONTROL_STOP, &sstat) ){
657	fossil_fatal(zErrFmt, zSvcName, win32_get_last_errmsg());
658	}
659	}
660	while( sstat.dwCurrentState!=SERVICE_STOPPED ){
661	Sleep(100);
662	fossil_print(".");
663	QueryServiceStatus(hSvc, &sstat);
664	}
665	fossil_print("\nService '%s' stopped.\n", zSvcName);
666	}
667	if( !DeleteService(hSvc) ){
668	if( GetLastError()==ERROR_SERVICE_MARKED_FOR_DELETE ){
669	fossil_warning("Service '%s' already marked for delete.\n", zSvcName);
670	}else{
671	fossil_fatal(zErrFmt, zSvcName, win32_get_last_errmsg());
672	}
673	}else{
674	fossil_print("Service '%s' successfully deleted.\n", zSvcName);
675	}
676	CloseServiceHandle(hSvc);
677	CloseServiceHandle(hScm);
678	}else
679	if( strncmp(zMethod, "show", n)==0 ){
680	SC_HANDLE hScm;
681	SC_HANDLE hSvc;
682	SERVICE_STATUS sstat;
683	LPQUERY_SERVICE_CONFIG pSvcConfig;
684	LPSERVICE_DESCRIPTION pSvcDescr;
685	BOOL bStatus;
686	DWORD nRequired;
687	char *zErrFmt = "unable to show service '%s': %s";
688	static const char *zSvcTypes[] = {
689	"Driver service",
690	"File system driver service",
691	"Service runs in its own process",
692	"Service shares a process with other services",
693	"Service can interact with the desktop"
694	};
695	const char *zSvcType = "";
696	static char *zSvcStartTypes[] = {
697	"Started by the system loader",
698	"Started by the IoInitSystem function",
699	"Started automatically by the service control manager",
700	"Started manually",
701	"Service cannot be started"
702	};
703	const char *zSvcStartType = "";
704	static const char *zSvcStates[] = {
705	"Stopped", "Starting", "Stopping", "Running",
706	"Continue pending", "Pause pending", "Paused"
707	};
708	const char *zSvcState = "";
709
710	verify_all_options();
711	if( g.argc>4 ) fossil_fatal("to much arguments for show method.");
712	hScm = OpenSCManager(NULL, NULL, GENERIC_READ);
713	if( !hScm ) fossil_fatal(zErrFmt, zSvcName, win32_get_last_errmsg());
714	hSvc = OpenService(hScm, fossil_utf8_to_mbcs(zSvcName), GENERIC_READ);
715	if( !hSvc ) fossil_fatal(zErrFmt, zSvcName, win32_get_last_errmsg());
716	/* Get the service configuration */
717	bStatus = QueryServiceConfig(hSvc, NULL, 0, &nRequired);
718	if( !bStatus && GetLastError()!=ERROR_INSUFFICIENT_BUFFER ){
719	fossil_fatal(zErrFmt, zSvcName, win32_get_last_errmsg());
720	}
721	pSvcConfig = fossil_malloc(nRequired);
722	bStatus = QueryServiceConfig(hSvc, pSvcConfig, nRequired, &nRequired);
723	if( !bStatus ) fossil_fatal(zErrFmt, zSvcName, win32_get_last_errmsg());
724	/* Translate the service type */
725	switch( pSvcConfig->dwServiceType ){
726	case SERVICE_KERNEL_DRIVER: zSvcType = zSvcTypes[0]; break;
727	case SERVICE_FILE_SYSTEM_DRIVER: zSvcType = zSvcTypes[1]; break;
728	case SERVICE_WIN32_OWN_PROCESS: zSvcType = zSvcTypes[2]; break;
729	case SERVICE_WIN32_SHARE_PROCESS: zSvcType = zSvcTypes[3]; break;
730	case SERVICE_INTERACTIVE_PROCESS: zSvcType = zSvcTypes[4]; break;
731	}
732	/* Translate the service start type */
733	switch( pSvcConfig->dwStartType ){
734	case SERVICE_BOOT_START: zSvcStartType = zSvcStartTypes[0]; break;
735	case SERVICE_SYSTEM_START: zSvcStartType = zSvcStartTypes[1]; break;
736	case SERVICE_AUTO_START: zSvcStartType = zSvcStartTypes[2]; break;
737	case SERVICE_DEMAND_START: zSvcStartType = zSvcStartTypes[3]; break;
738	case SERVICE_DISABLED: zSvcStartType = zSvcStartTypes[4]; break;
739	}
740	/* Get the service description. */
741	bStatus = QueryServiceConfig2(hSvc, SERVICE_CONFIG_DESCRIPTION,
742	NULL, 0, &nRequired);
743	if( !bStatus && GetLastError()!=ERROR_INSUFFICIENT_BUFFER ){
744	fossil_fatal(zErrFmt, zSvcName, win32_get_last_errmsg());
745	}
746	pSvcDescr = fossil_malloc(nRequired);
747	bStatus = QueryServiceConfig2(hSvc, SERVICE_CONFIG_DESCRIPTION,
748	(LPBYTE)pSvcDescr, nRequired, &nRequired);
749	if( !bStatus ) fossil_fatal(zErrFmt, zSvcName, win32_get_last_errmsg());
750	/* Retrieves the current status of the specified service. */
751	bStatus = QueryServiceStatus(hSvc, &sstat);
752	if( !bStatus ) fossil_fatal(zErrFmt, zSvcName, win32_get_last_errmsg());
753	/* Translate the current state. */
754	switch( sstat.dwCurrentState ){
755	case SERVICE_STOPPED: zSvcState = zSvcStates[0]; break;
756	case SERVICE_START_PENDING: zSvcState = zSvcStates[1]; break;
757	case SERVICE_STOP_PENDING: zSvcState = zSvcStates[2]; break;
758	case SERVICE_RUNNING: zSvcState = zSvcStates[3]; break;
759	case SERVICE_CONTINUE_PENDING: zSvcState = zSvcStates[4]; break;
760	case SERVICE_PAUSE_PENDING: zSvcState = zSvcStates[5]; break;
761	case SERVICE_PAUSED: zSvcState = zSvcStates[6]; break;
762	}
763	/* Print service information to terminal */
764	fossil_print("Service name .......: %s\n", zSvcName);
765	fossil_print("Display name .......: %s\n",
766	fossil_mbcs_to_utf8(pSvcConfig->lpDisplayName));
767	fossil_print("Service description : %s\n",
768	fossil_mbcs_to_utf8(pSvcDescr->lpDescription));
769	fossil_print("Service type .......: %s.\n", zSvcType);
770	fossil_print("Service start type .: %s.\n", zSvcStartType);
771	fossil_print("Binary path name ...: %s\n",
772	fossil_mbcs_to_utf8(pSvcConfig->lpBinaryPathName));
773	fossil_print("Service username ...: %s\n",
774	fossil_mbcs_to_utf8(pSvcConfig->lpServiceStartName));
775	fossil_print("Current state ......: %s.\n", zSvcState);
776	/* Cleanup */
777	fossil_free(pSvcConfig);
778	fossil_free(pSvcDescr);
779	CloseServiceHandle(hSvc);
780	CloseServiceHandle(hScm);
781	}else
782	if( strncmp(zMethod, "start", n)==0 ){
783	SC_HANDLE hScm;
784	SC_HANDLE hSvc;
785	SERVICE_STATUS sstat;
786	char *zErrFmt = "unable to start service '%s': %s";
787
788	verify_all_options();
789	if( g.argc>4 ) fossil_fatal("to much arguments for start method.");
790	hScm = OpenSCManager(NULL, NULL, SC_MANAGER_ALL_ACCESS);
791	if( !hScm ) fossil_fatal(zErrFmt, zSvcName, win32_get_last_errmsg());
792	hSvc = OpenService(hScm, fossil_utf8_to_mbcs(zSvcName), SERVICE_ALL_ACCESS);
793	if( !hSvc ) fossil_fatal(zErrFmt, zSvcName, win32_get_last_errmsg());
794	QueryServiceStatus(hSvc, &sstat);
795	if( sstat.dwCurrentState!=SERVICE_RUNNING ){
796	fossil_print("Starting service '%s'", zSvcName);
797	if( sstat.dwCurrentState!=SERVICE_START_PENDING ){
798	if( !StartService(hSvc, 0, NULL) ){
799	fossil_fatal(zErrFmt, zSvcName, win32_get_last_errmsg());
800	}
801	}
802	while( sstat.dwCurrentState!=SERVICE_RUNNING ){
803	Sleep(100);
804	fossil_print(".");
805	QueryServiceStatus(hSvc, &sstat);
806	}
807	fossil_print("\nService '%s' started.\n", zSvcName);
808	}else{
809	fossil_print("Service '%s' is already started.\n", zSvcName);
810	}
811	CloseServiceHandle(hSvc);
812	CloseServiceHandle(hScm);
813	}else
814	if( strncmp(zMethod, "stop", n)==0 ){
815	SC_HANDLE hScm;
816	SC_HANDLE hSvc;
817	SERVICE_STATUS sstat;
818	char *zErrFmt = "unable to stop service '%s': %s";
819
820	verify_all_options();
821	if( g.argc>4 ) fossil_fatal("to much arguments for stop method.");
822	hScm = OpenSCManager(NULL, NULL, SC_MANAGER_ALL_ACCESS);
823	if( !hScm ) fossil_fatal(zErrFmt, zSvcName, win32_get_last_errmsg());
824	hSvc = OpenService(hScm, fossil_utf8_to_mbcs(zSvcName), SERVICE_ALL_ACCESS);
825	if( !hSvc ) fossil_fatal(zErrFmt, zSvcName, win32_get_last_errmsg());
826	QueryServiceStatus(hSvc, &sstat);
827	if( sstat.dwCurrentState!=SERVICE_STOPPED ){
828	fossil_print("Stopping service '%s'", zSvcName);
829	if( sstat.dwCurrentState!=SERVICE_STOP_PENDING ){
830	if( !ControlService(hSvc, SERVICE_CONTROL_STOP, &sstat) ){
831	fossil_fatal(zErrFmt, zSvcName, win32_get_last_errmsg());
832	}
833	}
834	while( sstat.dwCurrentState!=SERVICE_STOPPED ){
835	Sleep(100);
836	fossil_print(".");
837	QueryServiceStatus(hSvc, &sstat);
838	}
839	fossil_print("\nService '%s' stopped.\n", zSvcName);
840	}else{
841	fossil_print("Service '%s' is already stopped.\n", zSvcName);
842	}
843	CloseServiceHandle(hSvc);
844	CloseServiceHandle(hScm);
845	}else
846	{
847	fossil_fatal("METHOD should be one of:"
848	" create delete show start stop");
849	}
850	return;
851	}
852
853	#else /* _WIN32 -- This code is for win32 only */
854	#include "winhttp.h"
855
856	void cmd_win32_service(void){
857	fossil_fatal("The winsrv command is platform specific "
858	"and not available on this platform.");
859	return;
860	}
861
862	#endif /* _WIN32 -- This code is for win32 only */
863

M test/release-checklist.wiki

+7 -1

		--- test/release-checklist.wiki
		+++ test/release-checklist.wiki
		@@ -21,10 +21,11 @@
21	21	<li> Linux x86
22	22	<li> Linux x86_64
23	23	<li> Mac x86
24	24	<li> Mac x86_64
25	25	<li> Windows (mingw)
	26	+<li> Windows (vc++)
26	27	<li> OpenBSD
27	28	</ol>
28	29
29	30	<li><p>
30	31	Run at least one occurrence of the the following commands on every
		@@ -56,11 +57,16 @@
56	57
57	58	<li><p>
58	59	Use the release candidate version of fossil in production on the
59	60	[http://www.fossil-scm.org/] website for at least 48 hours (without
60	61	incident) prior to making the release official.
	62	+
	63	+<li><p>
	64	+Verify that the [../www/changes.wiki \| Change Log] is correct and
	65	+up-to-date.
61	66	</ol>
62	67
63	68	<hr>
64	69
65	70	Upon successful completion of all tests above, tag the release candidate
66		-with the "release" tag and set its background color to "#d0c0ff".
	71	+with the "release" tag and set its background color to "#d0c0ff". Update
	72	+the www/changes.wiki file to show the date of the release.
67	73

	--- test/release-checklist.wiki
	+++ test/release-checklist.wiki
	@@ -21,10 +21,11 @@
21	<li> Linux x86
22	<li> Linux x86_64
23	<li> Mac x86
24	<li> Mac x86_64
25	<li> Windows (mingw)

26	<li> OpenBSD
27	</ol>
28
29	<li><p>
30	Run at least one occurrence of the the following commands on every
	@@ -56,11 +57,16 @@
56
57	<li><p>
58	Use the release candidate version of fossil in production on the
59	[http://www.fossil-scm.org/] website for at least 48 hours (without
60	incident) prior to making the release official.




61	</ol>
62
63	<hr>
64
65	Upon successful completion of all tests above, tag the release candidate
66	with the "release" tag and set its background color to "#d0c0ff".

67

	--- test/release-checklist.wiki
	+++ test/release-checklist.wiki
	@@ -21,10 +21,11 @@
21	<li> Linux x86
22	<li> Linux x86_64
23	<li> Mac x86
24	<li> Mac x86_64
25	<li> Windows (mingw)
26	<li> Windows (vc++)
27	<li> OpenBSD
28	</ol>
29
30	<li><p>
31	Run at least one occurrence of the the following commands on every
	@@ -56,11 +57,16 @@
57
58	<li><p>
59	Use the release candidate version of fossil in production on the
60	[http://www.fossil-scm.org/] website for at least 48 hours (without
61	incident) prior to making the release official.
62
63	<li><p>
64	Verify that the [../www/changes.wiki \| Change Log] is correct and
65	up-to-date.
66	</ol>
67
68	<hr>
69
70	Upon successful completion of all tests above, tag the release candidate
71	with the "release" tag and set its background color to "#d0c0ff". Update
72	the www/changes.wiki file to show the date of the release.
73

M win/Makefile.PellesCGMake

+3 -3

		--- win/Makefile.PellesCGMake
		+++ win/Makefile.PellesCGMake
		@@ -126,11 +126,11 @@
126	126	# compiling standard fossil utils
127	127	$(UTILS_OBJ): %.obj: $(SRCDIR)%.c
128	128	$(CC) $(CCFLAGS) $(INCLUDE) "$<" -Fo"$@"
129	129
130	130	# compile special windows utils
131		-version.obj: $(WINDIR)version.c
	131	+version.obj: $(SRCDIR)mkversion.c
132	132	$(CC) $(CCFLAGS) $(INCLUDE) "$<" -Fo"$@"
133	133
134	134	# generate the translated c-source files
135	135	$(TRANSLATEDSRC): %_.c: $(SRCDIR)%.c translate.exe
136	136	translate.exe $< >$@
		@@ -138,12 +138,12 @@
138	138	# generate the index source, containing all web references,..
139	139	page_index.h: $(TRANSLATEDSRC) mkindex.exe
140	140	mkindex.exe $(TRANSLATEDSRC) >$@
141	141
142	142	# extracting version info from manifest
143		-VERSION.h: version.exe ..\manifest.uuid ..\manifest
144		- version.exe ..\manifest.uuid ..\manifest > $@
	143	+VERSION.h: version.exe ..\manifest.uuid ..\manifest ..\VERSION
	144	+ version.exe ..\manifest.uuid ..\manifest ..\VERSION > $@
145	145
146	146	# generate the simplified headers
147	147	headers: makeheaders.exe page_index.h VERSION.h ../src/sqlite3.h ../src/th.h VERSION.h
148	148	makeheaders.exe $(foreach ts,$(TRANSLATEDSRC),$(ts):$(ts:_.c=.h)) ../src/sqlite3.h ../src/th.h VERSION.h
149	149	echo Done >$@
150	150

	--- win/Makefile.PellesCGMake
	+++ win/Makefile.PellesCGMake
	@@ -126,11 +126,11 @@
126	# compiling standard fossil utils
127	$(UTILS_OBJ): %.obj: $(SRCDIR)%.c
128	$(CC) $(CCFLAGS) $(INCLUDE) "$<" -Fo"$@"
129
130	# compile special windows utils
131	version.obj: $(WINDIR)version.c
132	$(CC) $(CCFLAGS) $(INCLUDE) "$<" -Fo"$@"
133
134	# generate the translated c-source files
135	$(TRANSLATEDSRC): %_.c: $(SRCDIR)%.c translate.exe
136	translate.exe $< >$@
	@@ -138,12 +138,12 @@
138	# generate the index source, containing all web references,..
139	page_index.h: $(TRANSLATEDSRC) mkindex.exe
140	mkindex.exe $(TRANSLATEDSRC) >$@
141
142	# extracting version info from manifest
143	VERSION.h: version.exe ..\manifest.uuid ..\manifest
144	version.exe ..\manifest.uuid ..\manifest > $@
145
146	# generate the simplified headers
147	headers: makeheaders.exe page_index.h VERSION.h ../src/sqlite3.h ../src/th.h VERSION.h
148	makeheaders.exe $(foreach ts,$(TRANSLATEDSRC),$(ts):$(ts:_.c=.h)) ../src/sqlite3.h ../src/th.h VERSION.h
149	echo Done >$@
150

	--- win/Makefile.PellesCGMake
	+++ win/Makefile.PellesCGMake
	@@ -126,11 +126,11 @@
126	# compiling standard fossil utils
127	$(UTILS_OBJ): %.obj: $(SRCDIR)%.c
128	$(CC) $(CCFLAGS) $(INCLUDE) "$<" -Fo"$@"
129
130	# compile special windows utils
131	version.obj: $(SRCDIR)mkversion.c
132	$(CC) $(CCFLAGS) $(INCLUDE) "$<" -Fo"$@"
133
134	# generate the translated c-source files
135	$(TRANSLATEDSRC): %_.c: $(SRCDIR)%.c translate.exe
136	translate.exe $< >$@
	@@ -138,12 +138,12 @@
138	# generate the index source, containing all web references,..
139	page_index.h: $(TRANSLATEDSRC) mkindex.exe
140	mkindex.exe $(TRANSLATEDSRC) >$@
141
142	# extracting version info from manifest
143	VERSION.h: version.exe ..\manifest.uuid ..\manifest ..\VERSION
144	version.exe ..\manifest.uuid ..\manifest ..\VERSION > $@
145
146	# generate the simplified headers
147	headers: makeheaders.exe page_index.h VERSION.h ../src/sqlite3.h ../src/th.h VERSION.h
148	makeheaders.exe $(foreach ts,$(TRANSLATEDSRC),$(ts):$(ts:_.c=.h)) ../src/sqlite3.h ../src/th.h VERSION.h
149	echo Done >$@
150

M win/Makefile.dmc

+4 -4

		--- win/Makefile.dmc
		+++ win/Makefile.dmc
		@@ -20,11 +20,11 @@
20	20	CFLAGS = -o
21	21	BCC = $(DMDIR)\bin\dmc $(CFLAGS)
22	22	TCC = $(DMDIR)\bin\dmc $(CFLAGS) $(DMCDEF) $(SSL) $(INCL)
23	23	LIBS = $(DMDIR)\extra\lib\ zlib wsock32
24	24
25		-SQLITE_OPTIONS = -DSQLITE_OMIT_LOAD_EXTENSION=1 -DSQLITE_THREADSAFE=0 -DSQLITE_DEFAULT_FILE_FORMAT=4 -DSQLITE_ENABLE_STAT2 -Dlocaltime=fossil_localtime -DSQLITE_ENABLE_LOCKING_STYLE=0
	25	+SQLITE_OPTIONS = -DSQLITE_OMIT_LOAD_EXTENSION=1 -DSQLITE_THREADSAFE=0 -DSQLITE_DEFAULT_FILE_FORMAT=4 -DSQLITE_ENABLE_STAT3 -Dlocaltime=fossil_localtime -DSQLITE_ENABLE_LOCKING_STYLE=0
26	26
27	27	SRC = add_.c allrepo_.c attach_.c bag_.c bisect_.c blob_.c branch_.c browse_.c captcha_.c cgi_.c checkin_.c checkout_.c clearsign_.c clone_.c comformat_.c configure_.c content_.c db_.c delta_.c deltacmd_.c descendants_.c diff_.c diffcmd_.c doc_.c encode_.c event_.c export_.c file_.c finfo_.c glob_.c graph_.c gzip_.c http_.c http_socket_.c http_ssl_.c http_transport_.c import_.c info_.c leaf_.c login_.c main_.c manifest_.c md5_.c merge_.c merge3_.c name_.c path_.c pivot_.c popen_.c pqueue_.c printf_.c rebuild_.c report_.c rss_.c schema_.c search_.c setup_.c sha1_.c shun_.c skins_.c sqlcmd_.c stash_.c stat_.c style_.c sync_.c tag_.c tar_.c th_main_.c timeline_.c tkt_.c tktsetup_.c undo_.c update_.c url_.c user_.c verify_.c vfile_.c wiki_.c wikiformat_.c winhttp_.c xfer_.c zip_.c
28	28
29	29	OBJ = $(OBJDIR)\add$O $(OBJDIR)\allrepo$O $(OBJDIR)\attach$O $(OBJDIR)\bag$O $(OBJDIR)\bisect$O $(OBJDIR)\blob$O $(OBJDIR)\branch$O $(OBJDIR)\browse$O $(OBJDIR)\captcha$O $(OBJDIR)\cgi$O $(OBJDIR)\checkin$O $(OBJDIR)\checkout$O $(OBJDIR)\clearsign$O $(OBJDIR)\clone$O $(OBJDIR)\comformat$O $(OBJDIR)\configure$O $(OBJDIR)\content$O $(OBJDIR)\db$O $(OBJDIR)\delta$O $(OBJDIR)\deltacmd$O $(OBJDIR)\descendants$O $(OBJDIR)\diff$O $(OBJDIR)\diffcmd$O $(OBJDIR)\doc$O $(OBJDIR)\encode$O $(OBJDIR)\event$O $(OBJDIR)\export$O $(OBJDIR)\file$O $(OBJDIR)\finfo$O $(OBJDIR)\glob$O $(OBJDIR)\graph$O $(OBJDIR)\gzip$O $(OBJDIR)\http$O $(OBJDIR)\http_socket$O $(OBJDIR)\http_ssl$O $(OBJDIR)\http_transport$O $(OBJDIR)\import$O $(OBJDIR)\info$O $(OBJDIR)\leaf$O $(OBJDIR)\login$O $(OBJDIR)\main$O $(OBJDIR)\manifest$O $(OBJDIR)\md5$O $(OBJDIR)\merge$O $(OBJDIR)\merge3$O $(OBJDIR)\name$O $(OBJDIR)\path$O $(OBJDIR)\pivot$O $(OBJDIR)\popen$O $(OBJDIR)\pqueue$O $(OBJDIR)\printf$O $(OBJDIR)\rebuild$O $(OBJDIR)\report$O $(OBJDIR)\rss$O $(OBJDIR)\schema$O $(OBJDIR)\search$O $(OBJDIR)\setup$O $(OBJDIR)\sha1$O $(OBJDIR)\shun$O $(OBJDIR)\skins$O $(OBJDIR)\sqlcmd$O $(OBJDIR)\stash$O $(OBJDIR)\stat$O $(OBJDIR)\style$O $(OBJDIR)\sync$O $(OBJDIR)\tag$O $(OBJDIR)\tar$O $(OBJDIR)\th_main$O $(OBJDIR)\timeline$O $(OBJDIR)\tkt$O $(OBJDIR)\tktsetup$O $(OBJDIR)\undo$O $(OBJDIR)\update$O $(OBJDIR)\url$O $(OBJDIR)\user$O $(OBJDIR)\verify$O $(OBJDIR)\vfile$O $(OBJDIR)\wiki$O $(OBJDIR)\wikiformat$O $(OBJDIR)\winhttp$O $(OBJDIR)\xfer$O $(OBJDIR)\zip$O $(OBJDIR)\shell$O $(OBJDIR)\sqlite3$O $(OBJDIR)\th$O $(OBJDIR)\th_lang$O
30	30
		@@ -58,11 +58,11 @@
58	58	$(BCC) -o$@ $**
59	59
60	60	mkindex$E: $(SRCDIR)\mkindex.c
61	61	$(BCC) -o$@ $**
62	62
63		-version$E: $B\win\version.c
	63	+version$E: $B\src\mkversion.c
64	64	$(BCC) -o$@ $**
65	65
66	66	$(OBJDIR)\shell$O : $(SRCDIR)\shell.c
67	67	$(TCC) -o$@ -c -Dmain=sqlite3_shell $(SQLITE_OPTIONS) $**
68	68
		@@ -73,11 +73,11 @@
73	73	$(TCC) -o$@ -c $**
74	74
75	75	$(OBJDIR)\th_lang$O : $(SRCDIR)\th_lang.c
76	76	$(TCC) -o$@ -c $**
77	77
78		-VERSION.h : version$E $B\manifest.uuid $B\manifest
	78	+VERSION.h : version$E $B\manifest.uuid $B\manifest $B\VERSION
79	79	+$** > $@
80	80
81	81	page_index.h: mkindex$E $(SRC)
82	82	+$** > $@
83	83
		@@ -84,11 +84,11 @@
84	84	clean:
85	85	-del $(OBJDIR)\*.obj
86	86	-del .obj _.c .h .map
87	87
88	88	realclean:
89		- -del $(APPNAME) translate$E mkindex$E makeheaders$E version$E
	89	+ -del $(APPNAME) translate$E mkindex$E makeheaders$E mkversion$E
90	90
91	91
92	92	$(OBJDIR)\add$O : add_.c add.h
93	93	$(TCC) -o$@ -c add_.c
94	94
95	95

	--- win/Makefile.dmc
	+++ win/Makefile.dmc
	@@ -20,11 +20,11 @@
20	CFLAGS = -o
21	BCC = $(DMDIR)\bin\dmc $(CFLAGS)
22	TCC = $(DMDIR)\bin\dmc $(CFLAGS) $(DMCDEF) $(SSL) $(INCL)
23	LIBS = $(DMDIR)\extra\lib\ zlib wsock32
24
25	SQLITE_OPTIONS = -DSQLITE_OMIT_LOAD_EXTENSION=1 -DSQLITE_THREADSAFE=0 -DSQLITE_DEFAULT_FILE_FORMAT=4 -DSQLITE_ENABLE_STAT2 -Dlocaltime=fossil_localtime -DSQLITE_ENABLE_LOCKING_STYLE=0
26
27	SRC = add_.c allrepo_.c attach_.c bag_.c bisect_.c blob_.c branch_.c browse_.c captcha_.c cgi_.c checkin_.c checkout_.c clearsign_.c clone_.c comformat_.c configure_.c content_.c db_.c delta_.c deltacmd_.c descendants_.c diff_.c diffcmd_.c doc_.c encode_.c event_.c export_.c file_.c finfo_.c glob_.c graph_.c gzip_.c http_.c http_socket_.c http_ssl_.c http_transport_.c import_.c info_.c leaf_.c login_.c main_.c manifest_.c md5_.c merge_.c merge3_.c name_.c path_.c pivot_.c popen_.c pqueue_.c printf_.c rebuild_.c report_.c rss_.c schema_.c search_.c setup_.c sha1_.c shun_.c skins_.c sqlcmd_.c stash_.c stat_.c style_.c sync_.c tag_.c tar_.c th_main_.c timeline_.c tkt_.c tktsetup_.c undo_.c update_.c url_.c user_.c verify_.c vfile_.c wiki_.c wikiformat_.c winhttp_.c xfer_.c zip_.c
28
29	OBJ = $(OBJDIR)\add$O $(OBJDIR)\allrepo$O $(OBJDIR)\attach$O $(OBJDIR)\bag$O $(OBJDIR)\bisect$O $(OBJDIR)\blob$O $(OBJDIR)\branch$O $(OBJDIR)\browse$O $(OBJDIR)\captcha$O $(OBJDIR)\cgi$O $(OBJDIR)\checkin$O $(OBJDIR)\checkout$O $(OBJDIR)\clearsign$O $(OBJDIR)\clone$O $(OBJDIR)\comformat$O $(OBJDIR)\configure$O $(OBJDIR)\content$O $(OBJDIR)\db$O $(OBJDIR)\delta$O $(OBJDIR)\deltacmd$O $(OBJDIR)\descendants$O $(OBJDIR)\diff$O $(OBJDIR)\diffcmd$O $(OBJDIR)\doc$O $(OBJDIR)\encode$O $(OBJDIR)\event$O $(OBJDIR)\export$O $(OBJDIR)\file$O $(OBJDIR)\finfo$O $(OBJDIR)\glob$O $(OBJDIR)\graph$O $(OBJDIR)\gzip$O $(OBJDIR)\http$O $(OBJDIR)\http_socket$O $(OBJDIR)\http_ssl$O $(OBJDIR)\http_transport$O $(OBJDIR)\import$O $(OBJDIR)\info$O $(OBJDIR)\leaf$O $(OBJDIR)\login$O $(OBJDIR)\main$O $(OBJDIR)\manifest$O $(OBJDIR)\md5$O $(OBJDIR)\merge$O $(OBJDIR)\merge3$O $(OBJDIR)\name$O $(OBJDIR)\path$O $(OBJDIR)\pivot$O $(OBJDIR)\popen$O $(OBJDIR)\pqueue$O $(OBJDIR)\printf$O $(OBJDIR)\rebuild$O $(OBJDIR)\report$O $(OBJDIR)\rss$O $(OBJDIR)\schema$O $(OBJDIR)\search$O $(OBJDIR)\setup$O $(OBJDIR)\sha1$O $(OBJDIR)\shun$O $(OBJDIR)\skins$O $(OBJDIR)\sqlcmd$O $(OBJDIR)\stash$O $(OBJDIR)\stat$O $(OBJDIR)\style$O $(OBJDIR)\sync$O $(OBJDIR)\tag$O $(OBJDIR)\tar$O $(OBJDIR)\th_main$O $(OBJDIR)\timeline$O $(OBJDIR)\tkt$O $(OBJDIR)\tktsetup$O $(OBJDIR)\undo$O $(OBJDIR)\update$O $(OBJDIR)\url$O $(OBJDIR)\user$O $(OBJDIR)\verify$O $(OBJDIR)\vfile$O $(OBJDIR)\wiki$O $(OBJDIR)\wikiformat$O $(OBJDIR)\winhttp$O $(OBJDIR)\xfer$O $(OBJDIR)\zip$O $(OBJDIR)\shell$O $(OBJDIR)\sqlite3$O $(OBJDIR)\th$O $(OBJDIR)\th_lang$O
30
	@@ -58,11 +58,11 @@
58	$(BCC) -o$@ $**
59
60	mkindex$E: $(SRCDIR)\mkindex.c
61	$(BCC) -o$@ $**
62
63	version$E: $B\win\version.c
64	$(BCC) -o$@ $**
65
66	$(OBJDIR)\shell$O : $(SRCDIR)\shell.c
67	$(TCC) -o$@ -c -Dmain=sqlite3_shell $(SQLITE_OPTIONS) $**
68
	@@ -73,11 +73,11 @@
73	$(TCC) -o$@ -c $**
74
75	$(OBJDIR)\th_lang$O : $(SRCDIR)\th_lang.c
76	$(TCC) -o$@ -c $**
77
78	VERSION.h : version$E $B\manifest.uuid $B\manifest
79	+$** > $@
80
81	page_index.h: mkindex$E $(SRC)
82	+$** > $@
83
	@@ -84,11 +84,11 @@
84	clean:
85	-del $(OBJDIR)\*.obj
86	-del .obj _.c .h .map
87
88	realclean:
89	-del $(APPNAME) translate$E mkindex$E makeheaders$E version$E
90
91
92	$(OBJDIR)\add$O : add_.c add.h
93	$(TCC) -o$@ -c add_.c
94
95

	--- win/Makefile.dmc
	+++ win/Makefile.dmc
	@@ -20,11 +20,11 @@
20	CFLAGS = -o
21	BCC = $(DMDIR)\bin\dmc $(CFLAGS)
22	TCC = $(DMDIR)\bin\dmc $(CFLAGS) $(DMCDEF) $(SSL) $(INCL)
23	LIBS = $(DMDIR)\extra\lib\ zlib wsock32
24
25	SQLITE_OPTIONS = -DSQLITE_OMIT_LOAD_EXTENSION=1 -DSQLITE_THREADSAFE=0 -DSQLITE_DEFAULT_FILE_FORMAT=4 -DSQLITE_ENABLE_STAT3 -Dlocaltime=fossil_localtime -DSQLITE_ENABLE_LOCKING_STYLE=0
26
27	SRC = add_.c allrepo_.c attach_.c bag_.c bisect_.c blob_.c branch_.c browse_.c captcha_.c cgi_.c checkin_.c checkout_.c clearsign_.c clone_.c comformat_.c configure_.c content_.c db_.c delta_.c deltacmd_.c descendants_.c diff_.c diffcmd_.c doc_.c encode_.c event_.c export_.c file_.c finfo_.c glob_.c graph_.c gzip_.c http_.c http_socket_.c http_ssl_.c http_transport_.c import_.c info_.c leaf_.c login_.c main_.c manifest_.c md5_.c merge_.c merge3_.c name_.c path_.c pivot_.c popen_.c pqueue_.c printf_.c rebuild_.c report_.c rss_.c schema_.c search_.c setup_.c sha1_.c shun_.c skins_.c sqlcmd_.c stash_.c stat_.c style_.c sync_.c tag_.c tar_.c th_main_.c timeline_.c tkt_.c tktsetup_.c undo_.c update_.c url_.c user_.c verify_.c vfile_.c wiki_.c wikiformat_.c winhttp_.c xfer_.c zip_.c
28
29	OBJ = $(OBJDIR)\add$O $(OBJDIR)\allrepo$O $(OBJDIR)\attach$O $(OBJDIR)\bag$O $(OBJDIR)\bisect$O $(OBJDIR)\blob$O $(OBJDIR)\branch$O $(OBJDIR)\browse$O $(OBJDIR)\captcha$O $(OBJDIR)\cgi$O $(OBJDIR)\checkin$O $(OBJDIR)\checkout$O $(OBJDIR)\clearsign$O $(OBJDIR)\clone$O $(OBJDIR)\comformat$O $(OBJDIR)\configure$O $(OBJDIR)\content$O $(OBJDIR)\db$O $(OBJDIR)\delta$O $(OBJDIR)\deltacmd$O $(OBJDIR)\descendants$O $(OBJDIR)\diff$O $(OBJDIR)\diffcmd$O $(OBJDIR)\doc$O $(OBJDIR)\encode$O $(OBJDIR)\event$O $(OBJDIR)\export$O $(OBJDIR)\file$O $(OBJDIR)\finfo$O $(OBJDIR)\glob$O $(OBJDIR)\graph$O $(OBJDIR)\gzip$O $(OBJDIR)\http$O $(OBJDIR)\http_socket$O $(OBJDIR)\http_ssl$O $(OBJDIR)\http_transport$O $(OBJDIR)\import$O $(OBJDIR)\info$O $(OBJDIR)\leaf$O $(OBJDIR)\login$O $(OBJDIR)\main$O $(OBJDIR)\manifest$O $(OBJDIR)\md5$O $(OBJDIR)\merge$O $(OBJDIR)\merge3$O $(OBJDIR)\name$O $(OBJDIR)\path$O $(OBJDIR)\pivot$O $(OBJDIR)\popen$O $(OBJDIR)\pqueue$O $(OBJDIR)\printf$O $(OBJDIR)\rebuild$O $(OBJDIR)\report$O $(OBJDIR)\rss$O $(OBJDIR)\schema$O $(OBJDIR)\search$O $(OBJDIR)\setup$O $(OBJDIR)\sha1$O $(OBJDIR)\shun$O $(OBJDIR)\skins$O $(OBJDIR)\sqlcmd$O $(OBJDIR)\stash$O $(OBJDIR)\stat$O $(OBJDIR)\style$O $(OBJDIR)\sync$O $(OBJDIR)\tag$O $(OBJDIR)\tar$O $(OBJDIR)\th_main$O $(OBJDIR)\timeline$O $(OBJDIR)\tkt$O $(OBJDIR)\tktsetup$O $(OBJDIR)\undo$O $(OBJDIR)\update$O $(OBJDIR)\url$O $(OBJDIR)\user$O $(OBJDIR)\verify$O $(OBJDIR)\vfile$O $(OBJDIR)\wiki$O $(OBJDIR)\wikiformat$O $(OBJDIR)\winhttp$O $(OBJDIR)\xfer$O $(OBJDIR)\zip$O $(OBJDIR)\shell$O $(OBJDIR)\sqlite3$O $(OBJDIR)\th$O $(OBJDIR)\th_lang$O
30
	@@ -58,11 +58,11 @@
58	$(BCC) -o$@ $**
59
60	mkindex$E: $(SRCDIR)\mkindex.c
61	$(BCC) -o$@ $**
62
63	version$E: $B\src\mkversion.c
64	$(BCC) -o$@ $**
65
66	$(OBJDIR)\shell$O : $(SRCDIR)\shell.c
67	$(TCC) -o$@ -c -Dmain=sqlite3_shell $(SQLITE_OPTIONS) $**
68
	@@ -73,11 +73,11 @@
73	$(TCC) -o$@ -c $**
74
75	$(OBJDIR)\th_lang$O : $(SRCDIR)\th_lang.c
76	$(TCC) -o$@ -c $**
77
78	VERSION.h : version$E $B\manifest.uuid $B\manifest $B\VERSION
79	+$** > $@
80
81	page_index.h: mkindex$E $(SRC)
82	+$** > $@
83
	@@ -84,11 +84,11 @@
84	clean:
85	-del $(OBJDIR)\*.obj
86	-del .obj _.c .h .map
87
88	realclean:
89	-del $(APPNAME) translate$E mkindex$E makeheaders$E mkversion$E
90
91
92	$(OBJDIR)\add$O : add_.c add.h
93	$(TCC) -o$@ -c add_.c
94
95

M win/Makefile.mingw

+5 -5

		--- win/Makefile.mingw
		+++ win/Makefile.mingw
		@@ -351,21 +351,21 @@
351	351	$(BCC) -o $(OBJDIR)/makeheaders $(SRCDIR)/makeheaders.c
352	352
353	353	$(OBJDIR)/mkindex: $(SRCDIR)/mkindex.c
354	354	$(BCC) -o $(OBJDIR)/mkindex $(SRCDIR)/mkindex.c
355	355
356		-$(VERSION): $(SRCDIR)/../win/version.c
357		- $(BCC) -o $(OBJDIR)/version $(SRCDIR)/../win/version.c
	356	+$(VERSION): $(SRCDIR)/mkversion.c
	357	+ $(BCC) -o $(OBJDIR)/version $(SRCDIR)/mkversion.c
358	358
359	359	# WARNING. DANGER. Running the testsuite modifies the repository the
360	360	# build is done from, i.e. the checkout belongs to. Do not sync/push
361	361	# the repository after running the tests.
362	362	test: $(APPNAME)
363	363	$(TCLSH) test/tester.tcl $(APPNAME)
364	364
365	365	$(OBJDIR)/VERSION.h: $(SRCDIR)/../manifest.uuid $(SRCDIR)/../manifest $(VERSION)
366		- $(VERSION) $(SRCDIR)/../manifest.uuid $(SRCDIR)/../manifest >$(OBJDIR)/VERSION.h
	366	+ $(VERSION) $(SRCDIR)/../manifest.uuid $(SRCDIR)/../manifest $(SRCDIR)/../VERSION >$(OBJDIR)/VERSION.h
367	367
368	368	EXTRAOBJ = $(OBJDIR)/sqlite3.o $(OBJDIR)/shell.o $(OBJDIR)/th.o $(OBJDIR)/th_lang.o
369	369
370	370	$(APPNAME): $(OBJDIR)/headers $(OBJ) $(EXTRAOBJ) $(OBJDIR)/icon.o
371	371	$(TCC) -o $(APPNAME) $(OBJ) $(EXTRAOBJ) $(LIB) $(OBJDIR)/icon.o
		@@ -968,16 +968,16 @@
968	968	$(OBJDIR)/zip.o: $(OBJDIR)/zip_.c $(OBJDIR)/zip.h $(SRCDIR)/config.h
969	969	$(XTCC) -o $(OBJDIR)/zip.o -c $(OBJDIR)/zip_.c
970	970
971	971	zip.h: $(OBJDIR)/headers
972	972	$(OBJDIR)/sqlite3.o: $(SRCDIR)/sqlite3.c
973		- $(XTCC) -DSQLITE_OMIT_LOAD_EXTENSION=1 -DSQLITE_THREADSAFE=0 -DSQLITE_DEFAULT_FILE_FORMAT=4 -DSQLITE_ENABLE_STAT2 -Dlocaltime=fossil_localtime -DSQLITE_ENABLE_LOCKING_STYLE=0 -c $(SRCDIR)/sqlite3.c -o $(OBJDIR)/sqlite3.o
	973	+ $(XTCC) -DSQLITE_OMIT_LOAD_EXTENSION=1 -DSQLITE_THREADSAFE=0 -DSQLITE_DEFAULT_FILE_FORMAT=4 -DSQLITE_ENABLE_STAT3 -Dlocaltime=fossil_localtime -DSQLITE_ENABLE_LOCKING_STYLE=0 -c $(SRCDIR)/sqlite3.c -o $(OBJDIR)/sqlite3.o
974	974
975		-$(OBJDIR)/shell.o: $(SRCDIR)/shell.c
	975	+$(OBJDIR)/shell.o: $(SRCDIR)/shell.c $(SRCDIR)/sqlite3.h
976	976	$(XTCC) -Dmain=sqlite3_shell -DSQLITE_OMIT_LOAD_EXTENSION=1 -c $(SRCDIR)/shell.c -o $(OBJDIR)/shell.o
977	977
978	978	$(OBJDIR)/th.o: $(SRCDIR)/th.c
979	979	$(XTCC) -I$(SRCDIR) -c $(SRCDIR)/th.c -o $(OBJDIR)/th.o
980	980
981	981	$(OBJDIR)/th_lang.o: $(SRCDIR)/th_lang.c
982	982	$(XTCC) -I$(SRCDIR) -c $(SRCDIR)/th_lang.c -o $(OBJDIR)/th_lang.o
983	983
984	984

	--- win/Makefile.mingw
	+++ win/Makefile.mingw
	@@ -351,21 +351,21 @@
351	$(BCC) -o $(OBJDIR)/makeheaders $(SRCDIR)/makeheaders.c
352
353	$(OBJDIR)/mkindex: $(SRCDIR)/mkindex.c
354	$(BCC) -o $(OBJDIR)/mkindex $(SRCDIR)/mkindex.c
355
356	$(VERSION): $(SRCDIR)/../win/version.c
357	$(BCC) -o $(OBJDIR)/version $(SRCDIR)/../win/version.c
358
359	# WARNING. DANGER. Running the testsuite modifies the repository the
360	# build is done from, i.e. the checkout belongs to. Do not sync/push
361	# the repository after running the tests.
362	test: $(APPNAME)
363	$(TCLSH) test/tester.tcl $(APPNAME)
364
365	$(OBJDIR)/VERSION.h: $(SRCDIR)/../manifest.uuid $(SRCDIR)/../manifest $(VERSION)
366	$(VERSION) $(SRCDIR)/../manifest.uuid $(SRCDIR)/../manifest >$(OBJDIR)/VERSION.h
367
368	EXTRAOBJ = $(OBJDIR)/sqlite3.o $(OBJDIR)/shell.o $(OBJDIR)/th.o $(OBJDIR)/th_lang.o
369
370	$(APPNAME): $(OBJDIR)/headers $(OBJ) $(EXTRAOBJ) $(OBJDIR)/icon.o
371	$(TCC) -o $(APPNAME) $(OBJ) $(EXTRAOBJ) $(LIB) $(OBJDIR)/icon.o
	@@ -968,16 +968,16 @@
968	$(OBJDIR)/zip.o: $(OBJDIR)/zip_.c $(OBJDIR)/zip.h $(SRCDIR)/config.h
969	$(XTCC) -o $(OBJDIR)/zip.o -c $(OBJDIR)/zip_.c
970
971	zip.h: $(OBJDIR)/headers
972	$(OBJDIR)/sqlite3.o: $(SRCDIR)/sqlite3.c
973	$(XTCC) -DSQLITE_OMIT_LOAD_EXTENSION=1 -DSQLITE_THREADSAFE=0 -DSQLITE_DEFAULT_FILE_FORMAT=4 -DSQLITE_ENABLE_STAT2 -Dlocaltime=fossil_localtime -DSQLITE_ENABLE_LOCKING_STYLE=0 -c $(SRCDIR)/sqlite3.c -o $(OBJDIR)/sqlite3.o
974
975	$(OBJDIR)/shell.o: $(SRCDIR)/shell.c
976	$(XTCC) -Dmain=sqlite3_shell -DSQLITE_OMIT_LOAD_EXTENSION=1 -c $(SRCDIR)/shell.c -o $(OBJDIR)/shell.o
977
978	$(OBJDIR)/th.o: $(SRCDIR)/th.c
979	$(XTCC) -I$(SRCDIR) -c $(SRCDIR)/th.c -o $(OBJDIR)/th.o
980
981	$(OBJDIR)/th_lang.o: $(SRCDIR)/th_lang.c
982	$(XTCC) -I$(SRCDIR) -c $(SRCDIR)/th_lang.c -o $(OBJDIR)/th_lang.o
983
984

	--- win/Makefile.mingw
	+++ win/Makefile.mingw
	@@ -351,21 +351,21 @@
351	$(BCC) -o $(OBJDIR)/makeheaders $(SRCDIR)/makeheaders.c
352
353	$(OBJDIR)/mkindex: $(SRCDIR)/mkindex.c
354	$(BCC) -o $(OBJDIR)/mkindex $(SRCDIR)/mkindex.c
355
356	$(VERSION): $(SRCDIR)/mkversion.c
357	$(BCC) -o $(OBJDIR)/version $(SRCDIR)/mkversion.c
358
359	# WARNING. DANGER. Running the testsuite modifies the repository the
360	# build is done from, i.e. the checkout belongs to. Do not sync/push
361	# the repository after running the tests.
362	test: $(APPNAME)
363	$(TCLSH) test/tester.tcl $(APPNAME)
364
365	$(OBJDIR)/VERSION.h: $(SRCDIR)/../manifest.uuid $(SRCDIR)/../manifest $(VERSION)
366	$(VERSION) $(SRCDIR)/../manifest.uuid $(SRCDIR)/../manifest $(SRCDIR)/../VERSION >$(OBJDIR)/VERSION.h
367
368	EXTRAOBJ = $(OBJDIR)/sqlite3.o $(OBJDIR)/shell.o $(OBJDIR)/th.o $(OBJDIR)/th_lang.o
369
370	$(APPNAME): $(OBJDIR)/headers $(OBJ) $(EXTRAOBJ) $(OBJDIR)/icon.o
371	$(TCC) -o $(APPNAME) $(OBJ) $(EXTRAOBJ) $(LIB) $(OBJDIR)/icon.o
	@@ -968,16 +968,16 @@
968	$(OBJDIR)/zip.o: $(OBJDIR)/zip_.c $(OBJDIR)/zip.h $(SRCDIR)/config.h
969	$(XTCC) -o $(OBJDIR)/zip.o -c $(OBJDIR)/zip_.c
970
971	zip.h: $(OBJDIR)/headers
972	$(OBJDIR)/sqlite3.o: $(SRCDIR)/sqlite3.c
973	$(XTCC) -DSQLITE_OMIT_LOAD_EXTENSION=1 -DSQLITE_THREADSAFE=0 -DSQLITE_DEFAULT_FILE_FORMAT=4 -DSQLITE_ENABLE_STAT3 -Dlocaltime=fossil_localtime -DSQLITE_ENABLE_LOCKING_STYLE=0 -c $(SRCDIR)/sqlite3.c -o $(OBJDIR)/sqlite3.o
974
975	$(OBJDIR)/shell.o: $(SRCDIR)/shell.c $(SRCDIR)/sqlite3.h
976	$(XTCC) -Dmain=sqlite3_shell -DSQLITE_OMIT_LOAD_EXTENSION=1 -c $(SRCDIR)/shell.c -o $(OBJDIR)/shell.o
977
978	$(OBJDIR)/th.o: $(SRCDIR)/th.c
979	$(XTCC) -I$(SRCDIR) -c $(SRCDIR)/th.c -o $(OBJDIR)/th.o
980
981	$(OBJDIR)/th_lang.o: $(SRCDIR)/th_lang.c
982	$(XTCC) -I$(SRCDIR) -c $(SRCDIR)/th_lang.c -o $(OBJDIR)/th_lang.o
983
984

M win/Makefile.msc

+5 -5

		--- win/Makefile.msc
		+++ win/Makefile.msc
		@@ -31,14 +31,14 @@
31	31	INCL = -I. -I$(SRCDIR) -I$B\win\include -I$(MSCDIR)\extra\include -I$(ZINCDIR)
32	32
33	33	CFLAGS = -nologo -MT -O2
34	34	BCC = $(CC) $(CFLAGS)
35	35	TCC = $(CC) -c $(CFLAGS) $(MSCDEF) $(SSL) $(INCL)
36		-LIBS = $(ZLIB) ws2_32.lib $(SSLLIB)
	36	+LIBS = $(ZLIB) ws2_32.lib advapi32.lib $(SSLLIB)
37	37	LIBDIR = -LIBPATH:$(MSCDIR)\extra\lib -LIBPATH:$(ZLIBDIR)
38	38
39		-SQLITE_OPTIONS = /DSQLITE_OMIT_LOAD_EXTENSION=1 /DSQLITE_THREADSAFE=0 /DSQLITE_DEFAULT_FILE_FORMAT=4 /DSQLITE_ENABLE_STAT2 /Dlocaltime=fossil_localtime /DSQLITE_ENABLE_LOCKING_STYLE=0
	39	+SQLITE_OPTIONS = /DSQLITE_OMIT_LOAD_EXTENSION=1 /DSQLITE_THREADSAFE=0 /DSQLITE_DEFAULT_FILE_FORMAT=4 /DSQLITE_ENABLE_STAT3 /Dlocaltime=fossil_localtime /DSQLITE_ENABLE_LOCKING_STYLE=0
40	40
41	41	SRC = add_.c allrepo_.c attach_.c bag_.c bisect_.c blob_.c branch_.c browse_.c captcha_.c cgi_.c checkin_.c checkout_.c clearsign_.c clone_.c comformat_.c configure_.c content_.c db_.c delta_.c deltacmd_.c descendants_.c diff_.c diffcmd_.c doc_.c encode_.c event_.c export_.c file_.c finfo_.c glob_.c graph_.c gzip_.c http_.c http_socket_.c http_ssl_.c http_transport_.c import_.c info_.c leaf_.c login_.c main_.c manifest_.c md5_.c merge_.c merge3_.c name_.c path_.c pivot_.c popen_.c pqueue_.c printf_.c rebuild_.c report_.c rss_.c schema_.c search_.c setup_.c sha1_.c shun_.c skins_.c sqlcmd_.c stash_.c stat_.c style_.c sync_.c tag_.c tar_.c th_main_.c timeline_.c tkt_.c tktsetup_.c undo_.c update_.c url_.c user_.c verify_.c vfile_.c wiki_.c wikiformat_.c winhttp_.c xfer_.c zip_.c
42	42
43	43	OBJ = $(OX)\add$O $(OX)\allrepo$O $(OX)\attach$O $(OX)\bag$O $(OX)\bisect$O $(OX)\blob$O $(OX)\branch$O $(OX)\browse$O $(OX)\captcha$O $(OX)\cgi$O $(OX)\checkin$O $(OX)\checkout$O $(OX)\clearsign$O $(OX)\clone$O $(OX)\comformat$O $(OX)\configure$O $(OX)\content$O $(OX)\db$O $(OX)\delta$O $(OX)\deltacmd$O $(OX)\descendants$O $(OX)\diff$O $(OX)\diffcmd$O $(OX)\doc$O $(OX)\encode$O $(OX)\event$O $(OX)\export$O $(OX)\file$O $(OX)\finfo$O $(OX)\glob$O $(OX)\graph$O $(OX)\gzip$O $(OX)\http$O $(OX)\http_socket$O $(OX)\http_ssl$O $(OX)\http_transport$O $(OX)\import$O $(OX)\info$O $(OX)\leaf$O $(OX)\login$O $(OX)\main$O $(OX)\manifest$O $(OX)\md5$O $(OX)\merge$O $(OX)\merge3$O $(OX)\name$O $(OX)\path$O $(OX)\pivot$O $(OX)\popen$O $(OX)\pqueue$O $(OX)\printf$O $(OX)\rebuild$O $(OX)\report$O $(OX)\rss$O $(OX)\schema$O $(OX)\search$O $(OX)\setup$O $(OX)\sha1$O $(OX)\shun$O $(OX)\skins$O $(OX)\sqlcmd$O $(OX)\stash$O $(OX)\stat$O $(OX)\style$O $(OX)\sync$O $(OX)\tag$O $(OX)\tar$O $(OX)\th_main$O $(OX)\timeline$O $(OX)\tkt$O $(OX)\tktsetup$O $(OX)\undo$O $(OX)\update$O $(OX)\url$O $(OX)\user$O $(OX)\verify$O $(OX)\vfile$O $(OX)\wiki$O $(OX)\wikiformat$O $(OX)\winhttp$O $(OX)\xfer$O $(OX)\zip$O $(OX)\shell$O $(OX)\sqlite3$O $(OX)\th$O $(OX)\th_lang$O
44	44
		@@ -153,11 +153,11 @@
153	153	$(BCC) $**
154	154
155	155	mkindex$E: $(SRCDIR)\mkindex.c
156	156	$(BCC) $**
157	157
158		-version$E: $B\win\version.c
	158	+mkversion$E: $B\src\mkversion.c
159	159	$(BCC) $**
160	160
161	161	$(OX)\shell$O : $(SRCDIR)\shell.c
162	162	$(TCC) /Fo$@ /Dmain=sqlite3_shell $(SQLITE_OPTIONS) -c $(SRCDIR)\shell.c
163	163
		@@ -168,11 +168,11 @@
168	168	$(TCC) /Fo$@ -c $**
169	169
170	170	$(OX)\th_lang$O : $(SRCDIR)\th_lang.c
171	171	$(TCC) /Fo$@ -c $**
172	172
173		-VERSION.h : version$E $B\manifest.uuid $B\manifest
	173	+VERSION.h : mkversion$E $B\manifest.uuid $B\manifest $B\VERSION
174	174	$** > $@
175	175
176	176	page_index.h: mkindex$E $(SRC)
177	177	$** > $@
178	178
		@@ -180,11 +180,11 @@
180	180	-del $(OX)\*.obj
181	181	-del .obj _.c .h .map
182	182	-del headers linkopts
183	183
184	184	realclean:
185		- -del $(APPNAME) translate$E mkindex$E makeheaders$E version$E
	185	+ -del $(APPNAME) translate$E mkindex$E makeheaders$E mkversion$E
186	186
187	187
188	188	$(OX)\add$O : add_.c add.h
189	189	$(TCC) /Fo$@ -c add_.c
190	190
191	191
192	192	DELETED win/version.c

	--- win/Makefile.msc
	+++ win/Makefile.msc
	@@ -31,14 +31,14 @@
31	INCL = -I. -I$(SRCDIR) -I$B\win\include -I$(MSCDIR)\extra\include -I$(ZINCDIR)
32
33	CFLAGS = -nologo -MT -O2
34	BCC = $(CC) $(CFLAGS)
35	TCC = $(CC) -c $(CFLAGS) $(MSCDEF) $(SSL) $(INCL)
36	LIBS = $(ZLIB) ws2_32.lib $(SSLLIB)
37	LIBDIR = -LIBPATH:$(MSCDIR)\extra\lib -LIBPATH:$(ZLIBDIR)
38
39	SQLITE_OPTIONS = /DSQLITE_OMIT_LOAD_EXTENSION=1 /DSQLITE_THREADSAFE=0 /DSQLITE_DEFAULT_FILE_FORMAT=4 /DSQLITE_ENABLE_STAT2 /Dlocaltime=fossil_localtime /DSQLITE_ENABLE_LOCKING_STYLE=0
40
41	SRC = add_.c allrepo_.c attach_.c bag_.c bisect_.c blob_.c branch_.c browse_.c captcha_.c cgi_.c checkin_.c checkout_.c clearsign_.c clone_.c comformat_.c configure_.c content_.c db_.c delta_.c deltacmd_.c descendants_.c diff_.c diffcmd_.c doc_.c encode_.c event_.c export_.c file_.c finfo_.c glob_.c graph_.c gzip_.c http_.c http_socket_.c http_ssl_.c http_transport_.c import_.c info_.c leaf_.c login_.c main_.c manifest_.c md5_.c merge_.c merge3_.c name_.c path_.c pivot_.c popen_.c pqueue_.c printf_.c rebuild_.c report_.c rss_.c schema_.c search_.c setup_.c sha1_.c shun_.c skins_.c sqlcmd_.c stash_.c stat_.c style_.c sync_.c tag_.c tar_.c th_main_.c timeline_.c tkt_.c tktsetup_.c undo_.c update_.c url_.c user_.c verify_.c vfile_.c wiki_.c wikiformat_.c winhttp_.c xfer_.c zip_.c
42
43	OBJ = $(OX)\add$O $(OX)\allrepo$O $(OX)\attach$O $(OX)\bag$O $(OX)\bisect$O $(OX)\blob$O $(OX)\branch$O $(OX)\browse$O $(OX)\captcha$O $(OX)\cgi$O $(OX)\checkin$O $(OX)\checkout$O $(OX)\clearsign$O $(OX)\clone$O $(OX)\comformat$O $(OX)\configure$O $(OX)\content$O $(OX)\db$O $(OX)\delta$O $(OX)\deltacmd$O $(OX)\descendants$O $(OX)\diff$O $(OX)\diffcmd$O $(OX)\doc$O $(OX)\encode$O $(OX)\event$O $(OX)\export$O $(OX)\file$O $(OX)\finfo$O $(OX)\glob$O $(OX)\graph$O $(OX)\gzip$O $(OX)\http$O $(OX)\http_socket$O $(OX)\http_ssl$O $(OX)\http_transport$O $(OX)\import$O $(OX)\info$O $(OX)\leaf$O $(OX)\login$O $(OX)\main$O $(OX)\manifest$O $(OX)\md5$O $(OX)\merge$O $(OX)\merge3$O $(OX)\name$O $(OX)\path$O $(OX)\pivot$O $(OX)\popen$O $(OX)\pqueue$O $(OX)\printf$O $(OX)\rebuild$O $(OX)\report$O $(OX)\rss$O $(OX)\schema$O $(OX)\search$O $(OX)\setup$O $(OX)\sha1$O $(OX)\shun$O $(OX)\skins$O $(OX)\sqlcmd$O $(OX)\stash$O $(OX)\stat$O $(OX)\style$O $(OX)\sync$O $(OX)\tag$O $(OX)\tar$O $(OX)\th_main$O $(OX)\timeline$O $(OX)\tkt$O $(OX)\tktsetup$O $(OX)\undo$O $(OX)\update$O $(OX)\url$O $(OX)\user$O $(OX)\verify$O $(OX)\vfile$O $(OX)\wiki$O $(OX)\wikiformat$O $(OX)\winhttp$O $(OX)\xfer$O $(OX)\zip$O $(OX)\shell$O $(OX)\sqlite3$O $(OX)\th$O $(OX)\th_lang$O
44
	@@ -153,11 +153,11 @@
153	$(BCC) $**
154
155	mkindex$E: $(SRCDIR)\mkindex.c
156	$(BCC) $**
157
158	version$E: $B\win\version.c
159	$(BCC) $**
160
161	$(OX)\shell$O : $(SRCDIR)\shell.c
162	$(TCC) /Fo$@ /Dmain=sqlite3_shell $(SQLITE_OPTIONS) -c $(SRCDIR)\shell.c
163
	@@ -168,11 +168,11 @@
168	$(TCC) /Fo$@ -c $**
169
170	$(OX)\th_lang$O : $(SRCDIR)\th_lang.c
171	$(TCC) /Fo$@ -c $**
172
173	VERSION.h : version$E $B\manifest.uuid $B\manifest
174	$** > $@
175
176	page_index.h: mkindex$E $(SRC)
177	$** > $@
178
	@@ -180,11 +180,11 @@
180	-del $(OX)\*.obj
181	-del .obj _.c .h .map
182	-del headers linkopts
183
184	realclean:
185	-del $(APPNAME) translate$E mkindex$E makeheaders$E version$E
186
187
188	$(OX)\add$O : add_.c add.h
189	$(TCC) /Fo$@ -c add_.c
190
191
192	ELETED win/version.c

	--- win/Makefile.msc
	+++ win/Makefile.msc
	@@ -31,14 +31,14 @@
31	INCL = -I. -I$(SRCDIR) -I$B\win\include -I$(MSCDIR)\extra\include -I$(ZINCDIR)
32
33	CFLAGS = -nologo -MT -O2
34	BCC = $(CC) $(CFLAGS)
35	TCC = $(CC) -c $(CFLAGS) $(MSCDEF) $(SSL) $(INCL)
36	LIBS = $(ZLIB) ws2_32.lib advapi32.lib $(SSLLIB)
37	LIBDIR = -LIBPATH:$(MSCDIR)\extra\lib -LIBPATH:$(ZLIBDIR)
38
39	SQLITE_OPTIONS = /DSQLITE_OMIT_LOAD_EXTENSION=1 /DSQLITE_THREADSAFE=0 /DSQLITE_DEFAULT_FILE_FORMAT=4 /DSQLITE_ENABLE_STAT3 /Dlocaltime=fossil_localtime /DSQLITE_ENABLE_LOCKING_STYLE=0
40
41	SRC = add_.c allrepo_.c attach_.c bag_.c bisect_.c blob_.c branch_.c browse_.c captcha_.c cgi_.c checkin_.c checkout_.c clearsign_.c clone_.c comformat_.c configure_.c content_.c db_.c delta_.c deltacmd_.c descendants_.c diff_.c diffcmd_.c doc_.c encode_.c event_.c export_.c file_.c finfo_.c glob_.c graph_.c gzip_.c http_.c http_socket_.c http_ssl_.c http_transport_.c import_.c info_.c leaf_.c login_.c main_.c manifest_.c md5_.c merge_.c merge3_.c name_.c path_.c pivot_.c popen_.c pqueue_.c printf_.c rebuild_.c report_.c rss_.c schema_.c search_.c setup_.c sha1_.c shun_.c skins_.c sqlcmd_.c stash_.c stat_.c style_.c sync_.c tag_.c tar_.c th_main_.c timeline_.c tkt_.c tktsetup_.c undo_.c update_.c url_.c user_.c verify_.c vfile_.c wiki_.c wikiformat_.c winhttp_.c xfer_.c zip_.c
42
43	OBJ = $(OX)\add$O $(OX)\allrepo$O $(OX)\attach$O $(OX)\bag$O $(OX)\bisect$O $(OX)\blob$O $(OX)\branch$O $(OX)\browse$O $(OX)\captcha$O $(OX)\cgi$O $(OX)\checkin$O $(OX)\checkout$O $(OX)\clearsign$O $(OX)\clone$O $(OX)\comformat$O $(OX)\configure$O $(OX)\content$O $(OX)\db$O $(OX)\delta$O $(OX)\deltacmd$O $(OX)\descendants$O $(OX)\diff$O $(OX)\diffcmd$O $(OX)\doc$O $(OX)\encode$O $(OX)\event$O $(OX)\export$O $(OX)\file$O $(OX)\finfo$O $(OX)\glob$O $(OX)\graph$O $(OX)\gzip$O $(OX)\http$O $(OX)\http_socket$O $(OX)\http_ssl$O $(OX)\http_transport$O $(OX)\import$O $(OX)\info$O $(OX)\leaf$O $(OX)\login$O $(OX)\main$O $(OX)\manifest$O $(OX)\md5$O $(OX)\merge$O $(OX)\merge3$O $(OX)\name$O $(OX)\path$O $(OX)\pivot$O $(OX)\popen$O $(OX)\pqueue$O $(OX)\printf$O $(OX)\rebuild$O $(OX)\report$O $(OX)\rss$O $(OX)\schema$O $(OX)\search$O $(OX)\setup$O $(OX)\sha1$O $(OX)\shun$O $(OX)\skins$O $(OX)\sqlcmd$O $(OX)\stash$O $(OX)\stat$O $(OX)\style$O $(OX)\sync$O $(OX)\tag$O $(OX)\tar$O $(OX)\th_main$O $(OX)\timeline$O $(OX)\tkt$O $(OX)\tktsetup$O $(OX)\undo$O $(OX)\update$O $(OX)\url$O $(OX)\user$O $(OX)\verify$O $(OX)\vfile$O $(OX)\wiki$O $(OX)\wikiformat$O $(OX)\winhttp$O $(OX)\xfer$O $(OX)\zip$O $(OX)\shell$O $(OX)\sqlite3$O $(OX)\th$O $(OX)\th_lang$O
44
	@@ -153,11 +153,11 @@
153	$(BCC) $**
154
155	mkindex$E: $(SRCDIR)\mkindex.c
156	$(BCC) $**
157
158	mkversion$E: $B\src\mkversion.c
159	$(BCC) $**
160
161	$(OX)\shell$O : $(SRCDIR)\shell.c
162	$(TCC) /Fo$@ /Dmain=sqlite3_shell $(SQLITE_OPTIONS) -c $(SRCDIR)\shell.c
163
	@@ -168,11 +168,11 @@
168	$(TCC) /Fo$@ -c $**
169
170	$(OX)\th_lang$O : $(SRCDIR)\th_lang.c
171	$(TCC) /Fo$@ -c $**
172
173	VERSION.h : mkversion$E $B\manifest.uuid $B\manifest $B\VERSION
174	$** > $@
175
176	page_index.h: mkindex$E $(SRC)
177	$** > $@
178
	@@ -180,11 +180,11 @@
180	-del $(OX)\*.obj
181	-del .obj _.c .h .map
182	-del headers linkopts
183
184	realclean:
185	-del $(APPNAME) translate$E mkindex$E makeheaders$E mkversion$E
186
187
188	$(OX)\add$O : add_.c add.h
189	$(TCC) /Fo$@ -c add_.c
190
191
192	ELETED win/version.c

D win/version.c

-10

		--- a/win/version.c
		+++ b/win/version.c
		@@ -1,10 +0,0 @@
1		-fopen(argv[1],"gv[1],exists to do the job that AWK would do for the unix
2		-** makefile - to extract information from the "mainfest" and "manifest.uuid"
3		-** files for this p)
4		- d = _MSC_VER / 1 return 0;
5		-}
6		-b2b2,0,sizeof(b2)strcpy(b2b2b2vx[1000]return 0;
7		-}
8		-;
9		- char b[1024b[strlen(b)-1] =0; return 0"gv[1],"fMSC_VER)
10		- fopen(argv[1

	--- a/win/version.c
	+++ b/win/version.c
	@@ -1,10 +0,0 @@
1	fopen(argv[1],"gv[1],exists to do the job that AWK would do for the unix
2	** makefile - to extract information from the "mainfest" and "manifest.uuid"
3	** files for this p)
4	d = _MSC_VER / 1 return 0;
5	}
6	b2b2,0,sizeof(b2)strcpy(b2b2b2vx[1000]return 0;
7	}
8	;
9	char b[1024b[strlen(b)-1] =0; return 0"gv[1],"fMSC_VER)
10	fopen(argv[1

	--- a/win/version.c
	+++ b/win/version.c
	@@ -1,10 +0,0 @@

M www/build.wiki

+63 -26

		--- www/build.wiki
		+++ www/build.wiki
		@@ -1,16 +1,29 @@
1	1	<title>Building and Installing Fossil</title>
2	2
3		-<p>This page describes how to build and install Fossil. The
4		-whole process is designed to be very easy.</p>
5		-
6	3	<h2>0.0 Using A Pre-compiled Binary</h2>
7	4
8		-<p>You can skip steps 1.0 and 2.0 below by downloading
	5	+<p>You can skip all of the following by downloading
9	6	a <a href="http://www.fossil-scm.org/download.html">pre-compiled binary</a>
10		-appropriate for your platform. If you use a pre-compiled binary
11		-jump immediate to step 3.0.</p>
	7	+appropriate for your platform and putting that self-contained binary
	8	+someplace on your $PATH.
	9	+To uninstall, simply delete the binary.
	10	+To upgrade from an older release, just overwrite the older binary with
	11	+the newer one.</p>
	12	+
	13	+<h2>0.1 Executive Summary</h2>
	14	+
	15	+<p>Building and installing is very simple. Three steps:</p>
	16	+
	17	+<ol>
	18	+<li> Download and unpack a source tarball or ZIP.
	19	+<li> <b>./configure; make</b>
	20	+<li> Move or copy the resulting "fossil" executable to someplace
	21	+ on your $PATH.
	22	+</ol>
	23	+
	24	+<p><hr>
12	25
13	26	<h2>1.0 Obtaining The Source Code</h2>
14	27
15	28	<p>Fossil is self-hosting, so you can obtain a ZIP archive containing
16	29	a snapshot of the latest version directly from fossil's own fossil
		@@ -33,42 +46,66 @@
33	46
34	47	<li><p>Select a version of of fossil you want to download. Click on its
35	48	link. Note that you must successfully log in as "anonymous" in step 1
36	49	above in order to see the link to the detailed version information.</p></li>
37	50
38		-<li><p>Finally, click on the
39		-"Zip Archive" link. This link will build a ZIP archive of the
	51	+<li><p>Finally, click on one of the
	52	+"Zip Archive" or "Tarball" links, according to your preference.
	53	+These link will build a ZIP archive or a gzip-compressed tarball of the
40	54	complete source code and download it to your browser.
41	55	</ol>
42	56
43	57	<h2>2.0 Compiling</h2>
44	58
45		-<p>Follow these steps to compile on a unix platform
46		-(Linux, *BSD, MacOS, etc):</p>
47		-
48	59	<ol>
49	60	<li value="6">
50		-<p>Create a directory to hold the source code. Then unzip the
51		-ZIP archive you downloaded into that directory. You should be
	61	+<p>Create a directory to hold the source code. Then unarchive the
	62	+ZIP or tarball you downloaded into that directory. You should be
52	63	in the top-level folder of that directory</p></li>
53	64
54		-<li><p><b>(Optional:)</b>
55		-Edit the Makefile to set it up like you want. You probably do not
56		-need to do anything. Do not be intimidated: There are less than 10
57		-variables in the makefile that can be changed. The whole Makefile
58		-is only a few dozen lines long and most of those lines are comments.</p>
	65	+<li><i>(Optional, unix only)</i>
	66	+Run <b>./configure</b> to construct a makefile.
	67	+
	68	+<ol type="a">
	69	+<li><p>
	70	+If you do not have the OpenSSL library installed on your system, then
	71	+add <b>--with-openssl=none</b> to omit the https functionality.
	72	+
	73	+<li><p>
	74	+To build a statically linked binary (suitable for use inside a chroot
	75	+jail) add the <b>--static</b> option.
	76	+
	77	+<li><p>
	78	+Other configuration options can be seen by running
	79	+<b>./configure --help</b>
	80	+</ol>
	81	+
	82	+<li><p>Run make to build the "fossil" or "fossil.exe" executable. The
	83	+details depend on your platform and compiler.
	84	+
	85	+<ol type="a">
	86	+<li><p><i>Unix</i> → the configure-generated Makefile should work on
	87	+all unix and unix-like systems. Simply type "<b>make</b>".
	88	+
	89	+<li><p><i>Unix without running "configure"</i> → if you prefer to avoid running configure, you
	90	+can also use: <b>make -f Makefile.classic</b>. You may want to make minor
	91	+edits to Makefile.classic to configure the build for your system.
	92	+
	93	+<li><p><i>Msys/MinGW</i> → Use the
	94	+mingw makefile: "<b>make -f win/Makefile.mingw</b>"
59	95
60		-<li><p>Type "<b>make</b>"
	96	+<li><p><i>VC++</i> → Use the msc makefile. First
	97	+change to the "win/" subdirectory ("<b>cd win</b>") then run
	98	+"<b>nmake /f Makefile.msc</b>".
	99	+</ol>
61	100	</ol>
62	101
63		-<p>To build on windows, use an alternative makefile suitable for your
64		-particular build environment. The alternative windows makefiles are
65		-all found in the win/ subdirectory of the source tree. So, for example,
66		-if you want build using the
67		-[http://www.mingw.org/ \| mingw/msys compiler package] for windows, then
68		-run "<b>make -f win/Makefile.mingw</b>" instead of just "<b>make</b>"
69		-in step 8 above.</p>
	102	+<p>Note that Fossil requires the "zlib" compression library. This library
	103	+is available by default on most unix systems, but it will typically have to
	104	+be installed separately on windows systems. For windows builds, you may
	105	+need to edit the makefile to tell it exactly where zlib is located on your
	106	+system.</p>
70	107
71	108	<h2>3.0 Installing</h2>
72	109
73	110	<ol>
74	111	<li value="9">
75	112
76	113	ADDED www/changes.wiki

	--- www/build.wiki
	+++ www/build.wiki
	@@ -1,16 +1,29 @@
1	<title>Building and Installing Fossil</title>
2
3	<p>This page describes how to build and install Fossil. The
4	whole process is designed to be very easy.</p>
5
6	<h2>0.0 Using A Pre-compiled Binary</h2>
7
8	<p>You can skip steps 1.0 and 2.0 below by downloading
9	a <a href="http://www.fossil-scm.org/download.html">pre-compiled binary</a>
10	appropriate for your platform. If you use a pre-compiled binary
11	jump immediate to step 3.0.</p>
















12
13	<h2>1.0 Obtaining The Source Code</h2>
14
15	<p>Fossil is self-hosting, so you can obtain a ZIP archive containing
16	a snapshot of the latest version directly from fossil's own fossil
	@@ -33,42 +46,66 @@
33
34	<li><p>Select a version of of fossil you want to download. Click on its
35	link. Note that you must successfully log in as "anonymous" in step 1
36	above in order to see the link to the detailed version information.</p></li>
37
38	<li><p>Finally, click on the
39	"Zip Archive" link. This link will build a ZIP archive of the

40	complete source code and download it to your browser.
41	</ol>
42
43	<h2>2.0 Compiling</h2>
44
45	<p>Follow these steps to compile on a unix platform
46	(Linux, *BSD, MacOS, etc):</p>
47
48	<ol>
49	<li value="6">
50	<p>Create a directory to hold the source code. Then unzip the
51	ZIP archive you downloaded into that directory. You should be
52	in the top-level folder of that directory</p></li>
53
54	<li><p><b>(Optional:)</b>
55	Edit the Makefile to set it up like you want. You probably do not
56	need to do anything. Do not be intimidated: There are less than 10
57	variables in the makefile that can be changed. The whole Makefile
58	is only a few dozen lines long and most of those lines are comments.</p>

























59
60	<li><p>Type "<b>make</b>"



61	</ol>
62
63	<p>To build on windows, use an alternative makefile suitable for your
64	particular build environment. The alternative windows makefiles are
65	all found in the win/ subdirectory of the source tree. So, for example,
66	if you want build using the
67	[http://www.mingw.org/ \| mingw/msys compiler package] for windows, then
68	run "<b>make -f win/Makefile.mingw</b>" instead of just "<b>make</b>"
69	in step 8 above.</p>
70
71	<h2>3.0 Installing</h2>
72
73	<ol>
74	<li value="9">
75
76	DDED www/changes.wiki

	--- www/build.wiki
	+++ www/build.wiki
	@@ -1,16 +1,29 @@
1	<title>Building and Installing Fossil</title>
2



3	<h2>0.0 Using A Pre-compiled Binary</h2>
4
5	<p>You can skip all of the following by downloading
6	a <a href="http://www.fossil-scm.org/download.html">pre-compiled binary</a>
7	appropriate for your platform and putting that self-contained binary
8	someplace on your $PATH.
9	To uninstall, simply delete the binary.
10	To upgrade from an older release, just overwrite the older binary with
11	the newer one.</p>
12
13	<h2>0.1 Executive Summary</h2>
14
15	<p>Building and installing is very simple. Three steps:</p>
16
17	<ol>
18	<li> Download and unpack a source tarball or ZIP.
19	<li> <b>./configure; make</b>
20	<li> Move or copy the resulting "fossil" executable to someplace
21	on your $PATH.
22	</ol>
23
24	<p><hr>
25
26	<h2>1.0 Obtaining The Source Code</h2>
27
28	<p>Fossil is self-hosting, so you can obtain a ZIP archive containing
29	a snapshot of the latest version directly from fossil's own fossil
	@@ -33,42 +46,66 @@
46
47	<li><p>Select a version of of fossil you want to download. Click on its
48	link. Note that you must successfully log in as "anonymous" in step 1
49	above in order to see the link to the detailed version information.</p></li>
50
51	<li><p>Finally, click on one of the
52	"Zip Archive" or "Tarball" links, according to your preference.
53	These link will build a ZIP archive or a gzip-compressed tarball of the
54	complete source code and download it to your browser.
55	</ol>
56
57	<h2>2.0 Compiling</h2>
58



59	<ol>
60	<li value="6">
61	<p>Create a directory to hold the source code. Then unarchive the
62	ZIP or tarball you downloaded into that directory. You should be
63	in the top-level folder of that directory</p></li>
64
65	<li><i>(Optional, unix only)</i>
66	Run <b>./configure</b> to construct a makefile.
67
68	<ol type="a">
69	<li><p>
70	If you do not have the OpenSSL library installed on your system, then
71	add <b>--with-openssl=none</b> to omit the https functionality.
72
73	<li><p>
74	To build a statically linked binary (suitable for use inside a chroot
75	jail) add the <b>--static</b> option.
76
77	<li><p>
78	Other configuration options can be seen by running
79	<b>./configure --help</b>
80	</ol>
81
82	<li><p>Run make to build the "fossil" or "fossil.exe" executable. The
83	details depend on your platform and compiler.
84
85	<ol type="a">
86	<li><p><i>Unix</i> → the configure-generated Makefile should work on
87	all unix and unix-like systems. Simply type "<b>make</b>".
88
89	<li><p><i>Unix without running "configure"</i> → if you prefer to avoid running configure, you
90	can also use: <b>make -f Makefile.classic</b>. You may want to make minor
91	edits to Makefile.classic to configure the build for your system.
92
93	<li><p><i>Msys/MinGW</i> → Use the
94	mingw makefile: "<b>make -f win/Makefile.mingw</b>"
95
96	<li><p><i>VC++</i> → Use the msc makefile. First
97	change to the "win/" subdirectory ("<b>cd win</b>") then run
98	"<b>nmake /f Makefile.msc</b>".
99	</ol>
100	</ol>
101
102	<p>Note that Fossil requires the "zlib" compression library. This library
103	is available by default on most unix systems, but it will typically have to
104	be installed separately on windows systems. For windows builds, you may
105	need to edit the makefile to tell it exactly where zlib is located on your
106	system.</p>


107
108	<h2>3.0 Installing</h2>
109
110	<ol>
111	<li value="9">
112
113	DDED www/changes.wiki

M www/changes.wiki

		--- a/www/changes.wiki
		+++ b/www/changes.wiki
		@@ -0,0 +1 @@
	1	+2-xxVarious minor bug fixepending

	--- a/www/changes.wiki
	+++ b/www/changes.wiki
	@@ -0,0 +1 @@

	--- a/www/changes.wiki
	+++ b/www/changes.wiki
	@@ -0,0 +1 @@
1	2-xxVarious minor bug fixepending

M www/checkin.wiki

+80 -16

		--- www/checkin.wiki
		+++ www/checkin.wiki
		@@ -1,21 +1,85 @@
1		-<title>Checkin Checklist</title>
	1	+<title>Check-in Checklist</title>
	2	+
	3	+<h2><u>Always</u> run the following checklist prior to <u>every</u>
	4	+check-in or commit to the Fossil repository:</h2>
	5	+
	6	+Before every check-in:
2	7
3		-Before every checkins:
	8	+ 0. <b>fossil user default</b> → your username is correct.
4	9
5		- 1. <b>fossil diff</b> → no stray changes
	10	+ 1. <b>fossil diff</b> →
	11	+ <ol type="a">
	12	+ <li> No stray changes
	13	+ <li> All changes comply with the license
	14	+ <li> All inputs are scrubbed before use
	15	+ <li> No injection attacks via %s formats
	16	+ </ol>
6	17
7	18	2. <b>fossil extra</b> → no unmanaged files need to be added.
8	19
9		- 3. The checkin will go onto the desired branch.
10		-
11		- 4. "Autosync" is enabled →
12		- <ol>
13		- <li> The checkin will not cause a unintentional fork.
14		- <li> The local system clock is set correctly.
15		- </ol>
16		-
17		-Before every checkin to <b>trunk</b>:
18		-
19		- 5. No compiler warnings on the development machine.
20		-
21		- 6. Changes will not cause problems on a future <b>bisect</b>.
	20	+ 3. The check-in will go onto the desired branch.
	21	+ → Do <u>not</u> check into trunk without prior approval from
	22	+ the lead programmer (drh)!
	23	+
	24	+ 4. auto-sync is on, or the system clock is verified
	25	+
	26	+ 5. If sources files have been added or removed, ensure all makefiles
	27	+ and configure scripts have been updated accordingly.
	28	+
	29	+Before every check-in to <b>trunk</b>:
	30	+
	31	+ 6. No compiler warnings on the development machine.
	32	+
	33	+ 7. The fossil executable that results from a build actually works.
	34	+
	35	+
	36	+<hr>
	37	+<h2>Commentary</h2>
	38	+
	39	+Before you go ahead and push content back to the servers, make sure
	40	+that the username you are using by default matches your username
	41	+within the project. Also remember to enable the localauth setting
	42	+if you intend to make changes via a locally served web UI.
	43	+
	44	+Item 1 is the most important step. Consider using <b>gdiff</b>
	45	+instead of <b>diff</b> if you have a graphical differ configured. Or,
	46	+pipe the output of "<b>fossil diff</b>" into "<b>open -f</b>" (on a mac) to
	47	+get the diff output in a separate text window for easier viewing.
	48	+Look carefully at every changed line in item 1.
	49	+Make sure that you are not about to commit unrelated changes.
	50	+If there are two or more unrelated changes present, consider
	51	+breaking up the commit into two or more separate commits.
	52	+Always make 100% sure that all changes are compatible with the
	53	+BSD license, that you have the authority to commit the code in accordance
	54	+with the [/doc/trunk/www/copyright-release.html \| CLA] that you have
	55	+signed and have on file, and that
	56	+no NDAs, copyrights, patents, or trademarks are infringed by the check-in.
	57	+Also check very carefully to make sure that
	58	+you are not introducing security vulnerabilities. Pay particular attention
	59	+to the possibility of SQL or HTML injection attacks.
	60	+
	61	+Item 2 verifies that you have not added source files but failed to
	62	+do the necessary "<b>fossil add</b>" to manage them. If you commit
	63	+without bringing the new file under source control, the check-in will
	64	+be broken. That, in turn, can cause complications far in the future
	65	+when we are bisecting for a bug.
	66	+
	67	+For item 3, Run "<b>fossil status</b>" or the equivalent to
	68	+make sure your changes are going into the branch you think they are.
	69	+Also double-check the branch name when entering change comments.
	70	+Never check into trunk unless you are expressly authorized to do so.
	71	+
	72	+For Item 4, if you are on-network, make sure auto-sync is enabled. This
	73	+will minimize the risk of an unintended fork. It will also give you a
	74	+warning if you system clock is set incorrectly. If you are off-network,
	75	+make sure that your system clock is correct or at least close to correct
	76	+so that your check-in does not appear out-of-sequence on timelines.
	77	+On-network commits are preferred, especially for trunk commits.
	78	+
	79	+Items 6 and 7 help to ensure that check-ins on the trunk always work.
	80	+Knowing that the trunk always works makes bisecting much easier. Items
	81	+6 and 7 are recommended for all check-ins, even those that are on a branch.
	82	+But they are especially important for trunk check-ins. We desire that
	83	+all trunk check-ins work at all times. Any experimental, unstable, or
	84	+questionable changes should go on a branch and be merged into trunk after
	85	+further testing.
22	86
23	87	ADDED www/foss-cklist.wiki

	--- www/checkin.wiki
	+++ www/checkin.wiki
	@@ -1,21 +1,85 @@
1	<title>Checkin Checklist</title>





2
3	Before every checkins:
4
5	1. <b>fossil diff</b> → no stray changes






6
7	2. <b>fossil extra</b> → no unmanaged files need to be added.
8
9	3. The checkin will go onto the desired branch.
10
11	4. "Autosync" is enabled →
12	<ol>
13	<li> The checkin will not cause a unintentional fork.
14	<li> The local system clock is set correctly.
15	</ol>
16
17	Before every checkin to <b>trunk</b>:
18
19	5. No compiler warnings on the development machine.
20
21	6. Changes will not cause problems on a future <b>bisect</b>.





















































22
23	DDED www/foss-cklist.wiki

	--- www/checkin.wiki
	+++ www/checkin.wiki
	@@ -1,21 +1,85 @@
1	<title>Check-in Checklist</title>
2
3	<h2><u>Always</u> run the following checklist prior to <u>every</u>
4	check-in or commit to the Fossil repository:</h2>
5
6	Before every check-in:
7
8	0. <b>fossil user default</b> → your username is correct.
9
10	1. <b>fossil diff</b> →
11	<ol type="a">
12	<li> No stray changes
13	<li> All changes comply with the license
14	<li> All inputs are scrubbed before use
15	<li> No injection attacks via %s formats
16	</ol>
17
18	2. <b>fossil extra</b> → no unmanaged files need to be added.
19
20	3. The check-in will go onto the desired branch.
21	→ Do <u>not</u> check into trunk without prior approval from
22	the lead programmer (drh)!
23
24	4. auto-sync is on, or the system clock is verified
25
26	5. If sources files have been added or removed, ensure all makefiles
27	and configure scripts have been updated accordingly.
28
29	Before every check-in to <b>trunk</b>:
30
31	6. No compiler warnings on the development machine.
32
33	7. The fossil executable that results from a build actually works.
34
35
36	<hr>
37	<h2>Commentary</h2>
38
39	Before you go ahead and push content back to the servers, make sure
40	that the username you are using by default matches your username
41	within the project. Also remember to enable the localauth setting
42	if you intend to make changes via a locally served web UI.
43
44	Item 1 is the most important step. Consider using <b>gdiff</b>
45	instead of <b>diff</b> if you have a graphical differ configured. Or,
46	pipe the output of "<b>fossil diff</b>" into "<b>open -f</b>" (on a mac) to
47	get the diff output in a separate text window for easier viewing.
48	Look carefully at every changed line in item 1.
49	Make sure that you are not about to commit unrelated changes.
50	If there are two or more unrelated changes present, consider
51	breaking up the commit into two or more separate commits.
52	Always make 100% sure that all changes are compatible with the
53	BSD license, that you have the authority to commit the code in accordance
54	with the [/doc/trunk/www/copyright-release.html \| CLA] that you have
55	signed and have on file, and that
56	no NDAs, copyrights, patents, or trademarks are infringed by the check-in.
57	Also check very carefully to make sure that
58	you are not introducing security vulnerabilities. Pay particular attention
59	to the possibility of SQL or HTML injection attacks.
60
61	Item 2 verifies that you have not added source files but failed to
62	do the necessary "<b>fossil add</b>" to manage them. If you commit
63	without bringing the new file under source control, the check-in will
64	be broken. That, in turn, can cause complications far in the future
65	when we are bisecting for a bug.
66
67	For item 3, Run "<b>fossil status</b>" or the equivalent to
68	make sure your changes are going into the branch you think they are.
69	Also double-check the branch name when entering change comments.
70	Never check into trunk unless you are expressly authorized to do so.
71
72	For Item 4, if you are on-network, make sure auto-sync is enabled. This
73	will minimize the risk of an unintended fork. It will also give you a
74	warning if you system clock is set incorrectly. If you are off-network,
75	make sure that your system clock is correct or at least close to correct
76	so that your check-in does not appear out-of-sequence on timelines.
77	On-network commits are preferred, especially for trunk commits.
78
79	Items 6 and 7 help to ensure that check-ins on the trunk always work.
80	Knowing that the trunk always works makes bisecting much easier. Items
81	6 and 7 are recommended for all check-ins, even those that are on a branch.
82	But they are especially important for trunk check-ins. We desire that
83	all trunk check-ins work at all times. Any experimental, unstable, or
84	questionable changes should go on a branch and be merged into trunk after
85	further testing.
86
87	DDED www/foss-cklist.wiki

M www/foss-cklist.wiki

+60

		--- a/www/foss-cklist.wiki
		+++ b/www/foss-cklist.wiki
		@@ -0,0 +1,60 @@
	1	+<title>Checklist For Successful Open-Source Projects</title>
	2	+<nowiki>
	3	+
	4	+<p>This checklist is loosely derived from Tom "Spot" Callaway's Fail Score
	5	+blog post <a href="http://spot.livejournal.com/308370.html">
	6	+http://spot.livejournal.com/308370.html</a> (see also
	7	+<a href="http://www.theopensourceway.org/book/The_Open_Source_Way-How_to_tell_if_a_FLOSS_project_is_doomed_to_FAIL.html">[1]</a> and
	8	+<a href="https://wwwHow_to_tell_if_a_FLOSS_project_is_doomed_to_FAIL.html">[1]</a>).
	9	+Tom's original post assigned point scores to the various elements and
	10	+by adding together the individual points, the reader is suposed to be able
	11	+to judge the likelihood that the project will fail.
	12	+The point scores, and the items on the list, clearly reflect Tom's
	13	+biases and are not necessarily those of the larger open-source community.
	14	+Nevertheless, the policy of the Fossil shall be to strive for a perfect
	15	+score.</p>
	16	+
	17	+<p>rive for a perfect
	18	+score.
	19	+
	20	+This checklist is an inversion of Tom's original post in that it strives to
	21	+say what the project should do in order to succeed rather than what it
	22	+should not do to avoid failure. Th</p>
	23	+
	24	+<p>point values are omitted.
	25	+
	26	+See also:
	27	+<ul>
	28	+<li><a href="http://offog.org/articles/packaging/">
	29	+http://offog.org/articles/packaging/</a>
	30	+<li>
	31	+<a href="http://www.gnu.org/prep/standards/standards.html#Managing-Releases">
	32	+http://www.gnu.org/prep/standards/standards.html#Managing-Releases<p>The source code size is less than <p>The project uses a Version Control System (VCS).
	33	+<ol type="a">
	34	+<li>The VCS has a working web interface.
	35	+<li>There is documentation on how to use the VCS.
	36	+<li>The VCS is general-purpose, not something hacked together for this
	37	+ <p>The project comes with documentation on how to build from source
	38	+and that documentation is lucid, correct, and up-to-date.
	39	+
	40	+<li><p>The project is configurable using an autoconf-generated configure
	41	+script, or the equivalent, and does not require:
	42	+<ol type="a">
	43	+<li>Manually editing flat files
	44	+<li>Editing code header files
	45	+</ol>
	46	+
	47	+<li><pes a Version les
	48	+</ol>
	49	+
	50	+<li>The project should be buildable using commonly available and
	51	+ standa<p Open-Source Projects</title>
	52	+<nowiki>
	53	+
	54	+This checklist is loosely derive<t>The project is configurable using an autoconf-generated configure
	55	+script, or the equivalent, and does not requir essful Open-Source Projects</title>
	56	+<nowiki>
	57	+
	58	+<p>This checklist is loosely derived from Tom "Spot" Callaway's Fail Score
	59	+blog post <a href="http://spot.livejournal.com/308370.html">
	60	+http://sp

	--- a/www/foss-cklist.wiki
	+++ b/www/foss-cklist.wiki
	@@ -0,0 +1,60 @@

	--- a/www/foss-cklist.wiki
	+++ b/www/foss-cklist.wiki
	@@ -0,0 +1,60 @@
1	<title>Checklist For Successful Open-Source Projects</title>
2	<nowiki>
3
4	<p>This checklist is loosely derived from Tom "Spot" Callaway's Fail Score
5	blog post <a href="http://spot.livejournal.com/308370.html">
6	http://spot.livejournal.com/308370.html</a> (see also
7	<a href="http://www.theopensourceway.org/book/The_Open_Source_Way-How_to_tell_if_a_FLOSS_project_is_doomed_to_FAIL.html">[1]</a> and
8	<a href="https://wwwHow_to_tell_if_a_FLOSS_project_is_doomed_to_FAIL.html">[1]</a>).
9	Tom's original post assigned point scores to the various elements and
10	by adding together the individual points, the reader is suposed to be able
11	to judge the likelihood that the project will fail.
12	The point scores, and the items on the list, clearly reflect Tom's
13	biases and are not necessarily those of the larger open-source community.
14	Nevertheless, the policy of the Fossil shall be to strive for a perfect
15	score.</p>
16
17	<p>rive for a perfect
18	score.
19
20	This checklist is an inversion of Tom's original post in that it strives to
21	say what the project should do in order to succeed rather than what it
22	should not do to avoid failure. Th</p>
23
24	<p>point values are omitted.
25
26	See also:
27	<ul>
28	<li><a href="http://offog.org/articles/packaging/">
29	http://offog.org/articles/packaging/</a>
30	<li>
31	<a href="http://www.gnu.org/prep/standards/standards.html#Managing-Releases">
32	http://www.gnu.org/prep/standards/standards.html#Managing-Releases<p>The source code size is less than <p>The project uses a Version Control System (VCS).
33	<ol type="a">
34	<li>The VCS has a working web interface.
35	<li>There is documentation on how to use the VCS.
36	<li>The VCS is general-purpose, not something hacked together for this
37	<p>The project comes with documentation on how to build from source
38	and that documentation is lucid, correct, and up-to-date.
39
40	<li><p>The project is configurable using an autoconf-generated configure
41	script, or the equivalent, and does not require:
42	<ol type="a">
43	<li>Manually editing flat files
44	<li>Editing code header files
45	</ol>
46
47	<li><pes a Version les
48	</ol>
49
50	<li>The project should be buildable using commonly available and
51	standa<p Open-Source Projects</title>
52	<nowiki>
53
54	This checklist is loosely derive<t>The project is configurable using an autoconf-generated configure
55	script, or the equivalent, and does not requir essful Open-Source Projects</title>
56	<nowiki>
57
58	<p>This checklist is loosely derived from Tom "Spot" Callaway's Fail Score
59	blog post <a href="http://spot.livejournal.com/308370.html">
60	http://sp

M www/index.wiki

		--- www/index.wiki
		+++ www/index.wiki
		@@ -20,10 +20,12 @@
20	20	<li> [./quickstart.wiki \| Quick Start]
21	21	<li> [./build.wiki \| Install]
22	22	<li> [../COPYRIGHT-BSD2.txt \| License]
23	23	<li> [/timeline \| Recent changes]
24	24	<li> [./faq.wiki \| FAQ]
	25	+<li> [./contribute.wiki \| Contributing]
	26	+<li> [./changes.wiki \| Change Log]
25	27	<li> [./permutedindex.wiki \| Doc Index]
26	28	<li> Mailing list
27	29	<ul>
28	30	<li> [http://lists.fossil-scm.org:8080/cgi-bin/mailman/listinfo/fossil-users \| sign-up]
29	31	<li> [http://www.mail-archive.com/[email protected] \| archives]
		@@ -125,10 +127,12 @@
125	127	* The [./selfcheck.wiki \| automatic self-check] mechanism
126	128	helps insure project integrity.
127	129	* Fossil contains a [./wikitheory.wiki \| built-in wiki].
128	130	* An [./event.wiki \| Event] is a special kind of wiki page associated
129	131	with a point in time rather than a name.
	132	+ * [./settings.wiki \| Settings] control the behaviour of fossil.
	133	+ * [./ssl.wiki \| Use SSL] to encrypt communication with the server.
130	134	* There is a
131	135	[http://lists.fossil-scm.org:8080/cgi-bin/mailman/listinfo/fossil-users \| mailing list] (with publicly readable
132	136	[http://www.mail-archive.com/[email protected] \| archives]
133	137	available for discussing fossil issues.
134	138	* [./stats.wiki \| Performance statistics] taken from real-world projects
135	139

	--- www/index.wiki
	+++ www/index.wiki
	@@ -20,10 +20,12 @@
20	<li> [./quickstart.wiki \| Quick Start]
21	<li> [./build.wiki \| Install]
22	<li> [../COPYRIGHT-BSD2.txt \| License]
23	<li> [/timeline \| Recent changes]
24	<li> [./faq.wiki \| FAQ]


25	<li> [./permutedindex.wiki \| Doc Index]
26	<li> Mailing list
27	<ul>
28	<li> [http://lists.fossil-scm.org:8080/cgi-bin/mailman/listinfo/fossil-users \| sign-up]
29	<li> [http://www.mail-archive.com/[email protected] \| archives]
	@@ -125,10 +127,12 @@
125	* The [./selfcheck.wiki \| automatic self-check] mechanism
126	helps insure project integrity.
127	* Fossil contains a [./wikitheory.wiki \| built-in wiki].
128	* An [./event.wiki \| Event] is a special kind of wiki page associated
129	with a point in time rather than a name.


130	* There is a
131	[http://lists.fossil-scm.org:8080/cgi-bin/mailman/listinfo/fossil-users \| mailing list] (with publicly readable
132	[http://www.mail-archive.com/[email protected] \| archives]
133	available for discussing fossil issues.
134	* [./stats.wiki \| Performance statistics] taken from real-world projects
135

	--- www/index.wiki
	+++ www/index.wiki
	@@ -20,10 +20,12 @@
20	<li> [./quickstart.wiki \| Quick Start]
21	<li> [./build.wiki \| Install]
22	<li> [../COPYRIGHT-BSD2.txt \| License]
23	<li> [/timeline \| Recent changes]
24	<li> [./faq.wiki \| FAQ]
25	<li> [./contribute.wiki \| Contributing]
26	<li> [./changes.wiki \| Change Log]
27	<li> [./permutedindex.wiki \| Doc Index]
28	<li> Mailing list
29	<ul>
30	<li> [http://lists.fossil-scm.org:8080/cgi-bin/mailman/listinfo/fossil-users \| sign-up]
31	<li> [http://www.mail-archive.com/[email protected] \| archives]
	@@ -125,10 +127,12 @@
127	* The [./selfcheck.wiki \| automatic self-check] mechanism
128	helps insure project integrity.
129	* Fossil contains a [./wikitheory.wiki \| built-in wiki].
130	* An [./event.wiki \| Event] is a special kind of wiki page associated
131	with a point in time rather than a name.
132	* [./settings.wiki \| Settings] control the behaviour of fossil.
133	* [./ssl.wiki \| Use SSL] to encrypt communication with the server.
134	* There is a
135	[http://lists.fossil-scm.org:8080/cgi-bin/mailman/listinfo/fossil-users \| mailing list] (with publicly readable
136	[http://www.mail-archive.com/[email protected] \| archives]
137	available for discussing fossil issues.
138	* [./stats.wiki \| Performance statistics] taken from real-world projects
139

M www/makefile.wiki

+21 -15

		--- www/makefile.wiki
		+++ www/makefile.wiki
		@@ -52,29 +52,34 @@
52	52	7. th.c
53	53	8. th.h
54	54
55	55	These two files are imports like the SQLite source files,
56	56	and so are not preprocessed.
	57	+
	58	+The VERSION.h header file is generated from other information sources
	59	+using a small program called:
	60	+
	61	+ 9. mkversion.c
57	62
58	63	The src/ subdirectory also contains documentation about the
59	64	makeheaders preprocessor program:
60	65
61		- 9. [../src/makeheaders.html \| makeheaders.html]
	66	+ 10. [../src/makeheaders.html \| makeheaders.html]
62	67
63	68	Click on the link to read this documentation. In addition there is
64	69	a [http://www.tcl.tk/ \| Tcl] script used to build the various makefiles:
65	70
66		- 10. makemake.tcl
	71	+ 11. makemake.tcl
67	72
68	73	Running this Tcl script will automatically regenerate all makefiles.
69	74	In order to add a new source file to the Fossil implementation, simply
70	75	edit makemake.tcl to add the new filename, then rerun the script, and
71	76	all of the makefiles for all targets will be rebuild.
72	77
73	78	Finally, there is one of the makefiles generated by makemake.tcl:
74	79
75		- 11. main.mk
	80	+ 12. main.mk
76	81
77	82	The main.mk makefile is invoked from the Makefile in the top-level
78	83	directory. The main.mk is generated by makemake.tcl and should not
79	84	be hand edited. Other makefiles generated by makemake.tcl are in
80	85	other subdirectories (currently all in the win/ subdirectory).
		@@ -89,31 +94,32 @@
89	94
90	95	<h1>3.0 Automatically generated files</h1>
91	96
92	97	The "VERSION.h" header file contains some C preprocessor macros that
93	98	identify the version of Fossil that is to be build. The VERSION.h file is
94		-generated automatically from information extracted from the "manifest"
95		-and "manifest.uuid" source files in the root directory of the source tree.
	99	+generated automatically from information extracted from the "manifest",
	100	+"manifest.uuid", and "VERSION" source files in the root directory of the
	101	+source tree.
96	102	(The "manifest" and "manifest.uuid" files are automatically generated and
97	103	updated by Fossil itself. See the [/help/setting \| fossil set manifest]
98	104	command for additional information.)
99	105
100		-Under unix, there is an AWK script that converts manifest and manifest.uuid
101		-into VERSION.h. For windows, there is a C program: win/version.c.
102		-(Unix builds could also used the version.c convert, if desired, but AWK
103		-seems easier there.) To run the VERSION.h generator, first compile
104		-the win/version.c source file into a command-line program (named "version.exe")
	106	+The VERSION.h header file is generated by
	107	+a C program: src/mkversion.c.
	108	+To run the VERSION.h generator, first compile the src/mkversion.c
	109	+ source file into a command-line program (named "mkversion.exe")
105	110	than run:
106	111
107	112	<blockquote><pre>
108		-version.exe manifest.uuid manifest VERSION.h
	113	+mkversion.exe manifest.uuid manifest VERSION >VERSION.h
109	114	</pre></blockquote>
110	115
111	116	The pathnames in the above command might need to be adjusted to get the
112		-directories right. The point is that the manifest.uuid and manifest files
113		-in the root of the source tree are the first two arguments and the name of
114		-the generated VERSION.h file is the third and final argument.
	117	+directories right. The point is that the manifest.uuid, manifest, and
	118	+VERSION files
	119	+in the root of the source tree are the three arguments and
	120	+the generated VERSION.h file appears on standard output.
115	121
116	122	<h1>4.0 Preprocessing</h1>
117	123
118	124	There are three preprocessors for the Fossil sources. The mkindex
119	125	and translate preprocessors can be run in any order. The makeheaders
		@@ -120,11 +126,11 @@
120	126	preprocessor must be run after translate.
121	127
122	128	<h2>4.1 The mkindex preprocessor</h2>
123	129
124	130	The mkindex program scans the "src.c" source files looking for special
125		-comments that identify routines that implement of various Fossil commands,
	131	+comments that identify routines that implement various Fossil commands,
126	132	web interface methods, and help text comments. The mkindex program
127	133	generates some C code that Fossil uses in order to dispatch commands and
128	134	HTTP requests and to show on-line help. Compile the mkindex program
129	135	from the mkindex.c source file. Then run:
130	136
131	137

	--- www/makefile.wiki
	+++ www/makefile.wiki
	@@ -52,29 +52,34 @@
52	7. th.c
53	8. th.h
54
55	These two files are imports like the SQLite source files,
56	and so are not preprocessed.





57
58	The src/ subdirectory also contains documentation about the
59	makeheaders preprocessor program:
60
61	9. [../src/makeheaders.html \| makeheaders.html]
62
63	Click on the link to read this documentation. In addition there is
64	a [http://www.tcl.tk/ \| Tcl] script used to build the various makefiles:
65
66	10. makemake.tcl
67
68	Running this Tcl script will automatically regenerate all makefiles.
69	In order to add a new source file to the Fossil implementation, simply
70	edit makemake.tcl to add the new filename, then rerun the script, and
71	all of the makefiles for all targets will be rebuild.
72
73	Finally, there is one of the makefiles generated by makemake.tcl:
74
75	11. main.mk
76
77	The main.mk makefile is invoked from the Makefile in the top-level
78	directory. The main.mk is generated by makemake.tcl and should not
79	be hand edited. Other makefiles generated by makemake.tcl are in
80	other subdirectories (currently all in the win/ subdirectory).
	@@ -89,31 +94,32 @@
89
90	<h1>3.0 Automatically generated files</h1>
91
92	The "VERSION.h" header file contains some C preprocessor macros that
93	identify the version of Fossil that is to be build. The VERSION.h file is
94	generated automatically from information extracted from the "manifest"
95	and "manifest.uuid" source files in the root directory of the source tree.

96	(The "manifest" and "manifest.uuid" files are automatically generated and
97	updated by Fossil itself. See the [/help/setting \| fossil set manifest]
98	command for additional information.)
99
100	Under unix, there is an AWK script that converts manifest and manifest.uuid
101	into VERSION.h. For windows, there is a C program: win/version.c.
102	(Unix builds could also used the version.c convert, if desired, but AWK
103	seems easier there.) To run the VERSION.h generator, first compile
104	the win/version.c source file into a command-line program (named "version.exe")
105	than run:
106
107	<blockquote><pre>
108	version.exe manifest.uuid manifest VERSION.h
109	</pre></blockquote>
110
111	The pathnames in the above command might need to be adjusted to get the
112	directories right. The point is that the manifest.uuid and manifest files
113	in the root of the source tree are the first two arguments and the name of
114	the generated VERSION.h file is the third and final argument.

115
116	<h1>4.0 Preprocessing</h1>
117
118	There are three preprocessors for the Fossil sources. The mkindex
119	and translate preprocessors can be run in any order. The makeheaders
	@@ -120,11 +126,11 @@
120	preprocessor must be run after translate.
121
122	<h2>4.1 The mkindex preprocessor</h2>
123
124	The mkindex program scans the "src.c" source files looking for special
125	comments that identify routines that implement of various Fossil commands,
126	web interface methods, and help text comments. The mkindex program
127	generates some C code that Fossil uses in order to dispatch commands and
128	HTTP requests and to show on-line help. Compile the mkindex program
129	from the mkindex.c source file. Then run:
130
131

	--- www/makefile.wiki
	+++ www/makefile.wiki
	@@ -52,29 +52,34 @@
52	7. th.c
53	8. th.h
54
55	These two files are imports like the SQLite source files,
56	and so are not preprocessed.
57
58	The VERSION.h header file is generated from other information sources
59	using a small program called:
60
61	9. mkversion.c
62
63	The src/ subdirectory also contains documentation about the
64	makeheaders preprocessor program:
65
66	10. [../src/makeheaders.html \| makeheaders.html]
67
68	Click on the link to read this documentation. In addition there is
69	a [http://www.tcl.tk/ \| Tcl] script used to build the various makefiles:
70
71	11. makemake.tcl
72
73	Running this Tcl script will automatically regenerate all makefiles.
74	In order to add a new source file to the Fossil implementation, simply
75	edit makemake.tcl to add the new filename, then rerun the script, and
76	all of the makefiles for all targets will be rebuild.
77
78	Finally, there is one of the makefiles generated by makemake.tcl:
79
80	12. main.mk
81
82	The main.mk makefile is invoked from the Makefile in the top-level
83	directory. The main.mk is generated by makemake.tcl and should not
84	be hand edited. Other makefiles generated by makemake.tcl are in
85	other subdirectories (currently all in the win/ subdirectory).
	@@ -89,31 +94,32 @@
94
95	<h1>3.0 Automatically generated files</h1>
96
97	The "VERSION.h" header file contains some C preprocessor macros that
98	identify the version of Fossil that is to be build. The VERSION.h file is
99	generated automatically from information extracted from the "manifest",
100	"manifest.uuid", and "VERSION" source files in the root directory of the
101	source tree.
102	(The "manifest" and "manifest.uuid" files are automatically generated and
103	updated by Fossil itself. See the [/help/setting \| fossil set manifest]
104	command for additional information.)
105
106	The VERSION.h header file is generated by
107	a C program: src/mkversion.c.
108	To run the VERSION.h generator, first compile the src/mkversion.c
109	source file into a command-line program (named "mkversion.exe")

110	than run:
111
112	<blockquote><pre>
113	mkversion.exe manifest.uuid manifest VERSION >VERSION.h
114	</pre></blockquote>
115
116	The pathnames in the above command might need to be adjusted to get the
117	directories right. The point is that the manifest.uuid, manifest, and
118	VERSION files
119	in the root of the source tree are the three arguments and
120	the generated VERSION.h file appears on standard output.
121
122	<h1>4.0 Preprocessing</h1>
123
124	There are three preprocessors for the Fossil sources. The mkindex
125	and translate preprocessors can be run in any order. The makeheaders
	@@ -120,11 +126,11 @@
126	preprocessor must be run after translate.
127
128	<h2>4.1 The mkindex preprocessor</h2>
129
130	The mkindex program scans the "src.c" source files looking for special
131	comments that identify routines that implement various Fossil commands,
132	web interface methods, and help text comments. The mkindex program
133	generates some C code that Fossil uses in order to dispatch commands and
134	HTTP requests and to show on-line help. Compile the mkindex program
135	from the mkindex.c source file. Then run:
136
137

M www/mkindex.tcl

+11 -6

		--- www/mkindex.tcl
		+++ www/mkindex.tcl
		@@ -9,10 +9,11 @@
9	9	set doclist {
10	10	bugtheory.wiki {Bug Tracking In Fossil}
11	11	branching.wiki {Branching, Forking, Merging, and Tagging}
12	12	build.wiki {Building and Installing Fossil}
13	13	checkin_names.wiki {Checkin And Version Names}
	14	+ changes.wiki {Fossil Changelog}
14	15	copyright-release.html {Contributor License Agreement}
15	16	concepts.wiki {Fossil Core Concepts}
16	17	contribute.wiki {Contributing Code or Documentation To The Fossil Project}
17	18	custom_ticket.wiki {Customizing The Ticket System}
18	19	delta_encoder_algorithm.wiki {Fossil Delta Encoding Algorithm}
		@@ -19,10 +20,11 @@
19	20	delta_format.wiki {Fossil Delta Format}
20	21	embeddeddoc.wiki {Embedded Project Documentation}
21	22	event.wiki {Events}
22	23	faq.wiki {Frequently Asked Questions}
23	24	fileformat.wiki {Fossil File Format}
	25	+ foss-cklist.wiki {Checklist For Successful Open-Source Projects}
24	26	fossil-v-git.wiki {Fossil Versus Git}
25	27	index.wiki {Home Page}
26	28	inout.wiki {Import And Export To And From Git}
27	29	makefile.wiki {The Fossil Build Process}
28	30	password.wiki {Password Management And Authentication}
		@@ -34,13 +36,15 @@
34	36	{Quotes: What People Are Saying About Fossil, Git, and DVCSes in General}
35	37	../test/release-checklist.wiki {Pre-Release Testing Checklist}
36	38	selfcheck.wiki {Fossil Repository Integrity Self Checks}
37	39	selfhost.wiki {Fossil Self Hosting Repositories}
38	40	server.wiki {How To Configure A Fossil Server}
	41	+ settings.wiki {Fossil Settings}
39	42	shunning.wiki {Shunning: Deleting Content From Fossil}
40	43	stats.wiki {Performance Statistics}
41	44	style.wiki {Source Code Style Guidelines}
	45	+ ssl.wiki {Using SSL with Fossil}
42	46	sync.wiki {The Fossil Sync Protocol}
43	47	tech_overview.wiki {A Technical Overview Of The Design And Implementation
44	48	Of Fossil}
45	49	tech_overview.wiki {SQLite Databases Used By Fossil}
46	50	theory1.wiki {Thoughts On The Design Of The Fossil DVCS}
		@@ -47,11 +51,11 @@
47	51	webui.wiki {The Fossil Web Interface}
48	52	wikitheory.wiki {Wiki In Fossil}
49	53	}
50	54
51	55	set permindex {}
52		-set stopwords {fossil and a in of on the to are about used by}
	56	+set stopwords {fossil and a in of on the to are about used by for or}
53	57	foreach {file title} $doclist {
54	58	set n [llength $title]
55	59	lappend permindex [list $title $file]
56	60	for {set i 0} {$i<$n-1} {incr i} {
57	61	set prefix [lrange $title 0 $i]
		@@ -61,13 +65,14 @@
61	65	lappend permindex [list "$suffix $prefix" $file]
62	66	}
63	67	}
64	68	}
65	69	set permindex [lsort -dict $permindex]
66		-puts "<title>Permuted Index Of Fossil Documentation</title>"
67		-puts "<nowiki>"
68		-puts "<ul>"
	70	+set out [open permutedindex.wiki w]
	71	+puts $out "<title>Permuted Index Of Fossil Documentation</title>"
	72	+puts $out "<nowiki>"
	73	+puts $out "<ul>"
69	74	foreach entry $permindex {
70	75	foreach {title file} $entry break
71		- puts "<li><a href=\"$file\">$title</a></li>"
	76	+ puts $out "<li><a href=\"$file\">$title</a></li>"
72	77	}
73		-puts "</ul>"
	78	+puts $out "</ul>"
74	79

	--- www/mkindex.tcl
	+++ www/mkindex.tcl
	@@ -9,10 +9,11 @@
9	set doclist {
10	bugtheory.wiki {Bug Tracking In Fossil}
11	branching.wiki {Branching, Forking, Merging, and Tagging}
12	build.wiki {Building and Installing Fossil}
13	checkin_names.wiki {Checkin And Version Names}

14	copyright-release.html {Contributor License Agreement}
15	concepts.wiki {Fossil Core Concepts}
16	contribute.wiki {Contributing Code or Documentation To The Fossil Project}
17	custom_ticket.wiki {Customizing The Ticket System}
18	delta_encoder_algorithm.wiki {Fossil Delta Encoding Algorithm}
	@@ -19,10 +20,11 @@
19	delta_format.wiki {Fossil Delta Format}
20	embeddeddoc.wiki {Embedded Project Documentation}
21	event.wiki {Events}
22	faq.wiki {Frequently Asked Questions}
23	fileformat.wiki {Fossil File Format}

24	fossil-v-git.wiki {Fossil Versus Git}
25	index.wiki {Home Page}
26	inout.wiki {Import And Export To And From Git}
27	makefile.wiki {The Fossil Build Process}
28	password.wiki {Password Management And Authentication}
	@@ -34,13 +36,15 @@
34	{Quotes: What People Are Saying About Fossil, Git, and DVCSes in General}
35	../test/release-checklist.wiki {Pre-Release Testing Checklist}
36	selfcheck.wiki {Fossil Repository Integrity Self Checks}
37	selfhost.wiki {Fossil Self Hosting Repositories}
38	server.wiki {How To Configure A Fossil Server}

39	shunning.wiki {Shunning: Deleting Content From Fossil}
40	stats.wiki {Performance Statistics}
41	style.wiki {Source Code Style Guidelines}

42	sync.wiki {The Fossil Sync Protocol}
43	tech_overview.wiki {A Technical Overview Of The Design And Implementation
44	Of Fossil}
45	tech_overview.wiki {SQLite Databases Used By Fossil}
46	theory1.wiki {Thoughts On The Design Of The Fossil DVCS}
	@@ -47,11 +51,11 @@
47	webui.wiki {The Fossil Web Interface}
48	wikitheory.wiki {Wiki In Fossil}
49	}
50
51	set permindex {}
52	set stopwords {fossil and a in of on the to are about used by}
53	foreach {file title} $doclist {
54	set n [llength $title]
55	lappend permindex [list $title $file]
56	for {set i 0} {$i<$n-1} {incr i} {
57	set prefix [lrange $title 0 $i]
	@@ -61,13 +65,14 @@
61	lappend permindex [list "$suffix $prefix" $file]
62	}
63	}
64	}
65	set permindex [lsort -dict $permindex]
66	puts "<title>Permuted Index Of Fossil Documentation</title>"
67	puts "<nowiki>"
68	puts "<ul>"

69	foreach entry $permindex {
70	foreach {title file} $entry break
71	puts "<li><a href=\"$file\">$title</a></li>"
72	}
73	puts "</ul>"
74

	--- www/mkindex.tcl
	+++ www/mkindex.tcl
	@@ -9,10 +9,11 @@
9	set doclist {
10	bugtheory.wiki {Bug Tracking In Fossil}
11	branching.wiki {Branching, Forking, Merging, and Tagging}
12	build.wiki {Building and Installing Fossil}
13	checkin_names.wiki {Checkin And Version Names}
14	changes.wiki {Fossil Changelog}
15	copyright-release.html {Contributor License Agreement}
16	concepts.wiki {Fossil Core Concepts}
17	contribute.wiki {Contributing Code or Documentation To The Fossil Project}
18	custom_ticket.wiki {Customizing The Ticket System}
19	delta_encoder_algorithm.wiki {Fossil Delta Encoding Algorithm}
	@@ -19,10 +20,11 @@
20	delta_format.wiki {Fossil Delta Format}
21	embeddeddoc.wiki {Embedded Project Documentation}
22	event.wiki {Events}
23	faq.wiki {Frequently Asked Questions}
24	fileformat.wiki {Fossil File Format}
25	foss-cklist.wiki {Checklist For Successful Open-Source Projects}
26	fossil-v-git.wiki {Fossil Versus Git}
27	index.wiki {Home Page}
28	inout.wiki {Import And Export To And From Git}
29	makefile.wiki {The Fossil Build Process}
30	password.wiki {Password Management And Authentication}
	@@ -34,13 +36,15 @@
36	{Quotes: What People Are Saying About Fossil, Git, and DVCSes in General}
37	../test/release-checklist.wiki {Pre-Release Testing Checklist}
38	selfcheck.wiki {Fossil Repository Integrity Self Checks}
39	selfhost.wiki {Fossil Self Hosting Repositories}
40	server.wiki {How To Configure A Fossil Server}
41	settings.wiki {Fossil Settings}
42	shunning.wiki {Shunning: Deleting Content From Fossil}
43	stats.wiki {Performance Statistics}
44	style.wiki {Source Code Style Guidelines}
45	ssl.wiki {Using SSL with Fossil}
46	sync.wiki {The Fossil Sync Protocol}
47	tech_overview.wiki {A Technical Overview Of The Design And Implementation
48	Of Fossil}
49	tech_overview.wiki {SQLite Databases Used By Fossil}
50	theory1.wiki {Thoughts On The Design Of The Fossil DVCS}
	@@ -47,11 +51,11 @@
51	webui.wiki {The Fossil Web Interface}
52	wikitheory.wiki {Wiki In Fossil}
53	}
54
55	set permindex {}
56	set stopwords {fossil and a in of on the to are about used by for or}
57	foreach {file title} $doclist {
58	set n [llength $title]
59	lappend permindex [list $title $file]
60	for {set i 0} {$i<$n-1} {incr i} {
61	set prefix [lrange $title 0 $i]
	@@ -61,13 +65,14 @@
65	lappend permindex [list "$suffix $prefix" $file]
66	}
67	}
68	}
69	set permindex [lsort -dict $permindex]
70	set out [open permutedindex.wiki w]
71	puts $out "<title>Permuted Index Of Fossil Documentation</title>"
72	puts $out "<nowiki>"
73	puts $out "<ul>"
74	foreach entry $permindex {
75	foreach {title file} $entry break
76	puts $out "<li><a href=\"$file\">$title</a></li>"
77	}
78	puts $out "</ul>"
79

M www/permutedindex.wiki

+11 -1

		--- www/permutedindex.wiki
		+++ www/permutedindex.wiki
		@@ -11,12 +11,14 @@
11	11	<li><a href="private.wiki">Branches Creating, Syncing, and Deleting Private</a></li>
12	12	<li><a href="branching.wiki">Branching, Forking, Merging, and Tagging</a></li>
13	13	<li><a href="bugtheory.wiki">Bug Tracking In Fossil</a></li>
14	14	<li><a href="makefile.wiki">Build Process The Fossil</a></li>
15	15	<li><a href="build.wiki">Building and Installing Fossil</a></li>
	16	+<li><a href="changes.wiki">Changelog Fossil</a></li>
16	17	<li><a href="checkin_names.wiki">Checkin And Version Names</a></li>
17	18	<li><a href="../test/release-checklist.wiki">Checklist Pre-Release Testing</a></li>
	19	+<li><a href="foss-cklist.wiki">Checklist For Successful Open-Source Projects</a></li>
18	20	<li><a href="selfcheck.wiki">Checks Fossil Repository Integrity Self</a></li>
19	21	<li><a href="contribute.wiki">Code or Documentation To The Fossil Project Contributing</a></li>
20	22	<li><a href="style.wiki">Code Style Guidelines Source</a></li>
21	23	<li><a href="concepts.wiki">Concepts Fossil Core</a></li>
22	24	<li><a href="server.wiki">Configure A Fossil Server How To</a></li>
		@@ -43,17 +45,19 @@
43	45	<li><a href="inout.wiki">Export To And From Git Import And</a></li>
44	46	<li><a href="fileformat.wiki">File Format Fossil</a></li>
45	47	<li><a href="branching.wiki">Forking, Merging, and Tagging Branching,</a></li>
46	48	<li><a href="delta_format.wiki">Format Fossil Delta</a></li>
47	49	<li><a href="fileformat.wiki">Format Fossil File</a></li>
	50	+<li><a href="changes.wiki">Fossil Changelog</a></li>
48	51	<li><a href="concepts.wiki">Fossil Core Concepts</a></li>
49	52	<li><a href="delta_encoder_algorithm.wiki">Fossil Delta Encoding Algorithm</a></li>
50	53	<li><a href="delta_format.wiki">Fossil Delta Format</a></li>
51	54	<li><a href="fileformat.wiki">Fossil File Format</a></li>
52	55	<li><a href="quickstart.wiki">Fossil Quick Start Guide</a></li>
53	56	<li><a href="selfcheck.wiki">Fossil Repository Integrity Self Checks</a></li>
54	57	<li><a href="selfhost.wiki">Fossil Self Hosting Repositories</a></li>
	58	+<li><a href="settings.wiki">Fossil Settings</a></li>
55	59	<li><a href="fossil-v-git.wiki">Fossil Versus Git</a></li>
56	60	<li><a href="quotes.wiki">Fossil, Git, and DVCSes in General Quotes: What People Are Saying About</a></li>
57	61	<li><a href="faq.wiki">Frequently Asked Questions</a></li>
58	62	<li><a href="shunning.wiki">From Fossil Shunning: Deleting Content</a></li>
59	63	<li><a href="inout.wiki">From Git Import And Export To And</a></li>
		@@ -73,12 +77,12 @@
73	77	<li><a href="webui.wiki">Interface The Fossil Web</a></li>
74	78	<li><a href="copyright-release.html">License Agreement Contributor</a></li>
75	79	<li><a href="password.wiki">Management And Authentication Password</a></li>
76	80	<li><a href="branching.wiki">Merging, and Tagging Branching, Forking,</a></li>
77	81	<li><a href="checkin_names.wiki">Names Checkin And Version</a></li>
	82	+<li><a href="foss-cklist.wiki">Open-Source Projects Checklist For Successful</a></li>
78	83	<li><a href="pop.wiki">Operations Principles Of</a></li>
79		-<li><a href="contribute.wiki">or Documentation To The Fossil Project Contributing Code</a></li>
80	84	<li><a href="tech_overview.wiki">Overview Of The Design And Implementation Of Fossil A Technical</a></li>
81	85	<li><a href="index.wiki">Page Home</a></li>
82	86	<li><a href="password.wiki">Password Management And Authentication</a></li>
83	87	<li><a href="quotes.wiki">People Are Saying About Fossil, Git, and DVCSes in General Quotes: What</a></li>
84	88	<li><a href="stats.wiki">Performance Statistics</a></li>
		@@ -86,10 +90,11 @@
86	90	<li><a href="pop.wiki">Principles Of Operations</a></li>
87	91	<li><a href="private.wiki">Private Branches Creating, Syncing, and Deleting</a></li>
88	92	<li><a href="makefile.wiki">Process The Fossil Build</a></li>
89	93	<li><a href="contribute.wiki">Project Contributing Code or Documentation To The Fossil</a></li>
90	94	<li><a href="embeddeddoc.wiki">Project Documentation Embedded</a></li>
	95	+<li><a href="foss-cklist.wiki">Projects Checklist For Successful Open-Source</a></li>
91	96	<li><a href="sync.wiki">Protocol The Fossil Sync</a></li>
92	97	<li><a href="faq.wiki">Questions Frequently Asked</a></li>
93	98	<li><a href="qandc.wiki">Questions And Criticisms</a></li>
94	99	<li><a href="quickstart.wiki">Quick Start Guide Fossil</a></li>
95	100	<li><a href="quotes.wiki">Quotes: What People Are Saying About Fossil, Git, and DVCSes in General</a></li>
		@@ -97,16 +102,19 @@
97	102	<li><a href="selfcheck.wiki">Repository Integrity Self Checks Fossil</a></li>
98	103	<li><a href="quotes.wiki">Saying About Fossil, Git, and DVCSes in General Quotes: What People Are</a></li>
99	104	<li><a href="selfcheck.wiki">Self Checks Fossil Repository Integrity</a></li>
100	105	<li><a href="selfhost.wiki">Self Hosting Repositories Fossil</a></li>
101	106	<li><a href="server.wiki">Server How To Configure A Fossil</a></li>
	107	+<li><a href="settings.wiki">Settings Fossil</a></li>
102	108	<li><a href="shunning.wiki">Shunning: Deleting Content From Fossil</a></li>
103	109	<li><a href="style.wiki">Source Code Style Guidelines</a></li>
104	110	<li><a href="tech_overview.wiki">SQLite Databases Used By Fossil</a></li>
	111	+<li><a href="ssl.wiki">SSL with Fossil Using</a></li>
105	112	<li><a href="quickstart.wiki">Start Guide Fossil Quick</a></li>
106	113	<li><a href="stats.wiki">Statistics Performance</a></li>
107	114	<li><a href="style.wiki">Style Guidelines Source Code</a></li>
	115	+<li><a href="foss-cklist.wiki">Successful Open-Source Projects Checklist For</a></li>
108	116	<li><a href="sync.wiki">Sync Protocol The Fossil</a></li>
109	117	<li><a href="private.wiki">Syncing, and Deleting Private Branches Creating,</a></li>
110	118	<li><a href="custom_ticket.wiki">System Customizing The Ticket</a></li>
111	119	<li><a href="branching.wiki">Tagging Branching, Forking, Merging, and</a></li>
112	120	<li><a href="tech_overview.wiki">Technical Overview Of The Design And Implementation Of Fossil A</a></li>
		@@ -115,11 +123,13 @@
115	123	<li><a href="sync.wiki">The Fossil Sync Protocol</a></li>
116	124	<li><a href="webui.wiki">The Fossil Web Interface</a></li>
117	125	<li><a href="theory1.wiki">Thoughts On The Design Of The Fossil DVCS</a></li>
118	126	<li><a href="custom_ticket.wiki">Ticket System Customizing The</a></li>
119	127	<li><a href="bugtheory.wiki">Tracking In Fossil Bug</a></li>
	128	+<li><a href="ssl.wiki">Using SSL with Fossil</a></li>
120	129	<li><a href="checkin_names.wiki">Version Names Checkin And</a></li>
121	130	<li><a href="fossil-v-git.wiki">Versus Git Fossil</a></li>
122	131	<li><a href="webui.wiki">Web Interface The Fossil</a></li>
123	132	<li><a href="quotes.wiki">What People Are Saying About Fossil, Git, and DVCSes in General Quotes:</a></li>
124	133	<li><a href="wikitheory.wiki">Wiki In Fossil</a></li>
	134	+<li><a href="ssl.wiki">with Fossil Using SSL</a></li>
125	135	</ul>
126	136

	--- www/permutedindex.wiki
	+++ www/permutedindex.wiki
	@@ -11,12 +11,14 @@
11	<li><a href="private.wiki">Branches Creating, Syncing, and Deleting Private</a></li>
12	<li><a href="branching.wiki">Branching, Forking, Merging, and Tagging</a></li>
13	<li><a href="bugtheory.wiki">Bug Tracking In Fossil</a></li>
14	<li><a href="makefile.wiki">Build Process The Fossil</a></li>
15	<li><a href="build.wiki">Building and Installing Fossil</a></li>

16	<li><a href="checkin_names.wiki">Checkin And Version Names</a></li>
17	<li><a href="../test/release-checklist.wiki">Checklist Pre-Release Testing</a></li>

18	<li><a href="selfcheck.wiki">Checks Fossil Repository Integrity Self</a></li>
19	<li><a href="contribute.wiki">Code or Documentation To The Fossil Project Contributing</a></li>
20	<li><a href="style.wiki">Code Style Guidelines Source</a></li>
21	<li><a href="concepts.wiki">Concepts Fossil Core</a></li>
22	<li><a href="server.wiki">Configure A Fossil Server How To</a></li>
	@@ -43,17 +45,19 @@
43	<li><a href="inout.wiki">Export To And From Git Import And</a></li>
44	<li><a href="fileformat.wiki">File Format Fossil</a></li>
45	<li><a href="branching.wiki">Forking, Merging, and Tagging Branching,</a></li>
46	<li><a href="delta_format.wiki">Format Fossil Delta</a></li>
47	<li><a href="fileformat.wiki">Format Fossil File</a></li>

48	<li><a href="concepts.wiki">Fossil Core Concepts</a></li>
49	<li><a href="delta_encoder_algorithm.wiki">Fossil Delta Encoding Algorithm</a></li>
50	<li><a href="delta_format.wiki">Fossil Delta Format</a></li>
51	<li><a href="fileformat.wiki">Fossil File Format</a></li>
52	<li><a href="quickstart.wiki">Fossil Quick Start Guide</a></li>
53	<li><a href="selfcheck.wiki">Fossil Repository Integrity Self Checks</a></li>
54	<li><a href="selfhost.wiki">Fossil Self Hosting Repositories</a></li>

55	<li><a href="fossil-v-git.wiki">Fossil Versus Git</a></li>
56	<li><a href="quotes.wiki">Fossil, Git, and DVCSes in General Quotes: What People Are Saying About</a></li>
57	<li><a href="faq.wiki">Frequently Asked Questions</a></li>
58	<li><a href="shunning.wiki">From Fossil Shunning: Deleting Content</a></li>
59	<li><a href="inout.wiki">From Git Import And Export To And</a></li>
	@@ -73,12 +77,12 @@
73	<li><a href="webui.wiki">Interface The Fossil Web</a></li>
74	<li><a href="copyright-release.html">License Agreement Contributor</a></li>
75	<li><a href="password.wiki">Management And Authentication Password</a></li>
76	<li><a href="branching.wiki">Merging, and Tagging Branching, Forking,</a></li>
77	<li><a href="checkin_names.wiki">Names Checkin And Version</a></li>

78	<li><a href="pop.wiki">Operations Principles Of</a></li>
79	<li><a href="contribute.wiki">or Documentation To The Fossil Project Contributing Code</a></li>
80	<li><a href="tech_overview.wiki">Overview Of The Design And Implementation Of Fossil A Technical</a></li>
81	<li><a href="index.wiki">Page Home</a></li>
82	<li><a href="password.wiki">Password Management And Authentication</a></li>
83	<li><a href="quotes.wiki">People Are Saying About Fossil, Git, and DVCSes in General Quotes: What</a></li>
84	<li><a href="stats.wiki">Performance Statistics</a></li>
	@@ -86,10 +90,11 @@
86	<li><a href="pop.wiki">Principles Of Operations</a></li>
87	<li><a href="private.wiki">Private Branches Creating, Syncing, and Deleting</a></li>
88	<li><a href="makefile.wiki">Process The Fossil Build</a></li>
89	<li><a href="contribute.wiki">Project Contributing Code or Documentation To The Fossil</a></li>
90	<li><a href="embeddeddoc.wiki">Project Documentation Embedded</a></li>

91	<li><a href="sync.wiki">Protocol The Fossil Sync</a></li>
92	<li><a href="faq.wiki">Questions Frequently Asked</a></li>
93	<li><a href="qandc.wiki">Questions And Criticisms</a></li>
94	<li><a href="quickstart.wiki">Quick Start Guide Fossil</a></li>
95	<li><a href="quotes.wiki">Quotes: What People Are Saying About Fossil, Git, and DVCSes in General</a></li>
	@@ -97,16 +102,19 @@
97	<li><a href="selfcheck.wiki">Repository Integrity Self Checks Fossil</a></li>
98	<li><a href="quotes.wiki">Saying About Fossil, Git, and DVCSes in General Quotes: What People Are</a></li>
99	<li><a href="selfcheck.wiki">Self Checks Fossil Repository Integrity</a></li>
100	<li><a href="selfhost.wiki">Self Hosting Repositories Fossil</a></li>
101	<li><a href="server.wiki">Server How To Configure A Fossil</a></li>

102	<li><a href="shunning.wiki">Shunning: Deleting Content From Fossil</a></li>
103	<li><a href="style.wiki">Source Code Style Guidelines</a></li>
104	<li><a href="tech_overview.wiki">SQLite Databases Used By Fossil</a></li>

105	<li><a href="quickstart.wiki">Start Guide Fossil Quick</a></li>
106	<li><a href="stats.wiki">Statistics Performance</a></li>
107	<li><a href="style.wiki">Style Guidelines Source Code</a></li>

108	<li><a href="sync.wiki">Sync Protocol The Fossil</a></li>
109	<li><a href="private.wiki">Syncing, and Deleting Private Branches Creating,</a></li>
110	<li><a href="custom_ticket.wiki">System Customizing The Ticket</a></li>
111	<li><a href="branching.wiki">Tagging Branching, Forking, Merging, and</a></li>
112	<li><a href="tech_overview.wiki">Technical Overview Of The Design And Implementation Of Fossil A</a></li>
	@@ -115,11 +123,13 @@
115	<li><a href="sync.wiki">The Fossil Sync Protocol</a></li>
116	<li><a href="webui.wiki">The Fossil Web Interface</a></li>
117	<li><a href="theory1.wiki">Thoughts On The Design Of The Fossil DVCS</a></li>
118	<li><a href="custom_ticket.wiki">Ticket System Customizing The</a></li>
119	<li><a href="bugtheory.wiki">Tracking In Fossil Bug</a></li>

120	<li><a href="checkin_names.wiki">Version Names Checkin And</a></li>
121	<li><a href="fossil-v-git.wiki">Versus Git Fossil</a></li>
122	<li><a href="webui.wiki">Web Interface The Fossil</a></li>
123	<li><a href="quotes.wiki">What People Are Saying About Fossil, Git, and DVCSes in General Quotes:</a></li>
124	<li><a href="wikitheory.wiki">Wiki In Fossil</a></li>

125	</ul>
126

	--- www/permutedindex.wiki
	+++ www/permutedindex.wiki
	@@ -11,12 +11,14 @@
11	<li><a href="private.wiki">Branches Creating, Syncing, and Deleting Private</a></li>
12	<li><a href="branching.wiki">Branching, Forking, Merging, and Tagging</a></li>
13	<li><a href="bugtheory.wiki">Bug Tracking In Fossil</a></li>
14	<li><a href="makefile.wiki">Build Process The Fossil</a></li>
15	<li><a href="build.wiki">Building and Installing Fossil</a></li>
16	<li><a href="changes.wiki">Changelog Fossil</a></li>
17	<li><a href="checkin_names.wiki">Checkin And Version Names</a></li>
18	<li><a href="../test/release-checklist.wiki">Checklist Pre-Release Testing</a></li>
19	<li><a href="foss-cklist.wiki">Checklist For Successful Open-Source Projects</a></li>
20	<li><a href="selfcheck.wiki">Checks Fossil Repository Integrity Self</a></li>
21	<li><a href="contribute.wiki">Code or Documentation To The Fossil Project Contributing</a></li>
22	<li><a href="style.wiki">Code Style Guidelines Source</a></li>
23	<li><a href="concepts.wiki">Concepts Fossil Core</a></li>
24	<li><a href="server.wiki">Configure A Fossil Server How To</a></li>
	@@ -43,17 +45,19 @@
45	<li><a href="inout.wiki">Export To And From Git Import And</a></li>
46	<li><a href="fileformat.wiki">File Format Fossil</a></li>
47	<li><a href="branching.wiki">Forking, Merging, and Tagging Branching,</a></li>
48	<li><a href="delta_format.wiki">Format Fossil Delta</a></li>
49	<li><a href="fileformat.wiki">Format Fossil File</a></li>
50	<li><a href="changes.wiki">Fossil Changelog</a></li>
51	<li><a href="concepts.wiki">Fossil Core Concepts</a></li>
52	<li><a href="delta_encoder_algorithm.wiki">Fossil Delta Encoding Algorithm</a></li>
53	<li><a href="delta_format.wiki">Fossil Delta Format</a></li>
54	<li><a href="fileformat.wiki">Fossil File Format</a></li>
55	<li><a href="quickstart.wiki">Fossil Quick Start Guide</a></li>
56	<li><a href="selfcheck.wiki">Fossil Repository Integrity Self Checks</a></li>
57	<li><a href="selfhost.wiki">Fossil Self Hosting Repositories</a></li>
58	<li><a href="settings.wiki">Fossil Settings</a></li>
59	<li><a href="fossil-v-git.wiki">Fossil Versus Git</a></li>
60	<li><a href="quotes.wiki">Fossil, Git, and DVCSes in General Quotes: What People Are Saying About</a></li>
61	<li><a href="faq.wiki">Frequently Asked Questions</a></li>
62	<li><a href="shunning.wiki">From Fossil Shunning: Deleting Content</a></li>
63	<li><a href="inout.wiki">From Git Import And Export To And</a></li>
	@@ -73,12 +77,12 @@
77	<li><a href="webui.wiki">Interface The Fossil Web</a></li>
78	<li><a href="copyright-release.html">License Agreement Contributor</a></li>
79	<li><a href="password.wiki">Management And Authentication Password</a></li>
80	<li><a href="branching.wiki">Merging, and Tagging Branching, Forking,</a></li>
81	<li><a href="checkin_names.wiki">Names Checkin And Version</a></li>
82	<li><a href="foss-cklist.wiki">Open-Source Projects Checklist For Successful</a></li>
83	<li><a href="pop.wiki">Operations Principles Of</a></li>

84	<li><a href="tech_overview.wiki">Overview Of The Design And Implementation Of Fossil A Technical</a></li>
85	<li><a href="index.wiki">Page Home</a></li>
86	<li><a href="password.wiki">Password Management And Authentication</a></li>
87	<li><a href="quotes.wiki">People Are Saying About Fossil, Git, and DVCSes in General Quotes: What</a></li>
88	<li><a href="stats.wiki">Performance Statistics</a></li>
	@@ -86,10 +90,11 @@
90	<li><a href="pop.wiki">Principles Of Operations</a></li>
91	<li><a href="private.wiki">Private Branches Creating, Syncing, and Deleting</a></li>
92	<li><a href="makefile.wiki">Process The Fossil Build</a></li>
93	<li><a href="contribute.wiki">Project Contributing Code or Documentation To The Fossil</a></li>
94	<li><a href="embeddeddoc.wiki">Project Documentation Embedded</a></li>
95	<li><a href="foss-cklist.wiki">Projects Checklist For Successful Open-Source</a></li>
96	<li><a href="sync.wiki">Protocol The Fossil Sync</a></li>
97	<li><a href="faq.wiki">Questions Frequently Asked</a></li>
98	<li><a href="qandc.wiki">Questions And Criticisms</a></li>
99	<li><a href="quickstart.wiki">Quick Start Guide Fossil</a></li>
100	<li><a href="quotes.wiki">Quotes: What People Are Saying About Fossil, Git, and DVCSes in General</a></li>
	@@ -97,16 +102,19 @@
102	<li><a href="selfcheck.wiki">Repository Integrity Self Checks Fossil</a></li>
103	<li><a href="quotes.wiki">Saying About Fossil, Git, and DVCSes in General Quotes: What People Are</a></li>
104	<li><a href="selfcheck.wiki">Self Checks Fossil Repository Integrity</a></li>
105	<li><a href="selfhost.wiki">Self Hosting Repositories Fossil</a></li>
106	<li><a href="server.wiki">Server How To Configure A Fossil</a></li>
107	<li><a href="settings.wiki">Settings Fossil</a></li>
108	<li><a href="shunning.wiki">Shunning: Deleting Content From Fossil</a></li>
109	<li><a href="style.wiki">Source Code Style Guidelines</a></li>
110	<li><a href="tech_overview.wiki">SQLite Databases Used By Fossil</a></li>
111	<li><a href="ssl.wiki">SSL with Fossil Using</a></li>
112	<li><a href="quickstart.wiki">Start Guide Fossil Quick</a></li>
113	<li><a href="stats.wiki">Statistics Performance</a></li>
114	<li><a href="style.wiki">Style Guidelines Source Code</a></li>
115	<li><a href="foss-cklist.wiki">Successful Open-Source Projects Checklist For</a></li>
116	<li><a href="sync.wiki">Sync Protocol The Fossil</a></li>
117	<li><a href="private.wiki">Syncing, and Deleting Private Branches Creating,</a></li>
118	<li><a href="custom_ticket.wiki">System Customizing The Ticket</a></li>
119	<li><a href="branching.wiki">Tagging Branching, Forking, Merging, and</a></li>
120	<li><a href="tech_overview.wiki">Technical Overview Of The Design And Implementation Of Fossil A</a></li>
	@@ -115,11 +123,13 @@
123	<li><a href="sync.wiki">The Fossil Sync Protocol</a></li>
124	<li><a href="webui.wiki">The Fossil Web Interface</a></li>
125	<li><a href="theory1.wiki">Thoughts On The Design Of The Fossil DVCS</a></li>
126	<li><a href="custom_ticket.wiki">Ticket System Customizing The</a></li>
127	<li><a href="bugtheory.wiki">Tracking In Fossil Bug</a></li>
128	<li><a href="ssl.wiki">Using SSL with Fossil</a></li>
129	<li><a href="checkin_names.wiki">Version Names Checkin And</a></li>
130	<li><a href="fossil-v-git.wiki">Versus Git Fossil</a></li>
131	<li><a href="webui.wiki">Web Interface The Fossil</a></li>
132	<li><a href="quotes.wiki">What People Are Saying About Fossil, Git, and DVCSes in General Quotes:</a></li>
133	<li><a href="wikitheory.wiki">Wiki In Fossil</a></li>
134	<li><a href="ssl.wiki">with Fossil Using SSL</a></li>
135	</ul>
136

M www/quickstart.wiki

+15 -7

		--- www/quickstart.wiki
		+++ www/quickstart.wiki
		@@ -9,11 +9,11 @@
9	9	<p>Fossil is a single self-contained C program. You need to
10	10	either download a
11	11	<a href="http://www.fossil-scm.org/download.html">precompiled binary</a>
12	12	or <a href="build.wiki">build it yourself</a> from sources.
13	13	Install fossil by putting the fossil binary
14		- someplace on your PATH environment variable.</p>
	14	+ someplace on your $PATH.</p>
15	15
16	16	<a name="fslclone"></a>
17	17	<h2>General Work Flow</h2>
18	18
19	19	<p>Fossil works with repository files (a database with the project's
		@@ -20,29 +20,27 @@
20	20	complete history) and with checked-out local trees (the working directory
21	21	you use to do your work).
22	22	The workflow looks like this:</p>
23	23
24	24	<ul>
25		- <li>Create or clone a repository file. ([/help/new\|fossil new] or
	25	+ <li>Create or clone a repository file. ([/help/init\|fossil init] or
26	26	[/help/clone \| fossil clone])
27	27	<li>Check out a local tree. ([/help/open \| fossil open])
28	28	<li>Perform operations on the repository (including repository
29	29	configuration).
30		- <li><em>Optionally</em> close the local tree.
31		- ([/help/close \| fossil close], but this is rarely used.)
32	30	</ul>
33	31
34	32	<p>The following sections will give you a brief overview of these
35	33	operations.</p>
36	34
37	35	<h2>Starting A New Project</h2>
38	36
39	37	<p>To start a new project with fossil, create a new empty repository
40		- this way: ([/help/new \| more info]) </p>
	38	+ this way: ([/help/init \| more info]) </p>
41	39
42	40	<blockquote>
43		- <b>fossil new </b><i> repository-filename</i>
	41	+ <b>fossil init </b><i> repository-filename</i>
44	42	</blockquote>
45	43
46	44	<h2>Cloning An Existing Repository</h2>
47	45
48	46	<p>Most fossil operations interact with a repository that is on the
		@@ -73,10 +71,19 @@
73	71	You can name your repositories anything you want. The ".fossil" suffix
74	72	is not required.</p>
75	73
76	74	<p>Note: If you are behind a restrictive firewall, you might need
77	75	to <a href="#proxy">specify an HTTP proxy</a> to use.</p>
	76	+
	77	+ <p>A Fossil repository is a single disk file. Instead of cloning,
	78	+ you can just make a copy of the repository file (for example, using
	79	+ "scp"). Note, however, that the repository file contains auxiliary
	80	+ information above and beyond the versioned files, including some
	81	+ sensitive information such as passwords and email addresses. If you
	82	+ want to share Fossil repositories directly, consider running the
	83	+ [/help/scrub\|fossil scrub] command to remove sensitive information
	84	+ before transmitting the file.
78	85
79	86	<h2>Importing From Another Version Control System</h2>
80	87
81	88	<p>Rather than start a new project, or clone an existing Fossil project,
82	89	you might prefer to
		@@ -147,11 +154,12 @@
147	154	<p>To add new files to your project, or remove old files, use these
148	155	commands:</p>
149	156
150	157	<blockquote>
151	158	<b>[/help/add \| fossil add]</b> <i>file...</i><br>
152		- <b>[/help/rm \| fossil rm]</b> <i>file...</i>
	159	+ <b>[/help/rm \| fossil rm]</b> <i>file...</i><br>
	160	+ <b>[/help/addremove \| fossil addremove]</b> <i>file...</i><br>
153	161	</blockquote>
154	162
155	163	<p>You can also edit files freely. Once you are ready to commit
156	164	your changes, type:</p>
157	165
158	166

	--- www/quickstart.wiki
	+++ www/quickstart.wiki
	@@ -9,11 +9,11 @@
9	<p>Fossil is a single self-contained C program. You need to
10	either download a
11	<a href="http://www.fossil-scm.org/download.html">precompiled binary</a>
12	or <a href="build.wiki">build it yourself</a> from sources.
13	Install fossil by putting the fossil binary
14	someplace on your PATH environment variable.</p>
15
16	<a name="fslclone"></a>
17	<h2>General Work Flow</h2>
18
19	<p>Fossil works with repository files (a database with the project's
	@@ -20,29 +20,27 @@
20	complete history) and with checked-out local trees (the working directory
21	you use to do your work).
22	The workflow looks like this:</p>
23
24	<ul>
25	<li>Create or clone a repository file. ([/help/new\|fossil new] or
26	[/help/clone \| fossil clone])
27	<li>Check out a local tree. ([/help/open \| fossil open])
28	<li>Perform operations on the repository (including repository
29	configuration).
30	<li><em>Optionally</em> close the local tree.
31	([/help/close \| fossil close], but this is rarely used.)
32	</ul>
33
34	<p>The following sections will give you a brief overview of these
35	operations.</p>
36
37	<h2>Starting A New Project</h2>
38
39	<p>To start a new project with fossil, create a new empty repository
40	this way: ([/help/new \| more info]) </p>
41
42	<blockquote>
43	<b>fossil new </b><i> repository-filename</i>
44	</blockquote>
45
46	<h2>Cloning An Existing Repository</h2>
47
48	<p>Most fossil operations interact with a repository that is on the
	@@ -73,10 +71,19 @@
73	You can name your repositories anything you want. The ".fossil" suffix
74	is not required.</p>
75
76	<p>Note: If you are behind a restrictive firewall, you might need
77	to <a href="#proxy">specify an HTTP proxy</a> to use.</p>









78
79	<h2>Importing From Another Version Control System</h2>
80
81	<p>Rather than start a new project, or clone an existing Fossil project,
82	you might prefer to
	@@ -147,11 +154,12 @@
147	<p>To add new files to your project, or remove old files, use these
148	commands:</p>
149
150	<blockquote>
151	<b>[/help/add \| fossil add]</b> <i>file...</i><br>
152	<b>[/help/rm \| fossil rm]</b> <i>file...</i>

153	</blockquote>
154
155	<p>You can also edit files freely. Once you are ready to commit
156	your changes, type:</p>
157
158

	--- www/quickstart.wiki
	+++ www/quickstart.wiki
	@@ -9,11 +9,11 @@
9	<p>Fossil is a single self-contained C program. You need to
10	either download a
11	<a href="http://www.fossil-scm.org/download.html">precompiled binary</a>
12	or <a href="build.wiki">build it yourself</a> from sources.
13	Install fossil by putting the fossil binary
14	someplace on your $PATH.</p>
15
16	<a name="fslclone"></a>
17	<h2>General Work Flow</h2>
18
19	<p>Fossil works with repository files (a database with the project's
	@@ -20,29 +20,27 @@
20	complete history) and with checked-out local trees (the working directory
21	you use to do your work).
22	The workflow looks like this:</p>
23
24	<ul>
25	<li>Create or clone a repository file. ([/help/init\|fossil init] or
26	[/help/clone \| fossil clone])
27	<li>Check out a local tree. ([/help/open \| fossil open])
28	<li>Perform operations on the repository (including repository
29	configuration).


30	</ul>
31
32	<p>The following sections will give you a brief overview of these
33	operations.</p>
34
35	<h2>Starting A New Project</h2>
36
37	<p>To start a new project with fossil, create a new empty repository
38	this way: ([/help/init \| more info]) </p>
39
40	<blockquote>
41	<b>fossil init </b><i> repository-filename</i>
42	</blockquote>
43
44	<h2>Cloning An Existing Repository</h2>
45
46	<p>Most fossil operations interact with a repository that is on the
	@@ -73,10 +71,19 @@
71	You can name your repositories anything you want. The ".fossil" suffix
72	is not required.</p>
73
74	<p>Note: If you are behind a restrictive firewall, you might need
75	to <a href="#proxy">specify an HTTP proxy</a> to use.</p>
76
77	<p>A Fossil repository is a single disk file. Instead of cloning,
78	you can just make a copy of the repository file (for example, using
79	"scp"). Note, however, that the repository file contains auxiliary
80	information above and beyond the versioned files, including some
81	sensitive information such as passwords and email addresses. If you
82	want to share Fossil repositories directly, consider running the
83	[/help/scrub\|fossil scrub] command to remove sensitive information
84	before transmitting the file.
85
86	<h2>Importing From Another Version Control System</h2>
87
88	<p>Rather than start a new project, or clone an existing Fossil project,
89	you might prefer to
	@@ -147,11 +154,12 @@
154	<p>To add new files to your project, or remove old files, use these
155	commands:</p>
156
157	<blockquote>
158	<b>[/help/add \| fossil add]</b> <i>file...</i><br>
159	<b>[/help/rm \| fossil rm]</b> <i>file...</i><br>
160	<b>[/help/addremove \| fossil addremove]</b> <i>file...</i><br>
161	</blockquote>
162
163	<p>You can also edit files freely. Once you are ready to commit
164	your changes, type:</p>
165
166

M www/server.wiki

		--- www/server.wiki
		+++ www/server.wiki
		@@ -97,10 +97,13 @@
97	97	If you are using "inetd" to serve your repository, then you simply need to add "/usr/bin/stunnel" (perhaps on a different path, depending on your setup) before the command line to launch Fossil.
98	98	</p>
99	99	<p>
100	100	At this stage, the standalone server (e.g. "fossil server") does not support SSL.
101	101	</p>
	102	+<p>
	103	+For more information, see <a href="./ssl.wiki">Using SSL with Fossil</a>.
	104	+</p>
102	105	</blockquote>
103	106
104	107	<h2>Various security concerns with hosted repositories</h2><blockquote>
105	108	<p>
106	109	There are two main concerns relating to usage of Fossil for sharing sensitive information (source or any other data):
107	110
108	111	ADDED www/settings.wiki
109	112	ADDED www/ssl.wiki

	--- www/server.wiki
	+++ www/server.wiki
	@@ -97,10 +97,13 @@
97	If you are using "inetd" to serve your repository, then you simply need to add "/usr/bin/stunnel" (perhaps on a different path, depending on your setup) before the command line to launch Fossil.
98	</p>
99	<p>
100	At this stage, the standalone server (e.g. "fossil server") does not support SSL.
101	</p>



102	</blockquote>
103
104	<h2>Various security concerns with hosted repositories</h2><blockquote>
105	<p>
106	There are two main concerns relating to usage of Fossil for sharing sensitive information (source or any other data):
107
108	DDED www/settings.wiki
109	DDED www/ssl.wiki

	--- www/server.wiki
	+++ www/server.wiki
	@@ -97,10 +97,13 @@
97	If you are using "inetd" to serve your repository, then you simply need to add "/usr/bin/stunnel" (perhaps on a different path, depending on your setup) before the command line to launch Fossil.
98	</p>
99	<p>
100	At this stage, the standalone server (e.g. "fossil server") does not support SSL.
101	</p>
102	<p>
103	For more information, see <a href="./ssl.wiki">Using SSL with Fossil</a>.
104	</p>
105	</blockquote>
106
107	<h2>Various security concerns with hosted repositories</h2><blockquote>
108	<p>
109	There are two main concerns relating to usage of Fossil for sharing sensitive information (source or any other data):
110
111	DDED www/settings.wiki
112	DDED www/ssl.wiki

M www/settings.wiki

		--- a/www/settings.wiki
		+++ b/www/settings.wiki
		@@ -0,0 +1,6 @@
	1	+ command.
	2	+ a repository on the web. b</tt>,
	3	+ � �
	4	+ddirectory < <title>Fossil S nl-glob</tt>,
	5	+ � �
	6	+ddirectory t ignorealso

	--- a/www/settings.wiki
	+++ b/www/settings.wiki
	@@ -0,0 +1,6 @@

	--- a/www/settings.wiki
	+++ b/www/settings.wiki
	@@ -0,0 +1,6 @@
1	command.
2	a repository on the web. b</tt>,
3	� �
4	ddirectory < <title>Fossil S nl-glob</tt>,
5	� �
6	ddirectory t ignorealso

M www/ssl.wiki

+22

		--- a/www/ssl.wiki
		+++ b/www/ssl.wiki
		@@ -0,0 +1,22 @@
	1	+<title>SSL and Fossiluring a Repository wSSLyour repositge." This setting is not . This will protectessed via HTTP,
	2	+sas well preventing eavesdropping of the contents of your repository.
	3	+
	4	+To host a repository with SSL, you need to use an web server which supports SSL in front of the Fossil server. You can host it using the CGI option or by proxying Fossil's built in HTTP server.
	5	+
	6	+Your fossil client must system Make sure scheme to ensure you're using SSL. from http as w important S front end proxy up again for access, which
	7	+causes Fossil to see that it's being accessed via HTTP, so it redirects
	8	+the client to detects
	9	+
	10	+
	11	+If you wishto a Fossil web server, iayer, not by use of Fossil-
	12	+shows one way to`ssl`SSL uses certificates.certificates you trustyou wishto a Fossil web server, iayer, not by use of Fossil-
	13	+shows one way e of Fossil-
	14	+shows one way to`ssl` using a certificate signed by a certificate authority, you need to specifyil-
	15	+shows one was you trust with the <tt>ssl-ca-location</tt> setting. Set this globally with the <tt>-global</tt> option for convenience.
	16	+
	17	+This should be set to the location of a file containing all the PEM encoded certificates you trust. You can obtain a certificate using a web browser, for example, Firefox, or just refer to your system's trusted CA roots which are usually stored somewhere in <tt>/etc</tt>.
	18	+
	19	+
	20	+<h2>C nications with Fossil</h2>
	21	+
	22	+If you are storing sensitive information T <title title>TLS and Fossiluring a Repository with TLS</title>

	--- a/www/ssl.wiki
	+++ b/www/ssl.wiki
	@@ -0,0 +1,22 @@

	--- a/www/ssl.wiki
	+++ b/www/ssl.wiki
	@@ -0,0 +1,22 @@
1	<title>SSL and Fossiluring a Repository wSSLyour repositge." This setting is not . This will protectessed via HTTP,
2	sas well preventing eavesdropping of the contents of your repository.
3
4	To host a repository with SSL, you need to use an web server which supports SSL in front of the Fossil server. You can host it using the CGI option or by proxying Fossil's built in HTTP server.
5
6	Your fossil client must system Make sure scheme to ensure you're using SSL. from http as w important S front end proxy up again for access, which
7	causes Fossil to see that it's being accessed via HTTP, so it redirects
8	the client to detects
9
10
11	If you wishto a Fossil web server, iayer, not by use of Fossil-
12	shows one way to`ssl`SSL uses certificates.certificates you trustyou wishto a Fossil web server, iayer, not by use of Fossil-
13	shows one way e of Fossil-
14	shows one way to`ssl` using a certificate signed by a certificate authority, you need to specifyil-
15	shows one was you trust with the <tt>ssl-ca-location</tt> setting. Set this globally with the <tt>-global</tt> option for convenience.
16
17	This should be set to the location of a file containing all the PEM encoded certificates you trust. You can obtain a certificate using a web browser, for example, Firefox, or just refer to your system's trusted CA roots which are usually stored somewhere in <tt>/etc</tt>.
18
19
20	<h2>C nications with Fossil</h2>
21
22	If you are storing sensitive information T <title title>TLS and Fossiluring a Repository with TLS</title>

Fossil SCM

Keyboard Shortcuts