Fossil SCM

Merge trunk

andygoth 2016-05-23 15:37 UTC reparent merge

Commit 1d2e740798a43ce498a952ae3f4a1ba5b6d90da0

Parent 6b4565002b827e3…

7 files changed +36 -26 +2 -2 -3 +6 -7 +226 -82 +1 -1 +20

~ src/checkin.c ~ src/db.c ~ src/export.c ~ src/foci.c ~ src/sqlite3.c ~ src/sqlite3.h ~ www/encryptedrepos.wiki

M src/checkin.c

+36 -26

		--- src/checkin.c
		+++ src/checkin.c
		@@ -549,20 +549,20 @@
549	549	/*
550	550	** COMMAND: extras
551	551	**
552	552	** Usage: %fossil extras ?OPTIONS? ?PATH1 ...?
553	553	**
554		-** Print a list of all files in the source tree that are not part of
555		-** the current checkout. See also the "clean" command. If paths are
556		-** specified, only files in the given directories will be listed.
	554	+** Print a list of all files in the source tree that are not part of the
	555	+** current checkout. See also the "clean" command. If paths are specified,
	556	+** only files in the given directories will be listed.
557	557	**
558	558	** Files and subdirectories whose names begin with "." are normally
559	559	** ignored but can be included by adding the --dotfiles option.
560	560	**
561		-** The GLOBPATTERN is a comma-separated list of GLOB expressions for
562		-** files that are ignored. The GLOBPATTERN specified by the "ignore-glob"
563		-** is used if the --ignore option is omitted.
	561	+** Files whose names match any of the glob patterns in the "ignore-glob"
	562	+** setting are ignored. This setting can be overridden by the --ignore
	563	+** option, whose CSG argument is a comma-separated list of glob patterns.
564	564	**
565	565	** Pathnames are displayed according to the "relative-paths" setting,
566	566	** unless overridden by the --abs-paths or --rel-paths options.
567	567	**
568	568	** Options:
		@@ -635,29 +635,38 @@
635	635	/*
636	636	** COMMAND: clean
637	637	**
638	638	** Usage: %fossil clean ?OPTIONS? ?PATH ...?
639	639	**
640		-** Delete all "extra" files in the source tree. "Extra" files are
641		-** files that are not officially part of the checkout. This operation
642		-** cannot be undone. If one or more PATH arguments appear, then only
643		-** the files named, or files contained with directories named, will be
644		-** removed.
645		-**
646		-** Prompted are issued to confirm the removal of each file, unless
647		-** the --force flag is used or unless the file matches glob pattern
648		-** specified by the --clean option. No file that matches glob patterns
649		-** specified by --ignore or --keep will ever be deleted. The default
650		-** values for --clean, --ignore, and --keep are determined by the
651		-** (versionable) clean-glob, ignore-glob, and keep-glob settings.
652		-** Files and subdirectories whose names begin with "." are automatically
653		-** ignored unless the --dotfiles option is used.
654		-**
655		-** The --verily option ignores the keep-glob and ignore-glob settings
656		-** and turns on --force, --dotfiles, and --emptydirs. Use the --verily
657		-** option when you really want to clean up everything. Extreme care
658		-** should be exercised when using the --verily option.
	640	+** Delete all "extra" files in the source tree. "Extra" files are files
	641	+** that are not officially part of the checkout. If one or more PATH
	642	+** arguments appear, then only the files named, or files contained with
	643	+** directories named, will be removed.
	644	+**
	645	+** If the --prompt option is used, prompts are issued to confirm the
	646	+** permanent removal of each file. Otherwise, files are backed up to the
	647	+** undo buffer prior to removal, and prompts are issued only for files
	648	+** whose removal cannot be undone due to their large size or due to
	649	+** --disable-undo being used.
	650	+**
	651	+** The --force option treats all prompts as having been answered yes,
	652	+** whereas --no-prompt treats them as having been answered no.
	653	+**
	654	+** Files matching any glob pattern specified by the --clean option are
	655	+** deleted without prompting, and the removal cannot be undone.
	656	+**
	657	+** No file that matches glob patterns specified by --ignore or --keep will
	658	+** ever be deleted. Files and subdirectories whose names begin with "."
	659	+** are automatically ignored unless the --dotfiles option is used.
	660	+**
	661	+** The default values for --clean, --ignore, and --keep are determined by
	662	+** the (versionable) clean-glob, ignore-glob, and keep-glob settings.
	663	+**
	664	+** The --verily option ignores the keep-glob and ignore-glob settings and
	665	+** turns on --force, --emptydirs, --dotfiles, and --disable-undo. Use the
	666	+** --verily option when you really want to clean up everything. Extreme
	667	+** care should be exercised when using the --verily option.
659	668	**
660	669	** Options:
661	670	** --allckouts Check for empty directories within any checkouts
662	671	** that may be nested within the current one. This
663	672	** option should be used with great care because the
		@@ -678,11 +687,12 @@
678	687	** argument. Matching files, if any, are removed
679	688	** prior to checking for any empty directories;
680	689	** therefore, directories that contain only files
681	690	** that were removed will be removed as well.
682	691	** -f\|--force Remove files without prompting.
683		-** -i\|--prompt Prompt before removing each file.
	692	+** -i\|--prompt Prompt before removing each file. This option
	693	+** implies the --disable-undo option.
684	694	** -x\|--verily WARNING: Removes everything that is not a managed
685	695	** file or the repository itself. This option
686	696	** implies the --force, --emptydirs, --dotfiles, and
687	697	** --disable-undo options. Furthermore, it completely
688	698	** disregards the keep-glob and ignore-glob settings.
689	699

	--- src/checkin.c
	+++ src/checkin.c
	@@ -549,20 +549,20 @@
549	/*
550	** COMMAND: extras
551	**
552	** Usage: %fossil extras ?OPTIONS? ?PATH1 ...?
553	**
554	** Print a list of all files in the source tree that are not part of
555	** the current checkout. See also the "clean" command. If paths are
556	** specified, only files in the given directories will be listed.
557	**
558	** Files and subdirectories whose names begin with "." are normally
559	** ignored but can be included by adding the --dotfiles option.
560	**
561	** The GLOBPATTERN is a comma-separated list of GLOB expressions for
562	** files that are ignored. The GLOBPATTERN specified by the "ignore-glob"
563	** is used if the --ignore option is omitted.
564	**
565	** Pathnames are displayed according to the "relative-paths" setting,
566	** unless overridden by the --abs-paths or --rel-paths options.
567	**
568	** Options:
	@@ -635,29 +635,38 @@
635	/*
636	** COMMAND: clean
637	**
638	** Usage: %fossil clean ?OPTIONS? ?PATH ...?
639	**
640	** Delete all "extra" files in the source tree. "Extra" files are
641	** files that are not officially part of the checkout. This operation
642	** cannot be undone. If one or more PATH arguments appear, then only
643	** the files named, or files contained with directories named, will be
644	** removed.
645	**
646	** Prompted are issued to confirm the removal of each file, unless
647	** the --force flag is used or unless the file matches glob pattern
648	** specified by the --clean option. No file that matches glob patterns
649	** specified by --ignore or --keep will ever be deleted. The default
650	** values for --clean, --ignore, and --keep are determined by the
651	** (versionable) clean-glob, ignore-glob, and keep-glob settings.
652	** Files and subdirectories whose names begin with "." are automatically
653	** ignored unless the --dotfiles option is used.
654	**
655	** The --verily option ignores the keep-glob and ignore-glob settings
656	** and turns on --force, --dotfiles, and --emptydirs. Use the --verily
657	** option when you really want to clean up everything. Extreme care
658	** should be exercised when using the --verily option.









659	**
660	** Options:
661	** --allckouts Check for empty directories within any checkouts
662	** that may be nested within the current one. This
663	** option should be used with great care because the
	@@ -678,11 +687,12 @@
678	** argument. Matching files, if any, are removed
679	** prior to checking for any empty directories;
680	** therefore, directories that contain only files
681	** that were removed will be removed as well.
682	** -f\|--force Remove files without prompting.
683	** -i\|--prompt Prompt before removing each file.

684	** -x\|--verily WARNING: Removes everything that is not a managed
685	** file or the repository itself. This option
686	** implies the --force, --emptydirs, --dotfiles, and
687	** --disable-undo options. Furthermore, it completely
688	** disregards the keep-glob and ignore-glob settings.
689

	--- src/checkin.c
	+++ src/checkin.c
	@@ -549,20 +549,20 @@
549	/*
550	** COMMAND: extras
551	**
552	** Usage: %fossil extras ?OPTIONS? ?PATH1 ...?
553	**
554	** Print a list of all files in the source tree that are not part of the
555	** current checkout. See also the "clean" command. If paths are specified,
556	** only files in the given directories will be listed.
557	**
558	** Files and subdirectories whose names begin with "." are normally
559	** ignored but can be included by adding the --dotfiles option.
560	**
561	** Files whose names match any of the glob patterns in the "ignore-glob"
562	** setting are ignored. This setting can be overridden by the --ignore
563	** option, whose CSG argument is a comma-separated list of glob patterns.
564	**
565	** Pathnames are displayed according to the "relative-paths" setting,
566	** unless overridden by the --abs-paths or --rel-paths options.
567	**
568	** Options:
	@@ -635,29 +635,38 @@
635	/*
636	** COMMAND: clean
637	**
638	** Usage: %fossil clean ?OPTIONS? ?PATH ...?
639	**
640	** Delete all "extra" files in the source tree. "Extra" files are files
641	** that are not officially part of the checkout. If one or more PATH
642	** arguments appear, then only the files named, or files contained with
643	** directories named, will be removed.
644	**
645	** If the --prompt option is used, prompts are issued to confirm the
646	** permanent removal of each file. Otherwise, files are backed up to the
647	** undo buffer prior to removal, and prompts are issued only for files
648	** whose removal cannot be undone due to their large size or due to
649	** --disable-undo being used.
650	**
651	** The --force option treats all prompts as having been answered yes,
652	** whereas --no-prompt treats them as having been answered no.
653	**
654	** Files matching any glob pattern specified by the --clean option are
655	** deleted without prompting, and the removal cannot be undone.
656	**
657	** No file that matches glob patterns specified by --ignore or --keep will
658	** ever be deleted. Files and subdirectories whose names begin with "."
659	** are automatically ignored unless the --dotfiles option is used.
660	**
661	** The default values for --clean, --ignore, and --keep are determined by
662	** the (versionable) clean-glob, ignore-glob, and keep-glob settings.
663	**
664	** The --verily option ignores the keep-glob and ignore-glob settings and
665	** turns on --force, --emptydirs, --dotfiles, and --disable-undo. Use the
666	** --verily option when you really want to clean up everything. Extreme
667	** care should be exercised when using the --verily option.
668	**
669	** Options:
670	** --allckouts Check for empty directories within any checkouts
671	** that may be nested within the current one. This
672	** option should be used with great care because the
	@@ -678,11 +687,12 @@
687	** argument. Matching files, if any, are removed
688	** prior to checking for any empty directories;
689	** therefore, directories that contain only files
690	** that were removed will be removed as well.
691	** -f\|--force Remove files without prompting.
692	** -i\|--prompt Prompt before removing each file. This option
693	** implies the --disable-undo option.
694	** -x\|--verily WARNING: Removes everything that is not a managed
695	** file or the repository itself. This option
696	** implies the --force, --emptydirs, --dotfiles, and
697	** --disable-undo options. Furthermore, it completely
698	** disregards the keep-glob and ignore-glob settings.
699

M src/db.c

+2 -2

		--- src/db.c
		+++ src/db.c
		@@ -2677,12 +2677,12 @@
2677	2677	** differ only in case are the same file. Defaults to
2678	2678	** TRUE for unix and FALSE for Cygwin, Mac and Windows.
2679	2679	**
2680	2680	** clean-glob The VALUE is a comma or newline-separated list of GLOB
2681	2681	** (versionable) patterns specifying files that the "clean" command will
2682		-** delete without prompting even when the -force flag has
2683		-** not been used. Example: .a .lib *.o
	2682	+** delete without prompting or allowing undo.
	2683	+** Example: .a,.lib,*.o
2684	2684	**
2685	2685	** clearsign When enabled, fossil will attempt to sign all commits
2686	2686	** with gpg. When disabled (the default), commits will
2687	2687	** be unsigned. Default: off
2688	2688	**
2689	2689

	--- src/db.c
	+++ src/db.c
	@@ -2677,12 +2677,12 @@
2677	** differ only in case are the same file. Defaults to
2678	** TRUE for unix and FALSE for Cygwin, Mac and Windows.
2679	**
2680	** clean-glob The VALUE is a comma or newline-separated list of GLOB
2681	** (versionable) patterns specifying files that the "clean" command will
2682	** delete without prompting even when the -force flag has
2683	** not been used. Example: .a .lib *.o
2684	**
2685	** clearsign When enabled, fossil will attempt to sign all commits
2686	** with gpg. When disabled (the default), commits will
2687	** be unsigned. Default: off
2688	**
2689

	--- src/db.c
	+++ src/db.c
	@@ -2677,12 +2677,12 @@
2677	** differ only in case are the same file. Defaults to
2678	** TRUE for unix and FALSE for Cygwin, Mac and Windows.
2679	**
2680	** clean-glob The VALUE is a comma or newline-separated list of GLOB
2681	** (versionable) patterns specifying files that the "clean" command will
2682	** delete without prompting or allowing undo.
2683	** Example: .a,.lib,*.o
2684	**
2685	** clearsign When enabled, fossil will attempt to sign all commits
2686	** with gpg. When disabled (the default), commits will
2687	** be unsigned. Default: off
2688	**
2689

M src/export.c

-3

		--- src/export.c
		+++ src/export.c
		@@ -185,11 +185,10 @@
185	185	** database. Otherwise, 0 is returned.
186	186	** mark->name is dynamically allocated, and owned by the caller.
187	187	*/
188	188	int parse_mark(char line, struct mark_t mark){
189	189	char *cur_tok;
190		- char type;
191	190	cur_tok = strtok(line, " \t");
192	191	if(!cur_tok\|\|strlen(cur_tok)<2){
193	192	return -1;
194	193	}
195	194	mark->rid = atoi(&cur_tok[1]);
		@@ -243,11 +242,10 @@
243	242	** or the rid/uuid information doesn't match what is in the repository
244	243	** database. Otherwise, 0 is returned.
245	244	*/
246	245	int import_marks(FILE* f, Bag blobs, Bag vers){
247	246	char line[101];
248		- size_t len;
249	247	while(fgets(line, sizeof(line), f)){
250	248	struct mark_t mark;
251	249	if(strlen(line)==100&&line[99]!='\n'){
252	250	/* line too long */
253	251	return -1;
		@@ -357,11 +355,10 @@
357	355	db_multi_exec("CREATE TEMPORARY TABLE oldblob(rid INTEGER PRIMARY KEY)");
358	356	db_multi_exec("CREATE TEMPORARY TABLE oldcommit(rid INTEGER PRIMARY KEY)");
359	357	db_multi_exec("CREATE TEMP TABLE xmark(tname TEXT UNIQUE, trid INT, tuuid TEXT)");
360	358	if( markfile_in!=0 ){
361	359	Stmt qb,qc;
362		- char line[100];
363	360	FILE *f;
364	361	int rid;
365	362
366	363	f = fossil_fopen(markfile_in, "r");
367	364	if( f==0 ){
368	365

	--- src/export.c
	+++ src/export.c
	@@ -185,11 +185,10 @@
185	** database. Otherwise, 0 is returned.
186	** mark->name is dynamically allocated, and owned by the caller.
187	*/
188	int parse_mark(char line, struct mark_t mark){
189	char *cur_tok;
190	char type;
191	cur_tok = strtok(line, " \t");
192	if(!cur_tok\|\|strlen(cur_tok)<2){
193	return -1;
194	}
195	mark->rid = atoi(&cur_tok[1]);
	@@ -243,11 +242,10 @@
243	** or the rid/uuid information doesn't match what is in the repository
244	** database. Otherwise, 0 is returned.
245	*/
246	int import_marks(FILE* f, Bag blobs, Bag vers){
247	char line[101];
248	size_t len;
249	while(fgets(line, sizeof(line), f)){
250	struct mark_t mark;
251	if(strlen(line)==100&&line[99]!='\n'){
252	/* line too long */
253	return -1;
	@@ -357,11 +355,10 @@
357	db_multi_exec("CREATE TEMPORARY TABLE oldblob(rid INTEGER PRIMARY KEY)");
358	db_multi_exec("CREATE TEMPORARY TABLE oldcommit(rid INTEGER PRIMARY KEY)");
359	db_multi_exec("CREATE TEMP TABLE xmark(tname TEXT UNIQUE, trid INT, tuuid TEXT)");
360	if( markfile_in!=0 ){
361	Stmt qb,qc;
362	char line[100];
363	FILE *f;
364	int rid;
365
366	f = fossil_fopen(markfile_in, "r");
367	if( f==0 ){
368

	--- src/export.c
	+++ src/export.c
	@@ -185,11 +185,10 @@
185	** database. Otherwise, 0 is returned.
186	** mark->name is dynamically allocated, and owned by the caller.
187	*/
188	int parse_mark(char line, struct mark_t mark){
189	char *cur_tok;

190	cur_tok = strtok(line, " \t");
191	if(!cur_tok\|\|strlen(cur_tok)<2){
192	return -1;
193	}
194	mark->rid = atoi(&cur_tok[1]);
	@@ -243,11 +242,10 @@
242	** or the rid/uuid information doesn't match what is in the repository
243	** database. Otherwise, 0 is returned.
244	*/
245	int import_marks(FILE* f, Bag blobs, Bag vers){
246	char line[101];

247	while(fgets(line, sizeof(line), f)){
248	struct mark_t mark;
249	if(strlen(line)==100&&line[99]!='\n'){
250	/* line too long */
251	return -1;
	@@ -357,11 +355,10 @@
355	db_multi_exec("CREATE TEMPORARY TABLE oldblob(rid INTEGER PRIMARY KEY)");
356	db_multi_exec("CREATE TEMPORARY TABLE oldcommit(rid INTEGER PRIMARY KEY)");
357	db_multi_exec("CREATE TEMP TABLE xmark(tname TEXT UNIQUE, trid INT, tuuid TEXT)");
358	if( markfile_in!=0 ){
359	Stmt qb,qc;

360	FILE *f;
361	int rid;
362
363	f = fossil_fopen(markfile_in, "r");
364	if( f==0 ){
365

M src/foci.c

+6 -7

		--- src/foci.c
		+++ src/foci.c
		@@ -13,23 +13,22 @@
13	13	** [email protected]
14	14	** http://www.hwaci.com/drh/
15	15	**
16	16	*******************************************************************************
17	17	**
18		-** This routine implements an SQLite virtual table that gives all of the
19		-** files associated with a single check-in.
	18	+** This routine implements eponymous virtual table for SQLite that gives
	19	+** all of the files associated with a single check-in.
20	20	**
21		-** The filename "foci" is short for "Files of Check-in".
	21	+** The source code filename "foci" is short for "Files of Check-in".
22	22	**
23	23	** Usage example:
24	24	**
25		-** CREATE VIRTUAL TABLE temp.foci USING files_of_checkin;
26		-** -- ^^^^--- important!
27		-** SELECT * FROM foci WHERE checkinID=symbolic_name_to_rid('trunk');
	25	+** SELECT * FROM files_of_checkin
	26	+** WHERE checkinID=symbolic_name_to_rid('trunk');
28	27	**
29	28	** The symbolic_name_to_rid('trunk') function finds the BLOB.RID value
30		-** corresponding to the 'trunk' tag. Then the files_of_checkin virtual table
	29	+** corresponding to the 'trunk' tag. Then the foci virtual table
31	30	** decodes the manifest defined by that BLOB and returns all files described
32	31	** by that manifest. The "schema" for the temp.foci table is:
33	32	**
34	33	** CREATE TABLE files_of_checkin(
35	34	** checkinID INTEGER, -- RID for the check-in manifest
36	35

	--- src/foci.c
	+++ src/foci.c
	@@ -13,23 +13,22 @@
13	** [email protected]
14	** http://www.hwaci.com/drh/
15	**
16	*******************************************************************************
17	**
18	** This routine implements an SQLite virtual table that gives all of the
19	** files associated with a single check-in.
20	**
21	** The filename "foci" is short for "Files of Check-in".
22	**
23	** Usage example:
24	**
25	** CREATE VIRTUAL TABLE temp.foci USING files_of_checkin;
26	** -- ^^^^--- important!
27	** SELECT * FROM foci WHERE checkinID=symbolic_name_to_rid('trunk');
28	**
29	** The symbolic_name_to_rid('trunk') function finds the BLOB.RID value
30	** corresponding to the 'trunk' tag. Then the files_of_checkin virtual table
31	** decodes the manifest defined by that BLOB and returns all files described
32	** by that manifest. The "schema" for the temp.foci table is:
33	**
34	** CREATE TABLE files_of_checkin(
35	** checkinID INTEGER, -- RID for the check-in manifest
36

	--- src/foci.c
	+++ src/foci.c
	@@ -13,23 +13,22 @@
13	** [email protected]
14	** http://www.hwaci.com/drh/
15	**
16	*******************************************************************************
17	**
18	** This routine implements eponymous virtual table for SQLite that gives
19	** all of the files associated with a single check-in.
20	**
21	** The source code filename "foci" is short for "Files of Check-in".
22	**
23	** Usage example:
24	**
25	** SELECT * FROM files_of_checkin
26	** WHERE checkinID=symbolic_name_to_rid('trunk');

27	**
28	** The symbolic_name_to_rid('trunk') function finds the BLOB.RID value
29	** corresponding to the 'trunk' tag. Then the foci virtual table
30	** decodes the manifest defined by that BLOB and returns all files described
31	** by that manifest. The "schema" for the temp.foci table is:
32	**
33	** CREATE TABLE files_of_checkin(
34	** checkinID INTEGER, -- RID for the check-in manifest
35

M src/sqlite3.c

+226 -82

		--- src/sqlite3.c
		+++ src/sqlite3.c
		@@ -363,11 +363,11 @@
363	363	** [sqlite3_libversion_number()], [sqlite3_sourceid()],
364	364	** [sqlite_version()] and [sqlite_source_id()].
365	365	*/
366	366	#define SQLITE_VERSION "3.13.0"
367	367	#define SQLITE_VERSION_NUMBER 3013000
368		-#define SQLITE_SOURCE_ID "2016-05-09 19:03:42 14e53d0e2f62d82ae1d64a72fd9711548e3bf5ea"
	368	+#define SQLITE_SOURCE_ID "2016-05-18 10:57:30 fc49f556e48970561d7ab6a2f24fdd7d9eb81ff2"
369	369
370	370	/*
371	371	** CAPI3REF: Run-Time Library Version Numbers
372	372	** KEYWORDS: sqlite3_version, sqlite3_sourceid
373	373	**
		@@ -10917,11 +10917,11 @@
10917	10917	**
10918	10918	** When doing coverage testing ALWAYS and NEVER are hard-coded to
10919	10919	** be true and false so that the unreachable code they specify will
10920	10920	** not be counted as untested code.
10921	10921	*/
10922		-#if defined(SQLITE_COVERAGE_TEST)
	10922	+#if defined(SQLITE_COVERAGE_TEST) \|\| defined(SQLITE_MUTATION_TEST)
10923	10923	# define ALWAYS(X) (1)
10924	10924	# define NEVER(X) (0)
10925	10925	#elif !defined(NDEBUG)
10926	10926	# define ALWAYS(X) ((X)?1:(assert(0),0))
10927	10927	# define NEVER(X) ((X)?(assert(0),1):0)
		@@ -12954,11 +12954,11 @@
12954	12954	*/
12955	12955	struct PgHdr {
12956	12956	sqlite3_pcache_page pPage; / Pcache object page handle */
12957	12957	void pData; / Page data */
12958	12958	void pExtra; / Extra content */
12959		- PgHdr pDirty; / Transient list of dirty pages */
	12959	+ PgHdr pDirty; / Transient list of dirty sorted by pgno */
12960	12960	Pager pPager; / The pager this page is part of */
12961	12961	Pgno pgno; /* Page number for this page */
12962	12962	#ifdef SQLITE_CHECK_PAGES
12963	12963	u32 pageHash; /* Hash of page content */
12964	12964	#endif
		@@ -12979,15 +12979,14 @@
12979	12979	#define PGHDR_CLEAN 0x001 /* Page not on the PCache.pDirty list */
12980	12980	#define PGHDR_DIRTY 0x002 /* Page is on the PCache.pDirty list */
12981	12981	#define PGHDR_WRITEABLE 0x004 /* Journaled and ready to modify */
12982	12982	#define PGHDR_NEED_SYNC 0x008 /* Fsync the rollback journal before
12983	12983	** writing this page to the database */
12984		-#define PGHDR_NEED_READ 0x010 /* Content is unread */
12985		-#define PGHDR_DONT_WRITE 0x020 /* Do not write content to disk */
12986		-#define PGHDR_MMAP 0x040 /* This is an mmap page object */
	12984	+#define PGHDR_DONT_WRITE 0x010 /* Do not write content to disk */
	12985	+#define PGHDR_MMAP 0x020 /* This is an mmap page object */
12987	12986
12988		-#define PGHDR_WAL_APPEND 0x080 /* Appended to wal file */
	12987	+#define PGHDR_WAL_APPEND 0x040 /* Appended to wal file */
12989	12988
12990	12989	/* Initialize and shutdown the page cache subsystem */
12991	12990	SQLITE_PRIVATE int sqlite3PcacheInitialize(void);
12992	12991	SQLITE_PRIVATE void sqlite3PcacheShutdown(void);
12993	12992
		@@ -13065,10 +13064,15 @@
13065	13064	** interface is only available if SQLITE_CHECK_PAGES is defined when the
13066	13065	** library is built.
13067	13066	*/
13068	13067	SQLITE_PRIVATE void sqlite3PcacheIterateDirty(PCache pCache, void (xIter)(PgHdr *));
13069	13068	#endif
	13069	+
	13070	+#if defined(SQLITE_DEBUG)
	13071	+/* Check invariants on a PgHdr object */
	13072	+SQLITE_PRIVATE int sqlite3PcachePageSanity(PgHdr*);
	13073	+#endif
13070	13074
13071	13075	/* Set and get the suggested cache-size for the specified pager-cache.
13072	13076	**
13073	13077	** If no global maximum is configured, then the system attempts to limit
13074	13078	** the total number of pages cached by purgeable pager-caches to the sum
		@@ -43115,11 +43119,33 @@
43115	43119	** This file implements that page cache.
43116	43120	*/
43117	43121	/* #include "sqliteInt.h" */
43118	43122
43119	43123	/*
43120		-** A complete page cache is an instance of this structure.
	43124	+** A complete page cache is an instance of this structure. Every
	43125	+** entry in the cache holds a single page of the database file. The
	43126	+** btree layer only operates on the cached copy of the database pages.
	43127	+**
	43128	+** A page cache entry is "clean" if it exactly matches what is currently
	43129	+** on disk. A page is "dirty" if it has been modified and needs to be
	43130	+** persisted to disk.
	43131	+**
	43132	+** pDirty, pDirtyTail, pSynced:
	43133	+** All dirty pages are linked into the doubly linked list using
	43134	+** PgHdr.pDirtyNext and pDirtyPrev. The list is maintained in LRU order
	43135	+** such that p was added to the list more recently than p->pDirtyNext.
	43136	+** PCache.pDirty points to the first (newest) element in the list and
	43137	+** pDirtyTail to the last (oldest).
	43138	+**
	43139	+** The PCache.pSynced variable is used to optimize searching for a dirty
	43140	+** page to eject from the cache mid-transaction. It is better to eject
	43141	+** a page that does not require a journal sync than one that does.
	43142	+** Therefore, pSynced is maintained to that it almost always points
	43143	+** to either the oldest page in the pDirty/pDirtyTail list that has a
	43144	+** clear PGHDR_NEED_SYNC flag or to a page that is older than this one
	43145	+** (so that the right page to eject can be found by following pDirtyPrev
	43146	+** pointers).
43121	43147	*/
43122	43148	struct PCache {
43123	43149	PgHdr pDirty, pDirtyTail; /* List of dirty pages in LRU order */
43124	43150	PgHdr pSynced; / Last synced page in dirty page list */
43125	43151	int nRefSum; /* Sum of ref counts over all pages */
		@@ -43131,10 +43157,99 @@
43131	43157	u8 eCreate; /* eCreate value for for xFetch() */
43132	43158	int (xStress)(void,PgHdr); / Call to try make a page clean */
43133	43159	void pStress; / Argument to xStress */
43134	43160	sqlite3_pcache pCache; / Pluggable cache module */
43135	43161	};
	43162	+
	43163	+/******************************** Test and Debug Logic ********************/
	43164	+/*
	43165	+** Debug tracing macros. Enable by by changing the "0" to "1" and
	43166	+** recompiling.
	43167	+**
	43168	+** When sqlite3PcacheTrace is 1, single line trace messages are issued.
	43169	+** When sqlite3PcacheTrace is 2, a dump of the pcache showing all cache entries
	43170	+** is displayed for many operations, resulting in a lot of output.
	43171	+*/
	43172	+#if defined(SQLITE_DEBUG) && 0
	43173	+ int sqlite3PcacheTrace = 2; /* 0: off 1: simple 2: cache dumps */
	43174	+ int sqlite3PcacheMxDump = 9999; /* Max cache entries for pcacheDump() */
	43175	+# define pcacheTrace(X) if(sqlite3PcacheTrace){sqlite3DebugPrintf X;}
	43176	+ void pcacheDump(PCache *pCache){
	43177	+ int N;
	43178	+ int i, j;
	43179	+ sqlite3_pcache_page *pLower;
	43180	+ PgHdr *pPg;
	43181	+ unsigned char *a;
	43182	+
	43183	+ if( sqlite3PcacheTrace<2 ) return;
	43184	+ if( pCache->pCache==0 ) return;
	43185	+ N = sqlite3PcachePagecount(pCache);
	43186	+ if( N>sqlite3PcacheMxDump ) N = sqlite3PcacheMxDump;
	43187	+ for(i=1; i<=N; i++){
	43188	+ pLower = sqlite3GlobalConfig.pcache2.xFetch(pCache->pCache, i, 0);
	43189	+ if( pLower==0 ) continue;
	43190	+ pPg = (PgHdr*)pLower->pExtra;
	43191	+ printf("%3d: nRef %2d flgs %02x data ", i, pPg->nRef, pPg->flags);
	43192	+ a = (unsigned char *)pLower->pBuf;
	43193	+ for(j=0; j<12; j++) printf("%02x", a[j]);
	43194	+ printf("\n");
	43195	+ if( pPg->pPage==0 ){
	43196	+ sqlite3GlobalConfig.pcache2.xUnpin(pCache->pCache, pLower, 0);
	43197	+ }
	43198	+ }
	43199	+ }
	43200	+ #else
	43201	+# define pcacheTrace(X)
	43202	+# define pcacheDump(X)
	43203	+#endif
	43204	+
	43205	+/*
	43206	+** Check invariants on a PgHdr entry. Return true if everything is OK.
	43207	+** Return false if any invariant is violated.
	43208	+**
	43209	+** This routine is for use inside of assert() statements only. For
	43210	+** example:
	43211	+**
	43212	+** assert( sqlite3PcachePageSanity(pPg) );
	43213	+*/
	43214	+#if SQLITE_DEBUG
	43215	+SQLITE_PRIVATE int sqlite3PcachePageSanity(PgHdr *pPg){
	43216	+ PCache *pCache;
	43217	+ assert( pPg!=0 );
	43218	+ assert( pPg->pgno>0 ); /* Page number is 1 or more */
	43219	+ pCache = pPg->pCache;
	43220	+ assert( pCache!=0 ); /* Every page has an associated PCache */
	43221	+ if( pPg->flags & PGHDR_CLEAN ){
	43222	+ assert( (pPg->flags & PGHDR_DIRTY)==0 );/* Cannot be both CLEAN and DIRTY */
	43223	+ assert( pCache->pDirty!=pPg ); /* CLEAN pages not on dirty list */
	43224	+ assert( pCache->pDirtyTail!=pPg );
	43225	+ }
	43226	+ /* WRITEABLE pages must also be DIRTY */
	43227	+ if( pPg->flags & PGHDR_WRITEABLE ){
	43228	+ assert( pPg->flags & PGHDR_DIRTY ); /* WRITEABLE implies DIRTY */
	43229	+ }
	43230	+ /* NEED_SYNC can be set independently of WRITEABLE. This can happen,
	43231	+ ** for example, when using the sqlite3PagerDontWrite() optimization:
	43232	+ ** (1) Page X is journalled, and gets WRITEABLE and NEED_SEEK.
	43233	+ ** (2) Page X moved to freelist, WRITEABLE is cleared
	43234	+ ** (3) Page X reused, WRITEABLE is set again
	43235	+ ** If NEED_SYNC had been cleared in step 2, then it would not be reset
	43236	+ ** in step 3, and page might be written into the database without first
	43237	+ ** syncing the rollback journal, which might cause corruption on a power
	43238	+ ** loss.
	43239	+ **
	43240	+ ** Another example is when the database page size is smaller than the
	43241	+ ** disk sector size. When any page of a sector is journalled, all pages
	43242	+ ** in that sector are marked NEED_SYNC even if they are still CLEAN, just
	43243	+ ** in case they are later modified, since all pages in the same sector
	43244	+ ** must be journalled and synced before any of those pages can be safely
	43245	+ ** written.
	43246	+ */
	43247	+ return 1;
	43248	+}
	43249	+#endif /* SQLITE_DEBUG */
	43250	+
43136	43251
43137	43252	/******************************** Linked List Management ******************/
43138	43253
43139	43254	/* Allowed values for second argument to pcacheManageDirtyList() */
43140	43255	#define PCACHE_DIRTYLIST_REMOVE 1 /* Remove pPage from dirty list */
		@@ -43148,21 +43263,20 @@
43148	43263	** the dirty list. Doing both moves pPage to the front of the dirty list.
43149	43264	*/
43150	43265	static void pcacheManageDirtyList(PgHdr *pPage, u8 addRemove){
43151	43266	PCache *p = pPage->pCache;
43152	43267
	43268	+ pcacheTrace(("%p.DIRTYLIST.%s %d\n", p,
	43269	+ addRemove==1 ? "REMOVE" : addRemove==2 ? "ADD" : "FRONT",
	43270	+ pPage->pgno));
43153	43271	if( addRemove & PCACHE_DIRTYLIST_REMOVE ){
43154	43272	assert( pPage->pDirtyNext \|\| pPage==p->pDirtyTail );
43155	43273	assert( pPage->pDirtyPrev \|\| pPage==p->pDirty );
43156	43274
43157	43275	/* Update the PCache1.pSynced variable if necessary. */
43158	43276	if( p->pSynced==pPage ){
43159		- PgHdr *pSynced = pPage->pDirtyPrev;
43160		- while( pSynced && (pSynced->flags&PGHDR_NEED_SYNC) ){
43161		- pSynced = pSynced->pDirtyPrev;
43162		- }
43163		- p->pSynced = pSynced;
	43277	+ p->pSynced = pPage->pDirtyPrev;
43164	43278	}
43165	43279
43166	43280	if( pPage->pDirtyNext ){
43167	43281	pPage->pDirtyNext->pDirtyPrev = pPage->pDirtyPrev;
43168	43282	}else{
		@@ -43170,14 +43284,19 @@
43170	43284	p->pDirtyTail = pPage->pDirtyPrev;
43171	43285	}
43172	43286	if( pPage->pDirtyPrev ){
43173	43287	pPage->pDirtyPrev->pDirtyNext = pPage->pDirtyNext;
43174	43288	}else{
	43289	+ /* If there are now no dirty pages in the cache, set eCreate to 2.
	43290	+ ** This is an optimization that allows sqlite3PcacheFetch() to skip
	43291	+ ** searching for a dirty page to eject from the cache when it might
	43292	+ ** otherwise have to. */
43175	43293	assert( pPage==p->pDirty );
43176	43294	p->pDirty = pPage->pDirtyNext;
43177		- if( p->pDirty==0 && p->bPurgeable ){
43178		- assert( p->eCreate==1 );
	43295	+ assert( p->bPurgeable \|\| p->eCreate==2 );
	43296	+ if( p->pDirty==0 ){ /OPTIMIZATION-IF-TRUE/
	43297	+ assert( p->bPurgeable==0 \|\| p->eCreate==1 );
43179	43298	p->eCreate = 2;
43180	43299	}
43181	43300	}
43182	43301	pPage->pDirtyNext = 0;
43183	43302	pPage->pDirtyPrev = 0;
		@@ -43195,23 +43314,34 @@
43195	43314	assert( p->eCreate==2 );
43196	43315	p->eCreate = 1;
43197	43316	}
43198	43317	}
43199	43318	p->pDirty = pPage;
43200		- if( !p->pSynced && 0==(pPage->flags&PGHDR_NEED_SYNC) ){
	43319	+
	43320	+ /* If pSynced is NULL and this page has a clear NEED_SYNC flag, set
	43321	+ ** pSynced to point to it. Checking the NEED_SYNC flag is an
	43322	+ ** optimization, as if pSynced points to a page with the NEED_SYNC
	43323	+ ** flag set sqlite3PcacheFetchStress() searches through all newer
	43324	+ ** entries of the dirty-list for a page with NEED_SYNC clear anyway. */
	43325	+ if( !p->pSynced
	43326	+ && 0==(pPage->flags&PGHDR_NEED_SYNC) /OPTIMIZATION-IF-FALSE/
	43327	+ ){
43201	43328	p->pSynced = pPage;
43202	43329	}
43203	43330	}
	43331	+ pcacheDump(p);
43204	43332	}
43205	43333
43206	43334	/*
43207	43335	** Wrapper around the pluggable caches xUnpin method. If the cache is
43208	43336	** being used for an in-memory database, this function is a no-op.
43209	43337	*/
43210	43338	static void pcacheUnpin(PgHdr *p){
43211	43339	if( p->pCache->bPurgeable ){
	43340	+ pcacheTrace(("%p.UNPIN %d\n", p->pCache, p->pgno));
43212	43341	sqlite3GlobalConfig.pcache2.xUnpin(p->pCache->pCache, p->pPage, 0);
	43342	+ pcacheDump(p->pCache);
43213	43343	}
43214	43344	}
43215	43345
43216	43346	/*
43217	43347	** Compute the number of pages of cache requested. p->szCache is the
		@@ -43277,10 +43407,11 @@
43277	43407	p->eCreate = 2;
43278	43408	p->xStress = xStress;
43279	43409	p->pStress = pStress;
43280	43410	p->szCache = 100;
43281	43411	p->szSpill = 1;
	43412	+ pcacheTrace(("%p.OPEN szPage %d bPurgeable %d\n",p,szPage,bPurgeable));
43282	43413	return sqlite3PcacheSetPageSize(p, szPage);
43283	43414	}
43284	43415
43285	43416	/*
43286	43417	** Change the page size for PCache object. The caller must ensure that there
		@@ -43299,10 +43430,11 @@
43299	43430	if( pCache->pCache ){
43300	43431	sqlite3GlobalConfig.pcache2.xDestroy(pCache->pCache);
43301	43432	}
43302	43433	pCache->pCache = pNew;
43303	43434	pCache->szPage = szPage;
	43435	+ pcacheTrace(("%p.PAGESIZE %d\n",pCache,szPage));
43304	43436	}
43305	43437	return SQLITE_OK;
43306	43438	}
43307	43439
43308	43440	/*
		@@ -43333,15 +43465,17 @@
43333	43465	PCache pCache, / Obtain the page from this cache */
43334	43466	Pgno pgno, /* Page number to obtain */
43335	43467	int createFlag /* If true, create page if it does not exist already */
43336	43468	){
43337	43469	int eCreate;
	43470	+ sqlite3_pcache_page *pRes;
43338	43471
43339	43472	assert( pCache!=0 );
43340	43473	assert( pCache->pCache!=0 );
43341	43474	assert( createFlag==3 \|\| createFlag==0 );
43342	43475	assert( pgno>0 );
	43476	+ assert( pCache->eCreate==((pCache->bPurgeable && pCache->pDirty) ? 1 : 2) );
43343	43477
43344	43478	/* eCreate defines what to do if the page does not exist.
43345	43479	** 0 Do not allocate a new page. (createFlag==0)
43346	43480	** 1 Allocate a new page if doing so is inexpensive.
43347	43481	** (createFlag==1 AND bPurgeable AND pDirty)
		@@ -43350,11 +43484,14 @@
43350	43484	*/
43351	43485	eCreate = createFlag & pCache->eCreate;
43352	43486	assert( eCreate==0 \|\| eCreate==1 \|\| eCreate==2 );
43353	43487	assert( createFlag==0 \|\| pCache->eCreate==eCreate );
43354	43488	assert( createFlag==0 \|\| eCreate==1+(!pCache->bPurgeable\|\|!pCache->pDirty) );
43355		- return sqlite3GlobalConfig.pcache2.xFetch(pCache->pCache, pgno, eCreate);
	43489	+ pRes = sqlite3GlobalConfig.pcache2.xFetch(pCache->pCache, pgno, eCreate);
	43490	+ pcacheTrace(("%p.FETCH %d%s (result: %p)\n",pCache,pgno,
	43491	+ createFlag?" create":"",pRes));
	43492	+ return pRes;
43356	43493	}
43357	43494
43358	43495	/*
43359	43496	** If the sqlite3PcacheFetch() routine is unable to allocate a new
43360	43497	** page because no clean pages are available for reuse and the cache
		@@ -43377,11 +43514,15 @@
43377	43514	if( sqlite3PcachePagecount(pCache)>pCache->szSpill ){
43378	43515	/* Find a dirty page to write-out and recycle. First try to find a
43379	43516	** page that does not require a journal-sync (one with PGHDR_NEED_SYNC
43380	43517	** cleared), but if that is not possible settle for any other
43381	43518	** unreferenced dirty page.
43382		- */
	43519	+ **
	43520	+ ** If the LRU page in the dirty list that has a clear PGHDR_NEED_SYNC
	43521	+ ** flag is currently referenced, then the following may leave pSynced
	43522	+ ** set incorrectly (pointing to other than the LRU page with NEED_SYNC
	43523	+ ** cleared). This is Ok, as pSynced is just an optimization. */
43383	43524	for(pPg=pCache->pSynced;
43384	43525	pPg && (pPg->nRef \|\| (pPg->flags&PGHDR_NEED_SYNC));
43385	43526	pPg=pPg->pDirtyPrev
43386	43527	);
43387	43528	pCache->pSynced = pPg;
		@@ -43395,11 +43536,13 @@
43395	43536	"spill page %d making room for %d - cache used: %d/%d",
43396	43537	pPg->pgno, pgno,
43397	43538	sqlite3GlobalConfig.pcache.xPagecount(pCache->pCache),
43398	43539	numberOfCachePages(pCache));
43399	43540	#endif
	43541	+ pcacheTrace(("%p.SPILL %d\n",pCache,pPg->pgno));
43400	43542	rc = pCache->xStress(pCache->pStress, pPg);
	43543	+ pcacheDump(pCache);
43401	43544	if( rc!=SQLITE_OK && rc!=SQLITE_BUSY ){
43402	43545	return rc;
43403	43546	}
43404	43547	}
43405	43548	}
		@@ -43455,10 +43598,11 @@
43455	43598	if( !pPgHdr->pPage ){
43456	43599	return pcacheFetchFinishWithInit(pCache, pgno, pPage);
43457	43600	}
43458	43601	pCache->nRefSum++;
43459	43602	pPgHdr->nRef++;
	43603	+ assert( sqlite3PcachePageSanity(pPgHdr) );
43460	43604	return pPgHdr;
43461	43605	}
43462	43606
43463	43607	/*
43464	43608	** Decrement the reference count on a page. If the page is clean and the
		@@ -43468,12 +43612,15 @@
43468	43612	assert( p->nRef>0 );
43469	43613	p->pCache->nRefSum--;
43470	43614	if( (--p->nRef)==0 ){
43471	43615	if( p->flags&PGHDR_CLEAN ){
43472	43616	pcacheUnpin(p);
43473		- }else if( p->pDirtyPrev!=0 ){
43474		- /* Move the page to the head of the dirty list. */
	43617	+ }else if( p->pDirtyPrev!=0 ){ /OPTIMIZATION-IF-FALSE/
	43618	+ /* Move the page to the head of the dirty list. If p->pDirtyPrev==0,
	43619	+ ** then page p is already at the head of the dirty list and the
	43620	+ ** following call would be a no-op. Hence the OPTIMIZATION-IF-FALSE
	43621	+ ** tag above. */
43475	43622	pcacheManageDirtyList(p, PCACHE_DIRTYLIST_FRONT);
43476	43623	}
43477	43624	}
43478	43625	}
43479	43626
		@@ -43480,10 +43627,11 @@
43480	43627	/*
43481	43628	** Increase the reference count of a supplied page by 1.
43482	43629	*/
43483	43630	SQLITE_PRIVATE void sqlite3PcacheRef(PgHdr *p){
43484	43631	assert(p->nRef>0);
	43632	+ assert( sqlite3PcachePageSanity(p) );
43485	43633	p->nRef++;
43486	43634	p->pCache->nRefSum++;
43487	43635	}
43488	43636
43489	43637	/*
		@@ -43491,10 +43639,11 @@
43491	43639	** page. This function deletes that reference, so after it returns the
43492	43640	** page pointed to by p is invalid.
43493	43641	*/
43494	43642	SQLITE_PRIVATE void sqlite3PcacheDrop(PgHdr *p){
43495	43643	assert( p->nRef==1 );
	43644	+ assert( sqlite3PcachePageSanity(p) );
43496	43645	if( p->flags&PGHDR_DIRTY ){
43497	43646	pcacheManageDirtyList(p, PCACHE_DIRTYLIST_REMOVE);
43498	43647	}
43499	43648	p->pCache->nRefSum--;
43500	43649	sqlite3GlobalConfig.pcache2.xUnpin(p->pCache->pCache, p->pPage, 1);
		@@ -43504,30 +43653,36 @@
43504	43653	** Make sure the page is marked as dirty. If it isn't dirty already,
43505	43654	** make it so.
43506	43655	*/
43507	43656	SQLITE_PRIVATE void sqlite3PcacheMakeDirty(PgHdr *p){
43508	43657	assert( p->nRef>0 );
43509		- if( p->flags & (PGHDR_CLEAN\|PGHDR_DONT_WRITE) ){
	43658	+ assert( sqlite3PcachePageSanity(p) );
	43659	+ if( p->flags & (PGHDR_CLEAN\|PGHDR_DONT_WRITE) ){ /OPTIMIZATION-IF-FALSE/
43510	43660	p->flags &= ~PGHDR_DONT_WRITE;
43511	43661	if( p->flags & PGHDR_CLEAN ){
43512	43662	p->flags ^= (PGHDR_DIRTY\|PGHDR_CLEAN);
	43663	+ pcacheTrace(("%p.DIRTY %d\n",p->pCache,p->pgno));
43513	43664	assert( (p->flags & (PGHDR_DIRTY\|PGHDR_CLEAN))==PGHDR_DIRTY );
43514	43665	pcacheManageDirtyList(p, PCACHE_DIRTYLIST_ADD);
43515	43666	}
	43667	+ assert( sqlite3PcachePageSanity(p) );
43516	43668	}
43517	43669	}
43518	43670
43519	43671	/*
43520	43672	** Make sure the page is marked as clean. If it isn't clean already,
43521	43673	** make it so.
43522	43674	*/
43523	43675	SQLITE_PRIVATE void sqlite3PcacheMakeClean(PgHdr *p){
43524		- if( (p->flags & PGHDR_DIRTY) ){
	43676	+ assert( sqlite3PcachePageSanity(p) );
	43677	+ if( ALWAYS((p->flags & PGHDR_DIRTY)!=0) ){
43525	43678	assert( (p->flags & PGHDR_CLEAN)==0 );
43526	43679	pcacheManageDirtyList(p, PCACHE_DIRTYLIST_REMOVE);
43527	43680	p->flags &= ~(PGHDR_DIRTY\|PGHDR_NEED_SYNC\|PGHDR_WRITEABLE);
43528	43681	p->flags \|= PGHDR_CLEAN;
	43682	+ pcacheTrace(("%p.CLEAN %d\n",p->pCache,p->pgno));
	43683	+ assert( sqlite3PcachePageSanity(p) );
43529	43684	if( p->nRef==0 ){
43530	43685	pcacheUnpin(p);
43531	43686	}
43532	43687	}
43533	43688	}
		@@ -43535,10 +43690,11 @@
43535	43690	/*
43536	43691	** Make every page in the cache clean.
43537	43692	*/
43538	43693	SQLITE_PRIVATE void sqlite3PcacheCleanAll(PCache *pCache){
43539	43694	PgHdr *p;
	43695	+ pcacheTrace(("%p.CLEAN-ALL\n",pCache));
43540	43696	while( (p = pCache->pDirty)!=0 ){
43541	43697	sqlite3PcacheMakeClean(p);
43542	43698	}
43543	43699	}
43544	43700
		@@ -43545,10 +43701,11 @@
43545	43701	/*
43546	43702	** Clear the PGHDR_NEED_SYNC and PGHDR_WRITEABLE flag from all dirty pages.
43547	43703	*/
43548	43704	SQLITE_PRIVATE void sqlite3PcacheClearWritable(PCache *pCache){
43549	43705	PgHdr *p;
	43706	+ pcacheTrace(("%p.CLEAR-WRITEABLE\n",pCache));
43550	43707	for(p=pCache->pDirty; p; p=p->pDirtyNext){
43551	43708	p->flags &= ~(PGHDR_NEED_SYNC\|PGHDR_WRITEABLE);
43552	43709	}
43553	43710	pCache->pSynced = pCache->pDirtyTail;
43554	43711	}
		@@ -43569,10 +43726,12 @@
43569	43726	*/
43570	43727	SQLITE_PRIVATE void sqlite3PcacheMove(PgHdr *p, Pgno newPgno){
43571	43728	PCache *pCache = p->pCache;
43572	43729	assert( p->nRef>0 );
43573	43730	assert( newPgno>0 );
	43731	+ assert( sqlite3PcachePageSanity(p) );
	43732	+ pcacheTrace(("%p.MOVE %d -> %d\n",pCache,p->pgno,newPgno));
43574	43733	sqlite3GlobalConfig.pcache2.xRekey(pCache->pCache, p->pPage, p->pgno,newPgno);
43575	43734	p->pgno = newPgno;
43576	43735	if( (p->flags&PGHDR_DIRTY) && (p->flags&PGHDR_NEED_SYNC) ){
43577	43736	pcacheManageDirtyList(p, PCACHE_DIRTYLIST_FRONT);
43578	43737	}
		@@ -43589,10 +43748,11 @@
43589	43748	*/
43590	43749	SQLITE_PRIVATE void sqlite3PcacheTruncate(PCache *pCache, Pgno pgno){
43591	43750	if( pCache->pCache ){
43592	43751	PgHdr *p;
43593	43752	PgHdr *pNext;
	43753	+ pcacheTrace(("%p.TRUNCATE %d\n",pCache,pgno));
43594	43754	for(p=pCache->pDirty; p; p=pNext){
43595	43755	pNext = p->pDirtyNext;
43596	43756	/* This routine never gets call with a positive pgno except right
43597	43757	** after sqlite3PcacheCleanAll(). So if there are dirty pages,
43598	43758	** it must be that pgno==0.
		@@ -43619,10 +43779,11 @@
43619	43779	/*
43620	43780	** Close a cache.
43621	43781	*/
43622	43782	SQLITE_PRIVATE void sqlite3PcacheClose(PCache *pCache){
43623	43783	assert( pCache->pCache!=0 );
	43784	+ pcacheTrace(("%p.CLOSE\n",pCache));
43624	43785	sqlite3GlobalConfig.pcache2.xDestroy(pCache->pCache);
43625	43786	}
43626	43787
43627	43788	/*
43628	43789	** Discard the contents of the cache.
		@@ -47585,24 +47746,28 @@
47585	47746	}
47586	47747
47587	47748	static int pager_truncate(Pager *pPager, Pgno nPage);
47588	47749
47589	47750	/*
47590		-** The write transaction open on the pager passed as the only argument is
47591		-** being committed. This function returns true if all dirty pages should
47592		-** be flushed to disk, or false otherwise. Pages should be flushed to disk
47593		-** unless one of the following is true:
47594		-**
47595		-** * The db is an in-memory database.
47596		-**
47597		-** * The db is a temporary database and the db file has not been opened.
47598		-**
47599		-** * The db is a temporary database and the cache contains less than
47600		-** C/4 dirty pages, where C is the configured cache-size.
47601		-*/
47602		-static int pagerFlushOnCommit(Pager *pPager){
	47751	+** The write transaction open on pPager is being committed (bCommit==1)
	47752	+** or rolled back (bCommit==0).
	47753	+**
	47754	+** Return TRUE if and only if all dirty pages should be flushed to disk.
	47755	+**
	47756	+** Rules:
	47757	+**
	47758	+** * For non-TEMP databases, always sync to disk. This is necessary
	47759	+** for transactions to be durable.
	47760	+**
	47761	+** * Sync TEMP database only on a COMMIT (not a ROLLBACK) when the backing
	47762	+** file has been created already (via a spill on pagerStress()) and
	47763	+** when the number of dirty pages in memory exceeds 25% of the total
	47764	+** cache size.
	47765	+*/
	47766	+static int pagerFlushOnCommit(Pager *pPager, int bCommit){
47603	47767	if( pPager->tempFile==0 ) return 1;
	47768	+ if( !bCommit ) return 0;
47604	47769	if( !isOpen(pPager->fd) ) return 0;
47605	47770	return (sqlite3PCachePercentDirty(pPager->pPCache)>=25);
47606	47771	}
47607	47772
47608	47773	/*
		@@ -47706,11 +47871,11 @@
47706	47871	}
47707	47872	pPager->journalOff = 0;
47708	47873	}else if( pPager->journalMode==PAGER_JOURNALMODE_PERSIST
47709	47874	\|\| (pPager->exclusiveMode && pPager->journalMode!=PAGER_JOURNALMODE_WAL)
47710	47875	){
47711		- rc = zeroJournalHdr(pPager, hasMaster);
	47876	+ rc = zeroJournalHdr(pPager, hasMaster\|\|pPager->tempFile);
47712	47877	pPager->journalOff = 0;
47713	47878	}else{
47714	47879	/* This branch may be executed with Pager.journalMode==MEMORY if
47715	47880	** a hot-journal was just rolled back. In this case the journal
47716	47881	** file should be closed and deleted. If this connection writes to
		@@ -47741,16 +47906,18 @@
47741	47906	#endif
47742	47907
47743	47908	sqlite3BitvecDestroy(pPager->pInJournal);
47744	47909	pPager->pInJournal = 0;
47745	47910	pPager->nRec = 0;
47746		- if( MEMDB \|\| pagerFlushOnCommit(pPager) ){
47747		- sqlite3PcacheCleanAll(pPager->pPCache);
47748		- }else{
47749		- sqlite3PcacheClearWritable(pPager->pPCache);
	47911	+ if( rc==SQLITE_OK ){
	47912	+ if( pagerFlushOnCommit(pPager, bCommit) ){
	47913	+ sqlite3PcacheCleanAll(pPager->pPCache);
	47914	+ }else{
	47915	+ sqlite3PcacheClearWritable(pPager->pPCache);
	47916	+ }
	47917	+ sqlite3PcacheTruncate(pPager->pPCache, pPager->dbSize);
47750	47918	}
47751		- sqlite3PcacheTruncate(pPager->pPCache, pPager->dbSize);
47752	47919
47753	47920	if( pagerUseWal(pPager) ){
47754	47921	/* Drop the WAL write-lock, if any. Also, if the connection was in
47755	47922	** locking_mode=exclusive mode but is no longer, drop the EXCLUSIVE
47756	47923	** lock held on the database file.
		@@ -48080,11 +48247,10 @@
48080	48247	pPager->doNotSpill \|= SPILLFLAG_ROLLBACK;
48081	48248	rc = sqlite3PagerGet(pPager, pgno, &pPg, 1);
48082	48249	assert( (pPager->doNotSpill & SPILLFLAG_ROLLBACK)!=0 );
48083	48250	pPager->doNotSpill &= ~SPILLFLAG_ROLLBACK;
48084	48251	if( rc!=SQLITE_OK ) return rc;
48085		- pPg->flags &= ~PGHDR_NEED_READ;
48086	48252	sqlite3PcacheMakeDirty(pPg);
48087	48253	}
48088	48254	if( pPg ){
48089	48255	/* No page should ever be explicitly rolled back that is in use, except
48090	48256	** for page 1 which is held in use in order to keep the lock on the
		@@ -48094,37 +48260,14 @@
48094	48260	*/
48095	48261	void *pData;
48096	48262	pData = pPg->pData;
48097	48263	memcpy(pData, (u8*)aData, pPager->pageSize);
48098	48264	pPager->xReiniter(pPg);
48099		- if( isMainJrnl && (!isSavepnt \|\| *pOffset<=pPager->journalHdr) ){
48100		- /* If the contents of this page were just restored from the main
48101		- ** journal file, then its content must be as they were when the
48102		- ** transaction was first opened. In this case we can mark the page
48103		- ** as clean, since there will be no need to write it out to the
48104		- ** database.
48105		- **
48106		- ** There is one exception to this rule. If the page is being rolled
48107		- ** back as part of a savepoint (or statement) rollback from an
48108		- ** unsynced portion of the main journal file, then it is not safe
48109		- ** to mark the page as clean. This is because marking the page as
48110		- ** clean will clear the PGHDR_NEED_SYNC flag. Since the page is
48111		- ** already in the journal file (recorded in Pager.pInJournal) and
48112		- ** the PGHDR_NEED_SYNC flag is cleared, if the page is written to
48113		- ** again within this transaction, it will be marked as dirty but
48114		- ** the PGHDR_NEED_SYNC flag will not be set. It could then potentially
48115		- ** be written out into the database file before its journal file
48116		- ** segment is synced. If a crash occurs during or following this,
48117		- ** database corruption may ensue.
48118		- **
48119		- ** Update: Another exception is for temp files that are not
48120		- ** in-memory databases. In this case the page may have been dirty
48121		- ** at the start of the transaction.
48122		- */
48123		- assert( !pagerUseWal(pPager) );
48124		- if( pPager->tempFile==0 ) sqlite3PcacheMakeClean(pPg);
48125		- }
	48265	+ /* It used to be that sqlite3PcacheMakeClean(pPg) was called here. But
	48266	+ ** that call was dangerous and had no detectable benefit since the cache
	48267	+ ** is normally cleaned by sqlite3PcacheCleanAll() after rollback and so
	48268	+ ** has been removed. */
48126	48269	pager_set_pagehash(pPg);
48127	48270
48128	48271	/* If this was page 1, then restore the value of Pager.dbFileVers.
48129	48272	** Do this before any decoding. */
48130	48273	if( pgno==1 ){
		@@ -51731,10 +51874,11 @@
51731	51874	if( !pPager->tempFile && (pPg->flags&PGHDR_DIRTY) && pPager->nSavepoint==0 ){
51732	51875	PAGERTRACE(("DONT_WRITE page %d of %d\n", pPg->pgno, PAGERID(pPager)));
51733	51876	IOTRACE(("CLEAN %p %d\n", pPager, pPg->pgno))
51734	51877	pPg->flags \|= PGHDR_DONT_WRITE;
51735	51878	pPg->flags &= ~PGHDR_WRITEABLE;
	51879	+ testcase( pPg->flags & PGHDR_NEED_SYNC );
51736	51880	pager_set_pagehash(pPg);
51737	51881	}
51738	51882	}
51739	51883
51740	51884	/*
		@@ -51926,21 +52070,21 @@
51926	52070
51927	52071	/* If a prior error occurred, report that error again. */
51928	52072	if( NEVER(pPager->errCode) ) return pPager->errCode;
51929	52073
51930	52074	/* Provide the ability to easily simulate an I/O error during testing */
51931		- if( (rc = sqlite3FaultSim(400))!=SQLITE_OK ) return rc;
	52075	+ if( sqlite3FaultSim(400) ) return SQLITE_IOERR;
51932	52076
51933	52077	PAGERTRACE(("DATABASE SYNC: File=%s zMaster=%s nSize=%d\n",
51934	52078	pPager->zFilename, zMaster, pPager->dbSize));
51935	52079
51936	52080	/* If no database changes have been made, return early. */
51937	52081	if( pPager->eState<PAGER_WRITER_CACHEMOD ) return SQLITE_OK;
51938	52082
51939	52083	assert( MEMDB==0 \|\| pPager->tempFile );
51940	52084	assert( isOpen(pPager->fd) \|\| pPager->tempFile );
51941		- if( 0==pagerFlushOnCommit(pPager) ){
	52085	+ if( 0==pagerFlushOnCommit(pPager, 1) ){
51942	52086	/* If this is an in-memory db, or no pages have been written to, or this
51943	52087	** function has already been called, it is mostly a no-op. However, any
51944	52088	** backup in progress needs to be restarted. */
51945	52089	sqlite3BackupRestart(pPager->pBackup);
51946	52090	}else{
		@@ -52561,10 +52705,11 @@
52561	52705	assert( assert_pager_state(pPager) );
52562	52706
52563	52707	/* In order to be able to rollback, an in-memory database must journal
52564	52708	** the page we are moving from.
52565	52709	*/
	52710	+ assert( pPager->tempFile \|\| !MEMDB );
52566	52711	if( pPager->tempFile ){
52567	52712	rc = sqlite3PagerWrite(pPg);
52568	52713	if( rc ) return rc;
52569	52714	}
52570	52715
		@@ -52635,12 +52780,11 @@
52635	52780
52636	52781	/* For an in-memory database, make sure the original page continues
52637	52782	** to exist, in case the transaction needs to roll back. Use pPgOld
52638	52783	** as the original page since it has already been allocated.
52639	52784	*/
52640		- if( pPager->tempFile ){
52641		- assert( pPgOld );
	52785	+ if( pPager->tempFile && pPgOld ){
52642	52786	sqlite3PcacheMove(pPgOld, origPgno);
52643	52787	sqlite3PagerUnrefNotNull(pPgOld);
52644	52788	}
52645	52789
52646	52790	if( needSyncPgno ){
		@@ -59256,15 +59400,15 @@
59256	59400	flagByte &= ~PTF_LEAF;
59257	59401	pPage->childPtrSize = 4-4*pPage->leaf;
59258	59402	pPage->xCellSize = cellSizePtr;
59259	59403	pBt = pPage->pBt;
59260	59404	if( flagByte==(PTF_LEAFDATA \| PTF_INTKEY) ){
59261		- /* EVIDENCE-OF: R-03640-13415 A value of 5 means the page is an interior
59262		- ** table b-tree page. */
	59405	+ /* EVIDENCE-OF: R-07291-35328 A value of 5 (0x05) means the page is an
	59406	+ ** interior table b-tree page. */
59263	59407	assert( (PTF_LEAFDATA\|PTF_INTKEY)==5 );
59264		- /* EVIDENCE-OF: R-20501-61796 A value of 13 means the page is a leaf
59265		- ** table b-tree page. */
	59408	+ /* EVIDENCE-OF: R-26900-09176 A value of 13 (0x0d) means the page is a
	59409	+ ** leaf table b-tree page. */
59266	59410	assert( (PTF_LEAFDATA\|PTF_INTKEY\|PTF_LEAF)==13 );
59267	59411	pPage->intKey = 1;
59268	59412	if( pPage->leaf ){
59269	59413	pPage->intKeyLeaf = 1;
59270	59414	pPage->xParseCell = btreeParseCellPtr;
		@@ -59274,15 +59418,15 @@
59274	59418	pPage->xParseCell = btreeParseCellPtrNoPayload;
59275	59419	}
59276	59420	pPage->maxLocal = pBt->maxLeaf;
59277	59421	pPage->minLocal = pBt->minLeaf;
59278	59422	}else if( flagByte==PTF_ZERODATA ){
59279		- /* EVIDENCE-OF: R-27225-53936 A value of 2 means the page is an interior
59280		- ** index b-tree page. */
	59423	+ /* EVIDENCE-OF: R-43316-37308 A value of 2 (0x02) means the page is an
	59424	+ ** interior index b-tree page. */
59281	59425	assert( (PTF_ZERODATA)==2 );
59282		- /* EVIDENCE-OF: R-16571-11615 A value of 10 means the page is a leaf
59283		- ** index b-tree page. */
	59426	+ /* EVIDENCE-OF: R-59615-42828 A value of 10 (0x0a) means the page is a
	59427	+ ** leaf index b-tree page. */
59284	59428	assert( (PTF_ZERODATA\|PTF_LEAF)==10 );
59285	59429	pPage->intKey = 0;
59286	59430	pPage->intKeyLeaf = 0;
59287	59431	pPage->xParseCell = btreeParseCellPtrIndex;
59288	59432	pPage->maxLocal = pBt->maxLocal;
		@@ -192850,11 +192994,11 @@
192850	192994	int nArg, /* Number of args */
192851	192995	sqlite3_value *apUnused / Function arguments */
192852	192996	){
192853	192997	assert( nArg==0 );
192854	192998	UNUSED_PARAM2(nArg, apUnused);
192855		- sqlite3_result_text(pCtx, "fts5: 2016-05-09 19:03:42 14e53d0e2f62d82ae1d64a72fd9711548e3bf5ea", -1, SQLITE_TRANSIENT);
	192999	+ sqlite3_result_text(pCtx, "fts5: 2016-05-18 10:57:30 fc49f556e48970561d7ab6a2f24fdd7d9eb81ff2", -1, SQLITE_TRANSIENT);
192856	193000	}
192857	193001
192858	193002	static int fts5Init(sqlite3 *db){
192859	193003	static const sqlite3_module fts5Mod = {
192860	193004	/* iVersion */ 2,
192861	193005

	--- src/sqlite3.c
	+++ src/sqlite3.c
	@@ -363,11 +363,11 @@
363	** [sqlite3_libversion_number()], [sqlite3_sourceid()],
364	** [sqlite_version()] and [sqlite_source_id()].
365	*/
366	#define SQLITE_VERSION "3.13.0"
367	#define SQLITE_VERSION_NUMBER 3013000
368	#define SQLITE_SOURCE_ID "2016-05-09 19:03:42 14e53d0e2f62d82ae1d64a72fd9711548e3bf5ea"
369
370	/*
371	** CAPI3REF: Run-Time Library Version Numbers
372	** KEYWORDS: sqlite3_version, sqlite3_sourceid
373	**
	@@ -10917,11 +10917,11 @@
10917	**
10918	** When doing coverage testing ALWAYS and NEVER are hard-coded to
10919	** be true and false so that the unreachable code they specify will
10920	** not be counted as untested code.
10921	*/
10922	#if defined(SQLITE_COVERAGE_TEST)
10923	# define ALWAYS(X) (1)
10924	# define NEVER(X) (0)
10925	#elif !defined(NDEBUG)
10926	# define ALWAYS(X) ((X)?1:(assert(0),0))
10927	# define NEVER(X) ((X)?(assert(0),1):0)
	@@ -12954,11 +12954,11 @@
12954	*/
12955	struct PgHdr {
12956	sqlite3_pcache_page pPage; / Pcache object page handle */
12957	void pData; / Page data */
12958	void pExtra; / Extra content */
12959	PgHdr pDirty; / Transient list of dirty pages */
12960	Pager pPager; / The pager this page is part of */
12961	Pgno pgno; /* Page number for this page */
12962	#ifdef SQLITE_CHECK_PAGES
12963	u32 pageHash; /* Hash of page content */
12964	#endif
	@@ -12979,15 +12979,14 @@
12979	#define PGHDR_CLEAN 0x001 /* Page not on the PCache.pDirty list */
12980	#define PGHDR_DIRTY 0x002 /* Page is on the PCache.pDirty list */
12981	#define PGHDR_WRITEABLE 0x004 /* Journaled and ready to modify */
12982	#define PGHDR_NEED_SYNC 0x008 /* Fsync the rollback journal before
12983	** writing this page to the database */
12984	#define PGHDR_NEED_READ 0x010 /* Content is unread */
12985	#define PGHDR_DONT_WRITE 0x020 /* Do not write content to disk */
12986	#define PGHDR_MMAP 0x040 /* This is an mmap page object */
12987
12988	#define PGHDR_WAL_APPEND 0x080 /* Appended to wal file */
12989
12990	/* Initialize and shutdown the page cache subsystem */
12991	SQLITE_PRIVATE int sqlite3PcacheInitialize(void);
12992	SQLITE_PRIVATE void sqlite3PcacheShutdown(void);
12993
	@@ -13065,10 +13064,15 @@
13065	** interface is only available if SQLITE_CHECK_PAGES is defined when the
13066	** library is built.
13067	*/
13068	SQLITE_PRIVATE void sqlite3PcacheIterateDirty(PCache pCache, void (xIter)(PgHdr *));
13069	#endif





13070
13071	/* Set and get the suggested cache-size for the specified pager-cache.
13072	**
13073	** If no global maximum is configured, then the system attempts to limit
13074	** the total number of pages cached by purgeable pager-caches to the sum
	@@ -43115,11 +43119,33 @@
43115	** This file implements that page cache.
43116	*/
43117	/* #include "sqliteInt.h" */
43118
43119	/*
43120	** A complete page cache is an instance of this structure.






















43121	*/
43122	struct PCache {
43123	PgHdr pDirty, pDirtyTail; /* List of dirty pages in LRU order */
43124	PgHdr pSynced; / Last synced page in dirty page list */
43125	int nRefSum; /* Sum of ref counts over all pages */
	@@ -43131,10 +43157,99 @@
43131	u8 eCreate; /* eCreate value for for xFetch() */
43132	int (xStress)(void,PgHdr); / Call to try make a page clean */
43133	void pStress; / Argument to xStress */
43134	sqlite3_pcache pCache; / Pluggable cache module */
43135	};

























































































43136
43137	/******************************** Linked List Management ******************/
43138
43139	/* Allowed values for second argument to pcacheManageDirtyList() */
43140	#define PCACHE_DIRTYLIST_REMOVE 1 /* Remove pPage from dirty list */
	@@ -43148,21 +43263,20 @@
43148	** the dirty list. Doing both moves pPage to the front of the dirty list.
43149	*/
43150	static void pcacheManageDirtyList(PgHdr *pPage, u8 addRemove){
43151	PCache *p = pPage->pCache;
43152



43153	if( addRemove & PCACHE_DIRTYLIST_REMOVE ){
43154	assert( pPage->pDirtyNext \|\| pPage==p->pDirtyTail );
43155	assert( pPage->pDirtyPrev \|\| pPage==p->pDirty );
43156
43157	/* Update the PCache1.pSynced variable if necessary. */
43158	if( p->pSynced==pPage ){
43159	PgHdr *pSynced = pPage->pDirtyPrev;
43160	while( pSynced && (pSynced->flags&PGHDR_NEED_SYNC) ){
43161	pSynced = pSynced->pDirtyPrev;
43162	}
43163	p->pSynced = pSynced;
43164	}
43165
43166	if( pPage->pDirtyNext ){
43167	pPage->pDirtyNext->pDirtyPrev = pPage->pDirtyPrev;
43168	}else{
	@@ -43170,14 +43284,19 @@
43170	p->pDirtyTail = pPage->pDirtyPrev;
43171	}
43172	if( pPage->pDirtyPrev ){
43173	pPage->pDirtyPrev->pDirtyNext = pPage->pDirtyNext;
43174	}else{




43175	assert( pPage==p->pDirty );
43176	p->pDirty = pPage->pDirtyNext;
43177	if( p->pDirty==0 && p->bPurgeable ){
43178	assert( p->eCreate==1 );

43179	p->eCreate = 2;
43180	}
43181	}
43182	pPage->pDirtyNext = 0;
43183	pPage->pDirtyPrev = 0;
	@@ -43195,23 +43314,34 @@
43195	assert( p->eCreate==2 );
43196	p->eCreate = 1;
43197	}
43198	}
43199	p->pDirty = pPage;
43200	if( !p->pSynced && 0==(pPage->flags&PGHDR_NEED_SYNC) ){








43201	p->pSynced = pPage;
43202	}
43203	}

43204	}
43205
43206	/*
43207	** Wrapper around the pluggable caches xUnpin method. If the cache is
43208	** being used for an in-memory database, this function is a no-op.
43209	*/
43210	static void pcacheUnpin(PgHdr *p){
43211	if( p->pCache->bPurgeable ){

43212	sqlite3GlobalConfig.pcache2.xUnpin(p->pCache->pCache, p->pPage, 0);

43213	}
43214	}
43215
43216	/*
43217	** Compute the number of pages of cache requested. p->szCache is the
	@@ -43277,10 +43407,11 @@
43277	p->eCreate = 2;
43278	p->xStress = xStress;
43279	p->pStress = pStress;
43280	p->szCache = 100;
43281	p->szSpill = 1;

43282	return sqlite3PcacheSetPageSize(p, szPage);
43283	}
43284
43285	/*
43286	** Change the page size for PCache object. The caller must ensure that there
	@@ -43299,10 +43430,11 @@
43299	if( pCache->pCache ){
43300	sqlite3GlobalConfig.pcache2.xDestroy(pCache->pCache);
43301	}
43302	pCache->pCache = pNew;
43303	pCache->szPage = szPage;

43304	}
43305	return SQLITE_OK;
43306	}
43307
43308	/*
	@@ -43333,15 +43465,17 @@
43333	PCache pCache, / Obtain the page from this cache */
43334	Pgno pgno, /* Page number to obtain */
43335	int createFlag /* If true, create page if it does not exist already */
43336	){
43337	int eCreate;

43338
43339	assert( pCache!=0 );
43340	assert( pCache->pCache!=0 );
43341	assert( createFlag==3 \|\| createFlag==0 );
43342	assert( pgno>0 );

43343
43344	/* eCreate defines what to do if the page does not exist.
43345	** 0 Do not allocate a new page. (createFlag==0)
43346	** 1 Allocate a new page if doing so is inexpensive.
43347	** (createFlag==1 AND bPurgeable AND pDirty)
	@@ -43350,11 +43484,14 @@
43350	*/
43351	eCreate = createFlag & pCache->eCreate;
43352	assert( eCreate==0 \|\| eCreate==1 \|\| eCreate==2 );
43353	assert( createFlag==0 \|\| pCache->eCreate==eCreate );
43354	assert( createFlag==0 \|\| eCreate==1+(!pCache->bPurgeable\|\|!pCache->pDirty) );
43355	return sqlite3GlobalConfig.pcache2.xFetch(pCache->pCache, pgno, eCreate);



43356	}
43357
43358	/*
43359	** If the sqlite3PcacheFetch() routine is unable to allocate a new
43360	** page because no clean pages are available for reuse and the cache
	@@ -43377,11 +43514,15 @@
43377	if( sqlite3PcachePagecount(pCache)>pCache->szSpill ){
43378	/* Find a dirty page to write-out and recycle. First try to find a
43379	** page that does not require a journal-sync (one with PGHDR_NEED_SYNC
43380	** cleared), but if that is not possible settle for any other
43381	** unreferenced dirty page.
43382	*/




43383	for(pPg=pCache->pSynced;
43384	pPg && (pPg->nRef \|\| (pPg->flags&PGHDR_NEED_SYNC));
43385	pPg=pPg->pDirtyPrev
43386	);
43387	pCache->pSynced = pPg;
	@@ -43395,11 +43536,13 @@
43395	"spill page %d making room for %d - cache used: %d/%d",
43396	pPg->pgno, pgno,
43397	sqlite3GlobalConfig.pcache.xPagecount(pCache->pCache),
43398	numberOfCachePages(pCache));
43399	#endif

43400	rc = pCache->xStress(pCache->pStress, pPg);

43401	if( rc!=SQLITE_OK && rc!=SQLITE_BUSY ){
43402	return rc;
43403	}
43404	}
43405	}
	@@ -43455,10 +43598,11 @@
43455	if( !pPgHdr->pPage ){
43456	return pcacheFetchFinishWithInit(pCache, pgno, pPage);
43457	}
43458	pCache->nRefSum++;
43459	pPgHdr->nRef++;

43460	return pPgHdr;
43461	}
43462
43463	/*
43464	** Decrement the reference count on a page. If the page is clean and the
	@@ -43468,12 +43612,15 @@
43468	assert( p->nRef>0 );
43469	p->pCache->nRefSum--;
43470	if( (--p->nRef)==0 ){
43471	if( p->flags&PGHDR_CLEAN ){
43472	pcacheUnpin(p);
43473	}else if( p->pDirtyPrev!=0 ){
43474	/* Move the page to the head of the dirty list. */



43475	pcacheManageDirtyList(p, PCACHE_DIRTYLIST_FRONT);
43476	}
43477	}
43478	}
43479
	@@ -43480,10 +43627,11 @@
43480	/*
43481	** Increase the reference count of a supplied page by 1.
43482	*/
43483	SQLITE_PRIVATE void sqlite3PcacheRef(PgHdr *p){
43484	assert(p->nRef>0);

43485	p->nRef++;
43486	p->pCache->nRefSum++;
43487	}
43488
43489	/*
	@@ -43491,10 +43639,11 @@
43491	** page. This function deletes that reference, so after it returns the
43492	** page pointed to by p is invalid.
43493	*/
43494	SQLITE_PRIVATE void sqlite3PcacheDrop(PgHdr *p){
43495	assert( p->nRef==1 );

43496	if( p->flags&PGHDR_DIRTY ){
43497	pcacheManageDirtyList(p, PCACHE_DIRTYLIST_REMOVE);
43498	}
43499	p->pCache->nRefSum--;
43500	sqlite3GlobalConfig.pcache2.xUnpin(p->pCache->pCache, p->pPage, 1);
	@@ -43504,30 +43653,36 @@
43504	** Make sure the page is marked as dirty. If it isn't dirty already,
43505	** make it so.
43506	*/
43507	SQLITE_PRIVATE void sqlite3PcacheMakeDirty(PgHdr *p){
43508	assert( p->nRef>0 );
43509	if( p->flags & (PGHDR_CLEAN\|PGHDR_DONT_WRITE) ){

43510	p->flags &= ~PGHDR_DONT_WRITE;
43511	if( p->flags & PGHDR_CLEAN ){
43512	p->flags ^= (PGHDR_DIRTY\|PGHDR_CLEAN);

43513	assert( (p->flags & (PGHDR_DIRTY\|PGHDR_CLEAN))==PGHDR_DIRTY );
43514	pcacheManageDirtyList(p, PCACHE_DIRTYLIST_ADD);
43515	}

43516	}
43517	}
43518
43519	/*
43520	** Make sure the page is marked as clean. If it isn't clean already,
43521	** make it so.
43522	*/
43523	SQLITE_PRIVATE void sqlite3PcacheMakeClean(PgHdr *p){
43524	if( (p->flags & PGHDR_DIRTY) ){

43525	assert( (p->flags & PGHDR_CLEAN)==0 );
43526	pcacheManageDirtyList(p, PCACHE_DIRTYLIST_REMOVE);
43527	p->flags &= ~(PGHDR_DIRTY\|PGHDR_NEED_SYNC\|PGHDR_WRITEABLE);
43528	p->flags \|= PGHDR_CLEAN;


43529	if( p->nRef==0 ){
43530	pcacheUnpin(p);
43531	}
43532	}
43533	}
	@@ -43535,10 +43690,11 @@
43535	/*
43536	** Make every page in the cache clean.
43537	*/
43538	SQLITE_PRIVATE void sqlite3PcacheCleanAll(PCache *pCache){
43539	PgHdr *p;

43540	while( (p = pCache->pDirty)!=0 ){
43541	sqlite3PcacheMakeClean(p);
43542	}
43543	}
43544
	@@ -43545,10 +43701,11 @@
43545	/*
43546	** Clear the PGHDR_NEED_SYNC and PGHDR_WRITEABLE flag from all dirty pages.
43547	*/
43548	SQLITE_PRIVATE void sqlite3PcacheClearWritable(PCache *pCache){
43549	PgHdr *p;

43550	for(p=pCache->pDirty; p; p=p->pDirtyNext){
43551	p->flags &= ~(PGHDR_NEED_SYNC\|PGHDR_WRITEABLE);
43552	}
43553	pCache->pSynced = pCache->pDirtyTail;
43554	}
	@@ -43569,10 +43726,12 @@
43569	*/
43570	SQLITE_PRIVATE void sqlite3PcacheMove(PgHdr *p, Pgno newPgno){
43571	PCache *pCache = p->pCache;
43572	assert( p->nRef>0 );
43573	assert( newPgno>0 );


43574	sqlite3GlobalConfig.pcache2.xRekey(pCache->pCache, p->pPage, p->pgno,newPgno);
43575	p->pgno = newPgno;
43576	if( (p->flags&PGHDR_DIRTY) && (p->flags&PGHDR_NEED_SYNC) ){
43577	pcacheManageDirtyList(p, PCACHE_DIRTYLIST_FRONT);
43578	}
	@@ -43589,10 +43748,11 @@
43589	*/
43590	SQLITE_PRIVATE void sqlite3PcacheTruncate(PCache *pCache, Pgno pgno){
43591	if( pCache->pCache ){
43592	PgHdr *p;
43593	PgHdr *pNext;

43594	for(p=pCache->pDirty; p; p=pNext){
43595	pNext = p->pDirtyNext;
43596	/* This routine never gets call with a positive pgno except right
43597	** after sqlite3PcacheCleanAll(). So if there are dirty pages,
43598	** it must be that pgno==0.
	@@ -43619,10 +43779,11 @@
43619	/*
43620	** Close a cache.
43621	*/
43622	SQLITE_PRIVATE void sqlite3PcacheClose(PCache *pCache){
43623	assert( pCache->pCache!=0 );

43624	sqlite3GlobalConfig.pcache2.xDestroy(pCache->pCache);
43625	}
43626
43627	/*
43628	** Discard the contents of the cache.
	@@ -47585,24 +47746,28 @@
47585	}
47586
47587	static int pager_truncate(Pager *pPager, Pgno nPage);
47588
47589	/*
47590	** The write transaction open on the pager passed as the only argument is
47591	** being committed. This function returns true if all dirty pages should
47592	** be flushed to disk, or false otherwise. Pages should be flushed to disk
47593	** unless one of the following is true:
47594	**
47595	** * The db is an in-memory database.
47596	**
47597	** * The db is a temporary database and the db file has not been opened.
47598	**
47599	** * The db is a temporary database and the cache contains less than
47600	** C/4 dirty pages, where C is the configured cache-size.
47601	*/
47602	static int pagerFlushOnCommit(Pager *pPager){



47603	if( pPager->tempFile==0 ) return 1;

47604	if( !isOpen(pPager->fd) ) return 0;
47605	return (sqlite3PCachePercentDirty(pPager->pPCache)>=25);
47606	}
47607
47608	/*
	@@ -47706,11 +47871,11 @@
47706	}
47707	pPager->journalOff = 0;
47708	}else if( pPager->journalMode==PAGER_JOURNALMODE_PERSIST
47709	\|\| (pPager->exclusiveMode && pPager->journalMode!=PAGER_JOURNALMODE_WAL)
47710	){
47711	rc = zeroJournalHdr(pPager, hasMaster);
47712	pPager->journalOff = 0;
47713	}else{
47714	/* This branch may be executed with Pager.journalMode==MEMORY if
47715	** a hot-journal was just rolled back. In this case the journal
47716	** file should be closed and deleted. If this connection writes to
	@@ -47741,16 +47906,18 @@
47741	#endif
47742
47743	sqlite3BitvecDestroy(pPager->pInJournal);
47744	pPager->pInJournal = 0;
47745	pPager->nRec = 0;
47746	if( MEMDB \|\| pagerFlushOnCommit(pPager) ){
47747	sqlite3PcacheCleanAll(pPager->pPCache);
47748	}else{
47749	sqlite3PcacheClearWritable(pPager->pPCache);



47750	}
47751	sqlite3PcacheTruncate(pPager->pPCache, pPager->dbSize);
47752
47753	if( pagerUseWal(pPager) ){
47754	/* Drop the WAL write-lock, if any. Also, if the connection was in
47755	** locking_mode=exclusive mode but is no longer, drop the EXCLUSIVE
47756	** lock held on the database file.
	@@ -48080,11 +48247,10 @@
48080	pPager->doNotSpill \|= SPILLFLAG_ROLLBACK;
48081	rc = sqlite3PagerGet(pPager, pgno, &pPg, 1);
48082	assert( (pPager->doNotSpill & SPILLFLAG_ROLLBACK)!=0 );
48083	pPager->doNotSpill &= ~SPILLFLAG_ROLLBACK;
48084	if( rc!=SQLITE_OK ) return rc;
48085	pPg->flags &= ~PGHDR_NEED_READ;
48086	sqlite3PcacheMakeDirty(pPg);
48087	}
48088	if( pPg ){
48089	/* No page should ever be explicitly rolled back that is in use, except
48090	** for page 1 which is held in use in order to keep the lock on the
	@@ -48094,37 +48260,14 @@
48094	*/
48095	void *pData;
48096	pData = pPg->pData;
48097	memcpy(pData, (u8*)aData, pPager->pageSize);
48098	pPager->xReiniter(pPg);
48099	if( isMainJrnl && (!isSavepnt \|\| *pOffset<=pPager->journalHdr) ){
48100	/* If the contents of this page were just restored from the main
48101	** journal file, then its content must be as they were when the
48102	** transaction was first opened. In this case we can mark the page
48103	** as clean, since there will be no need to write it out to the
48104	** database.
48105	**
48106	** There is one exception to this rule. If the page is being rolled
48107	** back as part of a savepoint (or statement) rollback from an
48108	** unsynced portion of the main journal file, then it is not safe
48109	** to mark the page as clean. This is because marking the page as
48110	** clean will clear the PGHDR_NEED_SYNC flag. Since the page is
48111	** already in the journal file (recorded in Pager.pInJournal) and
48112	** the PGHDR_NEED_SYNC flag is cleared, if the page is written to
48113	** again within this transaction, it will be marked as dirty but
48114	** the PGHDR_NEED_SYNC flag will not be set. It could then potentially
48115	** be written out into the database file before its journal file
48116	** segment is synced. If a crash occurs during or following this,
48117	** database corruption may ensue.
48118	**
48119	** Update: Another exception is for temp files that are not
48120	** in-memory databases. In this case the page may have been dirty
48121	** at the start of the transaction.
48122	*/
48123	assert( !pagerUseWal(pPager) );
48124	if( pPager->tempFile==0 ) sqlite3PcacheMakeClean(pPg);
48125	}
48126	pager_set_pagehash(pPg);
48127
48128	/* If this was page 1, then restore the value of Pager.dbFileVers.
48129	** Do this before any decoding. */
48130	if( pgno==1 ){
	@@ -51731,10 +51874,11 @@
51731	if( !pPager->tempFile && (pPg->flags&PGHDR_DIRTY) && pPager->nSavepoint==0 ){
51732	PAGERTRACE(("DONT_WRITE page %d of %d\n", pPg->pgno, PAGERID(pPager)));
51733	IOTRACE(("CLEAN %p %d\n", pPager, pPg->pgno))
51734	pPg->flags \|= PGHDR_DONT_WRITE;
51735	pPg->flags &= ~PGHDR_WRITEABLE;

51736	pager_set_pagehash(pPg);
51737	}
51738	}
51739
51740	/*
	@@ -51926,21 +52070,21 @@
51926
51927	/* If a prior error occurred, report that error again. */
51928	if( NEVER(pPager->errCode) ) return pPager->errCode;
51929
51930	/* Provide the ability to easily simulate an I/O error during testing */
51931	if( (rc = sqlite3FaultSim(400))!=SQLITE_OK ) return rc;
51932
51933	PAGERTRACE(("DATABASE SYNC: File=%s zMaster=%s nSize=%d\n",
51934	pPager->zFilename, zMaster, pPager->dbSize));
51935
51936	/* If no database changes have been made, return early. */
51937	if( pPager->eState<PAGER_WRITER_CACHEMOD ) return SQLITE_OK;
51938
51939	assert( MEMDB==0 \|\| pPager->tempFile );
51940	assert( isOpen(pPager->fd) \|\| pPager->tempFile );
51941	if( 0==pagerFlushOnCommit(pPager) ){
51942	/* If this is an in-memory db, or no pages have been written to, or this
51943	** function has already been called, it is mostly a no-op. However, any
51944	** backup in progress needs to be restarted. */
51945	sqlite3BackupRestart(pPager->pBackup);
51946	}else{
	@@ -52561,10 +52705,11 @@
52561	assert( assert_pager_state(pPager) );
52562
52563	/* In order to be able to rollback, an in-memory database must journal
52564	** the page we are moving from.
52565	*/

52566	if( pPager->tempFile ){
52567	rc = sqlite3PagerWrite(pPg);
52568	if( rc ) return rc;
52569	}
52570
	@@ -52635,12 +52780,11 @@
52635
52636	/* For an in-memory database, make sure the original page continues
52637	** to exist, in case the transaction needs to roll back. Use pPgOld
52638	** as the original page since it has already been allocated.
52639	*/
52640	if( pPager->tempFile ){
52641	assert( pPgOld );
52642	sqlite3PcacheMove(pPgOld, origPgno);
52643	sqlite3PagerUnrefNotNull(pPgOld);
52644	}
52645
52646	if( needSyncPgno ){
	@@ -59256,15 +59400,15 @@
59256	flagByte &= ~PTF_LEAF;
59257	pPage->childPtrSize = 4-4*pPage->leaf;
59258	pPage->xCellSize = cellSizePtr;
59259	pBt = pPage->pBt;
59260	if( flagByte==(PTF_LEAFDATA \| PTF_INTKEY) ){
59261	/* EVIDENCE-OF: R-03640-13415 A value of 5 means the page is an interior
59262	** table b-tree page. */
59263	assert( (PTF_LEAFDATA\|PTF_INTKEY)==5 );
59264	/* EVIDENCE-OF: R-20501-61796 A value of 13 means the page is a leaf
59265	** table b-tree page. */
59266	assert( (PTF_LEAFDATA\|PTF_INTKEY\|PTF_LEAF)==13 );
59267	pPage->intKey = 1;
59268	if( pPage->leaf ){
59269	pPage->intKeyLeaf = 1;
59270	pPage->xParseCell = btreeParseCellPtr;
	@@ -59274,15 +59418,15 @@
59274	pPage->xParseCell = btreeParseCellPtrNoPayload;
59275	}
59276	pPage->maxLocal = pBt->maxLeaf;
59277	pPage->minLocal = pBt->minLeaf;
59278	}else if( flagByte==PTF_ZERODATA ){
59279	/* EVIDENCE-OF: R-27225-53936 A value of 2 means the page is an interior
59280	** index b-tree page. */
59281	assert( (PTF_ZERODATA)==2 );
59282	/* EVIDENCE-OF: R-16571-11615 A value of 10 means the page is a leaf
59283	** index b-tree page. */
59284	assert( (PTF_ZERODATA\|PTF_LEAF)==10 );
59285	pPage->intKey = 0;
59286	pPage->intKeyLeaf = 0;
59287	pPage->xParseCell = btreeParseCellPtrIndex;
59288	pPage->maxLocal = pBt->maxLocal;
	@@ -192850,11 +192994,11 @@
192850	int nArg, /* Number of args */
192851	sqlite3_value *apUnused / Function arguments */
192852	){
192853	assert( nArg==0 );
192854	UNUSED_PARAM2(nArg, apUnused);
192855	sqlite3_result_text(pCtx, "fts5: 2016-05-09 19:03:42 14e53d0e2f62d82ae1d64a72fd9711548e3bf5ea", -1, SQLITE_TRANSIENT);
192856	}
192857
192858	static int fts5Init(sqlite3 *db){
192859	static const sqlite3_module fts5Mod = {
192860	/* iVersion */ 2,
192861

	--- src/sqlite3.c
	+++ src/sqlite3.c
	@@ -363,11 +363,11 @@
363	** [sqlite3_libversion_number()], [sqlite3_sourceid()],
364	** [sqlite_version()] and [sqlite_source_id()].
365	*/
366	#define SQLITE_VERSION "3.13.0"
367	#define SQLITE_VERSION_NUMBER 3013000
368	#define SQLITE_SOURCE_ID "2016-05-18 10:57:30 fc49f556e48970561d7ab6a2f24fdd7d9eb81ff2"
369
370	/*
371	** CAPI3REF: Run-Time Library Version Numbers
372	** KEYWORDS: sqlite3_version, sqlite3_sourceid
373	**
	@@ -10917,11 +10917,11 @@
10917	**
10918	** When doing coverage testing ALWAYS and NEVER are hard-coded to
10919	** be true and false so that the unreachable code they specify will
10920	** not be counted as untested code.
10921	*/
10922	#if defined(SQLITE_COVERAGE_TEST) \|\| defined(SQLITE_MUTATION_TEST)
10923	# define ALWAYS(X) (1)
10924	# define NEVER(X) (0)
10925	#elif !defined(NDEBUG)
10926	# define ALWAYS(X) ((X)?1:(assert(0),0))
10927	# define NEVER(X) ((X)?(assert(0),1):0)
	@@ -12954,11 +12954,11 @@
12954	*/
12955	struct PgHdr {
12956	sqlite3_pcache_page pPage; / Pcache object page handle */
12957	void pData; / Page data */
12958	void pExtra; / Extra content */
12959	PgHdr pDirty; / Transient list of dirty sorted by pgno */
12960	Pager pPager; / The pager this page is part of */
12961	Pgno pgno; /* Page number for this page */
12962	#ifdef SQLITE_CHECK_PAGES
12963	u32 pageHash; /* Hash of page content */
12964	#endif
	@@ -12979,15 +12979,14 @@
12979	#define PGHDR_CLEAN 0x001 /* Page not on the PCache.pDirty list */
12980	#define PGHDR_DIRTY 0x002 /* Page is on the PCache.pDirty list */
12981	#define PGHDR_WRITEABLE 0x004 /* Journaled and ready to modify */
12982	#define PGHDR_NEED_SYNC 0x008 /* Fsync the rollback journal before
12983	** writing this page to the database */
12984	#define PGHDR_DONT_WRITE 0x010 /* Do not write content to disk */
12985	#define PGHDR_MMAP 0x020 /* This is an mmap page object */

12986
12987	#define PGHDR_WAL_APPEND 0x040 /* Appended to wal file */
12988
12989	/* Initialize and shutdown the page cache subsystem */
12990	SQLITE_PRIVATE int sqlite3PcacheInitialize(void);
12991	SQLITE_PRIVATE void sqlite3PcacheShutdown(void);
12992
	@@ -13065,10 +13064,15 @@
13064	** interface is only available if SQLITE_CHECK_PAGES is defined when the
13065	** library is built.
13066	*/
13067	SQLITE_PRIVATE void sqlite3PcacheIterateDirty(PCache pCache, void (xIter)(PgHdr *));
13068	#endif
13069
13070	#if defined(SQLITE_DEBUG)
13071	/* Check invariants on a PgHdr object */
13072	SQLITE_PRIVATE int sqlite3PcachePageSanity(PgHdr*);
13073	#endif
13074
13075	/* Set and get the suggested cache-size for the specified pager-cache.
13076	**
13077	** If no global maximum is configured, then the system attempts to limit
13078	** the total number of pages cached by purgeable pager-caches to the sum
	@@ -43115,11 +43119,33 @@
43119	** This file implements that page cache.
43120	*/
43121	/* #include "sqliteInt.h" */
43122
43123	/*
43124	** A complete page cache is an instance of this structure. Every
43125	** entry in the cache holds a single page of the database file. The
43126	** btree layer only operates on the cached copy of the database pages.
43127	**
43128	** A page cache entry is "clean" if it exactly matches what is currently
43129	** on disk. A page is "dirty" if it has been modified and needs to be
43130	** persisted to disk.
43131	**
43132	** pDirty, pDirtyTail, pSynced:
43133	** All dirty pages are linked into the doubly linked list using
43134	** PgHdr.pDirtyNext and pDirtyPrev. The list is maintained in LRU order
43135	** such that p was added to the list more recently than p->pDirtyNext.
43136	** PCache.pDirty points to the first (newest) element in the list and
43137	** pDirtyTail to the last (oldest).
43138	**
43139	** The PCache.pSynced variable is used to optimize searching for a dirty
43140	** page to eject from the cache mid-transaction. It is better to eject
43141	** a page that does not require a journal sync than one that does.
43142	** Therefore, pSynced is maintained to that it almost always points
43143	** to either the oldest page in the pDirty/pDirtyTail list that has a
43144	** clear PGHDR_NEED_SYNC flag or to a page that is older than this one
43145	** (so that the right page to eject can be found by following pDirtyPrev
43146	** pointers).
43147	*/
43148	struct PCache {
43149	PgHdr pDirty, pDirtyTail; /* List of dirty pages in LRU order */
43150	PgHdr pSynced; / Last synced page in dirty page list */
43151	int nRefSum; /* Sum of ref counts over all pages */
	@@ -43131,10 +43157,99 @@
43157	u8 eCreate; /* eCreate value for for xFetch() */
43158	int (xStress)(void,PgHdr); / Call to try make a page clean */
43159	void pStress; / Argument to xStress */
43160	sqlite3_pcache pCache; / Pluggable cache module */
43161	};
43162
43163	/******************************** Test and Debug Logic ********************/
43164	/*
43165	** Debug tracing macros. Enable by by changing the "0" to "1" and
43166	** recompiling.
43167	**
43168	** When sqlite3PcacheTrace is 1, single line trace messages are issued.
43169	** When sqlite3PcacheTrace is 2, a dump of the pcache showing all cache entries
43170	** is displayed for many operations, resulting in a lot of output.
43171	*/
43172	#if defined(SQLITE_DEBUG) && 0
43173	int sqlite3PcacheTrace = 2; /* 0: off 1: simple 2: cache dumps */
43174	int sqlite3PcacheMxDump = 9999; /* Max cache entries for pcacheDump() */
43175	# define pcacheTrace(X) if(sqlite3PcacheTrace){sqlite3DebugPrintf X;}
43176	void pcacheDump(PCache *pCache){
43177	int N;
43178	int i, j;
43179	sqlite3_pcache_page *pLower;
43180	PgHdr *pPg;
43181	unsigned char *a;
43182
43183	if( sqlite3PcacheTrace<2 ) return;
43184	if( pCache->pCache==0 ) return;
43185	N = sqlite3PcachePagecount(pCache);
43186	if( N>sqlite3PcacheMxDump ) N = sqlite3PcacheMxDump;
43187	for(i=1; i<=N; i++){
43188	pLower = sqlite3GlobalConfig.pcache2.xFetch(pCache->pCache, i, 0);
43189	if( pLower==0 ) continue;
43190	pPg = (PgHdr*)pLower->pExtra;
43191	printf("%3d: nRef %2d flgs %02x data ", i, pPg->nRef, pPg->flags);
43192	a = (unsigned char *)pLower->pBuf;
43193	for(j=0; j<12; j++) printf("%02x", a[j]);
43194	printf("\n");
43195	if( pPg->pPage==0 ){
43196	sqlite3GlobalConfig.pcache2.xUnpin(pCache->pCache, pLower, 0);
43197	}
43198	}
43199	}
43200	#else
43201	# define pcacheTrace(X)
43202	# define pcacheDump(X)
43203	#endif
43204
43205	/*
43206	** Check invariants on a PgHdr entry. Return true if everything is OK.
43207	** Return false if any invariant is violated.
43208	**
43209	** This routine is for use inside of assert() statements only. For
43210	** example:
43211	**
43212	** assert( sqlite3PcachePageSanity(pPg) );
43213	*/
43214	#if SQLITE_DEBUG
43215	SQLITE_PRIVATE int sqlite3PcachePageSanity(PgHdr *pPg){
43216	PCache *pCache;
43217	assert( pPg!=0 );
43218	assert( pPg->pgno>0 ); /* Page number is 1 or more */
43219	pCache = pPg->pCache;
43220	assert( pCache!=0 ); /* Every page has an associated PCache */
43221	if( pPg->flags & PGHDR_CLEAN ){
43222	assert( (pPg->flags & PGHDR_DIRTY)==0 );/* Cannot be both CLEAN and DIRTY */
43223	assert( pCache->pDirty!=pPg ); /* CLEAN pages not on dirty list */
43224	assert( pCache->pDirtyTail!=pPg );
43225	}
43226	/* WRITEABLE pages must also be DIRTY */
43227	if( pPg->flags & PGHDR_WRITEABLE ){
43228	assert( pPg->flags & PGHDR_DIRTY ); /* WRITEABLE implies DIRTY */
43229	}
43230	/* NEED_SYNC can be set independently of WRITEABLE. This can happen,
43231	** for example, when using the sqlite3PagerDontWrite() optimization:
43232	** (1) Page X is journalled, and gets WRITEABLE and NEED_SEEK.
43233	** (2) Page X moved to freelist, WRITEABLE is cleared
43234	** (3) Page X reused, WRITEABLE is set again
43235	** If NEED_SYNC had been cleared in step 2, then it would not be reset
43236	** in step 3, and page might be written into the database without first
43237	** syncing the rollback journal, which might cause corruption on a power
43238	** loss.
43239	**
43240	** Another example is when the database page size is smaller than the
43241	** disk sector size. When any page of a sector is journalled, all pages
43242	** in that sector are marked NEED_SYNC even if they are still CLEAN, just
43243	** in case they are later modified, since all pages in the same sector
43244	** must be journalled and synced before any of those pages can be safely
43245	** written.
43246	*/
43247	return 1;
43248	}
43249	#endif /* SQLITE_DEBUG */
43250
43251
43252	/******************************** Linked List Management ******************/
43253
43254	/* Allowed values for second argument to pcacheManageDirtyList() */
43255	#define PCACHE_DIRTYLIST_REMOVE 1 /* Remove pPage from dirty list */
	@@ -43148,21 +43263,20 @@
43263	** the dirty list. Doing both moves pPage to the front of the dirty list.
43264	*/
43265	static void pcacheManageDirtyList(PgHdr *pPage, u8 addRemove){
43266	PCache *p = pPage->pCache;
43267
43268	pcacheTrace(("%p.DIRTYLIST.%s %d\n", p,
43269	addRemove==1 ? "REMOVE" : addRemove==2 ? "ADD" : "FRONT",
43270	pPage->pgno));
43271	if( addRemove & PCACHE_DIRTYLIST_REMOVE ){
43272	assert( pPage->pDirtyNext \|\| pPage==p->pDirtyTail );
43273	assert( pPage->pDirtyPrev \|\| pPage==p->pDirty );
43274
43275	/* Update the PCache1.pSynced variable if necessary. */
43276	if( p->pSynced==pPage ){
43277	p->pSynced = pPage->pDirtyPrev;




43278	}
43279
43280	if( pPage->pDirtyNext ){
43281	pPage->pDirtyNext->pDirtyPrev = pPage->pDirtyPrev;
43282	}else{
	@@ -43170,14 +43284,19 @@
43284	p->pDirtyTail = pPage->pDirtyPrev;
43285	}
43286	if( pPage->pDirtyPrev ){
43287	pPage->pDirtyPrev->pDirtyNext = pPage->pDirtyNext;
43288	}else{
43289	/* If there are now no dirty pages in the cache, set eCreate to 2.
43290	** This is an optimization that allows sqlite3PcacheFetch() to skip
43291	** searching for a dirty page to eject from the cache when it might
43292	** otherwise have to. */
43293	assert( pPage==p->pDirty );
43294	p->pDirty = pPage->pDirtyNext;
43295	assert( p->bPurgeable \|\| p->eCreate==2 );
43296	if( p->pDirty==0 ){ /OPTIMIZATION-IF-TRUE/
43297	assert( p->bPurgeable==0 \|\| p->eCreate==1 );
43298	p->eCreate = 2;
43299	}
43300	}
43301	pPage->pDirtyNext = 0;
43302	pPage->pDirtyPrev = 0;
	@@ -43195,23 +43314,34 @@
43314	assert( p->eCreate==2 );
43315	p->eCreate = 1;
43316	}
43317	}
43318	p->pDirty = pPage;
43319
43320	/* If pSynced is NULL and this page has a clear NEED_SYNC flag, set
43321	** pSynced to point to it. Checking the NEED_SYNC flag is an
43322	** optimization, as if pSynced points to a page with the NEED_SYNC
43323	** flag set sqlite3PcacheFetchStress() searches through all newer
43324	** entries of the dirty-list for a page with NEED_SYNC clear anyway. */
43325	if( !p->pSynced
43326	&& 0==(pPage->flags&PGHDR_NEED_SYNC) /OPTIMIZATION-IF-FALSE/
43327	){
43328	p->pSynced = pPage;
43329	}
43330	}
43331	pcacheDump(p);
43332	}
43333
43334	/*
43335	** Wrapper around the pluggable caches xUnpin method. If the cache is
43336	** being used for an in-memory database, this function is a no-op.
43337	*/
43338	static void pcacheUnpin(PgHdr *p){
43339	if( p->pCache->bPurgeable ){
43340	pcacheTrace(("%p.UNPIN %d\n", p->pCache, p->pgno));
43341	sqlite3GlobalConfig.pcache2.xUnpin(p->pCache->pCache, p->pPage, 0);
43342	pcacheDump(p->pCache);
43343	}
43344	}
43345
43346	/*
43347	** Compute the number of pages of cache requested. p->szCache is the
	@@ -43277,10 +43407,11 @@
43407	p->eCreate = 2;
43408	p->xStress = xStress;
43409	p->pStress = pStress;
43410	p->szCache = 100;
43411	p->szSpill = 1;
43412	pcacheTrace(("%p.OPEN szPage %d bPurgeable %d\n",p,szPage,bPurgeable));
43413	return sqlite3PcacheSetPageSize(p, szPage);
43414	}
43415
43416	/*
43417	** Change the page size for PCache object. The caller must ensure that there
	@@ -43299,10 +43430,11 @@
43430	if( pCache->pCache ){
43431	sqlite3GlobalConfig.pcache2.xDestroy(pCache->pCache);
43432	}
43433	pCache->pCache = pNew;
43434	pCache->szPage = szPage;
43435	pcacheTrace(("%p.PAGESIZE %d\n",pCache,szPage));
43436	}
43437	return SQLITE_OK;
43438	}
43439
43440	/*
	@@ -43333,15 +43465,17 @@
43465	PCache pCache, / Obtain the page from this cache */
43466	Pgno pgno, /* Page number to obtain */
43467	int createFlag /* If true, create page if it does not exist already */
43468	){
43469	int eCreate;
43470	sqlite3_pcache_page *pRes;
43471
43472	assert( pCache!=0 );
43473	assert( pCache->pCache!=0 );
43474	assert( createFlag==3 \|\| createFlag==0 );
43475	assert( pgno>0 );
43476	assert( pCache->eCreate==((pCache->bPurgeable && pCache->pDirty) ? 1 : 2) );
43477
43478	/* eCreate defines what to do if the page does not exist.
43479	** 0 Do not allocate a new page. (createFlag==0)
43480	** 1 Allocate a new page if doing so is inexpensive.
43481	** (createFlag==1 AND bPurgeable AND pDirty)
	@@ -43350,11 +43484,14 @@
43484	*/
43485	eCreate = createFlag & pCache->eCreate;
43486	assert( eCreate==0 \|\| eCreate==1 \|\| eCreate==2 );
43487	assert( createFlag==0 \|\| pCache->eCreate==eCreate );
43488	assert( createFlag==0 \|\| eCreate==1+(!pCache->bPurgeable\|\|!pCache->pDirty) );
43489	pRes = sqlite3GlobalConfig.pcache2.xFetch(pCache->pCache, pgno, eCreate);
43490	pcacheTrace(("%p.FETCH %d%s (result: %p)\n",pCache,pgno,
43491	createFlag?" create":"",pRes));
43492	return pRes;
43493	}
43494
43495	/*
43496	** If the sqlite3PcacheFetch() routine is unable to allocate a new
43497	** page because no clean pages are available for reuse and the cache
	@@ -43377,11 +43514,15 @@
43514	if( sqlite3PcachePagecount(pCache)>pCache->szSpill ){
43515	/* Find a dirty page to write-out and recycle. First try to find a
43516	** page that does not require a journal-sync (one with PGHDR_NEED_SYNC
43517	** cleared), but if that is not possible settle for any other
43518	** unreferenced dirty page.
43519	**
43520	** If the LRU page in the dirty list that has a clear PGHDR_NEED_SYNC
43521	** flag is currently referenced, then the following may leave pSynced
43522	** set incorrectly (pointing to other than the LRU page with NEED_SYNC
43523	** cleared). This is Ok, as pSynced is just an optimization. */
43524	for(pPg=pCache->pSynced;
43525	pPg && (pPg->nRef \|\| (pPg->flags&PGHDR_NEED_SYNC));
43526	pPg=pPg->pDirtyPrev
43527	);
43528	pCache->pSynced = pPg;
	@@ -43395,11 +43536,13 @@
43536	"spill page %d making room for %d - cache used: %d/%d",
43537	pPg->pgno, pgno,
43538	sqlite3GlobalConfig.pcache.xPagecount(pCache->pCache),
43539	numberOfCachePages(pCache));
43540	#endif
43541	pcacheTrace(("%p.SPILL %d\n",pCache,pPg->pgno));
43542	rc = pCache->xStress(pCache->pStress, pPg);
43543	pcacheDump(pCache);
43544	if( rc!=SQLITE_OK && rc!=SQLITE_BUSY ){
43545	return rc;
43546	}
43547	}
43548	}
	@@ -43455,10 +43598,11 @@
43598	if( !pPgHdr->pPage ){
43599	return pcacheFetchFinishWithInit(pCache, pgno, pPage);
43600	}
43601	pCache->nRefSum++;
43602	pPgHdr->nRef++;
43603	assert( sqlite3PcachePageSanity(pPgHdr) );
43604	return pPgHdr;
43605	}
43606
43607	/*
43608	** Decrement the reference count on a page. If the page is clean and the
	@@ -43468,12 +43612,15 @@
43612	assert( p->nRef>0 );
43613	p->pCache->nRefSum--;
43614	if( (--p->nRef)==0 ){
43615	if( p->flags&PGHDR_CLEAN ){
43616	pcacheUnpin(p);
43617	}else if( p->pDirtyPrev!=0 ){ /OPTIMIZATION-IF-FALSE/
43618	/* Move the page to the head of the dirty list. If p->pDirtyPrev==0,
43619	** then page p is already at the head of the dirty list and the
43620	** following call would be a no-op. Hence the OPTIMIZATION-IF-FALSE
43621	** tag above. */
43622	pcacheManageDirtyList(p, PCACHE_DIRTYLIST_FRONT);
43623	}
43624	}
43625	}
43626
	@@ -43480,10 +43627,11 @@
43627	/*
43628	** Increase the reference count of a supplied page by 1.
43629	*/
43630	SQLITE_PRIVATE void sqlite3PcacheRef(PgHdr *p){
43631	assert(p->nRef>0);
43632	assert( sqlite3PcachePageSanity(p) );
43633	p->nRef++;
43634	p->pCache->nRefSum++;
43635	}
43636
43637	/*
	@@ -43491,10 +43639,11 @@
43639	** page. This function deletes that reference, so after it returns the
43640	** page pointed to by p is invalid.
43641	*/
43642	SQLITE_PRIVATE void sqlite3PcacheDrop(PgHdr *p){
43643	assert( p->nRef==1 );
43644	assert( sqlite3PcachePageSanity(p) );
43645	if( p->flags&PGHDR_DIRTY ){
43646	pcacheManageDirtyList(p, PCACHE_DIRTYLIST_REMOVE);
43647	}
43648	p->pCache->nRefSum--;
43649	sqlite3GlobalConfig.pcache2.xUnpin(p->pCache->pCache, p->pPage, 1);
	@@ -43504,30 +43653,36 @@
43653	** Make sure the page is marked as dirty. If it isn't dirty already,
43654	** make it so.
43655	*/
43656	SQLITE_PRIVATE void sqlite3PcacheMakeDirty(PgHdr *p){
43657	assert( p->nRef>0 );
43658	assert( sqlite3PcachePageSanity(p) );
43659	if( p->flags & (PGHDR_CLEAN\|PGHDR_DONT_WRITE) ){ /OPTIMIZATION-IF-FALSE/
43660	p->flags &= ~PGHDR_DONT_WRITE;
43661	if( p->flags & PGHDR_CLEAN ){
43662	p->flags ^= (PGHDR_DIRTY\|PGHDR_CLEAN);
43663	pcacheTrace(("%p.DIRTY %d\n",p->pCache,p->pgno));
43664	assert( (p->flags & (PGHDR_DIRTY\|PGHDR_CLEAN))==PGHDR_DIRTY );
43665	pcacheManageDirtyList(p, PCACHE_DIRTYLIST_ADD);
43666	}
43667	assert( sqlite3PcachePageSanity(p) );
43668	}
43669	}
43670
43671	/*
43672	** Make sure the page is marked as clean. If it isn't clean already,
43673	** make it so.
43674	*/
43675	SQLITE_PRIVATE void sqlite3PcacheMakeClean(PgHdr *p){
43676	assert( sqlite3PcachePageSanity(p) );
43677	if( ALWAYS((p->flags & PGHDR_DIRTY)!=0) ){
43678	assert( (p->flags & PGHDR_CLEAN)==0 );
43679	pcacheManageDirtyList(p, PCACHE_DIRTYLIST_REMOVE);
43680	p->flags &= ~(PGHDR_DIRTY\|PGHDR_NEED_SYNC\|PGHDR_WRITEABLE);
43681	p->flags \|= PGHDR_CLEAN;
43682	pcacheTrace(("%p.CLEAN %d\n",p->pCache,p->pgno));
43683	assert( sqlite3PcachePageSanity(p) );
43684	if( p->nRef==0 ){
43685	pcacheUnpin(p);
43686	}
43687	}
43688	}
	@@ -43535,10 +43690,11 @@
43690	/*
43691	** Make every page in the cache clean.
43692	*/
43693	SQLITE_PRIVATE void sqlite3PcacheCleanAll(PCache *pCache){
43694	PgHdr *p;
43695	pcacheTrace(("%p.CLEAN-ALL\n",pCache));
43696	while( (p = pCache->pDirty)!=0 ){
43697	sqlite3PcacheMakeClean(p);
43698	}
43699	}
43700
	@@ -43545,10 +43701,11 @@
43701	/*
43702	** Clear the PGHDR_NEED_SYNC and PGHDR_WRITEABLE flag from all dirty pages.
43703	*/
43704	SQLITE_PRIVATE void sqlite3PcacheClearWritable(PCache *pCache){
43705	PgHdr *p;
43706	pcacheTrace(("%p.CLEAR-WRITEABLE\n",pCache));
43707	for(p=pCache->pDirty; p; p=p->pDirtyNext){
43708	p->flags &= ~(PGHDR_NEED_SYNC\|PGHDR_WRITEABLE);
43709	}
43710	pCache->pSynced = pCache->pDirtyTail;
43711	}
	@@ -43569,10 +43726,12 @@
43726	*/
43727	SQLITE_PRIVATE void sqlite3PcacheMove(PgHdr *p, Pgno newPgno){
43728	PCache *pCache = p->pCache;
43729	assert( p->nRef>0 );
43730	assert( newPgno>0 );
43731	assert( sqlite3PcachePageSanity(p) );
43732	pcacheTrace(("%p.MOVE %d -> %d\n",pCache,p->pgno,newPgno));
43733	sqlite3GlobalConfig.pcache2.xRekey(pCache->pCache, p->pPage, p->pgno,newPgno);
43734	p->pgno = newPgno;
43735	if( (p->flags&PGHDR_DIRTY) && (p->flags&PGHDR_NEED_SYNC) ){
43736	pcacheManageDirtyList(p, PCACHE_DIRTYLIST_FRONT);
43737	}
	@@ -43589,10 +43748,11 @@
43748	*/
43749	SQLITE_PRIVATE void sqlite3PcacheTruncate(PCache *pCache, Pgno pgno){
43750	if( pCache->pCache ){
43751	PgHdr *p;
43752	PgHdr *pNext;
43753	pcacheTrace(("%p.TRUNCATE %d\n",pCache,pgno));
43754	for(p=pCache->pDirty; p; p=pNext){
43755	pNext = p->pDirtyNext;
43756	/* This routine never gets call with a positive pgno except right
43757	** after sqlite3PcacheCleanAll(). So if there are dirty pages,
43758	** it must be that pgno==0.
	@@ -43619,10 +43779,11 @@
43779	/*
43780	** Close a cache.
43781	*/
43782	SQLITE_PRIVATE void sqlite3PcacheClose(PCache *pCache){
43783	assert( pCache->pCache!=0 );
43784	pcacheTrace(("%p.CLOSE\n",pCache));
43785	sqlite3GlobalConfig.pcache2.xDestroy(pCache->pCache);
43786	}
43787
43788	/*
43789	** Discard the contents of the cache.
	@@ -47585,24 +47746,28 @@
47746	}
47747
47748	static int pager_truncate(Pager *pPager, Pgno nPage);
47749
47750	/*
47751	** The write transaction open on pPager is being committed (bCommit==1)
47752	** or rolled back (bCommit==0).
47753	**
47754	** Return TRUE if and only if all dirty pages should be flushed to disk.
47755	**
47756	** Rules:
47757	**
47758	** * For non-TEMP databases, always sync to disk. This is necessary
47759	** for transactions to be durable.
47760	**
47761	** * Sync TEMP database only on a COMMIT (not a ROLLBACK) when the backing
47762	** file has been created already (via a spill on pagerStress()) and
47763	** when the number of dirty pages in memory exceeds 25% of the total
47764	** cache size.
47765	*/
47766	static int pagerFlushOnCommit(Pager *pPager, int bCommit){
47767	if( pPager->tempFile==0 ) return 1;
47768	if( !bCommit ) return 0;
47769	if( !isOpen(pPager->fd) ) return 0;
47770	return (sqlite3PCachePercentDirty(pPager->pPCache)>=25);
47771	}
47772
47773	/*
	@@ -47706,11 +47871,11 @@
47871	}
47872	pPager->journalOff = 0;
47873	}else if( pPager->journalMode==PAGER_JOURNALMODE_PERSIST
47874	\|\| (pPager->exclusiveMode && pPager->journalMode!=PAGER_JOURNALMODE_WAL)
47875	){
47876	rc = zeroJournalHdr(pPager, hasMaster\|\|pPager->tempFile);
47877	pPager->journalOff = 0;
47878	}else{
47879	/* This branch may be executed with Pager.journalMode==MEMORY if
47880	** a hot-journal was just rolled back. In this case the journal
47881	** file should be closed and deleted. If this connection writes to
	@@ -47741,16 +47906,18 @@
47906	#endif
47907
47908	sqlite3BitvecDestroy(pPager->pInJournal);
47909	pPager->pInJournal = 0;
47910	pPager->nRec = 0;
47911	if( rc==SQLITE_OK ){
47912	if( pagerFlushOnCommit(pPager, bCommit) ){
47913	sqlite3PcacheCleanAll(pPager->pPCache);
47914	}else{
47915	sqlite3PcacheClearWritable(pPager->pPCache);
47916	}
47917	sqlite3PcacheTruncate(pPager->pPCache, pPager->dbSize);
47918	}

47919
47920	if( pagerUseWal(pPager) ){
47921	/* Drop the WAL write-lock, if any. Also, if the connection was in
47922	** locking_mode=exclusive mode but is no longer, drop the EXCLUSIVE
47923	** lock held on the database file.
	@@ -48080,11 +48247,10 @@
48247	pPager->doNotSpill \|= SPILLFLAG_ROLLBACK;
48248	rc = sqlite3PagerGet(pPager, pgno, &pPg, 1);
48249	assert( (pPager->doNotSpill & SPILLFLAG_ROLLBACK)!=0 );
48250	pPager->doNotSpill &= ~SPILLFLAG_ROLLBACK;
48251	if( rc!=SQLITE_OK ) return rc;

48252	sqlite3PcacheMakeDirty(pPg);
48253	}
48254	if( pPg ){
48255	/* No page should ever be explicitly rolled back that is in use, except
48256	** for page 1 which is held in use in order to keep the lock on the
	@@ -48094,37 +48260,14 @@
48260	*/
48261	void *pData;
48262	pData = pPg->pData;
48263	memcpy(pData, (u8*)aData, pPager->pageSize);
48264	pPager->xReiniter(pPg);
48265	/* It used to be that sqlite3PcacheMakeClean(pPg) was called here. But
48266	** that call was dangerous and had no detectable benefit since the cache
48267	** is normally cleaned by sqlite3PcacheCleanAll() after rollback and so
48268	** has been removed. */























48269	pager_set_pagehash(pPg);
48270
48271	/* If this was page 1, then restore the value of Pager.dbFileVers.
48272	** Do this before any decoding. */
48273	if( pgno==1 ){
	@@ -51731,10 +51874,11 @@
51874	if( !pPager->tempFile && (pPg->flags&PGHDR_DIRTY) && pPager->nSavepoint==0 ){
51875	PAGERTRACE(("DONT_WRITE page %d of %d\n", pPg->pgno, PAGERID(pPager)));
51876	IOTRACE(("CLEAN %p %d\n", pPager, pPg->pgno))
51877	pPg->flags \|= PGHDR_DONT_WRITE;
51878	pPg->flags &= ~PGHDR_WRITEABLE;
51879	testcase( pPg->flags & PGHDR_NEED_SYNC );
51880	pager_set_pagehash(pPg);
51881	}
51882	}
51883
51884	/*
	@@ -51926,21 +52070,21 @@
52070
52071	/* If a prior error occurred, report that error again. */
52072	if( NEVER(pPager->errCode) ) return pPager->errCode;
52073
52074	/* Provide the ability to easily simulate an I/O error during testing */
52075	if( sqlite3FaultSim(400) ) return SQLITE_IOERR;
52076
52077	PAGERTRACE(("DATABASE SYNC: File=%s zMaster=%s nSize=%d\n",
52078	pPager->zFilename, zMaster, pPager->dbSize));
52079
52080	/* If no database changes have been made, return early. */
52081	if( pPager->eState<PAGER_WRITER_CACHEMOD ) return SQLITE_OK;
52082
52083	assert( MEMDB==0 \|\| pPager->tempFile );
52084	assert( isOpen(pPager->fd) \|\| pPager->tempFile );
52085	if( 0==pagerFlushOnCommit(pPager, 1) ){
52086	/* If this is an in-memory db, or no pages have been written to, or this
52087	** function has already been called, it is mostly a no-op. However, any
52088	** backup in progress needs to be restarted. */
52089	sqlite3BackupRestart(pPager->pBackup);
52090	}else{
	@@ -52561,10 +52705,11 @@
52705	assert( assert_pager_state(pPager) );
52706
52707	/* In order to be able to rollback, an in-memory database must journal
52708	** the page we are moving from.
52709	*/
52710	assert( pPager->tempFile \|\| !MEMDB );
52711	if( pPager->tempFile ){
52712	rc = sqlite3PagerWrite(pPg);
52713	if( rc ) return rc;
52714	}
52715
	@@ -52635,12 +52780,11 @@
52780
52781	/* For an in-memory database, make sure the original page continues
52782	** to exist, in case the transaction needs to roll back. Use pPgOld
52783	** as the original page since it has already been allocated.
52784	*/
52785	if( pPager->tempFile && pPgOld ){

52786	sqlite3PcacheMove(pPgOld, origPgno);
52787	sqlite3PagerUnrefNotNull(pPgOld);
52788	}
52789
52790	if( needSyncPgno ){
	@@ -59256,15 +59400,15 @@
59400	flagByte &= ~PTF_LEAF;
59401	pPage->childPtrSize = 4-4*pPage->leaf;
59402	pPage->xCellSize = cellSizePtr;
59403	pBt = pPage->pBt;
59404	if( flagByte==(PTF_LEAFDATA \| PTF_INTKEY) ){
59405	/* EVIDENCE-OF: R-07291-35328 A value of 5 (0x05) means the page is an
59406	** interior table b-tree page. */
59407	assert( (PTF_LEAFDATA\|PTF_INTKEY)==5 );
59408	/* EVIDENCE-OF: R-26900-09176 A value of 13 (0x0d) means the page is a
59409	** leaf table b-tree page. */
59410	assert( (PTF_LEAFDATA\|PTF_INTKEY\|PTF_LEAF)==13 );
59411	pPage->intKey = 1;
59412	if( pPage->leaf ){
59413	pPage->intKeyLeaf = 1;
59414	pPage->xParseCell = btreeParseCellPtr;
	@@ -59274,15 +59418,15 @@
59418	pPage->xParseCell = btreeParseCellPtrNoPayload;
59419	}
59420	pPage->maxLocal = pBt->maxLeaf;
59421	pPage->minLocal = pBt->minLeaf;
59422	}else if( flagByte==PTF_ZERODATA ){
59423	/* EVIDENCE-OF: R-43316-37308 A value of 2 (0x02) means the page is an
59424	** interior index b-tree page. */
59425	assert( (PTF_ZERODATA)==2 );
59426	/* EVIDENCE-OF: R-59615-42828 A value of 10 (0x0a) means the page is a
59427	** leaf index b-tree page. */
59428	assert( (PTF_ZERODATA\|PTF_LEAF)==10 );
59429	pPage->intKey = 0;
59430	pPage->intKeyLeaf = 0;
59431	pPage->xParseCell = btreeParseCellPtrIndex;
59432	pPage->maxLocal = pBt->maxLocal;
	@@ -192850,11 +192994,11 @@
192994	int nArg, /* Number of args */
192995	sqlite3_value *apUnused / Function arguments */
192996	){
192997	assert( nArg==0 );
192998	UNUSED_PARAM2(nArg, apUnused);
192999	sqlite3_result_text(pCtx, "fts5: 2016-05-18 10:57:30 fc49f556e48970561d7ab6a2f24fdd7d9eb81ff2", -1, SQLITE_TRANSIENT);
193000	}
193001
193002	static int fts5Init(sqlite3 *db){
193003	static const sqlite3_module fts5Mod = {
193004	/* iVersion */ 2,
193005

M src/sqlite3.h

+1 -1

		--- src/sqlite3.h
		+++ src/sqlite3.h
		@@ -111,11 +111,11 @@
111	111	** [sqlite3_libversion_number()], [sqlite3_sourceid()],
112	112	** [sqlite_version()] and [sqlite_source_id()].
113	113	*/
114	114	#define SQLITE_VERSION "3.13.0"
115	115	#define SQLITE_VERSION_NUMBER 3013000
116		-#define SQLITE_SOURCE_ID "2016-05-09 19:03:42 14e53d0e2f62d82ae1d64a72fd9711548e3bf5ea"
	116	+#define SQLITE_SOURCE_ID "2016-05-18 10:57:30 fc49f556e48970561d7ab6a2f24fdd7d9eb81ff2"
117	117
118	118	/*
119	119	** CAPI3REF: Run-Time Library Version Numbers
120	120	** KEYWORDS: sqlite3_version, sqlite3_sourceid
121	121	**
122	122
123	123	ADDED www/encryptedrepos.wiki

	--- src/sqlite3.h
	+++ src/sqlite3.h
	@@ -111,11 +111,11 @@
111	** [sqlite3_libversion_number()], [sqlite3_sourceid()],
112	** [sqlite_version()] and [sqlite_source_id()].
113	*/
114	#define SQLITE_VERSION "3.13.0"
115	#define SQLITE_VERSION_NUMBER 3013000
116	#define SQLITE_SOURCE_ID "2016-05-09 19:03:42 14e53d0e2f62d82ae1d64a72fd9711548e3bf5ea"
117
118	/*
119	** CAPI3REF: Run-Time Library Version Numbers
120	** KEYWORDS: sqlite3_version, sqlite3_sourceid
121	**
122
123	DDED www/encryptedrepos.wiki

	--- src/sqlite3.h
	+++ src/sqlite3.h
	@@ -111,11 +111,11 @@
111	** [sqlite3_libversion_number()], [sqlite3_sourceid()],
112	** [sqlite_version()] and [sqlite_source_id()].
113	*/
114	#define SQLITE_VERSION "3.13.0"
115	#define SQLITE_VERSION_NUMBER 3013000
116	#define SQLITE_SOURCE_ID "2016-05-18 10:57:30 fc49f556e48970561d7ab6a2f24fdd7d9eb81ff2"
117
118	/*
119	** CAPI3REF: Run-Time Library Version Numbers
120	** KEYWORDS: sqlite3_version, sqlite3_sourceid
121	**
122
123	DDED www/encryptedrepos.wiki

M www/encryptedrepos.wiki

+20

		--- a/www/encryptedrepos.wiki
		+++ b/www/encryptedrepos.wiki
		@@ -0,0 +1,20 @@
	1	+<title>How To Use Encr<h2>Int Settingcrypted repository.<title>How To Use Encr<h2>Introduction</h2><blockquote>
	2	+Fossil can be compiled so that it works with encrypted repositories using
	3	+the [https://www.sqlite.org/see/doc/trunk/www/readme.wiki\|SQLite Encryption Extension].
	4	+This technical note explains the process.
	5	+</blockquote>
	6	+<h2>Building An Enc<blockquote>
	7	+The SQLite Encryption ExtensionSQLite Encryption Extension (SEE) is proprietary software and requires
	8	+[http://www.hwaci.com/cgi-bin/see-step1
	9	+Assuming you have an SEE license, the first step of compiling Fossil to
	10	+use SEE is to create an SEE-enabled version of the SQLite database source code.
	11	+This alternative SQLite database source file should be called "sqlite3-see.c"
	12	+and should be placed in the src/ subfolder of the Fossil sourrces, right beside
	13	+the public-domain "sqlite3.c" source fitps://www.sqlite.org/see/doc/trunk/www/readme.wiki\|SQLite Encryption Extension].
	14	+This technical note explains the process.
	15	+</blockquote>
	16	+<h2>Building An Enc<blockquote>
	17	+The SQLite Encryption ExtensionSQLite Encryption Extension (SEE) is proprietary software and requires
	18	+[http://www.hwaci.com/cgi-bin/see-step1
	19	+Assuming you have an SEE license, the first step of compiling Fossil to
	20	+use SEE is to create an SEE-enabled version of the SQLite database source couse of requires retyping and "fossil ui

	--- a/www/encryptedrepos.wiki
	+++ b/www/encryptedrepos.wiki
	@@ -0,0 +1,20 @@

	--- a/www/encryptedrepos.wiki
	+++ b/www/encryptedrepos.wiki
	@@ -0,0 +1,20 @@
1	<title>How To Use Encr<h2>Int Settingcrypted repository.<title>How To Use Encr<h2>Introduction</h2><blockquote>
2	Fossil can be compiled so that it works with encrypted repositories using
3	the [https://www.sqlite.org/see/doc/trunk/www/readme.wiki\|SQLite Encryption Extension].
4	This technical note explains the process.
5	</blockquote>
6	<h2>Building An Enc<blockquote>
7	The SQLite Encryption ExtensionSQLite Encryption Extension (SEE) is proprietary software and requires
8	[http://www.hwaci.com/cgi-bin/see-step1
9	Assuming you have an SEE license, the first step of compiling Fossil to
10	use SEE is to create an SEE-enabled version of the SQLite database source code.
11	This alternative SQLite database source file should be called "sqlite3-see.c"
12	and should be placed in the src/ subfolder of the Fossil sourrces, right beside
13	the public-domain "sqlite3.c" source fitps://www.sqlite.org/see/doc/trunk/www/readme.wiki\|SQLite Encryption Extension].
14	This technical note explains the process.
15	</blockquote>
16	<h2>Building An Enc<blockquote>
17	The SQLite Encryption ExtensionSQLite Encryption Extension (SEE) is proprietary software and requires
18	[http://www.hwaci.com/cgi-bin/see-step1
19	Assuming you have an SEE license, the first step of compiling Fossil to
20	use SEE is to create an SEE-enabled version of the SQLite database source couse of requires retyping and "fossil ui

Fossil SCM

Keyboard Shortcuts