Fossil SCM

Merge the trunk changes into the retro-sbsdiff branch.

drh 2012-01-05 13:09 UTC retro-sbsdiff merge

Commit 4b432961acb733eb5a1d86dbb5f3001808afd995

Parent a1e667e6b5a2101…

3 files changed +2 -1 +119 -213 +19 -22

~ src/file.c ~ src/sqlite3.c ~ src/sqlite3.h

M src/file.c

+2 -1

		--- src/file.c
		+++ src/file.c
		@@ -997,11 +997,12 @@
997	997	** Translate MBCS to UTF8. Return a pointer. Call fossil_mbcs_free()
998	998	** to deallocate any memory used to store the returned pointer when done.
999	999	*/
1000	1000	void fossil_mbcs_free(char *zOld){
1001	1001	#ifdef _WIN32
1002		- free(zOld);
	1002	+ extern void sqlite3_free(void*);
	1003	+ sqlite3_free(zOld);
1003	1004	#else
1004	1005	/* No-op on unix */
1005	1006	#endif
1006	1007	}
1007	1008
1008	1009

	--- src/file.c
	+++ src/file.c
	@@ -997,11 +997,12 @@
997	** Translate MBCS to UTF8. Return a pointer. Call fossil_mbcs_free()
998	** to deallocate any memory used to store the returned pointer when done.
999	*/
1000	void fossil_mbcs_free(char *zOld){
1001	#ifdef _WIN32
1002	free(zOld);

1003	#else
1004	/* No-op on unix */
1005	#endif
1006	}
1007
1008

	--- src/file.c
	+++ src/file.c
	@@ -997,11 +997,12 @@
997	** Translate MBCS to UTF8. Return a pointer. Call fossil_mbcs_free()
998	** to deallocate any memory used to store the returned pointer when done.
999	*/
1000	void fossil_mbcs_free(char *zOld){
1001	#ifdef _WIN32
1002	extern void sqlite3_free(void*);
1003	sqlite3_free(zOld);
1004	#else
1005	/* No-op on unix */
1006	#endif
1007	}
1008
1009

M src/sqlite3.c

+119 -213

		--- src/sqlite3.c
		+++ src/sqlite3.c
		@@ -657,11 +657,11 @@
657	657	** [sqlite3_libversion_number()], [sqlite3_sourceid()],
658	658	** [sqlite_version()] and [sqlite_source_id()].
659	659	*/
660	660	#define SQLITE_VERSION "3.7.10"
661	661	#define SQLITE_VERSION_NUMBER 3007010
662		-#define SQLITE_SOURCE_ID "2011-12-23 13:32:07 1f24ae716df6232de768e245ea990049deee3c22"
	662	+#define SQLITE_SOURCE_ID "2012-01-05 12:38:02 1378f905d37544701776d38987fe7a312b255983"
663	663
664	664	/*
665	665	** CAPI3REF: Run-Time Library Version Numbers
666	666	** KEYWORDS: sqlite3_version, sqlite3_sourceid
667	667	**
		@@ -725,11 +725,11 @@
725	725
726	726	/*
727	727	** CAPI3REF: Test To See If The Library Is Threadsafe
728	728	**
729	729	** ^The sqlite3_threadsafe() function returns zero if and only if
730		-** SQLite was compiled mutexing code omitted due to the
	730	+** SQLite was compiled with mutexing code omitted due to the
731	731	** [SQLITE_THREADSAFE] compile-time option being set to 0.
732	732	**
733	733	** SQLite can be compiled with or without mutexes. When
734	734	** the [SQLITE_THREADSAFE] C preprocessor macro is 1 or 2, mutexes
735	735	** are enabled and SQLite is threadsafe. When the
		@@ -919,11 +919,11 @@
919	919	** CAPI3REF: Result Codes
920	920	** KEYWORDS: SQLITE_OK {error code} {error codes}
921	921	** KEYWORDS: {result code} {result codes}
922	922	**
923	923	** Many SQLite functions return an integer result code from the set shown
924		-** here in order to indicates success or failure.
	924	+** here in order to indicate success or failure.
925	925	**
926	926	** New error codes may be added in future versions of SQLite.
927	927	**
928	928	** See also: [SQLITE_IOERR_READ \| extended result codes],
929	929	** [sqlite3_vtab_on_conflict()] [SQLITE_ROLLBACK \| result codes].
		@@ -1058,17 +1058,14 @@
1058	1058	** that when data is appended to a file, the data is appended
1059	1059	** first then the size of the file is extended, never the other
1060	1060	** way around. The SQLITE_IOCAP_SEQUENTIAL property means that
1061	1061	** information is written to disk in the same order as calls
1062	1062	** to xWrite(). The SQLITE_IOCAP_POWERSAFE_OVERWRITE property means that
1063		-** after reboot following a crash or power loss, the value of
1064		-** each byte in a file is a value that was actually written
1065		-** into that byte at some point. In other words, a crash will
1066		-** not cause unwritten bytes of the file to change nor introduce
1067		-** randomness into a file nor zero out parts of the file, and any byte of
1068		-** a file that are never written will not change values due to
1069		-** writes to nearby bytes.
	1063	+** after reboot following a crash or power loss, the only bytes in a
	1064	+** file that were written at the application level might have changed
	1065	+** and that adjacent bytes, even bytes within the same sector are
	1066	+** guaranteed to be unchanged.
1070	1067	*/
1071	1068	#define SQLITE_IOCAP_ATOMIC 0x00000001
1072	1069	#define SQLITE_IOCAP_ATOMIC512 0x00000002
1073	1070	#define SQLITE_IOCAP_ATOMIC1K 0x00000004
1074	1071	#define SQLITE_IOCAP_ATOMIC2K 0x00000008
		@@ -1300,16 +1297,16 @@
1300	1297	** opcode as doing so may disrupt the operation of the specialized VFSes
1301	1298	** that do require it.
1302	1299	**
1303	1300	** ^The [SQLITE_FCNTL_WIN32_AV_RETRY] opcode is used to configure automatic
1304	1301	** retry counts and intervals for certain disk I/O operations for the
1305		-** windows [VFS] in order to work to provide robustness against
	1302	+** windows [VFS] in order to provide robustness in the presence of
1306	1303	** anti-virus programs. By default, the windows VFS will retry file read,
1307	1304	** file write, and file delete operations up to 10 times, with a delay
1308	1305	** of 25 milliseconds before the first retry and with the delay increasing
1309	1306	** by an additional 25 milliseconds with each subsequent retry. This
1310		-** opcode allows those to values (10 retries and 25 milliseconds of delay)
	1307	+** opcode allows these two values (10 retries and 25 milliseconds of delay)
1311	1308	** to be adjusted. The values are changed for all database connections
1312	1309	** within the same process. The argument is a pointer to an array of two
1313	1310	** integers where the first integer i the new retry count and the second
1314	1311	** integer is the delay. If either integer is negative, then the setting
1315	1312	** is not changed but instead the prior value of that setting is written
		@@ -1420,11 +1417,11 @@
1420	1417	** ^SQLite guarantees that the zFilename parameter to xOpen
1421	1418	** is either a NULL pointer or string obtained
1422	1419	** from xFullPathname() with an optional suffix added.
1423	1420	** ^If a suffix is added to the zFilename parameter, it will
1424	1421	** consist of a single "-" character followed by no more than
1425		-** 10 alphanumeric and/or "-" characters.
	1422	+** 11 alphanumeric and/or "-" characters.
1426	1423	** ^SQLite further guarantees that
1427	1424	** the string will be valid and unchanged until xClose() is
1428	1425	** called. Because of the previous sentence,
1429	1426	** the [sqlite3_file] can safely store a pointer to the
1430	1427	** filename if it needs to remember the filename for some reason.
		@@ -2571,11 +2568,11 @@
2571	2568	** These routines all implement some additional formatting
2572	2569	** options that are useful for constructing SQL statements.
2573	2570	** All of the usual printf() formatting options apply. In addition, there
2574	2571	** is are "%q", "%Q", and "%z" options.
2575	2572	**
2576		-** ^(The %q option works like %s in that it substitutes a null-terminated
	2573	+** ^(The %q option works like %s in that it substitutes a nul-terminated
2577	2574	** string from the argument list. But %q also doubles every '\'' character.
2578	2575	** %q is designed for use inside a string literal.)^ By doubling each '\''
2579	2576	** character it escapes that character and allows it to be inserted into
2580	2577	** the string.
2581	2578	**
		@@ -4091,11 +4088,11 @@
4091	4088	** of the string. ^For clarity: the values returned by
4092	4089	** [sqlite3_column_bytes()] and [sqlite3_column_bytes16()] are the number of
4093	4090	** bytes in the string, not the number of characters.
4094	4091	**
4095	4092	** ^Strings returned by sqlite3_column_text() and sqlite3_column_text16(),
4096		-** even empty strings, are always zero terminated. ^The return
	4093	+** even empty strings, are always zero-terminated. ^The return
4097	4094	** value from sqlite3_column_blob() for a zero-length BLOB is a NULL pointer.
4098	4095	**
4099	4096	** ^The object returned by [sqlite3_column_value()] is an
4100	4097	** [unprotected sqlite3_value] object. An unprotected sqlite3_value object
4101	4098	** may only be used with [sqlite3_bind_value()] and [sqlite3_result_value()].
		@@ -5171,11 +5168,11 @@
5171	5168	SQLITE_API int sqlite3_release_memory(int);
5172	5169
5173	5170	/*
5174	5171	** CAPI3REF: Free Memory Used By A Database Connection
5175	5172	**
5176		-** ^The sqlite3_db_shrink(D) interface attempts to free as much heap
	5173	+** ^The sqlite3_db_release_memory(D) interface attempts to free as much heap
5177	5174	** memory as possible from database connection D. Unlike the
5178	5175	** [sqlite3_release_memory()] interface, this interface is effect even
5179	5176	** when then [SQLITE_ENABLE_MEMORY_MANAGEMENT] compile-time option is
5180	5177	** omitted.
5181	5178	**
		@@ -5954,19 +5951,19 @@
5954	5951	** is selected automatically at compile-time. ^(The following
5955	5952	** implementations are available in the SQLite core:
5956	5953	**
5957	5954	** <ul>
5958	5955	** <li> SQLITE_MUTEX_OS2
5959		-** <li> SQLITE_MUTEX_PTHREAD
	5956	+** <li> SQLITE_MUTEX_PTHREADS
5960	5957	** <li> SQLITE_MUTEX_W32
5961	5958	** <li> SQLITE_MUTEX_NOOP
5962	5959	** </ul>)^
5963	5960	**
5964	5961	** ^The SQLITE_MUTEX_NOOP implementation is a set of routines
5965	5962	** that does no real locking and is appropriate for use in
5966	5963	** a single-threaded application. ^The SQLITE_MUTEX_OS2,
5967		-** SQLITE_MUTEX_PTHREAD, and SQLITE_MUTEX_W32 implementations
	5964	+** SQLITE_MUTEX_PTHREADS, and SQLITE_MUTEX_W32 implementations
5968	5965	** are appropriate for use on OS/2, Unix, and Windows.
5969	5966	**
5970	5967	** ^(If SQLite is compiled with the SQLITE_MUTEX_APPDEF preprocessor
5971	5968	** macro defined (with "-DSQLITE_MUTEX_APPDEF=1"), then no mutex
5972	5969	** implementation is included with the library. In this case the
		@@ -6152,11 +6149,11 @@
6152	6149	** defined and if NDEBUG is not defined.
6153	6150	**
6154	6151	** ^These routines should return true if the mutex in their argument
6155	6152	** is held or not held, respectively, by the calling thread.
6156	6153	**
6157		-** ^The implementation is not required to provided versions of these
	6154	+** ^The implementation is not required to provide versions of these
6158	6155	** routines that actually work. If the implementation does not provide working
6159	6156	** versions of these routines, it should at least provide stubs that always
6160	6157	** return true so that one does not get spurious assertion failures.
6161	6158	**
6162	6159	** ^If the argument to sqlite3_mutex_held() is a NULL pointer then
		@@ -6634,11 +6631,11 @@
6634	6631	**
6635	6632	** [[the xInit() page cache method]]
6636	6633	** ^(The xInit() method is called once for each effective
6637	6634	** call to [sqlite3_initialize()])^
6638	6635	** (usually only once during the lifetime of the process). ^(The xInit()
6639		-** method is passed a copy of the sqlite3_pcache_methods.pArg value.)^
	6636	+** method is passed a copy of the sqlite3_pcache_methods2.pArg value.)^
6640	6637	** The intent of the xInit() method is to set up global data structures
6641	6638	** required by the custom page cache implementation.
6642	6639	** ^(If the xInit() method is NULL, then the
6643	6640	** built-in default page cache is used instead of the application defined
6644	6641	** page cache.)^
		@@ -6660,11 +6657,11 @@
6660	6657	**
6661	6658	** [[the xCreate() page cache methods]]
6662	6659	** ^SQLite invokes the xCreate() method to construct a new cache instance.
6663	6660	** SQLite will typically create one cache instance for each open database file,
6664	6661	** though this is not guaranteed. ^The
6665		-** parameter parameter, szPage, is the size in bytes of the pages that must
	6662	+** first parameter, szPage, is the size in bytes of the pages that must
6666	6663	** be allocated by the cache. ^szPage will always a power of two. ^The
6667	6664	** second parameter szExtra is a number of bytes of extra storage
6668	6665	** associated with each page cache entry. ^The szExtra parameter will
6669	6666	** a number less than 250. SQLite will use the
6670	6667	** extra szExtra bytes on each page to store metadata about the underlying
		@@ -6755,11 +6752,11 @@
6755	6752	**
6756	6753	** [[the xDestroy() page cache method]]
6757	6754	** ^The xDestroy() method is used to delete a cache allocated by xCreate().
6758	6755	** All resources associated with the specified cache should be freed. ^After
6759	6756	** calling the xDestroy() method, SQLite considers the [sqlite3_pcache*]
6760		-** handle invalid, and will not use it with any other sqlite3_pcache_methods
	6757	+** handle invalid, and will not use it with any other sqlite3_pcache_methods2
6761	6758	** functions.
6762	6759	**
6763	6760	** [[the xShrink() page cache method]]
6764	6761	** ^SQLite invokes the xShrink() method when it wants the page cache to
6765	6762	** free up as much of heap memory as possible. The page cache implementation
		@@ -9266,11 +9263,11 @@
9266	9263
9267	9264	/*
9268	9265	** The default size of a disk sector
9269	9266	*/
9270	9267	#ifndef SQLITE_DEFAULT_SECTOR_SIZE
9271		-# define SQLITE_DEFAULT_SECTOR_SIZE 512
	9268	+# define SQLITE_DEFAULT_SECTOR_SIZE 4096
9272	9269	#endif
9273	9270
9274	9271	/*
9275	9272	** Temporary files are named starting with this prefix followed by 16 random
9276	9273	** alphanumeric characters, and no file extension. They are stored in the
		@@ -13190,11 +13187,11 @@
13190	13187	SQLITE_PRIVATE int sqlite3VdbeMemRealify(Mem*);
13191	13188	SQLITE_PRIVATE int sqlite3VdbeMemNumerify(Mem*);
13192	13189	SQLITE_PRIVATE int sqlite3VdbeMemFromBtree(BtCursor,int,int,int,Mem);
13193	13190	SQLITE_PRIVATE void sqlite3VdbeMemRelease(Mem *p);
13194	13191	SQLITE_PRIVATE void sqlite3VdbeMemReleaseExternal(Mem *p);
13195		-#define MemReleaseExt(X) \
	13192	+#define VdbeMemRelease(X) \
13196	13193	if((X)->flags&(MEM_Agg\|MEM_Dyn\|MEM_RowSet\|MEM_Frame)) \
13197	13194	sqlite3VdbeMemReleaseExternal(X);
13198	13195	SQLITE_PRIVATE int sqlite3VdbeMemFinalize(Mem, FuncDef);
13199	13196	SQLITE_PRIVATE const char *sqlite3OpcodeName(int);
13200	13197	SQLITE_PRIVATE int sqlite3VdbeMemGrow(Mem *pMem, int n, int preserve);
		@@ -13229,11 +13226,11 @@
13229	13226	# define sqlite3VdbeEnter(X)
13230	13227	# define sqlite3VdbeLeave(X)
13231	13228	#endif
13232	13229
13233	13230	#ifdef SQLITE_DEBUG
13234		-SQLITE_PRIVATE void sqlite3VdbeMemPrepareToChange(Vdbe,Mem);
	13231	+SQLITE_PRIVATE void sqlite3VdbeMemAboutToChange(Vdbe,Mem);
13235	13232	#endif
13236	13233
13237	13234	#ifndef SQLITE_OMIT_FOREIGN_KEY
13238	13235	SQLITE_PRIVATE int sqlite3VdbeCheckFk(Vdbe *, int);
13239	13236	#else
		@@ -14762,11 +14759,11 @@
14762	14759	int flags,
14763	14760	int *pFlagsOut
14764	14761	){
14765	14762	int rc;
14766	14763	DO_OS_MALLOC_TEST(0);
14767		- /* 0x87f3f is a mask of SQLITE_OPEN_ flags that are valid to be passed
	14764	+ /* 0x87f7f is a mask of SQLITE_OPEN_ flags that are valid to be passed
14768	14765	** down into the VFS layer. Some SQLITE_OPEN_ flags (for example,
14769	14766	** SQLITE_OPEN_FULLMUTEX or SQLITE_OPEN_SHAREDCACHE) are blocked before
14770	14767	** reaching the VFS. */
14771	14768	rc = pVfs->xOpen(pVfs, zPath, pFile, flags & 0x87f7f, pFlagsOut);
14772	14769	assert( rc==SQLITE_OK \|\| pFile->pMethods==0 );
		@@ -17316,11 +17313,11 @@
17316	17313	SQLITE_API int sqlite3_mutex_notheld(sqlite3_mutex *p){
17317	17314	return p==0 \|\| sqlite3GlobalConfig.mutex.xMutexNotheld(p);
17318	17315	}
17319	17316	#endif
17320	17317
17321		-#endif /* SQLITE_MUTEX_OMIT */
	17318	+#endif /* !defined(SQLITE_MUTEX_OMIT) */
17322	17319
17323	17320	/************ End of mutex.c *********************************************/
17324	17321	/************ Begin file mutex_noop.c ************************************/
17325	17322	/*
17326	17323	** 2008 October 07
		@@ -17523,12 +17520,12 @@
17523	17520	*/
17524	17521	#ifdef SQLITE_MUTEX_NOOP
17525	17522	SQLITE_PRIVATE sqlite3_mutex_methods const *sqlite3DefaultMutex(void){
17526	17523	return sqlite3NoopMutex();
17527	17524	}
17528		-#endif /* SQLITE_MUTEX_NOOP */
17529		-#endif /* SQLITE_MUTEX_OMIT */
	17525	+#endif /* defined(SQLITE_MUTEX_NOOP) */
	17526	+#endif /* !defined(SQLITE_MUTEX_OMIT) */
17530	17527
17531	17528	/************ End of mutex_noop.c ****************************************/
17532	17529	/************ Begin file mutex_os2.c *************************************/
17533	17530	/*
17534	17531	** 2007 August 28
		@@ -18153,11 +18150,11 @@
18153	18150	};
18154	18151
18155	18152	return &sMutex;
18156	18153	}
18157	18154
18158		-#endif /* SQLITE_MUTEX_PTHREAD */
	18155	+#endif /* SQLITE_MUTEX_PTHREADS */
18159	18156
18160	18157	/************ End of mutex_unix.c ****************************************/
18161	18158	/************ Begin file mutex_w32.c *************************************/
18162	18159	/*
18163	18160	** 2007 August 14
		@@ -22116,26 +22113,31 @@
22116	22113	** characters of the original suffix.
22117	22114	**
22118	22115	** If SQLITE_ENABLE_8_3_NAMES is set to 2 at compile-time, then always
22119	22116	** do the suffix shortening regardless of URI parameter.
22120	22117	**
	22118	+** Assume that zBaseFilename contains two \000 terminator bytes (so that
	22119	+** it can be harmlessly passed into sqlite3_uri_parameter()) and copy both
	22120	+** zero terminator bytes into the end of the revised name.
	22121	+**
22121	22122	** Examples:
22122	22123	**
22123	22124	** test.db-journal => test.nal
22124	22125	** test.db-wal => test.wal
22125	22126	** test.db-shm => test.shm
22126	22127	** test.db-mj7f3319fa => test.9fa
22127	22128	*/
22128	22129	SQLITE_PRIVATE void sqlite3FileSuffix3(const char zBaseFilename, char z){
	22130	+ assert( zBaseFilename[strlen(zBaseFilename)+1]==0 );
22129	22131	#if SQLITE_ENABLE_8_3_NAMES<2
22130		- if( sqlite3_uri_boolean(zBaseFilename, "8_3_names") )
	22132	+ if( sqlite3_uri_boolean(zBaseFilename, "8_3_names", 0) )
22131	22133	#endif
22132	22134	{
22133	22135	int i, sz;
22134	22136	sz = sqlite3Strlen30(z);
22135	22137	for(i=sz-1; i>0 && z[i]!='/' && z[i]!='.'; i--){}
22136		- if( z[i]=='.' && ALWAYS(sz>i+4) ) memmove(&z[i+1], &z[sz-3], 4);
	22138	+ if( z[i]=='.' && ALWAYS(sz>i+4) ) memmove(&z[i+1], &z[sz-3], 5);
22137	22139	}
22138	22140	}
22139	22141	#endif
22140	22142
22141	22143	/************ End of util.c **********************************************/
		@@ -24836,13 +24838,10 @@
24836	24838	#include <sys/time.h>
24837	24839	#include <errno.h>
24838	24840	#ifndef SQLITE_OMIT_WAL
24839	24841	#include <sys/mman.h>
24840	24842	#endif
24841		-#ifndef MISSING_STATVFS
24842		-#include <sys/statvfs.h>
24843		-#endif
24844	24843
24845	24844
24846	24845	#if SQLITE_ENABLE_LOCKING_STYLE
24847	24846	# include <sys/ioctl.h>
24848	24847	# if OS_VXWORKS
		@@ -24929,11 +24928,10 @@
24929	24928	sqlite3_vfs pVfs; / The VFS that created this unixFile */
24930	24929	unixInodeInfo pInode; / Info about locks on this inode */
24931	24930	int h; /* The file descriptor */
24932	24931	unsigned char eFileLock; /* The type of lock held on this fd */
24933	24932	unsigned char ctrlFlags; /* Behavioral bits. UNIXFILE_* flags */
24934		- unsigned char szSector; /* Sectorsize/512 */
24935	24933	int lastErrno; /* The unix errno from last I/O error */
24936	24934	void lockingContext; / Locking style specific state */
24937	24935	UnixUnusedFd pUnused; / Pre-allocated UnixUnusedFd */
24938	24936	const char zPath; / Name of the file */
24939	24937	unixShm pShm; / Shared memory segment information */
		@@ -25342,18 +25340,10 @@
25342	25340	#define osMkdir ((int()(const char,mode_t))aSyscall[18].pCurrent)
25343	25341
25344	25342	{ "rmdir", (sqlite3_syscall_ptr)rmdir, 0 },
25345	25343	#define osRmdir ((int()(const char))aSyscall[19].pCurrent)
25346	25344
25347		-#if defined(MISSING_STATVFS)
25348		- { "statvfs", (sqlite3_syscall_ptr)0, 0 },
25349		-#define osStatvfs ((int()(const char,void*))aSyscall[20].pCurrent)
25350		-#else
25351		- { "statvfs", (sqlite3_syscall_ptr)statvfs, 0 },
25352		-#define osStatvfs ((int()(const char,struct statvfs*))aSyscall[20].pCurrent)
25353		-#endif
25354		-
25355	25345	}; /* End of the overrideable system calls */
25356	25346
25357	25347	/*
25358	25348	** This is the xSetSystemCall() method of sqlite3_vfs for all of the
25359	25349	** "unix" VFSes. Return SQLITE_OK opon successfully updating the
		@@ -28504,13 +28494,10 @@
28504	28494	case SQLITE_SET_LOCKPROXYFILE:
28505	28495	case SQLITE_GET_LOCKPROXYFILE: {
28506	28496	return proxyFileControl(id,op,pArg);
28507	28497	}
28508	28498	#endif /* SQLITE_ENABLE_LOCKING_STYLE && defined(__APPLE__) */
28509		- case SQLITE_FCNTL_SYNC_OMITTED: {
28510		- return SQLITE_OK; /* A no-op */
28511		- }
28512	28499	}
28513	28500	return SQLITE_NOTFOUND;
28514	28501	}
28515	28502
28516	28503	/*
		@@ -28522,27 +28509,12 @@
28522	28509	** if two files are created in the same file-system directory (i.e.
28523	28510	** a database and its journal file) that the sector size will be the
28524	28511	** same for both.
28525	28512	*/
28526	28513	static int unixSectorSize(sqlite3_file *pFile){
28527		- unixFile p = (unixFile)pFile;
28528		- if( p->szSector==0 ){
28529		-#ifdef MISSING_STATVFS
28530		- p->szSector = SQLITE_DEFAULT_SECTOR_SIZE/512;
28531		-#else
28532		- struct statvfs x;
28533		- int sz;
28534		- memset(&x, 0, sizeof(x));
28535		- osStatvfs(p->zPath, &x);
28536		- sz = (int)x.f_frsize;
28537		- if( sz<512 \|\| sz>65536 \|\| (sz&(sz-1))!=0 ){
28538		- sz = SQLITE_DEFAULT_SECTOR_SIZE;
28539		- }
28540		- p->szSector = sz/512;
28541		-#endif
28542		- }
28543		- return p->szSector*512;
	28514	+ (void)pFile;
	28515	+ return SQLITE_DEFAULT_SECTOR_SIZE;
28544	28516	}
28545	28517
28546	28518	/*
28547	28519	** Return the device characteristics for the file.
28548	28520	**
		@@ -28815,13 +28787,13 @@
28815	28787	rc = SQLITE_IOERR_FSTAT;
28816	28788	goto shm_open_err;
28817	28789	}
28818	28790
28819	28791	#ifdef SQLITE_SHM_DIRECTORY
28820		- nShmFilename = sizeof(SQLITE_SHM_DIRECTORY) + 30;
	28792	+ nShmFilename = sizeof(SQLITE_SHM_DIRECTORY) + 31;
28821	28793	#else
28822		- nShmFilename = 5 + (int)strlen(pDbFd->zPath);
	28794	+ nShmFilename = 6 + (int)strlen(pDbFd->zPath);
28823	28795	#endif
28824	28796	pShmNode = sqlite3_malloc( sizeof(*pShmNode) + nShmFilename );
28825	28797	if( pShmNode==0 ){
28826	28798	rc = SQLITE_NOMEM;
28827	28799	goto shm_open_err;
		@@ -31755,11 +31727,11 @@
31755	31727	};
31756	31728	unsigned int i; /* Loop counter */
31757	31729
31758	31730	/* Double-check that the aSyscall[] array has been constructed
31759	31731	** correctly. See ticket [bb3a86e890c8e96ab] */
31760		- assert( ArraySize(aSyscall)==21 );
	31732	+ assert( ArraySize(aSyscall)==20 );
31761	31733
31762	31734	/* Register all VFSes defined in the aVfs[] array */
31763	31735	for(i=0; i<(sizeof(aVfs)/sizeof(sqlite3_vfs)); i++){
31764	31736	sqlite3_vfs_register(&aVfs[i], i==0);
31765	31737	}
		@@ -32049,11 +32021,10 @@
32049	32021	HANDLE h; /* Handle for accessing the file */
32050	32022	u8 locktype; /* Type of lock currently held on this file */
32051	32023	short sharedLockByte; /* Randomly chosen byte used as a shared lock */
32052	32024	u8 ctrlFlags; /* Flags. See WINFILE_* below */
32053	32025	DWORD lastErrno; /* The Windows errno from the last I/O error */
32054		- DWORD sectorSize; /* Sector size of the device file is on */
32055	32026	winShm pShm; / Instance of shared memory on this file */
32056	32027	const char zPath; / Full pathname of this file */
32057	32028	int szChunk; /* Chunk size configured by FCNTL_CHUNK_SIZE */
32058	32029	#if SQLITE_OS_WINCE
32059	32030	LPWSTR zDeleteOnClose; /* Name of file to delete when closing */
		@@ -32139,18 +32110,10 @@
32139	32110	static void winMemShutdown(void *pAppData);
32140	32111
32141	32112	SQLITE_PRIVATE const sqlite3_mem_methods *sqlite3MemGetWin32(void);
32142	32113	#endif /* SQLITE_WIN32_MALLOC */
32143	32114
32144		-/*
32145		-** Forward prototypes.
32146		-*/
32147		-static int getSectorSize(
32148		- sqlite3_vfs *pVfs,
32149		- const char zRelative / UTF-8 file name */
32150		-);
32151		-
32152	32115	/*
32153	32116	** The following variable is (normally) set once and never changes
32154	32117	** thereafter. It records whether the operating system is Win9x
32155	32118	** or WinNT.
32156	32119	**
		@@ -34181,13 +34144,10 @@
34181	34144	}
34182	34145	case SQLITE_FCNTL_VFSNAME: {
34183	34146	(char*)pArg = sqlite3_mprintf("win32");
34184	34147	return SQLITE_OK;
34185	34148	}
34186		- case SQLITE_FCNTL_SYNC_OMITTED: {
34187		- return SQLITE_OK;
34188		- }
34189	34149	case SQLITE_FCNTL_WIN32_AV_RETRY: {
34190	34150	int a = (int)pArg;
34191	34151	if( a[0]>0 ){
34192	34152	win32IoerrRetry = a[0];
34193	34153	}else{
		@@ -34213,12 +34173,12 @@
34213	34173	** if two files are created in the same file-system directory (i.e.
34214	34174	** a database and its journal file) that the sector size will be the
34215	34175	** same for both.
34216	34176	*/
34217	34177	static int winSectorSize(sqlite3_file *id){
34218		- assert( id!=0 );
34219		- return (int)(((winFile*)id)->sectorSize);
	34178	+ (void)id;
	34179	+ return SQLITE_DEFAULT_SECTOR_SIZE;
34220	34180	}
34221	34181
34222	34182	/*
34223	34183	** Return a vector of device characteristics.
34224	34184	*/
		@@ -34463,11 +34423,11 @@
34463	34423	*/
34464	34424	p = sqlite3_malloc( sizeof(*p) );
34465	34425	if( p==0 ) return SQLITE_IOERR_NOMEM;
34466	34426	memset(p, 0, sizeof(*p));
34467	34427	nName = sqlite3Strlen30(pDbFd->zPath);
34468		- pNew = sqlite3_malloc( sizeof(*pShmNode) + nName + 15 );
	34428	+ pNew = sqlite3_malloc( sizeof(*pShmNode) + nName + 16 );
34469	34429	if( pNew==0 ){
34470	34430	sqlite3_free(p);
34471	34431	return SQLITE_IOERR_NOMEM;
34472	34432	}
34473	34433	memset(pNew, 0, sizeof(*pNew));
		@@ -35191,11 +35151,10 @@
35191	35151	pFile->pShm = 0;
35192	35152	pFile->zPath = zName;
35193	35153	if( sqlite3_uri_boolean(zName, "psow", SQLITE_POWERSAFE_OVERWRITE) ){
35194	35154	pFile->ctrlFlags \|= WINFILE_PSOW;
35195	35155	}
35196		- pFile->sectorSize = getSectorSize(pVfs, zUtf8Name);
35197	35156
35198	35157	#if SQLITE_OS_WINCE
35199	35158	if( isReadWrite && eType==SQLITE_OPEN_MAIN_DB
35200	35159	&& !winceCreateLock(zName, pFile)
35201	35160	){
		@@ -35436,85 +35395,10 @@
35436	35395	return SQLITE_IOERR_NOMEM;
35437	35396	}
35438	35397	#endif
35439	35398	}
35440	35399
35441		-/*
35442		-** Get the sector size of the device used to store
35443		-** file.
35444		-*/
35445		-static int getSectorSize(
35446		- sqlite3_vfs *pVfs,
35447		- const char zRelative / UTF-8 file name */
35448		-){
35449		- DWORD bytesPerSector = SQLITE_DEFAULT_SECTOR_SIZE;
35450		- /* GetDiskFreeSpace is not supported under WINCE */
35451		-#if SQLITE_OS_WINCE
35452		- UNUSED_PARAMETER(pVfs);
35453		- UNUSED_PARAMETER(zRelative);
35454		-#else
35455		- char zFullpath[MAX_PATH+1];
35456		- int rc;
35457		- DWORD dwRet = 0;
35458		- DWORD dwDummy;
35459		-
35460		- /*
35461		- ** We need to get the full path name of the file
35462		- ** to get the drive letter to look up the sector
35463		- ** size.
35464		- */
35465		- SimulateIOErrorBenign(1);
35466		- sqlite3BeginBenignMalloc();
35467		- rc = winFullPathname(pVfs, zRelative, MAX_PATH, zFullpath);
35468		- sqlite3EndBenignMalloc();
35469		- SimulateIOErrorBenign(0);
35470		- if( rc == SQLITE_OK )
35471		- {
35472		- void *zConverted;
35473		- sqlite3BeginBenignMalloc();
35474		- zConverted = convertUtf8Filename(zFullpath);
35475		- sqlite3EndBenignMalloc();
35476		- if( zConverted ){
35477		- if( isNT() ){
35478		- /* trim path to just drive reference */
35479		- LPWSTR p = zConverted;
35480		- for(;*p;p++){
35481		- if( *p == '\\' ){
35482		- *p = '\0';
35483		- break;
35484		- }
35485		- }
35486		- dwRet = osGetDiskFreeSpaceW((LPCWSTR)zConverted,
35487		- &dwDummy,
35488		- &bytesPerSector,
35489		- &dwDummy,
35490		- &dwDummy);
35491		- }else{
35492		- /* trim path to just drive reference */
35493		- char p = (char )zConverted;
35494		- for(;*p;p++){
35495		- if( *p == '\\' ){
35496		- *p = '\0';
35497		- break;
35498		- }
35499		- }
35500		- dwRet = osGetDiskFreeSpaceA((char*)zConverted,
35501		- &dwDummy,
35502		- &bytesPerSector,
35503		- &dwDummy,
35504		- &dwDummy);
35505		- }
35506		- sqlite3_free(zConverted);
35507		- }
35508		- if( !dwRet ){
35509		- bytesPerSector = SQLITE_DEFAULT_SECTOR_SIZE;
35510		- }
35511		- }
35512		-#endif
35513		- return (int) bytesPerSector;
35514		-}
35515		-
35516	35400	#ifndef SQLITE_OMIT_LOAD_EXTENSION
35517	35401	/*
35518	35402	** Interfaces for opening a shared library, finding entry points
35519	35403	** within the shared library, and closing the shared library.
35520	35404	*/
		@@ -36381,11 +36265,11 @@
36381	36265	*/
36382	36266	static int numberOfCachePages(PCache *p){
36383	36267	if( p->szCache>=0 ){
36384	36268	return p->szCache;
36385	36269	}else{
36386		- return (-1024*p->szCache)/(p->szPage+p->szExtra);
	36270	+ return (int)((-1024*(i64)p->szCache)/(p->szPage+p->szExtra));
36387	36271	}
36388	36272	}
36389	36273
36390	36274	/*
36391	36275	** Try to obtain a page from the cache.
		@@ -36845,14 +36729,14 @@
36845	36729	** PGroup which is the pcache1.grp global variable and its mutex is
36846	36730	** SQLITE_MUTEX_STATIC_LRU.
36847	36731	*/
36848	36732	struct PGroup {
36849	36733	sqlite3_mutex mutex; / MUTEX_STATIC_LRU or NULL */
36850		- int nMaxPage; /* Sum of nMax for purgeable caches */
36851		- int nMinPage; /* Sum of nMin for purgeable caches */
36852		- int mxPinned; /* nMaxpage + 10 - nMinPage */
36853		- int nCurrentPage; /* Number of purgeable pages allocated */
	36734	+ unsigned int nMaxPage; /* Sum of nMax for purgeable caches */
	36735	+ unsigned int nMinPage; /* Sum of nMin for purgeable caches */
	36736	+ unsigned int mxPinned; /* nMaxpage + 10 - nMinPage */
	36737	+ unsigned int nCurrentPage; /* Number of purgeable pages allocated */
36854	36738	PgHdr1 pLruHead, pLruTail; /* LRU list of unpinned pages */
36855	36739	};
36856	36740
36857	36741	/* Each page cache is an instance of the following object. Every
36858	36742	** open database file (including each in-memory database and each
		@@ -37511,11 +37395,11 @@
37511	37395	static sqlite3_pcache_page *pcache1Fetch(
37512	37396	sqlite3_pcache *p,
37513	37397	unsigned int iKey,
37514	37398	int createFlag
37515	37399	){
37516		- int nPinned;
	37400	+ unsigned int nPinned;
37517	37401	PCache1 pCache = (PCache1 )p;
37518	37402	PGroup *pGroup;
37519	37403	PgHdr1 *pPage = 0;
37520	37404
37521	37405	assert( pCache->bPurgeable \|\| createFlag!=1 );
		@@ -37546,17 +37430,17 @@
37546	37430	#ifdef SQLITE_MUTEX_OMIT
37547	37431	pGroup = pCache->pGroup;
37548	37432	#endif
37549	37433
37550	37434	/* Step 3: Abort if createFlag is 1 but the cache is nearly full */
	37435	+ assert( pCache->nPage >= pCache->nRecyclable );
37551	37436	nPinned = pCache->nPage - pCache->nRecyclable;
37552		- assert( nPinned>=0 );
37553	37437	assert( pGroup->mxPinned == pGroup->nMaxPage + 10 - pGroup->nMinPage );
37554	37438	assert( pCache->n90pct == pCache->nMax*9/10 );
37555	37439	if( createFlag==1 && (
37556	37440	nPinned>=pGroup->mxPinned
37557		- \|\| nPinned>=(int)pCache->n90pct
	37441	+ \|\| nPinned>=pCache->n90pct
37558	37442	\|\| pcache1UnderMemoryPressure(pCache)
37559	37443	)){
37560	37444	goto fetch_out;
37561	37445	}
37562	37446
		@@ -37725,11 +37609,13 @@
37725	37609	PCache1 pCache = (PCache1 )p;
37726	37610	PGroup *pGroup = pCache->pGroup;
37727	37611	assert( pCache->bPurgeable \|\| (pCache->nMax==0 && pCache->nMin==0) );
37728	37612	pcache1EnterMutex(pGroup);
37729	37613	pcache1TruncateUnsafe(pCache, 0);
	37614	+ assert( pGroup->nMaxPage >= pCache->nMax );
37730	37615	pGroup->nMaxPage -= pCache->nMax;
	37616	+ assert( pGroup->nMinPage >= pCache->nMin );
37731	37617	pGroup->nMinPage -= pCache->nMin;
37732	37618	pGroup->mxPinned = pGroup->nMaxPage + 10 - pGroup->nMinPage;
37733	37619	pcache1EnforceMaxPage(pGroup);
37734	37620	pcache1LeaveMutex(pGroup);
37735	37621	sqlite3_free(pCache->apHash);
		@@ -37807,12 +37693,12 @@
37807	37693	int nRecyclable = 0;
37808	37694	for(p=pcache1.grp.pLruHead; p; p=p->pLruNext){
37809	37695	nRecyclable++;
37810	37696	}
37811	37697	*pnCurrent = pcache1.grp.nCurrentPage;
37812		- *pnMax = pcache1.grp.nMaxPage;
37813		- *pnMin = pcache1.grp.nMinPage;
	37698	+ *pnMax = (int)pcache1.grp.nMaxPage;
	37699	+ *pnMin = (int)pcache1.grp.nMinPage;
37814	37700	*pnRecyclable = nRecyclable;
37815	37701	}
37816	37702	#endif
37817	37703
37818	37704	/************ End of pcache1.c *******************************************/
		@@ -38282,10 +38168,16 @@
38282	38168
38283	38169	#ifndef _WAL_H_
38284	38170	#define _WAL_H_
38285	38171
38286	38172
	38173	+/* Additional values that can be added to the sync_flags argument of
	38174	+** sqlite3WalFrames():
	38175	+*/
	38176	+#define WAL_SYNC_TRANSACTIONS 0x20 /* Sync at the end of each transaction */
	38177	+#define SQLITE_SYNC_MASK 0x13 /* Mask off the SQLITE_SYNC_* values */
	38178	+
38287	38179	#ifdef SQLITE_OMIT_WAL
38288	38180	# define sqlite3WalOpen(x,y,z) 0
38289	38181	# define sqlite3WalLimit(x,y)
38290	38182	# define sqlite3WalClose(w,x,y,z) 0
38291	38183	# define sqlite3WalBeginReadTransaction(y,z) 0
		@@ -38350,16 +38242,10 @@
38350	38242	SQLITE_PRIVATE int sqlite3WalSavepointUndo(Wal pWal, u32 aWalData);
38351	38243
38352	38244	/* Write a frame or frames to the log. */
38353	38245	SQLITE_PRIVATE int sqlite3WalFrames(Wal pWal, int, PgHdr , Pgno, int, int);
38354	38246
38355		-/* Additional values that can be added to the sync_flags argument of
38356		-** sqlite3WalFrames():
38357		-*/
38358		-#define WAL_SYNC_TRANSACTIONS 0x20 /* Sync at the end of each transaction */
38359		-#define SQLITE_SYNC_MASK 0x13 /* Mask off the SQLITE_SYNC_* values */
38360		-
38361	38247	/* Copy pages from the log to the database file */
38362	38248	SQLITE_PRIVATE int sqlite3WalCheckpoint(
38363	38249	Wal pWal, / Write-ahead log connection */
38364	38250	int eMode, /* One of PASSIVE, FULL and RESTART */
38365	38251	int (xBusy)(void), /* Function to call when busy */
		@@ -42400,11 +42286,13 @@
42400	42286	** file size will be.
42401	42287	*/
42402	42288	assert( rc!=SQLITE_OK \|\| isOpen(pPager->fd) );
42403	42289	if( rc==SQLITE_OK && pPager->dbSize>pPager->dbHintSize ){
42404	42290	sqlite3_int64 szFile = pPager->pageSize * (sqlite3_int64)pPager->dbSize;
	42291	+ sqlite3BeginBenignMalloc();
42405	42292	sqlite3OsFileControl(pPager->fd, SQLITE_FCNTL_SIZE_HINT, &szFile);
	42293	+ sqlite3EndBenignMalloc();
42406	42294	pPager->dbHintSize = pPager->dbSize;
42407	42295	}
42408	42296
42409	42297	while( rc==SQLITE_OK && pList ){
42410	42298	Pgno pgno = pList->pgno;
		@@ -42741,11 +42629,12 @@
42741	42629	z = zUri = &zFilename[sqlite3Strlen30(zFilename)+1];
42742	42630	while( *z ){
42743	42631	z += sqlite3Strlen30(z)+1;
42744	42632	z += sqlite3Strlen30(z)+1;
42745	42633	}
42746		- nUri = &z[1] - zUri;
	42634	+ nUri = (int)(&z[1] - zUri);
	42635	+ assert( nUri>=0 );
42747	42636	if( rc==SQLITE_OK && nPathname+8>pVfs->mxPathname ){
42748	42637	/* This branch is taken when the journal path required by
42749	42638	** the database being opened will be more than pVfs->mxPathname
42750	42639	** bytes in length. This means the database cannot be opened,
42751	42640	** as it will not be possible to open the journal file or even
		@@ -42775,13 +42664,13 @@
42775	42664	ROUND8(sizeof(pPager)) + / Pager structure */
42776	42665	ROUND8(pcacheSize) + /* PCache object */
42777	42666	ROUND8(pVfs->szOsFile) + /* The main db file */
42778	42667	journalFileSize * 2 + /* The two journal files */
42779	42668	nPathname + 1 + nUri + /* zFilename */
42780		- nPathname + 8 + 1 /* zJournal */
	42669	+ nPathname + 8 + 2 /* zJournal */
42781	42670	#ifndef SQLITE_OMIT_WAL
42782		- + nPathname + 4 + 1 /* zWal */
	42671	+ + nPathname + 4 + 2 /* zWal */
42783	42672	#endif
42784	42673	);
42785	42674	assert( EIGHT_BYTE_ALIGNMENT(SQLITE_INT_TO_PTR(journalFileSize)) );
42786	42675	if( !pPtr ){
42787	42676	sqlite3_free(zPathname);
		@@ -42800,16 +42689,16 @@
42800	42689	assert( nPathname>0 );
42801	42690	pPager->zJournal = (char*)(pPtr += nPathname + 1 + nUri);
42802	42691	memcpy(pPager->zFilename, zPathname, nPathname);
42803	42692	memcpy(&pPager->zFilename[nPathname+1], zUri, nUri);
42804	42693	memcpy(pPager->zJournal, zPathname, nPathname);
42805		- memcpy(&pPager->zJournal[nPathname], "-journal", 8);
	42694	+ memcpy(&pPager->zJournal[nPathname], "-journal\000", 8+1);
42806	42695	sqlite3FileSuffix3(pPager->zFilename, pPager->zJournal);
42807	42696	#ifndef SQLITE_OMIT_WAL
42808	42697	pPager->zWal = &pPager->zJournal[nPathname+8+1];
42809	42698	memcpy(pPager->zWal, zPathname, nPathname);
42810		- memcpy(&pPager->zWal[nPathname], "-wal", 4);
	42699	+ memcpy(&pPager->zWal[nPathname], "-wal\000", 4+1);
42811	42700	sqlite3FileSuffix3(pPager->zFilename, pPager->zWal);
42812	42701	#endif
42813	42702	sqlite3_free(zPathname);
42814	42703	}
42815	42704	pPager->pVfs = pVfs;
		@@ -44064,11 +43953,14 @@
44064	43953	if( !pPager->noSync ){
44065	43954	assert( !MEMDB );
44066	43955	rc = sqlite3OsSync(pPager->fd, pPager->syncFlags);
44067	43956	}else if( isOpen(pPager->fd) ){
44068	43957	assert( !MEMDB );
44069		- sqlite3OsFileControl(pPager->fd, SQLITE_FCNTL_SYNC_OMITTED, (void *)&rc);
	43958	+ rc = sqlite3OsFileControl(pPager->fd, SQLITE_FCNTL_SYNC_OMITTED, 0);
	43959	+ if( rc==SQLITE_NOTFOUND ){
	43960	+ rc = SQLITE_OK;
	43961	+ }
44070	43962	}
44071	43963	return rc;
44072	43964	}
44073	43965
44074	43966	/*
		@@ -48125,11 +48017,11 @@
48125	48017	iFrame = pWal->hdr.mxFrame;
48126	48018	for(p=pList; p && rc==SQLITE_OK; p=p->pDirty){
48127	48019	iFrame++;
48128	48020	rc = walIndexAppend(pWal, iFrame, p->pgno);
48129	48021	}
48130		- while( nExtra>0 && rc==SQLITE_OK ){
	48022	+ while( rc==SQLITE_OK && nExtra>0 ){
48131	48023	iFrame++;
48132	48024	nExtra--;
48133	48025	rc = walIndexAppend(pWal, iFrame, pLast->pgno);
48134	48026	}
48135	48027
		@@ -48726,11 +48618,11 @@
48726	48618	#define TRANS_WRITE 2
48727	48619
48728	48620	/*
48729	48621	** An instance of this object represents a single database file.
48730	48622	**
48731		-** A single database file can be in use as the same time by two
	48623	+** A single database file can be in use at the same time by two
48732	48624	** or more database connections. When two or more connections are
48733	48625	** sharing the same database file, each connection has it own
48734	48626	** private Btree object for the file and each of those Btrees points
48735	48627	** to this one BtShared object. BtShared.nRef is the number of
48736	48628	** connections currently sharing this database file.
		@@ -48832,11 +48724,11 @@
48832	48724	** b-tree within a database file.
48833	48725	**
48834	48726	** The entry is identified by its MemPage and the index in
48835	48727	** MemPage.aCell[] of the entry.
48836	48728	**
48837		-** A single database file can shared by two more database connections,
	48729	+** A single database file can be shared by two more database connections,
48838	48730	** but cursors cannot be shared. Each cursor is associated with a
48839	48731	** particular database connection identified BtCursor.pBtree.db.
48840	48732	**
48841	48733	** Fields in this structure are accessed under the BtShared.mutex
48842	48734	** found at self->pBt->mutex.
		@@ -48993,11 +48885,11 @@
48993	48885	int mallocFailed; /* A memory allocation error has occurred */
48994	48886	StrAccum errMsg; /* Accumulate the error message text here */
48995	48887	};
48996	48888
48997	48889	/*
48998		-** Read or write a two- and four-byte big-endian integer values.
	48890	+** Routines to read or write a two- and four-byte big-endian integer values.
48999	48891	*/
49000	48892	#define get2byte(x) ((x)[0]<<8 \| (x)[1])
49001	48893	#define put2byte(p,v) ((p)[0] = (u8)((v)>>8), (p)[1] = (u8)(v))
49002	48894	#define get4byte sqlite3Get4byte
49003	48895	#define put4byte sqlite3Put4byte
		@@ -55478,12 +55370,18 @@
55478	55370	}
55479	55371
55480	55372	/* Either we found one or more cells (cntnew[0])>0) or pPage is
55481	55373	** a virtual root page. A virtual root page is when the real root
55482	55374	** page is page 1 and we are the only child of that page.
	55375	+ **
	55376	+ ** UPDATE: The assert() below is not necessarily true if the database
	55377	+ ** file is corrupt. The corruption will be detected and reported later
	55378	+ ** in this procedure so there is no need to act upon it now.
55483	55379	*/
	55380	+#if 0
55484	55381	assert( cntNew[0]>0 \|\| (pParent->pgno==1 && pParent->nCell==0) );
	55382	+#endif
55485	55383
55486	55384	TRACE(("BALANCE: old: %d %d %d ",
55487	55385	apOld[0]->pgno,
55488	55386	nOld>=2 ? apOld[1]->pgno : 0,
55489	55387	nOld>=3 ? apOld[2]->pgno : 0
		@@ -58194,11 +58092,13 @@
58194	58092
58195	58093	assert( sqlite3BtreeIsInTrans(pTo) );
58196	58094	pFd = sqlite3PagerFile(sqlite3BtreePager(pTo));
58197	58095	if( pFd->pMethods ){
58198	58096	i64 nByte = sqlite3BtreeGetPageSize(pFrom)*(i64)sqlite3BtreeLastPage(pFrom);
	58097	+ sqlite3BeginBenignMalloc();
58199	58098	sqlite3OsFileControl(pFd, SQLITE_FCNTL_OVERWRITE, &nByte);
	58099	+ sqlite3EndBenignMalloc();
58200	58100	}
58201	58101
58202	58102	/* Set up an sqlite3_backup object. sqlite3_backup.pDestDb must be set
58203	58103	** to 0. This is used by the implementations of sqlite3_backup_step()
58204	58104	** and sqlite3_backup_finish() to detect that they are being called
		@@ -58525,11 +58425,11 @@
58525	58425	** Release any memory held by the Mem. This may leave the Mem in an
58526	58426	** inconsistent state, for example with (Mem.z==0) and
58527	58427	** (Mem.type==SQLITE_TEXT).
58528	58428	*/
58529	58429	SQLITE_PRIVATE void sqlite3VdbeMemRelease(Mem *p){
58530		- MemReleaseExt(p);
	58430	+ VdbeMemRelease(p);
58531	58431	sqlite3DbFree(p->db, p->zMalloc);
58532	58432	p->z = 0;
58533	58433	p->zMalloc = 0;
58534	58434	p->xDel = 0;
58535	58435	}
		@@ -58821,11 +58721,11 @@
58821	58721	** copies of this cell as invalid.
58822	58722	**
58823	58723	** This is used for testing and debugging only - to make sure shallow
58824	58724	** copies are not misused.
58825	58725	*/
58826		-SQLITE_PRIVATE void sqlite3VdbeMemPrepareToChange(Vdbe pVdbe, Mem pMem){
	58726	+SQLITE_PRIVATE void sqlite3VdbeMemAboutToChange(Vdbe pVdbe, Mem pMem){
58827	58727	int i;
58828	58728	Mem *pX;
58829	58729	for(i=1, pX=&pVdbe->aMem[1]; i<=pVdbe->nMem; i++, pX++){
58830	58730	if( pX->pScopyFrom==pMem ){
58831	58731	pX->flags \|= MEM_Invalid;
		@@ -58847,11 +58747,11 @@
58847	58747	** pFrom->z is used, then pTo->z points to the same thing as pFrom->z
58848	58748	** and flags gets srcType (either MEM_Ephem or MEM_Static).
58849	58749	*/
58850	58750	SQLITE_PRIVATE void sqlite3VdbeMemShallowCopy(Mem pTo, const Mem pFrom, int srcType){
58851	58751	assert( (pFrom->flags & MEM_RowSet)==0 );
58852		- MemReleaseExt(pTo);
	58752	+ VdbeMemRelease(pTo);
58853	58753	memcpy(pTo, pFrom, MEMCELLSIZE);
58854	58754	pTo->xDel = 0;
58855	58755	if( (pFrom->flags&MEM_Static)==0 ){
58856	58756	pTo->flags &= ~(MEM_Dyn\|MEM_Static\|MEM_Ephem);
58857	58757	assert( srcType==MEM_Ephem \|\| srcType==MEM_Static );
		@@ -58865,11 +58765,11 @@
58865	58765	*/
58866	58766	SQLITE_PRIVATE int sqlite3VdbeMemCopy(Mem pTo, const Mem pFrom){
58867	58767	int rc = SQLITE_OK;
58868	58768
58869	58769	assert( (pFrom->flags & MEM_RowSet)==0 );
58870		- MemReleaseExt(pTo);
	58770	+ VdbeMemRelease(pTo);
58871	58771	memcpy(pTo, pFrom, MEMCELLSIZE);
58872	58772	pTo->flags &= ~MEM_Dyn;
58873	58773
58874	58774	if( pTo->flags&(MEM_Str\|MEM_Blob) ){
58875	58775	if( 0==(pFrom->flags&MEM_Static) ){
		@@ -59202,11 +59102,11 @@
59202	59102	assert( (pVal->flags & (MEM_Ephem\|MEM_Static))!=0 );
59203	59103	if( sqlite3VdbeMemMakeWriteable(pVal)!=SQLITE_OK ){
59204	59104	return 0;
59205	59105	}
59206	59106	}
59207		- sqlite3VdbeMemNulTerminate(pVal); /* IMP: R-59893-45467 */
	59107	+ sqlite3VdbeMemNulTerminate(pVal); /* IMP: R-31275-44060 */
59208	59108	}else{
59209	59109	assert( (pVal->flags&MEM_Blob)==0 );
59210	59110	sqlite3VdbeMemStringify(pVal, enc);
59211	59111	assert( 0==(1&SQLITE_PTR_TO_INT(pVal->z)) );
59212	59112	}
		@@ -59583,11 +59483,12 @@
59583	59483	return addr;
59584	59484	}
59585	59485
59586	59486	/*
59587	59487	** Add an OP_ParseSchema opcode. This routine is broken out from
59588		-** sqlite3VdbeAddOp4() since it needs to also local all btrees.
	59488	+** sqlite3VdbeAddOp4() since it needs to also needs to mark all btrees
	59489	+** as having been used.
59589	59490	**
59590	59491	** The zWhere string must have been obtained from sqlite3_malloc().
59591	59492	** This routine will take ownership of the allocated memory.
59592	59493	*/
59593	59494	SQLITE_PRIVATE void sqlite3VdbeAddParseSchemaOp(Vdbe p, int iDb, char zWhere){
		@@ -60350,12 +60251,13 @@
60350	60251
60351	60252	/*
60352	60253	** Declare to the Vdbe that the BTree object at db->aDb[i] is used.
60353	60254	**
60354	60255	** The prepared statements need to know in advance the complete set of
60355		-** attached databases that they will be using. A mask of these databases
60356		-** is maintained in p->btreeMask and is used for locking and other purposes.
	60256	+** attached databases that will be use. A mask of these databases
	60257	+** is maintained in p->btreeMask. The p->lockMask value is the subset of
	60258	+** p->btreeMask of databases that will require a lock.
60357	60259	*/
60358	60260	SQLITE_PRIVATE void sqlite3VdbeUsesBtree(Vdbe *p, int i){
60359	60261	assert( i>=0 && i<p->db->nDb && i<(int)sizeof(yDbMask)*8 );
60360	60262	assert( i<(int)sizeof(p->btreeMask)*8 );
60361	60263	p->btreeMask \|= ((yDbMask)1)<<i;
		@@ -60874,10 +60776,11 @@
60874	60776	nVar = pParse->nVar;
60875	60777	nMem = pParse->nMem;
60876	60778	nCursor = pParse->nTab;
60877	60779	nArg = pParse->nMaxArg;
60878	60780	nOnce = pParse->nOnce;
	60781	+ if( nOnce==0 ) nOnce = 1; /* Ensure at least one byte in p->aOnceFlag[] */
60879	60782
60880	60783	/* For each cursor required, also allocate a memory cell. Memory
60881	60784	** cells (nMem+1-nCursor)..nMem, inclusive, will never be used by
60882	60785	** the vdbe program. Instead they are used to allocate space for
60883	60786	** VdbeCursor/BtCursor structures. The blob of memory associated with
		@@ -61225,11 +61128,11 @@
61225	61128	int retryCount = 0;
61226	61129	int nMainFile;
61227	61130
61228	61131	/* Select a master journal file name */
61229	61132	nMainFile = sqlite3Strlen30(zMainFile);
61230		- zMaster = sqlite3MPrintf(db, "%s-mjXXXXXX9XX", zMainFile);
	61133	+ zMaster = sqlite3MPrintf(db, "%s-mjXXXXXX9XXz", zMainFile);
61231	61134	if( zMaster==0 ) return SQLITE_NOMEM;
61232	61135	do {
61233	61136	u32 iRandom;
61234	61137	if( retryCount ){
61235	61138	if( retryCount>100 ){
		@@ -63013,11 +62916,11 @@
63013	62916	** since any application that receives an SQLITE_MISUSE is broken by
63014	62917	** definition.
63015	62918	**
63016	62919	** Nevertheless, some published applications that were originally written
63017	62920	** for version 3.6.23 or earlier do in fact depend on SQLITE_MISUSE
63018		- ** returns, and the so were broken by the automatic-reset change. As a
	62921	+ ** returns, and those were broken by the automatic-reset change. As a
63019	62922	** a work-around, the SQLITE_OMIT_AUTORESET compile-time restores the
63020	62923	** legacy behavior of returning SQLITE_MISUSE for cases where the
63021	62924	** previous sqlite3_step() returned something other than a SQLITE_LOCKED
63022	62925	** or SQLITE_BUSY error.
63023	62926	*/
		@@ -63359,17 +63262,17 @@
63359	63262	pOut = &pVm->pResultSet[i];
63360	63263	}else{
63361	63264	/* If the value passed as the second argument is out of range, return
63362	63265	** a pointer to the following static Mem object which contains the
63363	63266	** value SQL NULL. Even though the Mem structure contains an element
63364		- ** of type i64, on certain architecture (x86) with certain compiler
	63267	+ ** of type i64, on certain architectures (x86) with certain compiler
63365	63268	** switches (-Os), gcc may align this Mem object on a 4-byte boundary
63366	63269	** instead of an 8-byte one. This all works fine, except that when
63367	63270	** running with SQLITE_DEBUG defined the SQLite code sometimes assert()s
63368	63271	** that a Mem structure is located on an 8-byte boundary. To prevent
63369		- ** this assert() from failing, when building with SQLITE_DEBUG defined
63370		- ** using gcc, force nullMem to be 8-byte aligned using the magical
	63272	+ ** these assert()s from failing, when building with SQLITE_DEBUG defined
	63273	+ ** using gcc, we force nullMem to be 8-byte aligned using the magical
63371	63274	** __attribute__((aligned(8))) macro. */
63372	63275	static const Mem nullMem
63373	63276	#if defined(SQLITE_DEBUG) && defined(__GNUC__)
63374	63277	__attribute__((aligned(8)))
63375	63278	#endif
		@@ -64298,11 +64201,11 @@
64298	64201	** Invoke this macro on memory cells just prior to changing the
64299	64202	** value of the cell. This macro verifies that shallow copies are
64300	64203	** not misused.
64301	64204	*/
64302	64205	#ifdef SQLITE_DEBUG
64303		-# define memAboutToChange(P,M) sqlite3VdbeMemPrepareToChange(P,M)
	64206	+# define memAboutToChange(P,M) sqlite3VdbeMemAboutToChange(P,M)
64304	64207	#else
64305	64208	# define memAboutToChange(P,M)
64306	64209	#endif
64307	64210
64308	64211	/*
		@@ -64316,12 +64219,12 @@
64316	64219	SQLITE_API int sqlite3_search_count = 0;
64317	64220	#endif
64318	64221
64319	64222	/*
64320	64223	** When this global variable is positive, it gets decremented once before
64321		-** each instruction in the VDBE. When reaches zero, the u1.isInterrupted
64322		-** field of the sqlite3 structure is set in order to simulate and interrupt.
	64224	+** each instruction in the VDBE. When it reaches zero, the u1.isInterrupted
	64225	+** field of the sqlite3 structure is set in order to simulate an interrupt.
64323	64226	**
64324	64227	** This facility is used for testing purposes only. It does not function
64325	64228	** in an ordinary build.
64326	64229	*/
64327	64230	#ifdef SQLITE_TEST
		@@ -64442,11 +64345,11 @@
64442	64345	*/
64443	64346	static VdbeCursor *allocateCursor(
64444	64347	Vdbe p, / The virtual machine */
64445	64348	int iCur, /* Index of the new VdbeCursor */
64446	64349	int nField, /* Number of fields in the table or index */
64447		- int iDb, /* When database the cursor belongs to, or -1 */
	64350	+ int iDb, /* Database the cursor belongs to, or -1 */
64448	64351	int isBtreeCursor /* True for B-Tree. False for pseudo-table or vtab */
64449	64352	){
64450	64353	/* Find the memory cell that will be used to store the blob of memory
64451	64354	** required for this VdbeCursor structure. It is convenient to use a
64452	64355	** vdbe memory cell to manage the memory allocation required for a
		@@ -64813,11 +64716,11 @@
64813	64716	** sqlite3_interrupt() routine has been called. If it has been, then
64814	64717	** processing of the VDBE program is interrupted.
64815	64718	**
64816	64719	** This macro added to every instruction that does a jump in order to
64817	64720	** implement a loop. This test used to be on every single instruction,
64818		-** but that meant we more testing that we needed. By only testing the
	64721	+** but that meant we more testing than we needed. By only testing the
64819	64722	** flag on jump instructions, we get a (small) speed improvement.
64820	64723	*/
64821	64724	#define CHECK_FOR_INTERRUPT \
64822	64725	if( db->u1.isInterrupted ) goto abort_due_to_interrupt;
64823	64726
		@@ -65430,11 +65333,11 @@
65430	65333	if( pOp->opflags & OPFLG_OUT2_PRERELEASE ){
65431	65334	assert( pOp->p2>0 );
65432	65335	assert( pOp->p2<=p->nMem );
65433	65336	pOut = &aMem[pOp->p2];
65434	65337	memAboutToChange(p, pOut);
65435		- MemReleaseExt(pOut);
	65338	+ VdbeMemRelease(pOut);
65436	65339	pOut->flags = MEM_Int;
65437	65340	}
65438	65341
65439	65342	/* Sanity checking on other operands */
65440	65343	#ifdef SQLITE_DEBUG
		@@ -65738,11 +65641,11 @@
65738	65641	assert( pOp->p3<=p->nMem );
65739	65642	pOut->flags = MEM_Null;
65740	65643	while( u.ab.cnt>0 ){
65741	65644	pOut++;
65742	65645	memAboutToChange(p, pOut);
65743		- MemReleaseExt(pOut);
	65646	+ VdbeMemRelease(pOut);
65744	65647	pOut->flags = MEM_Null;
65745	65648	u.ab.cnt--;
65746	65649	}
65747	65650	break;
65748	65651	}
		@@ -65919,11 +65822,11 @@
65919	65822	/* Invalidate all ephemeral cursor row caches */
65920	65823	p->cacheCtr = (p->cacheCtr + 2)\|1;
65921	65824
65922	65825	/* Make sure the results of the current row are \000 terminated
65923	65826	** and have an assigned type. The results are de-ephemeralized as
65924		- ** as side effect.
	65827	+ ** a side effect.
65925	65828	*/
65926	65829	u.ae.pMem = p->pResultSet = &aMem[pOp->p1];
65927	65830	for(u.ae.i=0; u.ae.i<pOp->p2; u.ae.i++){
65928	65831	assert( memIsValid(&u.ae.pMem[u.ae.i]) );
65929	65832	Deephemeralize(&u.ae.pMem[u.ae.i]);
		@@ -67160,11 +67063,11 @@
67160	67063	** a pointer to a Mem object.
67161	67064	*/
67162	67065	if( u.an.aOffset[u.an.p2] ){
67163	67066	assert( rc==SQLITE_OK );
67164	67067	if( u.an.zRec ){
67165		- MemReleaseExt(u.an.pDest);
	67068	+ VdbeMemRelease(u.an.pDest);
67166	67069	sqlite3VdbeSerialGet((u8 *)&u.an.zRec[u.an.aOffset[u.an.p2]], u.an.aType[u.an.p2], u.an.pDest);
67167	67070	}else{
67168	67071	u.an.len = sqlite3VdbeSerialTypeLen(u.an.aType[u.an.p2]);
67169	67072	sqlite3VdbeMemMove(&u.an.sMem, u.an.pDest);
67170	67073	rc = sqlite3VdbeMemFromBtree(u.an.pCrsr, u.an.aOffset[u.an.p2], u.an.len, u.an.pC->isIndex, &u.an.sMem);
		@@ -83758,11 +83661,11 @@
83758	83661	** Allocate the index structure.
83759	83662	*/
83760	83663	nName = sqlite3Strlen30(zName);
83761	83664	nCol = pList->nExpr;
83762	83665	pIndex = sqlite3DbMallocZero(db,
83763		- sizeof(Index) + /* Index structure */
	83666	+ ROUND8(sizeof(Index)) + /* Index structure */
83764	83667	ROUND8(sizeof(tRowcnt)(nCol+1)) + / Index.aiRowEst */
83765	83668	sizeof(char )nCol + /* Index.azColl */
83766	83669	sizeof(int)nCol + / Index.aiColumn */
83767	83670	sizeof(u8)nCol + / Index.aSortOrder */
83768	83671	nName + 1 + /* Index.zName */
		@@ -83769,11 +83672,12 @@
83769	83672	nExtra /* Collation sequence names */
83770	83673	);
83771	83674	if( db->mallocFailed ){
83772	83675	goto exit_create_index;
83773	83676	}
83774		- pIndex->aiRowEst = (tRowcnt*)(&pIndex[1]);
	83677	+ zExtra = (char*)pIndex;
	83678	+ pIndex->aiRowEst = (tRowcnt*)&zExtra[ROUND8(sizeof(Index))];
83775	83679	pIndex->azColl = (char**)
83776	83680	((char)pIndex->aiRowEst + ROUND8(sizeof(tRowcnt)nCol+1));
83777	83681	assert( EIGHT_BYTE_ALIGNMENT(pIndex->aiRowEst) );
83778	83682	assert( EIGHT_BYTE_ALIGNMENT(pIndex->azColl) );
83779	83683	pIndex->aiColumn = (int *)(&pIndex->azColl[nCol]);
		@@ -89966,11 +89870,11 @@
89966	89870	** any ABORT Back out changes from the current command
89967	89871	** only (do not do a complete rollback) then
89968	89872	** cause sqlite3_exec() to return immediately
89969	89873	** with SQLITE_CONSTRAINT.
89970	89874	**
89971		-** any FAIL Sqlite_exec() returns immediately with a
	89875	+** any FAIL Sqlite3_exec() returns immediately with a
89972	89876	** return code of SQLITE_CONSTRAINT. The
89973	89877	** transaction is not rolled back and any
89974	89878	** prior changes are retained.
89975	89879	**
89976	89880	** any IGNORE The record number and data is popped from
		@@ -111450,11 +111354,11 @@
111450	111354	*tokenType = TK_SPACE;
111451	111355	return i;
111452	111356	}
111453	111357	case '-': {
111454	111358	if( z[1]=='-' ){
111455		- /* IMP: R-15891-05542 -- syntax diagram for comments */
	111359	+ /* IMP: R-50417-27976 -- syntax diagram for comments */
111456	111360	for(i=2; (c=z[i])!=0 && c!='\n'; i++){}
111457	111361	tokenType = TK_SPACE; / IMP: R-22934-25134 */
111458	111362	return i;
111459	111363	}
111460	111364	*tokenType = TK_MINUS;
		@@ -111483,11 +111387,11 @@
111483	111387	case '/': {
111484	111388	if( z[1]!='*' \|\| z[2]==0 ){
111485	111389	*tokenType = TK_SLASH;
111486	111390	return 1;
111487	111391	}
111488		- /* IMP: R-15891-05542 -- syntax diagram for comments */
	111392	+ /* IMP: R-50417-27976 -- syntax diagram for comments */
111489	111393	for(i=3, c=z[2]; (c!='*' \|\| z[i]!='/') && (c=z[i])!=0; i++){}
111490	111394	if( c ) i++;
111491	111395	tokenType = TK_SPACE; / IMP: R-22934-25134 */
111492	111396	return i;
111493	111397	}
		@@ -112277,12 +112181,12 @@
112277	112181	/* IMPLEMENTATION-OF: R-35210-63508 The sqlite3_libversion_number() function
112278	112182	** returns an integer equal to SQLITE_VERSION_NUMBER.
112279	112183	*/
112280	112184	SQLITE_API int sqlite3_libversion_number(void){ return SQLITE_VERSION_NUMBER; }
112281	112185
112282		-/* IMPLEMENTATION-OF: R-54823-41343 The sqlite3_threadsafe() function returns
112283		-** zero if and only if SQLite was compiled mutexing code omitted due to
	112186	+/* IMPLEMENTATION-OF: R-20790-14025 The sqlite3_threadsafe() function returns
	112187	+** zero if and only if SQLite was compiled with mutexing code omitted due to
112284	112188	** the SQLITE_THREADSAFE compile-time option being set to 0.
112285	112189	*/
112286	112190	SQLITE_API int sqlite3_threadsafe(void){ return SQLITE_THREADSAFE; }
112287	112191
112288	112192	#if !defined(SQLITE_OMIT_TRACE) && defined(SQLITE_ENABLE_IOTRACE)
		@@ -112768,19 +112672,21 @@
112768	112672	** Free up as much memory as we can from the given database
112769	112673	** connection.
112770	112674	*/
112771	112675	SQLITE_API int sqlite3_db_release_memory(sqlite3 *db){
112772	112676	int i;
	112677	+ sqlite3_mutex_enter(db->mutex);
112773	112678	sqlite3BtreeEnterAll(db);
112774	112679	for(i=0; i<db->nDb; i++){
112775	112680	Btree *pBt = db->aDb[i].pBt;
112776	112681	if( pBt ){
112777	112682	Pager *pPager = sqlite3BtreePager(pBt);
112778	112683	sqlite3PagerShrink(pPager);
112779	112684	}
112780	112685	}
112781	112686	sqlite3BtreeLeaveAll(db);
	112687	+ sqlite3_mutex_leave(db->mutex);
112782	112688	return SQLITE_OK;
112783	112689	}
112784	112690
112785	112691	/*
112786	112692	** Configuration settings for an individual database connection
112787	112693

	--- src/sqlite3.c
	+++ src/sqlite3.c
	@@ -657,11 +657,11 @@
657	** [sqlite3_libversion_number()], [sqlite3_sourceid()],
658	** [sqlite_version()] and [sqlite_source_id()].
659	*/
660	#define SQLITE_VERSION "3.7.10"
661	#define SQLITE_VERSION_NUMBER 3007010
662	#define SQLITE_SOURCE_ID "2011-12-23 13:32:07 1f24ae716df6232de768e245ea990049deee3c22"
663
664	/*
665	** CAPI3REF: Run-Time Library Version Numbers
666	** KEYWORDS: sqlite3_version, sqlite3_sourceid
667	**
	@@ -725,11 +725,11 @@
725
726	/*
727	** CAPI3REF: Test To See If The Library Is Threadsafe
728	**
729	** ^The sqlite3_threadsafe() function returns zero if and only if
730	** SQLite was compiled mutexing code omitted due to the
731	** [SQLITE_THREADSAFE] compile-time option being set to 0.
732	**
733	** SQLite can be compiled with or without mutexes. When
734	** the [SQLITE_THREADSAFE] C preprocessor macro is 1 or 2, mutexes
735	** are enabled and SQLite is threadsafe. When the
	@@ -919,11 +919,11 @@
919	** CAPI3REF: Result Codes
920	** KEYWORDS: SQLITE_OK {error code} {error codes}
921	** KEYWORDS: {result code} {result codes}
922	**
923	** Many SQLite functions return an integer result code from the set shown
924	** here in order to indicates success or failure.
925	**
926	** New error codes may be added in future versions of SQLite.
927	**
928	** See also: [SQLITE_IOERR_READ \| extended result codes],
929	** [sqlite3_vtab_on_conflict()] [SQLITE_ROLLBACK \| result codes].
	@@ -1058,17 +1058,14 @@
1058	** that when data is appended to a file, the data is appended
1059	** first then the size of the file is extended, never the other
1060	** way around. The SQLITE_IOCAP_SEQUENTIAL property means that
1061	** information is written to disk in the same order as calls
1062	** to xWrite(). The SQLITE_IOCAP_POWERSAFE_OVERWRITE property means that
1063	** after reboot following a crash or power loss, the value of
1064	** each byte in a file is a value that was actually written
1065	** into that byte at some point. In other words, a crash will
1066	** not cause unwritten bytes of the file to change nor introduce
1067	** randomness into a file nor zero out parts of the file, and any byte of
1068	** a file that are never written will not change values due to
1069	** writes to nearby bytes.
1070	*/
1071	#define SQLITE_IOCAP_ATOMIC 0x00000001
1072	#define SQLITE_IOCAP_ATOMIC512 0x00000002
1073	#define SQLITE_IOCAP_ATOMIC1K 0x00000004
1074	#define SQLITE_IOCAP_ATOMIC2K 0x00000008
	@@ -1300,16 +1297,16 @@
1300	** opcode as doing so may disrupt the operation of the specialized VFSes
1301	** that do require it.
1302	**
1303	** ^The [SQLITE_FCNTL_WIN32_AV_RETRY] opcode is used to configure automatic
1304	** retry counts and intervals for certain disk I/O operations for the
1305	** windows [VFS] in order to work to provide robustness against
1306	** anti-virus programs. By default, the windows VFS will retry file read,
1307	** file write, and file delete operations up to 10 times, with a delay
1308	** of 25 milliseconds before the first retry and with the delay increasing
1309	** by an additional 25 milliseconds with each subsequent retry. This
1310	** opcode allows those to values (10 retries and 25 milliseconds of delay)
1311	** to be adjusted. The values are changed for all database connections
1312	** within the same process. The argument is a pointer to an array of two
1313	** integers where the first integer i the new retry count and the second
1314	** integer is the delay. If either integer is negative, then the setting
1315	** is not changed but instead the prior value of that setting is written
	@@ -1420,11 +1417,11 @@
1420	** ^SQLite guarantees that the zFilename parameter to xOpen
1421	** is either a NULL pointer or string obtained
1422	** from xFullPathname() with an optional suffix added.
1423	** ^If a suffix is added to the zFilename parameter, it will
1424	** consist of a single "-" character followed by no more than
1425	** 10 alphanumeric and/or "-" characters.
1426	** ^SQLite further guarantees that
1427	** the string will be valid and unchanged until xClose() is
1428	** called. Because of the previous sentence,
1429	** the [sqlite3_file] can safely store a pointer to the
1430	** filename if it needs to remember the filename for some reason.
	@@ -2571,11 +2568,11 @@
2571	** These routines all implement some additional formatting
2572	** options that are useful for constructing SQL statements.
2573	** All of the usual printf() formatting options apply. In addition, there
2574	** is are "%q", "%Q", and "%z" options.
2575	**
2576	** ^(The %q option works like %s in that it substitutes a null-terminated
2577	** string from the argument list. But %q also doubles every '\'' character.
2578	** %q is designed for use inside a string literal.)^ By doubling each '\''
2579	** character it escapes that character and allows it to be inserted into
2580	** the string.
2581	**
	@@ -4091,11 +4088,11 @@
4091	** of the string. ^For clarity: the values returned by
4092	** [sqlite3_column_bytes()] and [sqlite3_column_bytes16()] are the number of
4093	** bytes in the string, not the number of characters.
4094	**
4095	** ^Strings returned by sqlite3_column_text() and sqlite3_column_text16(),
4096	** even empty strings, are always zero terminated. ^The return
4097	** value from sqlite3_column_blob() for a zero-length BLOB is a NULL pointer.
4098	**
4099	** ^The object returned by [sqlite3_column_value()] is an
4100	** [unprotected sqlite3_value] object. An unprotected sqlite3_value object
4101	** may only be used with [sqlite3_bind_value()] and [sqlite3_result_value()].
	@@ -5171,11 +5168,11 @@
5171	SQLITE_API int sqlite3_release_memory(int);
5172
5173	/*
5174	** CAPI3REF: Free Memory Used By A Database Connection
5175	**
5176	** ^The sqlite3_db_shrink(D) interface attempts to free as much heap
5177	** memory as possible from database connection D. Unlike the
5178	** [sqlite3_release_memory()] interface, this interface is effect even
5179	** when then [SQLITE_ENABLE_MEMORY_MANAGEMENT] compile-time option is
5180	** omitted.
5181	**
	@@ -5954,19 +5951,19 @@
5954	** is selected automatically at compile-time. ^(The following
5955	** implementations are available in the SQLite core:
5956	**
5957	** <ul>
5958	** <li> SQLITE_MUTEX_OS2
5959	** <li> SQLITE_MUTEX_PTHREAD
5960	** <li> SQLITE_MUTEX_W32
5961	** <li> SQLITE_MUTEX_NOOP
5962	** </ul>)^
5963	**
5964	** ^The SQLITE_MUTEX_NOOP implementation is a set of routines
5965	** that does no real locking and is appropriate for use in
5966	** a single-threaded application. ^The SQLITE_MUTEX_OS2,
5967	** SQLITE_MUTEX_PTHREAD, and SQLITE_MUTEX_W32 implementations
5968	** are appropriate for use on OS/2, Unix, and Windows.
5969	**
5970	** ^(If SQLite is compiled with the SQLITE_MUTEX_APPDEF preprocessor
5971	** macro defined (with "-DSQLITE_MUTEX_APPDEF=1"), then no mutex
5972	** implementation is included with the library. In this case the
	@@ -6152,11 +6149,11 @@
6152	** defined and if NDEBUG is not defined.
6153	**
6154	** ^These routines should return true if the mutex in their argument
6155	** is held or not held, respectively, by the calling thread.
6156	**
6157	** ^The implementation is not required to provided versions of these
6158	** routines that actually work. If the implementation does not provide working
6159	** versions of these routines, it should at least provide stubs that always
6160	** return true so that one does not get spurious assertion failures.
6161	**
6162	** ^If the argument to sqlite3_mutex_held() is a NULL pointer then
	@@ -6634,11 +6631,11 @@
6634	**
6635	** [[the xInit() page cache method]]
6636	** ^(The xInit() method is called once for each effective
6637	** call to [sqlite3_initialize()])^
6638	** (usually only once during the lifetime of the process). ^(The xInit()
6639	** method is passed a copy of the sqlite3_pcache_methods.pArg value.)^
6640	** The intent of the xInit() method is to set up global data structures
6641	** required by the custom page cache implementation.
6642	** ^(If the xInit() method is NULL, then the
6643	** built-in default page cache is used instead of the application defined
6644	** page cache.)^
	@@ -6660,11 +6657,11 @@
6660	**
6661	** [[the xCreate() page cache methods]]
6662	** ^SQLite invokes the xCreate() method to construct a new cache instance.
6663	** SQLite will typically create one cache instance for each open database file,
6664	** though this is not guaranteed. ^The
6665	** parameter parameter, szPage, is the size in bytes of the pages that must
6666	** be allocated by the cache. ^szPage will always a power of two. ^The
6667	** second parameter szExtra is a number of bytes of extra storage
6668	** associated with each page cache entry. ^The szExtra parameter will
6669	** a number less than 250. SQLite will use the
6670	** extra szExtra bytes on each page to store metadata about the underlying
	@@ -6755,11 +6752,11 @@
6755	**
6756	** [[the xDestroy() page cache method]]
6757	** ^The xDestroy() method is used to delete a cache allocated by xCreate().
6758	** All resources associated with the specified cache should be freed. ^After
6759	** calling the xDestroy() method, SQLite considers the [sqlite3_pcache*]
6760	** handle invalid, and will not use it with any other sqlite3_pcache_methods
6761	** functions.
6762	**
6763	** [[the xShrink() page cache method]]
6764	** ^SQLite invokes the xShrink() method when it wants the page cache to
6765	** free up as much of heap memory as possible. The page cache implementation
	@@ -9266,11 +9263,11 @@
9266
9267	/*
9268	** The default size of a disk sector
9269	*/
9270	#ifndef SQLITE_DEFAULT_SECTOR_SIZE
9271	# define SQLITE_DEFAULT_SECTOR_SIZE 512
9272	#endif
9273
9274	/*
9275	** Temporary files are named starting with this prefix followed by 16 random
9276	** alphanumeric characters, and no file extension. They are stored in the
	@@ -13190,11 +13187,11 @@
13190	SQLITE_PRIVATE int sqlite3VdbeMemRealify(Mem*);
13191	SQLITE_PRIVATE int sqlite3VdbeMemNumerify(Mem*);
13192	SQLITE_PRIVATE int sqlite3VdbeMemFromBtree(BtCursor,int,int,int,Mem);
13193	SQLITE_PRIVATE void sqlite3VdbeMemRelease(Mem *p);
13194	SQLITE_PRIVATE void sqlite3VdbeMemReleaseExternal(Mem *p);
13195	#define MemReleaseExt(X) \
13196	if((X)->flags&(MEM_Agg\|MEM_Dyn\|MEM_RowSet\|MEM_Frame)) \
13197	sqlite3VdbeMemReleaseExternal(X);
13198	SQLITE_PRIVATE int sqlite3VdbeMemFinalize(Mem, FuncDef);
13199	SQLITE_PRIVATE const char *sqlite3OpcodeName(int);
13200	SQLITE_PRIVATE int sqlite3VdbeMemGrow(Mem *pMem, int n, int preserve);
	@@ -13229,11 +13226,11 @@
13229	# define sqlite3VdbeEnter(X)
13230	# define sqlite3VdbeLeave(X)
13231	#endif
13232
13233	#ifdef SQLITE_DEBUG
13234	SQLITE_PRIVATE void sqlite3VdbeMemPrepareToChange(Vdbe,Mem);
13235	#endif
13236
13237	#ifndef SQLITE_OMIT_FOREIGN_KEY
13238	SQLITE_PRIVATE int sqlite3VdbeCheckFk(Vdbe *, int);
13239	#else
	@@ -14762,11 +14759,11 @@
14762	int flags,
14763	int *pFlagsOut
14764	){
14765	int rc;
14766	DO_OS_MALLOC_TEST(0);
14767	/* 0x87f3f is a mask of SQLITE_OPEN_ flags that are valid to be passed
14768	** down into the VFS layer. Some SQLITE_OPEN_ flags (for example,
14769	** SQLITE_OPEN_FULLMUTEX or SQLITE_OPEN_SHAREDCACHE) are blocked before
14770	** reaching the VFS. */
14771	rc = pVfs->xOpen(pVfs, zPath, pFile, flags & 0x87f7f, pFlagsOut);
14772	assert( rc==SQLITE_OK \|\| pFile->pMethods==0 );
	@@ -17316,11 +17313,11 @@
17316	SQLITE_API int sqlite3_mutex_notheld(sqlite3_mutex *p){
17317	return p==0 \|\| sqlite3GlobalConfig.mutex.xMutexNotheld(p);
17318	}
17319	#endif
17320
17321	#endif /* SQLITE_MUTEX_OMIT */
17322
17323	/************ End of mutex.c *********************************************/
17324	/************ Begin file mutex_noop.c ************************************/
17325	/*
17326	** 2008 October 07
	@@ -17523,12 +17520,12 @@
17523	*/
17524	#ifdef SQLITE_MUTEX_NOOP
17525	SQLITE_PRIVATE sqlite3_mutex_methods const *sqlite3DefaultMutex(void){
17526	return sqlite3NoopMutex();
17527	}
17528	#endif /* SQLITE_MUTEX_NOOP */
17529	#endif /* SQLITE_MUTEX_OMIT */
17530
17531	/************ End of mutex_noop.c ****************************************/
17532	/************ Begin file mutex_os2.c *************************************/
17533	/*
17534	** 2007 August 28
	@@ -18153,11 +18150,11 @@
18153	};
18154
18155	return &sMutex;
18156	}
18157
18158	#endif /* SQLITE_MUTEX_PTHREAD */
18159
18160	/************ End of mutex_unix.c ****************************************/
18161	/************ Begin file mutex_w32.c *************************************/
18162	/*
18163	** 2007 August 14
	@@ -22116,26 +22113,31 @@
22116	** characters of the original suffix.
22117	**
22118	** If SQLITE_ENABLE_8_3_NAMES is set to 2 at compile-time, then always
22119	** do the suffix shortening regardless of URI parameter.
22120	**




22121	** Examples:
22122	**
22123	** test.db-journal => test.nal
22124	** test.db-wal => test.wal
22125	** test.db-shm => test.shm
22126	** test.db-mj7f3319fa => test.9fa
22127	*/
22128	SQLITE_PRIVATE void sqlite3FileSuffix3(const char zBaseFilename, char z){

22129	#if SQLITE_ENABLE_8_3_NAMES<2
22130	if( sqlite3_uri_boolean(zBaseFilename, "8_3_names") )
22131	#endif
22132	{
22133	int i, sz;
22134	sz = sqlite3Strlen30(z);
22135	for(i=sz-1; i>0 && z[i]!='/' && z[i]!='.'; i--){}
22136	if( z[i]=='.' && ALWAYS(sz>i+4) ) memmove(&z[i+1], &z[sz-3], 4);
22137	}
22138	}
22139	#endif
22140
22141	/************ End of util.c **********************************************/
	@@ -24836,13 +24838,10 @@
24836	#include <sys/time.h>
24837	#include <errno.h>
24838	#ifndef SQLITE_OMIT_WAL
24839	#include <sys/mman.h>
24840	#endif
24841	#ifndef MISSING_STATVFS
24842	#include <sys/statvfs.h>
24843	#endif
24844
24845
24846	#if SQLITE_ENABLE_LOCKING_STYLE
24847	# include <sys/ioctl.h>
24848	# if OS_VXWORKS
	@@ -24929,11 +24928,10 @@
24929	sqlite3_vfs pVfs; / The VFS that created this unixFile */
24930	unixInodeInfo pInode; / Info about locks on this inode */
24931	int h; /* The file descriptor */
24932	unsigned char eFileLock; /* The type of lock held on this fd */
24933	unsigned char ctrlFlags; /* Behavioral bits. UNIXFILE_* flags */
24934	unsigned char szSector; /* Sectorsize/512 */
24935	int lastErrno; /* The unix errno from last I/O error */
24936	void lockingContext; / Locking style specific state */
24937	UnixUnusedFd pUnused; / Pre-allocated UnixUnusedFd */
24938	const char zPath; / Name of the file */
24939	unixShm pShm; / Shared memory segment information */
	@@ -25342,18 +25340,10 @@
25342	#define osMkdir ((int()(const char,mode_t))aSyscall[18].pCurrent)
25343
25344	{ "rmdir", (sqlite3_syscall_ptr)rmdir, 0 },
25345	#define osRmdir ((int()(const char))aSyscall[19].pCurrent)
25346
25347	#if defined(MISSING_STATVFS)
25348	{ "statvfs", (sqlite3_syscall_ptr)0, 0 },
25349	#define osStatvfs ((int()(const char,void*))aSyscall[20].pCurrent)
25350	#else
25351	{ "statvfs", (sqlite3_syscall_ptr)statvfs, 0 },
25352	#define osStatvfs ((int()(const char,struct statvfs*))aSyscall[20].pCurrent)
25353	#endif
25354
25355	}; /* End of the overrideable system calls */
25356
25357	/*
25358	** This is the xSetSystemCall() method of sqlite3_vfs for all of the
25359	** "unix" VFSes. Return SQLITE_OK opon successfully updating the
	@@ -28504,13 +28494,10 @@
28504	case SQLITE_SET_LOCKPROXYFILE:
28505	case SQLITE_GET_LOCKPROXYFILE: {
28506	return proxyFileControl(id,op,pArg);
28507	}
28508	#endif /* SQLITE_ENABLE_LOCKING_STYLE && defined(__APPLE__) */
28509	case SQLITE_FCNTL_SYNC_OMITTED: {
28510	return SQLITE_OK; /* A no-op */
28511	}
28512	}
28513	return SQLITE_NOTFOUND;
28514	}
28515
28516	/*
	@@ -28522,27 +28509,12 @@
28522	** if two files are created in the same file-system directory (i.e.
28523	** a database and its journal file) that the sector size will be the
28524	** same for both.
28525	*/
28526	static int unixSectorSize(sqlite3_file *pFile){
28527	unixFile p = (unixFile)pFile;
28528	if( p->szSector==0 ){
28529	#ifdef MISSING_STATVFS
28530	p->szSector = SQLITE_DEFAULT_SECTOR_SIZE/512;
28531	#else
28532	struct statvfs x;
28533	int sz;
28534	memset(&x, 0, sizeof(x));
28535	osStatvfs(p->zPath, &x);
28536	sz = (int)x.f_frsize;
28537	if( sz<512 \|\| sz>65536 \|\| (sz&(sz-1))!=0 ){
28538	sz = SQLITE_DEFAULT_SECTOR_SIZE;
28539	}
28540	p->szSector = sz/512;
28541	#endif
28542	}
28543	return p->szSector*512;
28544	}
28545
28546	/*
28547	** Return the device characteristics for the file.
28548	**
	@@ -28815,13 +28787,13 @@
28815	rc = SQLITE_IOERR_FSTAT;
28816	goto shm_open_err;
28817	}
28818
28819	#ifdef SQLITE_SHM_DIRECTORY
28820	nShmFilename = sizeof(SQLITE_SHM_DIRECTORY) + 30;
28821	#else
28822	nShmFilename = 5 + (int)strlen(pDbFd->zPath);
28823	#endif
28824	pShmNode = sqlite3_malloc( sizeof(*pShmNode) + nShmFilename );
28825	if( pShmNode==0 ){
28826	rc = SQLITE_NOMEM;
28827	goto shm_open_err;
	@@ -31755,11 +31727,11 @@
31755	};
31756	unsigned int i; /* Loop counter */
31757
31758	/* Double-check that the aSyscall[] array has been constructed
31759	** correctly. See ticket [bb3a86e890c8e96ab] */
31760	assert( ArraySize(aSyscall)==21 );
31761
31762	/* Register all VFSes defined in the aVfs[] array */
31763	for(i=0; i<(sizeof(aVfs)/sizeof(sqlite3_vfs)); i++){
31764	sqlite3_vfs_register(&aVfs[i], i==0);
31765	}
	@@ -32049,11 +32021,10 @@
32049	HANDLE h; /* Handle for accessing the file */
32050	u8 locktype; /* Type of lock currently held on this file */
32051	short sharedLockByte; /* Randomly chosen byte used as a shared lock */
32052	u8 ctrlFlags; /* Flags. See WINFILE_* below */
32053	DWORD lastErrno; /* The Windows errno from the last I/O error */
32054	DWORD sectorSize; /* Sector size of the device file is on */
32055	winShm pShm; / Instance of shared memory on this file */
32056	const char zPath; / Full pathname of this file */
32057	int szChunk; /* Chunk size configured by FCNTL_CHUNK_SIZE */
32058	#if SQLITE_OS_WINCE
32059	LPWSTR zDeleteOnClose; /* Name of file to delete when closing */
	@@ -32139,18 +32110,10 @@
32139	static void winMemShutdown(void *pAppData);
32140
32141	SQLITE_PRIVATE const sqlite3_mem_methods *sqlite3MemGetWin32(void);
32142	#endif /* SQLITE_WIN32_MALLOC */
32143
32144	/*
32145	** Forward prototypes.
32146	*/
32147	static int getSectorSize(
32148	sqlite3_vfs *pVfs,
32149	const char zRelative / UTF-8 file name */
32150	);
32151
32152	/*
32153	** The following variable is (normally) set once and never changes
32154	** thereafter. It records whether the operating system is Win9x
32155	** or WinNT.
32156	**
	@@ -34181,13 +34144,10 @@
34181	}
34182	case SQLITE_FCNTL_VFSNAME: {
34183	(char*)pArg = sqlite3_mprintf("win32");
34184	return SQLITE_OK;
34185	}
34186	case SQLITE_FCNTL_SYNC_OMITTED: {
34187	return SQLITE_OK;
34188	}
34189	case SQLITE_FCNTL_WIN32_AV_RETRY: {
34190	int a = (int)pArg;
34191	if( a[0]>0 ){
34192	win32IoerrRetry = a[0];
34193	}else{
	@@ -34213,12 +34173,12 @@
34213	** if two files are created in the same file-system directory (i.e.
34214	** a database and its journal file) that the sector size will be the
34215	** same for both.
34216	*/
34217	static int winSectorSize(sqlite3_file *id){
34218	assert( id!=0 );
34219	return (int)(((winFile*)id)->sectorSize);
34220	}
34221
34222	/*
34223	** Return a vector of device characteristics.
34224	*/
	@@ -34463,11 +34423,11 @@
34463	*/
34464	p = sqlite3_malloc( sizeof(*p) );
34465	if( p==0 ) return SQLITE_IOERR_NOMEM;
34466	memset(p, 0, sizeof(*p));
34467	nName = sqlite3Strlen30(pDbFd->zPath);
34468	pNew = sqlite3_malloc( sizeof(*pShmNode) + nName + 15 );
34469	if( pNew==0 ){
34470	sqlite3_free(p);
34471	return SQLITE_IOERR_NOMEM;
34472	}
34473	memset(pNew, 0, sizeof(*pNew));
	@@ -35191,11 +35151,10 @@
35191	pFile->pShm = 0;
35192	pFile->zPath = zName;
35193	if( sqlite3_uri_boolean(zName, "psow", SQLITE_POWERSAFE_OVERWRITE) ){
35194	pFile->ctrlFlags \|= WINFILE_PSOW;
35195	}
35196	pFile->sectorSize = getSectorSize(pVfs, zUtf8Name);
35197
35198	#if SQLITE_OS_WINCE
35199	if( isReadWrite && eType==SQLITE_OPEN_MAIN_DB
35200	&& !winceCreateLock(zName, pFile)
35201	){
	@@ -35436,85 +35395,10 @@
35436	return SQLITE_IOERR_NOMEM;
35437	}
35438	#endif
35439	}
35440
35441	/*
35442	** Get the sector size of the device used to store
35443	** file.
35444	*/
35445	static int getSectorSize(
35446	sqlite3_vfs *pVfs,
35447	const char zRelative / UTF-8 file name */
35448	){
35449	DWORD bytesPerSector = SQLITE_DEFAULT_SECTOR_SIZE;
35450	/* GetDiskFreeSpace is not supported under WINCE */
35451	#if SQLITE_OS_WINCE
35452	UNUSED_PARAMETER(pVfs);
35453	UNUSED_PARAMETER(zRelative);
35454	#else
35455	char zFullpath[MAX_PATH+1];
35456	int rc;
35457	DWORD dwRet = 0;
35458	DWORD dwDummy;
35459
35460	/*
35461	** We need to get the full path name of the file
35462	** to get the drive letter to look up the sector
35463	** size.
35464	*/
35465	SimulateIOErrorBenign(1);
35466	sqlite3BeginBenignMalloc();
35467	rc = winFullPathname(pVfs, zRelative, MAX_PATH, zFullpath);
35468	sqlite3EndBenignMalloc();
35469	SimulateIOErrorBenign(0);
35470	if( rc == SQLITE_OK )
35471	{
35472	void *zConverted;
35473	sqlite3BeginBenignMalloc();
35474	zConverted = convertUtf8Filename(zFullpath);
35475	sqlite3EndBenignMalloc();
35476	if( zConverted ){
35477	if( isNT() ){
35478	/* trim path to just drive reference */
35479	LPWSTR p = zConverted;
35480	for(;*p;p++){
35481	if( *p == '\\' ){
35482	*p = '\0';
35483	break;
35484	}
35485	}
35486	dwRet = osGetDiskFreeSpaceW((LPCWSTR)zConverted,
35487	&dwDummy,
35488	&bytesPerSector,
35489	&dwDummy,
35490	&dwDummy);
35491	}else{
35492	/* trim path to just drive reference */
35493	char p = (char )zConverted;
35494	for(;*p;p++){
35495	if( *p == '\\' ){
35496	*p = '\0';
35497	break;
35498	}
35499	}
35500	dwRet = osGetDiskFreeSpaceA((char*)zConverted,
35501	&dwDummy,
35502	&bytesPerSector,
35503	&dwDummy,
35504	&dwDummy);
35505	}
35506	sqlite3_free(zConverted);
35507	}
35508	if( !dwRet ){
35509	bytesPerSector = SQLITE_DEFAULT_SECTOR_SIZE;
35510	}
35511	}
35512	#endif
35513	return (int) bytesPerSector;
35514	}
35515
35516	#ifndef SQLITE_OMIT_LOAD_EXTENSION
35517	/*
35518	** Interfaces for opening a shared library, finding entry points
35519	** within the shared library, and closing the shared library.
35520	*/
	@@ -36381,11 +36265,11 @@
36381	*/
36382	static int numberOfCachePages(PCache *p){
36383	if( p->szCache>=0 ){
36384	return p->szCache;
36385	}else{
36386	return (-1024*p->szCache)/(p->szPage+p->szExtra);
36387	}
36388	}
36389
36390	/*
36391	** Try to obtain a page from the cache.
	@@ -36845,14 +36729,14 @@
36845	** PGroup which is the pcache1.grp global variable and its mutex is
36846	** SQLITE_MUTEX_STATIC_LRU.
36847	*/
36848	struct PGroup {
36849	sqlite3_mutex mutex; / MUTEX_STATIC_LRU or NULL */
36850	int nMaxPage; /* Sum of nMax for purgeable caches */
36851	int nMinPage; /* Sum of nMin for purgeable caches */
36852	int mxPinned; /* nMaxpage + 10 - nMinPage */
36853	int nCurrentPage; /* Number of purgeable pages allocated */
36854	PgHdr1 pLruHead, pLruTail; /* LRU list of unpinned pages */
36855	};
36856
36857	/* Each page cache is an instance of the following object. Every
36858	** open database file (including each in-memory database and each
	@@ -37511,11 +37395,11 @@
37511	static sqlite3_pcache_page *pcache1Fetch(
37512	sqlite3_pcache *p,
37513	unsigned int iKey,
37514	int createFlag
37515	){
37516	int nPinned;
37517	PCache1 pCache = (PCache1 )p;
37518	PGroup *pGroup;
37519	PgHdr1 *pPage = 0;
37520
37521	assert( pCache->bPurgeable \|\| createFlag!=1 );
	@@ -37546,17 +37430,17 @@
37546	#ifdef SQLITE_MUTEX_OMIT
37547	pGroup = pCache->pGroup;
37548	#endif
37549
37550	/* Step 3: Abort if createFlag is 1 but the cache is nearly full */

37551	nPinned = pCache->nPage - pCache->nRecyclable;
37552	assert( nPinned>=0 );
37553	assert( pGroup->mxPinned == pGroup->nMaxPage + 10 - pGroup->nMinPage );
37554	assert( pCache->n90pct == pCache->nMax*9/10 );
37555	if( createFlag==1 && (
37556	nPinned>=pGroup->mxPinned
37557	\|\| nPinned>=(int)pCache->n90pct
37558	\|\| pcache1UnderMemoryPressure(pCache)
37559	)){
37560	goto fetch_out;
37561	}
37562
	@@ -37725,11 +37609,13 @@
37725	PCache1 pCache = (PCache1 )p;
37726	PGroup *pGroup = pCache->pGroup;
37727	assert( pCache->bPurgeable \|\| (pCache->nMax==0 && pCache->nMin==0) );
37728	pcache1EnterMutex(pGroup);
37729	pcache1TruncateUnsafe(pCache, 0);

37730	pGroup->nMaxPage -= pCache->nMax;

37731	pGroup->nMinPage -= pCache->nMin;
37732	pGroup->mxPinned = pGroup->nMaxPage + 10 - pGroup->nMinPage;
37733	pcache1EnforceMaxPage(pGroup);
37734	pcache1LeaveMutex(pGroup);
37735	sqlite3_free(pCache->apHash);
	@@ -37807,12 +37693,12 @@
37807	int nRecyclable = 0;
37808	for(p=pcache1.grp.pLruHead; p; p=p->pLruNext){
37809	nRecyclable++;
37810	}
37811	*pnCurrent = pcache1.grp.nCurrentPage;
37812	*pnMax = pcache1.grp.nMaxPage;
37813	*pnMin = pcache1.grp.nMinPage;
37814	*pnRecyclable = nRecyclable;
37815	}
37816	#endif
37817
37818	/************ End of pcache1.c *******************************************/
	@@ -38282,10 +38168,16 @@
38282
38283	#ifndef _WAL_H_
38284	#define _WAL_H_
38285
38286






38287	#ifdef SQLITE_OMIT_WAL
38288	# define sqlite3WalOpen(x,y,z) 0
38289	# define sqlite3WalLimit(x,y)
38290	# define sqlite3WalClose(w,x,y,z) 0
38291	# define sqlite3WalBeginReadTransaction(y,z) 0
	@@ -38350,16 +38242,10 @@
38350	SQLITE_PRIVATE int sqlite3WalSavepointUndo(Wal pWal, u32 aWalData);
38351
38352	/* Write a frame or frames to the log. */
38353	SQLITE_PRIVATE int sqlite3WalFrames(Wal pWal, int, PgHdr , Pgno, int, int);
38354
38355	/* Additional values that can be added to the sync_flags argument of
38356	** sqlite3WalFrames():
38357	*/
38358	#define WAL_SYNC_TRANSACTIONS 0x20 /* Sync at the end of each transaction */
38359	#define SQLITE_SYNC_MASK 0x13 /* Mask off the SQLITE_SYNC_* values */
38360
38361	/* Copy pages from the log to the database file */
38362	SQLITE_PRIVATE int sqlite3WalCheckpoint(
38363	Wal pWal, / Write-ahead log connection */
38364	int eMode, /* One of PASSIVE, FULL and RESTART */
38365	int (xBusy)(void), /* Function to call when busy */
	@@ -42400,11 +42286,13 @@
42400	** file size will be.
42401	*/
42402	assert( rc!=SQLITE_OK \|\| isOpen(pPager->fd) );
42403	if( rc==SQLITE_OK && pPager->dbSize>pPager->dbHintSize ){
42404	sqlite3_int64 szFile = pPager->pageSize * (sqlite3_int64)pPager->dbSize;

42405	sqlite3OsFileControl(pPager->fd, SQLITE_FCNTL_SIZE_HINT, &szFile);

42406	pPager->dbHintSize = pPager->dbSize;
42407	}
42408
42409	while( rc==SQLITE_OK && pList ){
42410	Pgno pgno = pList->pgno;
	@@ -42741,11 +42629,12 @@
42741	z = zUri = &zFilename[sqlite3Strlen30(zFilename)+1];
42742	while( *z ){
42743	z += sqlite3Strlen30(z)+1;
42744	z += sqlite3Strlen30(z)+1;
42745	}
42746	nUri = &z[1] - zUri;

42747	if( rc==SQLITE_OK && nPathname+8>pVfs->mxPathname ){
42748	/* This branch is taken when the journal path required by
42749	** the database being opened will be more than pVfs->mxPathname
42750	** bytes in length. This means the database cannot be opened,
42751	** as it will not be possible to open the journal file or even
	@@ -42775,13 +42664,13 @@
42775	ROUND8(sizeof(pPager)) + / Pager structure */
42776	ROUND8(pcacheSize) + /* PCache object */
42777	ROUND8(pVfs->szOsFile) + /* The main db file */
42778	journalFileSize * 2 + /* The two journal files */
42779	nPathname + 1 + nUri + /* zFilename */
42780	nPathname + 8 + 1 /* zJournal */
42781	#ifndef SQLITE_OMIT_WAL
42782	+ nPathname + 4 + 1 /* zWal */
42783	#endif
42784	);
42785	assert( EIGHT_BYTE_ALIGNMENT(SQLITE_INT_TO_PTR(journalFileSize)) );
42786	if( !pPtr ){
42787	sqlite3_free(zPathname);
	@@ -42800,16 +42689,16 @@
42800	assert( nPathname>0 );
42801	pPager->zJournal = (char*)(pPtr += nPathname + 1 + nUri);
42802	memcpy(pPager->zFilename, zPathname, nPathname);
42803	memcpy(&pPager->zFilename[nPathname+1], zUri, nUri);
42804	memcpy(pPager->zJournal, zPathname, nPathname);
42805	memcpy(&pPager->zJournal[nPathname], "-journal", 8);
42806	sqlite3FileSuffix3(pPager->zFilename, pPager->zJournal);
42807	#ifndef SQLITE_OMIT_WAL
42808	pPager->zWal = &pPager->zJournal[nPathname+8+1];
42809	memcpy(pPager->zWal, zPathname, nPathname);
42810	memcpy(&pPager->zWal[nPathname], "-wal", 4);
42811	sqlite3FileSuffix3(pPager->zFilename, pPager->zWal);
42812	#endif
42813	sqlite3_free(zPathname);
42814	}
42815	pPager->pVfs = pVfs;
	@@ -44064,11 +43953,14 @@
44064	if( !pPager->noSync ){
44065	assert( !MEMDB );
44066	rc = sqlite3OsSync(pPager->fd, pPager->syncFlags);
44067	}else if( isOpen(pPager->fd) ){
44068	assert( !MEMDB );
44069	sqlite3OsFileControl(pPager->fd, SQLITE_FCNTL_SYNC_OMITTED, (void *)&rc);



44070	}
44071	return rc;
44072	}
44073
44074	/*
	@@ -48125,11 +48017,11 @@
48125	iFrame = pWal->hdr.mxFrame;
48126	for(p=pList; p && rc==SQLITE_OK; p=p->pDirty){
48127	iFrame++;
48128	rc = walIndexAppend(pWal, iFrame, p->pgno);
48129	}
48130	while( nExtra>0 && rc==SQLITE_OK ){
48131	iFrame++;
48132	nExtra--;
48133	rc = walIndexAppend(pWal, iFrame, pLast->pgno);
48134	}
48135
	@@ -48726,11 +48618,11 @@
48726	#define TRANS_WRITE 2
48727
48728	/*
48729	** An instance of this object represents a single database file.
48730	**
48731	** A single database file can be in use as the same time by two
48732	** or more database connections. When two or more connections are
48733	** sharing the same database file, each connection has it own
48734	** private Btree object for the file and each of those Btrees points
48735	** to this one BtShared object. BtShared.nRef is the number of
48736	** connections currently sharing this database file.
	@@ -48832,11 +48724,11 @@
48832	** b-tree within a database file.
48833	**
48834	** The entry is identified by its MemPage and the index in
48835	** MemPage.aCell[] of the entry.
48836	**
48837	** A single database file can shared by two more database connections,
48838	** but cursors cannot be shared. Each cursor is associated with a
48839	** particular database connection identified BtCursor.pBtree.db.
48840	**
48841	** Fields in this structure are accessed under the BtShared.mutex
48842	** found at self->pBt->mutex.
	@@ -48993,11 +48885,11 @@
48993	int mallocFailed; /* A memory allocation error has occurred */
48994	StrAccum errMsg; /* Accumulate the error message text here */
48995	};
48996
48997	/*
48998	** Read or write a two- and four-byte big-endian integer values.
48999	*/
49000	#define get2byte(x) ((x)[0]<<8 \| (x)[1])
49001	#define put2byte(p,v) ((p)[0] = (u8)((v)>>8), (p)[1] = (u8)(v))
49002	#define get4byte sqlite3Get4byte
49003	#define put4byte sqlite3Put4byte
	@@ -55478,12 +55370,18 @@
55478	}
55479
55480	/* Either we found one or more cells (cntnew[0])>0) or pPage is
55481	** a virtual root page. A virtual root page is when the real root
55482	** page is page 1 and we are the only child of that page.




55483	*/

55484	assert( cntNew[0]>0 \|\| (pParent->pgno==1 && pParent->nCell==0) );

55485
55486	TRACE(("BALANCE: old: %d %d %d ",
55487	apOld[0]->pgno,
55488	nOld>=2 ? apOld[1]->pgno : 0,
55489	nOld>=3 ? apOld[2]->pgno : 0
	@@ -58194,11 +58092,13 @@
58194
58195	assert( sqlite3BtreeIsInTrans(pTo) );
58196	pFd = sqlite3PagerFile(sqlite3BtreePager(pTo));
58197	if( pFd->pMethods ){
58198	i64 nByte = sqlite3BtreeGetPageSize(pFrom)*(i64)sqlite3BtreeLastPage(pFrom);

58199	sqlite3OsFileControl(pFd, SQLITE_FCNTL_OVERWRITE, &nByte);

58200	}
58201
58202	/* Set up an sqlite3_backup object. sqlite3_backup.pDestDb must be set
58203	** to 0. This is used by the implementations of sqlite3_backup_step()
58204	** and sqlite3_backup_finish() to detect that they are being called
	@@ -58525,11 +58425,11 @@
58525	** Release any memory held by the Mem. This may leave the Mem in an
58526	** inconsistent state, for example with (Mem.z==0) and
58527	** (Mem.type==SQLITE_TEXT).
58528	*/
58529	SQLITE_PRIVATE void sqlite3VdbeMemRelease(Mem *p){
58530	MemReleaseExt(p);
58531	sqlite3DbFree(p->db, p->zMalloc);
58532	p->z = 0;
58533	p->zMalloc = 0;
58534	p->xDel = 0;
58535	}
	@@ -58821,11 +58721,11 @@
58821	** copies of this cell as invalid.
58822	**
58823	** This is used for testing and debugging only - to make sure shallow
58824	** copies are not misused.
58825	*/
58826	SQLITE_PRIVATE void sqlite3VdbeMemPrepareToChange(Vdbe pVdbe, Mem pMem){
58827	int i;
58828	Mem *pX;
58829	for(i=1, pX=&pVdbe->aMem[1]; i<=pVdbe->nMem; i++, pX++){
58830	if( pX->pScopyFrom==pMem ){
58831	pX->flags \|= MEM_Invalid;
	@@ -58847,11 +58747,11 @@
58847	** pFrom->z is used, then pTo->z points to the same thing as pFrom->z
58848	** and flags gets srcType (either MEM_Ephem or MEM_Static).
58849	*/
58850	SQLITE_PRIVATE void sqlite3VdbeMemShallowCopy(Mem pTo, const Mem pFrom, int srcType){
58851	assert( (pFrom->flags & MEM_RowSet)==0 );
58852	MemReleaseExt(pTo);
58853	memcpy(pTo, pFrom, MEMCELLSIZE);
58854	pTo->xDel = 0;
58855	if( (pFrom->flags&MEM_Static)==0 ){
58856	pTo->flags &= ~(MEM_Dyn\|MEM_Static\|MEM_Ephem);
58857	assert( srcType==MEM_Ephem \|\| srcType==MEM_Static );
	@@ -58865,11 +58765,11 @@
58865	*/
58866	SQLITE_PRIVATE int sqlite3VdbeMemCopy(Mem pTo, const Mem pFrom){
58867	int rc = SQLITE_OK;
58868
58869	assert( (pFrom->flags & MEM_RowSet)==0 );
58870	MemReleaseExt(pTo);
58871	memcpy(pTo, pFrom, MEMCELLSIZE);
58872	pTo->flags &= ~MEM_Dyn;
58873
58874	if( pTo->flags&(MEM_Str\|MEM_Blob) ){
58875	if( 0==(pFrom->flags&MEM_Static) ){
	@@ -59202,11 +59102,11 @@
59202	assert( (pVal->flags & (MEM_Ephem\|MEM_Static))!=0 );
59203	if( sqlite3VdbeMemMakeWriteable(pVal)!=SQLITE_OK ){
59204	return 0;
59205	}
59206	}
59207	sqlite3VdbeMemNulTerminate(pVal); /* IMP: R-59893-45467 */
59208	}else{
59209	assert( (pVal->flags&MEM_Blob)==0 );
59210	sqlite3VdbeMemStringify(pVal, enc);
59211	assert( 0==(1&SQLITE_PTR_TO_INT(pVal->z)) );
59212	}
	@@ -59583,11 +59483,12 @@
59583	return addr;
59584	}
59585
59586	/*
59587	** Add an OP_ParseSchema opcode. This routine is broken out from
59588	** sqlite3VdbeAddOp4() since it needs to also local all btrees.

59589	**
59590	** The zWhere string must have been obtained from sqlite3_malloc().
59591	** This routine will take ownership of the allocated memory.
59592	*/
59593	SQLITE_PRIVATE void sqlite3VdbeAddParseSchemaOp(Vdbe p, int iDb, char zWhere){
	@@ -60350,12 +60251,13 @@
60350
60351	/*
60352	** Declare to the Vdbe that the BTree object at db->aDb[i] is used.
60353	**
60354	** The prepared statements need to know in advance the complete set of
60355	** attached databases that they will be using. A mask of these databases
60356	** is maintained in p->btreeMask and is used for locking and other purposes.

60357	*/
60358	SQLITE_PRIVATE void sqlite3VdbeUsesBtree(Vdbe *p, int i){
60359	assert( i>=0 && i<p->db->nDb && i<(int)sizeof(yDbMask)*8 );
60360	assert( i<(int)sizeof(p->btreeMask)*8 );
60361	p->btreeMask \|= ((yDbMask)1)<<i;
	@@ -60874,10 +60776,11 @@
60874	nVar = pParse->nVar;
60875	nMem = pParse->nMem;
60876	nCursor = pParse->nTab;
60877	nArg = pParse->nMaxArg;
60878	nOnce = pParse->nOnce;

60879
60880	/* For each cursor required, also allocate a memory cell. Memory
60881	** cells (nMem+1-nCursor)..nMem, inclusive, will never be used by
60882	** the vdbe program. Instead they are used to allocate space for
60883	** VdbeCursor/BtCursor structures. The blob of memory associated with
	@@ -61225,11 +61128,11 @@
61225	int retryCount = 0;
61226	int nMainFile;
61227
61228	/* Select a master journal file name */
61229	nMainFile = sqlite3Strlen30(zMainFile);
61230	zMaster = sqlite3MPrintf(db, "%s-mjXXXXXX9XX", zMainFile);
61231	if( zMaster==0 ) return SQLITE_NOMEM;
61232	do {
61233	u32 iRandom;
61234	if( retryCount ){
61235	if( retryCount>100 ){
	@@ -63013,11 +62916,11 @@
63013	** since any application that receives an SQLITE_MISUSE is broken by
63014	** definition.
63015	**
63016	** Nevertheless, some published applications that were originally written
63017	** for version 3.6.23 or earlier do in fact depend on SQLITE_MISUSE
63018	** returns, and the so were broken by the automatic-reset change. As a
63019	** a work-around, the SQLITE_OMIT_AUTORESET compile-time restores the
63020	** legacy behavior of returning SQLITE_MISUSE for cases where the
63021	** previous sqlite3_step() returned something other than a SQLITE_LOCKED
63022	** or SQLITE_BUSY error.
63023	*/
	@@ -63359,17 +63262,17 @@
63359	pOut = &pVm->pResultSet[i];
63360	}else{
63361	/* If the value passed as the second argument is out of range, return
63362	** a pointer to the following static Mem object which contains the
63363	** value SQL NULL. Even though the Mem structure contains an element
63364	** of type i64, on certain architecture (x86) with certain compiler
63365	** switches (-Os), gcc may align this Mem object on a 4-byte boundary
63366	** instead of an 8-byte one. This all works fine, except that when
63367	** running with SQLITE_DEBUG defined the SQLite code sometimes assert()s
63368	** that a Mem structure is located on an 8-byte boundary. To prevent
63369	** this assert() from failing, when building with SQLITE_DEBUG defined
63370	** using gcc, force nullMem to be 8-byte aligned using the magical
63371	** __attribute__((aligned(8))) macro. */
63372	static const Mem nullMem
63373	#if defined(SQLITE_DEBUG) && defined(__GNUC__)
63374	__attribute__((aligned(8)))
63375	#endif
	@@ -64298,11 +64201,11 @@
64298	** Invoke this macro on memory cells just prior to changing the
64299	** value of the cell. This macro verifies that shallow copies are
64300	** not misused.
64301	*/
64302	#ifdef SQLITE_DEBUG
64303	# define memAboutToChange(P,M) sqlite3VdbeMemPrepareToChange(P,M)
64304	#else
64305	# define memAboutToChange(P,M)
64306	#endif
64307
64308	/*
	@@ -64316,12 +64219,12 @@
64316	SQLITE_API int sqlite3_search_count = 0;
64317	#endif
64318
64319	/*
64320	** When this global variable is positive, it gets decremented once before
64321	** each instruction in the VDBE. When reaches zero, the u1.isInterrupted
64322	** field of the sqlite3 structure is set in order to simulate and interrupt.
64323	**
64324	** This facility is used for testing purposes only. It does not function
64325	** in an ordinary build.
64326	*/
64327	#ifdef SQLITE_TEST
	@@ -64442,11 +64345,11 @@
64442	*/
64443	static VdbeCursor *allocateCursor(
64444	Vdbe p, / The virtual machine */
64445	int iCur, /* Index of the new VdbeCursor */
64446	int nField, /* Number of fields in the table or index */
64447	int iDb, /* When database the cursor belongs to, or -1 */
64448	int isBtreeCursor /* True for B-Tree. False for pseudo-table or vtab */
64449	){
64450	/* Find the memory cell that will be used to store the blob of memory
64451	** required for this VdbeCursor structure. It is convenient to use a
64452	** vdbe memory cell to manage the memory allocation required for a
	@@ -64813,11 +64716,11 @@
64813	** sqlite3_interrupt() routine has been called. If it has been, then
64814	** processing of the VDBE program is interrupted.
64815	**
64816	** This macro added to every instruction that does a jump in order to
64817	** implement a loop. This test used to be on every single instruction,
64818	** but that meant we more testing that we needed. By only testing the
64819	** flag on jump instructions, we get a (small) speed improvement.
64820	*/
64821	#define CHECK_FOR_INTERRUPT \
64822	if( db->u1.isInterrupted ) goto abort_due_to_interrupt;
64823
	@@ -65430,11 +65333,11 @@
65430	if( pOp->opflags & OPFLG_OUT2_PRERELEASE ){
65431	assert( pOp->p2>0 );
65432	assert( pOp->p2<=p->nMem );
65433	pOut = &aMem[pOp->p2];
65434	memAboutToChange(p, pOut);
65435	MemReleaseExt(pOut);
65436	pOut->flags = MEM_Int;
65437	}
65438
65439	/* Sanity checking on other operands */
65440	#ifdef SQLITE_DEBUG
	@@ -65738,11 +65641,11 @@
65738	assert( pOp->p3<=p->nMem );
65739	pOut->flags = MEM_Null;
65740	while( u.ab.cnt>0 ){
65741	pOut++;
65742	memAboutToChange(p, pOut);
65743	MemReleaseExt(pOut);
65744	pOut->flags = MEM_Null;
65745	u.ab.cnt--;
65746	}
65747	break;
65748	}
	@@ -65919,11 +65822,11 @@
65919	/* Invalidate all ephemeral cursor row caches */
65920	p->cacheCtr = (p->cacheCtr + 2)\|1;
65921
65922	/* Make sure the results of the current row are \000 terminated
65923	** and have an assigned type. The results are de-ephemeralized as
65924	** as side effect.
65925	*/
65926	u.ae.pMem = p->pResultSet = &aMem[pOp->p1];
65927	for(u.ae.i=0; u.ae.i<pOp->p2; u.ae.i++){
65928	assert( memIsValid(&u.ae.pMem[u.ae.i]) );
65929	Deephemeralize(&u.ae.pMem[u.ae.i]);
	@@ -67160,11 +67063,11 @@
67160	** a pointer to a Mem object.
67161	*/
67162	if( u.an.aOffset[u.an.p2] ){
67163	assert( rc==SQLITE_OK );
67164	if( u.an.zRec ){
67165	MemReleaseExt(u.an.pDest);
67166	sqlite3VdbeSerialGet((u8 *)&u.an.zRec[u.an.aOffset[u.an.p2]], u.an.aType[u.an.p2], u.an.pDest);
67167	}else{
67168	u.an.len = sqlite3VdbeSerialTypeLen(u.an.aType[u.an.p2]);
67169	sqlite3VdbeMemMove(&u.an.sMem, u.an.pDest);
67170	rc = sqlite3VdbeMemFromBtree(u.an.pCrsr, u.an.aOffset[u.an.p2], u.an.len, u.an.pC->isIndex, &u.an.sMem);
	@@ -83758,11 +83661,11 @@
83758	** Allocate the index structure.
83759	*/
83760	nName = sqlite3Strlen30(zName);
83761	nCol = pList->nExpr;
83762	pIndex = sqlite3DbMallocZero(db,
83763	sizeof(Index) + /* Index structure */
83764	ROUND8(sizeof(tRowcnt)(nCol+1)) + / Index.aiRowEst */
83765	sizeof(char )nCol + /* Index.azColl */
83766	sizeof(int)nCol + / Index.aiColumn */
83767	sizeof(u8)nCol + / Index.aSortOrder */
83768	nName + 1 + /* Index.zName */
	@@ -83769,11 +83672,12 @@
83769	nExtra /* Collation sequence names */
83770	);
83771	if( db->mallocFailed ){
83772	goto exit_create_index;
83773	}
83774	pIndex->aiRowEst = (tRowcnt*)(&pIndex[1]);

83775	pIndex->azColl = (char**)
83776	((char)pIndex->aiRowEst + ROUND8(sizeof(tRowcnt)nCol+1));
83777	assert( EIGHT_BYTE_ALIGNMENT(pIndex->aiRowEst) );
83778	assert( EIGHT_BYTE_ALIGNMENT(pIndex->azColl) );
83779	pIndex->aiColumn = (int *)(&pIndex->azColl[nCol]);
	@@ -89966,11 +89870,11 @@
89966	** any ABORT Back out changes from the current command
89967	** only (do not do a complete rollback) then
89968	** cause sqlite3_exec() to return immediately
89969	** with SQLITE_CONSTRAINT.
89970	**
89971	** any FAIL Sqlite_exec() returns immediately with a
89972	** return code of SQLITE_CONSTRAINT. The
89973	** transaction is not rolled back and any
89974	** prior changes are retained.
89975	**
89976	** any IGNORE The record number and data is popped from
	@@ -111450,11 +111354,11 @@
111450	*tokenType = TK_SPACE;
111451	return i;
111452	}
111453	case '-': {
111454	if( z[1]=='-' ){
111455	/* IMP: R-15891-05542 -- syntax diagram for comments */
111456	for(i=2; (c=z[i])!=0 && c!='\n'; i++){}
111457	tokenType = TK_SPACE; / IMP: R-22934-25134 */
111458	return i;
111459	}
111460	*tokenType = TK_MINUS;
	@@ -111483,11 +111387,11 @@
111483	case '/': {
111484	if( z[1]!='*' \|\| z[2]==0 ){
111485	*tokenType = TK_SLASH;
111486	return 1;
111487	}
111488	/* IMP: R-15891-05542 -- syntax diagram for comments */
111489	for(i=3, c=z[2]; (c!='*' \|\| z[i]!='/') && (c=z[i])!=0; i++){}
111490	if( c ) i++;
111491	tokenType = TK_SPACE; / IMP: R-22934-25134 */
111492	return i;
111493	}
	@@ -112277,12 +112181,12 @@
112277	/* IMPLEMENTATION-OF: R-35210-63508 The sqlite3_libversion_number() function
112278	** returns an integer equal to SQLITE_VERSION_NUMBER.
112279	*/
112280	SQLITE_API int sqlite3_libversion_number(void){ return SQLITE_VERSION_NUMBER; }
112281
112282	/* IMPLEMENTATION-OF: R-54823-41343 The sqlite3_threadsafe() function returns
112283	** zero if and only if SQLite was compiled mutexing code omitted due to
112284	** the SQLITE_THREADSAFE compile-time option being set to 0.
112285	*/
112286	SQLITE_API int sqlite3_threadsafe(void){ return SQLITE_THREADSAFE; }
112287
112288	#if !defined(SQLITE_OMIT_TRACE) && defined(SQLITE_ENABLE_IOTRACE)
	@@ -112768,19 +112672,21 @@
112768	** Free up as much memory as we can from the given database
112769	** connection.
112770	*/
112771	SQLITE_API int sqlite3_db_release_memory(sqlite3 *db){
112772	int i;

112773	sqlite3BtreeEnterAll(db);
112774	for(i=0; i<db->nDb; i++){
112775	Btree *pBt = db->aDb[i].pBt;
112776	if( pBt ){
112777	Pager *pPager = sqlite3BtreePager(pBt);
112778	sqlite3PagerShrink(pPager);
112779	}
112780	}
112781	sqlite3BtreeLeaveAll(db);

112782	return SQLITE_OK;
112783	}
112784
112785	/*
112786	** Configuration settings for an individual database connection
112787

	--- src/sqlite3.c
	+++ src/sqlite3.c
	@@ -657,11 +657,11 @@
657	** [sqlite3_libversion_number()], [sqlite3_sourceid()],
658	** [sqlite_version()] and [sqlite_source_id()].
659	*/
660	#define SQLITE_VERSION "3.7.10"
661	#define SQLITE_VERSION_NUMBER 3007010
662	#define SQLITE_SOURCE_ID "2012-01-05 12:38:02 1378f905d37544701776d38987fe7a312b255983"
663
664	/*
665	** CAPI3REF: Run-Time Library Version Numbers
666	** KEYWORDS: sqlite3_version, sqlite3_sourceid
667	**
	@@ -725,11 +725,11 @@
725
726	/*
727	** CAPI3REF: Test To See If The Library Is Threadsafe
728	**
729	** ^The sqlite3_threadsafe() function returns zero if and only if
730	** SQLite was compiled with mutexing code omitted due to the
731	** [SQLITE_THREADSAFE] compile-time option being set to 0.
732	**
733	** SQLite can be compiled with or without mutexes. When
734	** the [SQLITE_THREADSAFE] C preprocessor macro is 1 or 2, mutexes
735	** are enabled and SQLite is threadsafe. When the
	@@ -919,11 +919,11 @@
919	** CAPI3REF: Result Codes
920	** KEYWORDS: SQLITE_OK {error code} {error codes}
921	** KEYWORDS: {result code} {result codes}
922	**
923	** Many SQLite functions return an integer result code from the set shown
924	** here in order to indicate success or failure.
925	**
926	** New error codes may be added in future versions of SQLite.
927	**
928	** See also: [SQLITE_IOERR_READ \| extended result codes],
929	** [sqlite3_vtab_on_conflict()] [SQLITE_ROLLBACK \| result codes].
	@@ -1058,17 +1058,14 @@
1058	** that when data is appended to a file, the data is appended
1059	** first then the size of the file is extended, never the other
1060	** way around. The SQLITE_IOCAP_SEQUENTIAL property means that
1061	** information is written to disk in the same order as calls
1062	** to xWrite(). The SQLITE_IOCAP_POWERSAFE_OVERWRITE property means that
1063	** after reboot following a crash or power loss, the only bytes in a
1064	** file that were written at the application level might have changed
1065	** and that adjacent bytes, even bytes within the same sector are
1066	** guaranteed to be unchanged.



1067	*/
1068	#define SQLITE_IOCAP_ATOMIC 0x00000001
1069	#define SQLITE_IOCAP_ATOMIC512 0x00000002
1070	#define SQLITE_IOCAP_ATOMIC1K 0x00000004
1071	#define SQLITE_IOCAP_ATOMIC2K 0x00000008
	@@ -1300,16 +1297,16 @@
1297	** opcode as doing so may disrupt the operation of the specialized VFSes
1298	** that do require it.
1299	**
1300	** ^The [SQLITE_FCNTL_WIN32_AV_RETRY] opcode is used to configure automatic
1301	** retry counts and intervals for certain disk I/O operations for the
1302	** windows [VFS] in order to provide robustness in the presence of
1303	** anti-virus programs. By default, the windows VFS will retry file read,
1304	** file write, and file delete operations up to 10 times, with a delay
1305	** of 25 milliseconds before the first retry and with the delay increasing
1306	** by an additional 25 milliseconds with each subsequent retry. This
1307	** opcode allows these two values (10 retries and 25 milliseconds of delay)
1308	** to be adjusted. The values are changed for all database connections
1309	** within the same process. The argument is a pointer to an array of two
1310	** integers where the first integer i the new retry count and the second
1311	** integer is the delay. If either integer is negative, then the setting
1312	** is not changed but instead the prior value of that setting is written
	@@ -1420,11 +1417,11 @@
1417	** ^SQLite guarantees that the zFilename parameter to xOpen
1418	** is either a NULL pointer or string obtained
1419	** from xFullPathname() with an optional suffix added.
1420	** ^If a suffix is added to the zFilename parameter, it will
1421	** consist of a single "-" character followed by no more than
1422	** 11 alphanumeric and/or "-" characters.
1423	** ^SQLite further guarantees that
1424	** the string will be valid and unchanged until xClose() is
1425	** called. Because of the previous sentence,
1426	** the [sqlite3_file] can safely store a pointer to the
1427	** filename if it needs to remember the filename for some reason.
	@@ -2571,11 +2568,11 @@
2568	** These routines all implement some additional formatting
2569	** options that are useful for constructing SQL statements.
2570	** All of the usual printf() formatting options apply. In addition, there
2571	** is are "%q", "%Q", and "%z" options.
2572	**
2573	** ^(The %q option works like %s in that it substitutes a nul-terminated
2574	** string from the argument list. But %q also doubles every '\'' character.
2575	** %q is designed for use inside a string literal.)^ By doubling each '\''
2576	** character it escapes that character and allows it to be inserted into
2577	** the string.
2578	**
	@@ -4091,11 +4088,11 @@
4088	** of the string. ^For clarity: the values returned by
4089	** [sqlite3_column_bytes()] and [sqlite3_column_bytes16()] are the number of
4090	** bytes in the string, not the number of characters.
4091	**
4092	** ^Strings returned by sqlite3_column_text() and sqlite3_column_text16(),
4093	** even empty strings, are always zero-terminated. ^The return
4094	** value from sqlite3_column_blob() for a zero-length BLOB is a NULL pointer.
4095	**
4096	** ^The object returned by [sqlite3_column_value()] is an
4097	** [unprotected sqlite3_value] object. An unprotected sqlite3_value object
4098	** may only be used with [sqlite3_bind_value()] and [sqlite3_result_value()].
	@@ -5171,11 +5168,11 @@
5168	SQLITE_API int sqlite3_release_memory(int);
5169
5170	/*
5171	** CAPI3REF: Free Memory Used By A Database Connection
5172	**
5173	** ^The sqlite3_db_release_memory(D) interface attempts to free as much heap
5174	** memory as possible from database connection D. Unlike the
5175	** [sqlite3_release_memory()] interface, this interface is effect even
5176	** when then [SQLITE_ENABLE_MEMORY_MANAGEMENT] compile-time option is
5177	** omitted.
5178	**
	@@ -5954,19 +5951,19 @@
5951	** is selected automatically at compile-time. ^(The following
5952	** implementations are available in the SQLite core:
5953	**
5954	** <ul>
5955	** <li> SQLITE_MUTEX_OS2
5956	** <li> SQLITE_MUTEX_PTHREADS
5957	** <li> SQLITE_MUTEX_W32
5958	** <li> SQLITE_MUTEX_NOOP
5959	** </ul>)^
5960	**
5961	** ^The SQLITE_MUTEX_NOOP implementation is a set of routines
5962	** that does no real locking and is appropriate for use in
5963	** a single-threaded application. ^The SQLITE_MUTEX_OS2,
5964	** SQLITE_MUTEX_PTHREADS, and SQLITE_MUTEX_W32 implementations
5965	** are appropriate for use on OS/2, Unix, and Windows.
5966	**
5967	** ^(If SQLite is compiled with the SQLITE_MUTEX_APPDEF preprocessor
5968	** macro defined (with "-DSQLITE_MUTEX_APPDEF=1"), then no mutex
5969	** implementation is included with the library. In this case the
	@@ -6152,11 +6149,11 @@
6149	** defined and if NDEBUG is not defined.
6150	**
6151	** ^These routines should return true if the mutex in their argument
6152	** is held or not held, respectively, by the calling thread.
6153	**
6154	** ^The implementation is not required to provide versions of these
6155	** routines that actually work. If the implementation does not provide working
6156	** versions of these routines, it should at least provide stubs that always
6157	** return true so that one does not get spurious assertion failures.
6158	**
6159	** ^If the argument to sqlite3_mutex_held() is a NULL pointer then
	@@ -6634,11 +6631,11 @@
6631	**
6632	** [[the xInit() page cache method]]
6633	** ^(The xInit() method is called once for each effective
6634	** call to [sqlite3_initialize()])^
6635	** (usually only once during the lifetime of the process). ^(The xInit()
6636	** method is passed a copy of the sqlite3_pcache_methods2.pArg value.)^
6637	** The intent of the xInit() method is to set up global data structures
6638	** required by the custom page cache implementation.
6639	** ^(If the xInit() method is NULL, then the
6640	** built-in default page cache is used instead of the application defined
6641	** page cache.)^
	@@ -6660,11 +6657,11 @@
6657	**
6658	** [[the xCreate() page cache methods]]
6659	** ^SQLite invokes the xCreate() method to construct a new cache instance.
6660	** SQLite will typically create one cache instance for each open database file,
6661	** though this is not guaranteed. ^The
6662	** first parameter, szPage, is the size in bytes of the pages that must
6663	** be allocated by the cache. ^szPage will always a power of two. ^The
6664	** second parameter szExtra is a number of bytes of extra storage
6665	** associated with each page cache entry. ^The szExtra parameter will
6666	** a number less than 250. SQLite will use the
6667	** extra szExtra bytes on each page to store metadata about the underlying
	@@ -6755,11 +6752,11 @@
6752	**
6753	** [[the xDestroy() page cache method]]
6754	** ^The xDestroy() method is used to delete a cache allocated by xCreate().
6755	** All resources associated with the specified cache should be freed. ^After
6756	** calling the xDestroy() method, SQLite considers the [sqlite3_pcache*]
6757	** handle invalid, and will not use it with any other sqlite3_pcache_methods2
6758	** functions.
6759	**
6760	** [[the xShrink() page cache method]]
6761	** ^SQLite invokes the xShrink() method when it wants the page cache to
6762	** free up as much of heap memory as possible. The page cache implementation
	@@ -9266,11 +9263,11 @@
9263
9264	/*
9265	** The default size of a disk sector
9266	*/
9267	#ifndef SQLITE_DEFAULT_SECTOR_SIZE
9268	# define SQLITE_DEFAULT_SECTOR_SIZE 4096
9269	#endif
9270
9271	/*
9272	** Temporary files are named starting with this prefix followed by 16 random
9273	** alphanumeric characters, and no file extension. They are stored in the
	@@ -13190,11 +13187,11 @@
13187	SQLITE_PRIVATE int sqlite3VdbeMemRealify(Mem*);
13188	SQLITE_PRIVATE int sqlite3VdbeMemNumerify(Mem*);
13189	SQLITE_PRIVATE int sqlite3VdbeMemFromBtree(BtCursor,int,int,int,Mem);
13190	SQLITE_PRIVATE void sqlite3VdbeMemRelease(Mem *p);
13191	SQLITE_PRIVATE void sqlite3VdbeMemReleaseExternal(Mem *p);
13192	#define VdbeMemRelease(X) \
13193	if((X)->flags&(MEM_Agg\|MEM_Dyn\|MEM_RowSet\|MEM_Frame)) \
13194	sqlite3VdbeMemReleaseExternal(X);
13195	SQLITE_PRIVATE int sqlite3VdbeMemFinalize(Mem, FuncDef);
13196	SQLITE_PRIVATE const char *sqlite3OpcodeName(int);
13197	SQLITE_PRIVATE int sqlite3VdbeMemGrow(Mem *pMem, int n, int preserve);
	@@ -13229,11 +13226,11 @@
13226	# define sqlite3VdbeEnter(X)
13227	# define sqlite3VdbeLeave(X)
13228	#endif
13229
13230	#ifdef SQLITE_DEBUG
13231	SQLITE_PRIVATE void sqlite3VdbeMemAboutToChange(Vdbe,Mem);
13232	#endif
13233
13234	#ifndef SQLITE_OMIT_FOREIGN_KEY
13235	SQLITE_PRIVATE int sqlite3VdbeCheckFk(Vdbe *, int);
13236	#else
	@@ -14762,11 +14759,11 @@
14759	int flags,
14760	int *pFlagsOut
14761	){
14762	int rc;
14763	DO_OS_MALLOC_TEST(0);
14764	/* 0x87f7f is a mask of SQLITE_OPEN_ flags that are valid to be passed
14765	** down into the VFS layer. Some SQLITE_OPEN_ flags (for example,
14766	** SQLITE_OPEN_FULLMUTEX or SQLITE_OPEN_SHAREDCACHE) are blocked before
14767	** reaching the VFS. */
14768	rc = pVfs->xOpen(pVfs, zPath, pFile, flags & 0x87f7f, pFlagsOut);
14769	assert( rc==SQLITE_OK \|\| pFile->pMethods==0 );
	@@ -17316,11 +17313,11 @@
17313	SQLITE_API int sqlite3_mutex_notheld(sqlite3_mutex *p){
17314	return p==0 \|\| sqlite3GlobalConfig.mutex.xMutexNotheld(p);
17315	}
17316	#endif
17317
17318	#endif /* !defined(SQLITE_MUTEX_OMIT) */
17319
17320	/************ End of mutex.c *********************************************/
17321	/************ Begin file mutex_noop.c ************************************/
17322	/*
17323	** 2008 October 07
	@@ -17523,12 +17520,12 @@
17520	*/
17521	#ifdef SQLITE_MUTEX_NOOP
17522	SQLITE_PRIVATE sqlite3_mutex_methods const *sqlite3DefaultMutex(void){
17523	return sqlite3NoopMutex();
17524	}
17525	#endif /* defined(SQLITE_MUTEX_NOOP) */
17526	#endif /* !defined(SQLITE_MUTEX_OMIT) */
17527
17528	/************ End of mutex_noop.c ****************************************/
17529	/************ Begin file mutex_os2.c *************************************/
17530	/*
17531	** 2007 August 28
	@@ -18153,11 +18150,11 @@
18150	};
18151
18152	return &sMutex;
18153	}
18154
18155	#endif /* SQLITE_MUTEX_PTHREADS */
18156
18157	/************ End of mutex_unix.c ****************************************/
18158	/************ Begin file mutex_w32.c *************************************/
18159	/*
18160	** 2007 August 14
	@@ -22116,26 +22113,31 @@
22113	** characters of the original suffix.
22114	**
22115	** If SQLITE_ENABLE_8_3_NAMES is set to 2 at compile-time, then always
22116	** do the suffix shortening regardless of URI parameter.
22117	**
22118	** Assume that zBaseFilename contains two \000 terminator bytes (so that
22119	** it can be harmlessly passed into sqlite3_uri_parameter()) and copy both
22120	** zero terminator bytes into the end of the revised name.
22121	**
22122	** Examples:
22123	**
22124	** test.db-journal => test.nal
22125	** test.db-wal => test.wal
22126	** test.db-shm => test.shm
22127	** test.db-mj7f3319fa => test.9fa
22128	*/
22129	SQLITE_PRIVATE void sqlite3FileSuffix3(const char zBaseFilename, char z){
22130	assert( zBaseFilename[strlen(zBaseFilename)+1]==0 );
22131	#if SQLITE_ENABLE_8_3_NAMES<2
22132	if( sqlite3_uri_boolean(zBaseFilename, "8_3_names", 0) )
22133	#endif
22134	{
22135	int i, sz;
22136	sz = sqlite3Strlen30(z);
22137	for(i=sz-1; i>0 && z[i]!='/' && z[i]!='.'; i--){}
22138	if( z[i]=='.' && ALWAYS(sz>i+4) ) memmove(&z[i+1], &z[sz-3], 5);
22139	}
22140	}
22141	#endif
22142
22143	/************ End of util.c **********************************************/
	@@ -24836,13 +24838,10 @@
24838	#include <sys/time.h>
24839	#include <errno.h>
24840	#ifndef SQLITE_OMIT_WAL
24841	#include <sys/mman.h>
24842	#endif



24843
24844
24845	#if SQLITE_ENABLE_LOCKING_STYLE
24846	# include <sys/ioctl.h>
24847	# if OS_VXWORKS
	@@ -24929,11 +24928,10 @@
24928	sqlite3_vfs pVfs; / The VFS that created this unixFile */
24929	unixInodeInfo pInode; / Info about locks on this inode */
24930	int h; /* The file descriptor */
24931	unsigned char eFileLock; /* The type of lock held on this fd */
24932	unsigned char ctrlFlags; /* Behavioral bits. UNIXFILE_* flags */

24933	int lastErrno; /* The unix errno from last I/O error */
24934	void lockingContext; / Locking style specific state */
24935	UnixUnusedFd pUnused; / Pre-allocated UnixUnusedFd */
24936	const char zPath; / Name of the file */
24937	unixShm pShm; / Shared memory segment information */
	@@ -25342,18 +25340,10 @@
25340	#define osMkdir ((int()(const char,mode_t))aSyscall[18].pCurrent)
25341
25342	{ "rmdir", (sqlite3_syscall_ptr)rmdir, 0 },
25343	#define osRmdir ((int()(const char))aSyscall[19].pCurrent)
25344








25345	}; /* End of the overrideable system calls */
25346
25347	/*
25348	** This is the xSetSystemCall() method of sqlite3_vfs for all of the
25349	** "unix" VFSes. Return SQLITE_OK opon successfully updating the
	@@ -28504,13 +28494,10 @@
28494	case SQLITE_SET_LOCKPROXYFILE:
28495	case SQLITE_GET_LOCKPROXYFILE: {
28496	return proxyFileControl(id,op,pArg);
28497	}
28498	#endif /* SQLITE_ENABLE_LOCKING_STYLE && defined(__APPLE__) */



28499	}
28500	return SQLITE_NOTFOUND;
28501	}
28502
28503	/*
	@@ -28522,27 +28509,12 @@
28509	** if two files are created in the same file-system directory (i.e.
28510	** a database and its journal file) that the sector size will be the
28511	** same for both.
28512	*/
28513	static int unixSectorSize(sqlite3_file *pFile){
28514	(void)pFile;
28515	return SQLITE_DEFAULT_SECTOR_SIZE;















28516	}
28517
28518	/*
28519	** Return the device characteristics for the file.
28520	**
	@@ -28815,13 +28787,13 @@
28787	rc = SQLITE_IOERR_FSTAT;
28788	goto shm_open_err;
28789	}
28790
28791	#ifdef SQLITE_SHM_DIRECTORY
28792	nShmFilename = sizeof(SQLITE_SHM_DIRECTORY) + 31;
28793	#else
28794	nShmFilename = 6 + (int)strlen(pDbFd->zPath);
28795	#endif
28796	pShmNode = sqlite3_malloc( sizeof(*pShmNode) + nShmFilename );
28797	if( pShmNode==0 ){
28798	rc = SQLITE_NOMEM;
28799	goto shm_open_err;
	@@ -31755,11 +31727,11 @@
31727	};
31728	unsigned int i; /* Loop counter */
31729
31730	/* Double-check that the aSyscall[] array has been constructed
31731	** correctly. See ticket [bb3a86e890c8e96ab] */
31732	assert( ArraySize(aSyscall)==20 );
31733
31734	/* Register all VFSes defined in the aVfs[] array */
31735	for(i=0; i<(sizeof(aVfs)/sizeof(sqlite3_vfs)); i++){
31736	sqlite3_vfs_register(&aVfs[i], i==0);
31737	}
	@@ -32049,11 +32021,10 @@
32021	HANDLE h; /* Handle for accessing the file */
32022	u8 locktype; /* Type of lock currently held on this file */
32023	short sharedLockByte; /* Randomly chosen byte used as a shared lock */
32024	u8 ctrlFlags; /* Flags. See WINFILE_* below */
32025	DWORD lastErrno; /* The Windows errno from the last I/O error */

32026	winShm pShm; / Instance of shared memory on this file */
32027	const char zPath; / Full pathname of this file */
32028	int szChunk; /* Chunk size configured by FCNTL_CHUNK_SIZE */
32029	#if SQLITE_OS_WINCE
32030	LPWSTR zDeleteOnClose; /* Name of file to delete when closing */
	@@ -32139,18 +32110,10 @@
32110	static void winMemShutdown(void *pAppData);
32111
32112	SQLITE_PRIVATE const sqlite3_mem_methods *sqlite3MemGetWin32(void);
32113	#endif /* SQLITE_WIN32_MALLOC */
32114








32115	/*
32116	** The following variable is (normally) set once and never changes
32117	** thereafter. It records whether the operating system is Win9x
32118	** or WinNT.
32119	**
	@@ -34181,13 +34144,10 @@
34144	}
34145	case SQLITE_FCNTL_VFSNAME: {
34146	(char*)pArg = sqlite3_mprintf("win32");
34147	return SQLITE_OK;
34148	}



34149	case SQLITE_FCNTL_WIN32_AV_RETRY: {
34150	int a = (int)pArg;
34151	if( a[0]>0 ){
34152	win32IoerrRetry = a[0];
34153	}else{
	@@ -34213,12 +34173,12 @@
34173	** if two files are created in the same file-system directory (i.e.
34174	** a database and its journal file) that the sector size will be the
34175	** same for both.
34176	*/
34177	static int winSectorSize(sqlite3_file *id){
34178	(void)id;
34179	return SQLITE_DEFAULT_SECTOR_SIZE;
34180	}
34181
34182	/*
34183	** Return a vector of device characteristics.
34184	*/
	@@ -34463,11 +34423,11 @@
34423	*/
34424	p = sqlite3_malloc( sizeof(*p) );
34425	if( p==0 ) return SQLITE_IOERR_NOMEM;
34426	memset(p, 0, sizeof(*p));
34427	nName = sqlite3Strlen30(pDbFd->zPath);
34428	pNew = sqlite3_malloc( sizeof(*pShmNode) + nName + 16 );
34429	if( pNew==0 ){
34430	sqlite3_free(p);
34431	return SQLITE_IOERR_NOMEM;
34432	}
34433	memset(pNew, 0, sizeof(*pNew));
	@@ -35191,11 +35151,10 @@
35151	pFile->pShm = 0;
35152	pFile->zPath = zName;
35153	if( sqlite3_uri_boolean(zName, "psow", SQLITE_POWERSAFE_OVERWRITE) ){
35154	pFile->ctrlFlags \|= WINFILE_PSOW;
35155	}

35156
35157	#if SQLITE_OS_WINCE
35158	if( isReadWrite && eType==SQLITE_OPEN_MAIN_DB
35159	&& !winceCreateLock(zName, pFile)
35160	){
	@@ -35436,85 +35395,10 @@
35395	return SQLITE_IOERR_NOMEM;
35396	}
35397	#endif
35398	}
35399











































































35400	#ifndef SQLITE_OMIT_LOAD_EXTENSION
35401	/*
35402	** Interfaces for opening a shared library, finding entry points
35403	** within the shared library, and closing the shared library.
35404	*/
	@@ -36381,11 +36265,11 @@
36265	*/
36266	static int numberOfCachePages(PCache *p){
36267	if( p->szCache>=0 ){
36268	return p->szCache;
36269	}else{
36270	return (int)((-1024*(i64)p->szCache)/(p->szPage+p->szExtra));
36271	}
36272	}
36273
36274	/*
36275	** Try to obtain a page from the cache.
	@@ -36845,14 +36729,14 @@
36729	** PGroup which is the pcache1.grp global variable and its mutex is
36730	** SQLITE_MUTEX_STATIC_LRU.
36731	*/
36732	struct PGroup {
36733	sqlite3_mutex mutex; / MUTEX_STATIC_LRU or NULL */
36734	unsigned int nMaxPage; /* Sum of nMax for purgeable caches */
36735	unsigned int nMinPage; /* Sum of nMin for purgeable caches */
36736	unsigned int mxPinned; /* nMaxpage + 10 - nMinPage */
36737	unsigned int nCurrentPage; /* Number of purgeable pages allocated */
36738	PgHdr1 pLruHead, pLruTail; /* LRU list of unpinned pages */
36739	};
36740
36741	/* Each page cache is an instance of the following object. Every
36742	** open database file (including each in-memory database and each
	@@ -37511,11 +37395,11 @@
37395	static sqlite3_pcache_page *pcache1Fetch(
37396	sqlite3_pcache *p,
37397	unsigned int iKey,
37398	int createFlag
37399	){
37400	unsigned int nPinned;
37401	PCache1 pCache = (PCache1 )p;
37402	PGroup *pGroup;
37403	PgHdr1 *pPage = 0;
37404
37405	assert( pCache->bPurgeable \|\| createFlag!=1 );
	@@ -37546,17 +37430,17 @@
37430	#ifdef SQLITE_MUTEX_OMIT
37431	pGroup = pCache->pGroup;
37432	#endif
37433
37434	/* Step 3: Abort if createFlag is 1 but the cache is nearly full */
37435	assert( pCache->nPage >= pCache->nRecyclable );
37436	nPinned = pCache->nPage - pCache->nRecyclable;

37437	assert( pGroup->mxPinned == pGroup->nMaxPage + 10 - pGroup->nMinPage );
37438	assert( pCache->n90pct == pCache->nMax*9/10 );
37439	if( createFlag==1 && (
37440	nPinned>=pGroup->mxPinned
37441	\|\| nPinned>=pCache->n90pct
37442	\|\| pcache1UnderMemoryPressure(pCache)
37443	)){
37444	goto fetch_out;
37445	}
37446
	@@ -37725,11 +37609,13 @@
37609	PCache1 pCache = (PCache1 )p;
37610	PGroup *pGroup = pCache->pGroup;
37611	assert( pCache->bPurgeable \|\| (pCache->nMax==0 && pCache->nMin==0) );
37612	pcache1EnterMutex(pGroup);
37613	pcache1TruncateUnsafe(pCache, 0);
37614	assert( pGroup->nMaxPage >= pCache->nMax );
37615	pGroup->nMaxPage -= pCache->nMax;
37616	assert( pGroup->nMinPage >= pCache->nMin );
37617	pGroup->nMinPage -= pCache->nMin;
37618	pGroup->mxPinned = pGroup->nMaxPage + 10 - pGroup->nMinPage;
37619	pcache1EnforceMaxPage(pGroup);
37620	pcache1LeaveMutex(pGroup);
37621	sqlite3_free(pCache->apHash);
	@@ -37807,12 +37693,12 @@
37693	int nRecyclable = 0;
37694	for(p=pcache1.grp.pLruHead; p; p=p->pLruNext){
37695	nRecyclable++;
37696	}
37697	*pnCurrent = pcache1.grp.nCurrentPage;
37698	*pnMax = (int)pcache1.grp.nMaxPage;
37699	*pnMin = (int)pcache1.grp.nMinPage;
37700	*pnRecyclable = nRecyclable;
37701	}
37702	#endif
37703
37704	/************ End of pcache1.c *******************************************/
	@@ -38282,10 +38168,16 @@
38168
38169	#ifndef _WAL_H_
38170	#define _WAL_H_
38171
38172
38173	/* Additional values that can be added to the sync_flags argument of
38174	** sqlite3WalFrames():
38175	*/
38176	#define WAL_SYNC_TRANSACTIONS 0x20 /* Sync at the end of each transaction */
38177	#define SQLITE_SYNC_MASK 0x13 /* Mask off the SQLITE_SYNC_* values */
38178
38179	#ifdef SQLITE_OMIT_WAL
38180	# define sqlite3WalOpen(x,y,z) 0
38181	# define sqlite3WalLimit(x,y)
38182	# define sqlite3WalClose(w,x,y,z) 0
38183	# define sqlite3WalBeginReadTransaction(y,z) 0
	@@ -38350,16 +38242,10 @@
38242	SQLITE_PRIVATE int sqlite3WalSavepointUndo(Wal pWal, u32 aWalData);
38243
38244	/* Write a frame or frames to the log. */
38245	SQLITE_PRIVATE int sqlite3WalFrames(Wal pWal, int, PgHdr , Pgno, int, int);
38246






38247	/* Copy pages from the log to the database file */
38248	SQLITE_PRIVATE int sqlite3WalCheckpoint(
38249	Wal pWal, / Write-ahead log connection */
38250	int eMode, /* One of PASSIVE, FULL and RESTART */
38251	int (xBusy)(void), /* Function to call when busy */
	@@ -42400,11 +42286,13 @@
42286	** file size will be.
42287	*/
42288	assert( rc!=SQLITE_OK \|\| isOpen(pPager->fd) );
42289	if( rc==SQLITE_OK && pPager->dbSize>pPager->dbHintSize ){
42290	sqlite3_int64 szFile = pPager->pageSize * (sqlite3_int64)pPager->dbSize;
42291	sqlite3BeginBenignMalloc();
42292	sqlite3OsFileControl(pPager->fd, SQLITE_FCNTL_SIZE_HINT, &szFile);
42293	sqlite3EndBenignMalloc();
42294	pPager->dbHintSize = pPager->dbSize;
42295	}
42296
42297	while( rc==SQLITE_OK && pList ){
42298	Pgno pgno = pList->pgno;
	@@ -42741,11 +42629,12 @@
42629	z = zUri = &zFilename[sqlite3Strlen30(zFilename)+1];
42630	while( *z ){
42631	z += sqlite3Strlen30(z)+1;
42632	z += sqlite3Strlen30(z)+1;
42633	}
42634	nUri = (int)(&z[1] - zUri);
42635	assert( nUri>=0 );
42636	if( rc==SQLITE_OK && nPathname+8>pVfs->mxPathname ){
42637	/* This branch is taken when the journal path required by
42638	** the database being opened will be more than pVfs->mxPathname
42639	** bytes in length. This means the database cannot be opened,
42640	** as it will not be possible to open the journal file or even
	@@ -42775,13 +42664,13 @@
42664	ROUND8(sizeof(pPager)) + / Pager structure */
42665	ROUND8(pcacheSize) + /* PCache object */
42666	ROUND8(pVfs->szOsFile) + /* The main db file */
42667	journalFileSize * 2 + /* The two journal files */
42668	nPathname + 1 + nUri + /* zFilename */
42669	nPathname + 8 + 2 /* zJournal */
42670	#ifndef SQLITE_OMIT_WAL
42671	+ nPathname + 4 + 2 /* zWal */
42672	#endif
42673	);
42674	assert( EIGHT_BYTE_ALIGNMENT(SQLITE_INT_TO_PTR(journalFileSize)) );
42675	if( !pPtr ){
42676	sqlite3_free(zPathname);
	@@ -42800,16 +42689,16 @@
42689	assert( nPathname>0 );
42690	pPager->zJournal = (char*)(pPtr += nPathname + 1 + nUri);
42691	memcpy(pPager->zFilename, zPathname, nPathname);
42692	memcpy(&pPager->zFilename[nPathname+1], zUri, nUri);
42693	memcpy(pPager->zJournal, zPathname, nPathname);
42694	memcpy(&pPager->zJournal[nPathname], "-journal\000", 8+1);
42695	sqlite3FileSuffix3(pPager->zFilename, pPager->zJournal);
42696	#ifndef SQLITE_OMIT_WAL
42697	pPager->zWal = &pPager->zJournal[nPathname+8+1];
42698	memcpy(pPager->zWal, zPathname, nPathname);
42699	memcpy(&pPager->zWal[nPathname], "-wal\000", 4+1);
42700	sqlite3FileSuffix3(pPager->zFilename, pPager->zWal);
42701	#endif
42702	sqlite3_free(zPathname);
42703	}
42704	pPager->pVfs = pVfs;
	@@ -44064,11 +43953,14 @@
43953	if( !pPager->noSync ){
43954	assert( !MEMDB );
43955	rc = sqlite3OsSync(pPager->fd, pPager->syncFlags);
43956	}else if( isOpen(pPager->fd) ){
43957	assert( !MEMDB );
43958	rc = sqlite3OsFileControl(pPager->fd, SQLITE_FCNTL_SYNC_OMITTED, 0);
43959	if( rc==SQLITE_NOTFOUND ){
43960	rc = SQLITE_OK;
43961	}
43962	}
43963	return rc;
43964	}
43965
43966	/*
	@@ -48125,11 +48017,11 @@
48017	iFrame = pWal->hdr.mxFrame;
48018	for(p=pList; p && rc==SQLITE_OK; p=p->pDirty){
48019	iFrame++;
48020	rc = walIndexAppend(pWal, iFrame, p->pgno);
48021	}
48022	while( rc==SQLITE_OK && nExtra>0 ){
48023	iFrame++;
48024	nExtra--;
48025	rc = walIndexAppend(pWal, iFrame, pLast->pgno);
48026	}
48027
	@@ -48726,11 +48618,11 @@
48618	#define TRANS_WRITE 2
48619
48620	/*
48621	** An instance of this object represents a single database file.
48622	**
48623	** A single database file can be in use at the same time by two
48624	** or more database connections. When two or more connections are
48625	** sharing the same database file, each connection has it own
48626	** private Btree object for the file and each of those Btrees points
48627	** to this one BtShared object. BtShared.nRef is the number of
48628	** connections currently sharing this database file.
	@@ -48832,11 +48724,11 @@
48724	** b-tree within a database file.
48725	**
48726	** The entry is identified by its MemPage and the index in
48727	** MemPage.aCell[] of the entry.
48728	**
48729	** A single database file can be shared by two more database connections,
48730	** but cursors cannot be shared. Each cursor is associated with a
48731	** particular database connection identified BtCursor.pBtree.db.
48732	**
48733	** Fields in this structure are accessed under the BtShared.mutex
48734	** found at self->pBt->mutex.
	@@ -48993,11 +48885,11 @@
48885	int mallocFailed; /* A memory allocation error has occurred */
48886	StrAccum errMsg; /* Accumulate the error message text here */
48887	};
48888
48889	/*
48890	** Routines to read or write a two- and four-byte big-endian integer values.
48891	*/
48892	#define get2byte(x) ((x)[0]<<8 \| (x)[1])
48893	#define put2byte(p,v) ((p)[0] = (u8)((v)>>8), (p)[1] = (u8)(v))
48894	#define get4byte sqlite3Get4byte
48895	#define put4byte sqlite3Put4byte
	@@ -55478,12 +55370,18 @@
55370	}
55371
55372	/* Either we found one or more cells (cntnew[0])>0) or pPage is
55373	** a virtual root page. A virtual root page is when the real root
55374	** page is page 1 and we are the only child of that page.
55375	**
55376	** UPDATE: The assert() below is not necessarily true if the database
55377	** file is corrupt. The corruption will be detected and reported later
55378	** in this procedure so there is no need to act upon it now.
55379	*/
55380	#if 0
55381	assert( cntNew[0]>0 \|\| (pParent->pgno==1 && pParent->nCell==0) );
55382	#endif
55383
55384	TRACE(("BALANCE: old: %d %d %d ",
55385	apOld[0]->pgno,
55386	nOld>=2 ? apOld[1]->pgno : 0,
55387	nOld>=3 ? apOld[2]->pgno : 0
	@@ -58194,11 +58092,13 @@
58092
58093	assert( sqlite3BtreeIsInTrans(pTo) );
58094	pFd = sqlite3PagerFile(sqlite3BtreePager(pTo));
58095	if( pFd->pMethods ){
58096	i64 nByte = sqlite3BtreeGetPageSize(pFrom)*(i64)sqlite3BtreeLastPage(pFrom);
58097	sqlite3BeginBenignMalloc();
58098	sqlite3OsFileControl(pFd, SQLITE_FCNTL_OVERWRITE, &nByte);
58099	sqlite3EndBenignMalloc();
58100	}
58101
58102	/* Set up an sqlite3_backup object. sqlite3_backup.pDestDb must be set
58103	** to 0. This is used by the implementations of sqlite3_backup_step()
58104	** and sqlite3_backup_finish() to detect that they are being called
	@@ -58525,11 +58425,11 @@
58425	** Release any memory held by the Mem. This may leave the Mem in an
58426	** inconsistent state, for example with (Mem.z==0) and
58427	** (Mem.type==SQLITE_TEXT).
58428	*/
58429	SQLITE_PRIVATE void sqlite3VdbeMemRelease(Mem *p){
58430	VdbeMemRelease(p);
58431	sqlite3DbFree(p->db, p->zMalloc);
58432	p->z = 0;
58433	p->zMalloc = 0;
58434	p->xDel = 0;
58435	}
	@@ -58821,11 +58721,11 @@
58721	** copies of this cell as invalid.
58722	**
58723	** This is used for testing and debugging only - to make sure shallow
58724	** copies are not misused.
58725	*/
58726	SQLITE_PRIVATE void sqlite3VdbeMemAboutToChange(Vdbe pVdbe, Mem pMem){
58727	int i;
58728	Mem *pX;
58729	for(i=1, pX=&pVdbe->aMem[1]; i<=pVdbe->nMem; i++, pX++){
58730	if( pX->pScopyFrom==pMem ){
58731	pX->flags \|= MEM_Invalid;
	@@ -58847,11 +58747,11 @@
58747	** pFrom->z is used, then pTo->z points to the same thing as pFrom->z
58748	** and flags gets srcType (either MEM_Ephem or MEM_Static).
58749	*/
58750	SQLITE_PRIVATE void sqlite3VdbeMemShallowCopy(Mem pTo, const Mem pFrom, int srcType){
58751	assert( (pFrom->flags & MEM_RowSet)==0 );
58752	VdbeMemRelease(pTo);
58753	memcpy(pTo, pFrom, MEMCELLSIZE);
58754	pTo->xDel = 0;
58755	if( (pFrom->flags&MEM_Static)==0 ){
58756	pTo->flags &= ~(MEM_Dyn\|MEM_Static\|MEM_Ephem);
58757	assert( srcType==MEM_Ephem \|\| srcType==MEM_Static );
	@@ -58865,11 +58765,11 @@
58765	*/
58766	SQLITE_PRIVATE int sqlite3VdbeMemCopy(Mem pTo, const Mem pFrom){
58767	int rc = SQLITE_OK;
58768
58769	assert( (pFrom->flags & MEM_RowSet)==0 );
58770	VdbeMemRelease(pTo);
58771	memcpy(pTo, pFrom, MEMCELLSIZE);
58772	pTo->flags &= ~MEM_Dyn;
58773
58774	if( pTo->flags&(MEM_Str\|MEM_Blob) ){
58775	if( 0==(pFrom->flags&MEM_Static) ){
	@@ -59202,11 +59102,11 @@
59102	assert( (pVal->flags & (MEM_Ephem\|MEM_Static))!=0 );
59103	if( sqlite3VdbeMemMakeWriteable(pVal)!=SQLITE_OK ){
59104	return 0;
59105	}
59106	}
59107	sqlite3VdbeMemNulTerminate(pVal); /* IMP: R-31275-44060 */
59108	}else{
59109	assert( (pVal->flags&MEM_Blob)==0 );
59110	sqlite3VdbeMemStringify(pVal, enc);
59111	assert( 0==(1&SQLITE_PTR_TO_INT(pVal->z)) );
59112	}
	@@ -59583,11 +59483,12 @@
59483	return addr;
59484	}
59485
59486	/*
59487	** Add an OP_ParseSchema opcode. This routine is broken out from
59488	** sqlite3VdbeAddOp4() since it needs to also needs to mark all btrees
59489	** as having been used.
59490	**
59491	** The zWhere string must have been obtained from sqlite3_malloc().
59492	** This routine will take ownership of the allocated memory.
59493	*/
59494	SQLITE_PRIVATE void sqlite3VdbeAddParseSchemaOp(Vdbe p, int iDb, char zWhere){
	@@ -60350,12 +60251,13 @@
60251
60252	/*
60253	** Declare to the Vdbe that the BTree object at db->aDb[i] is used.
60254	**
60255	** The prepared statements need to know in advance the complete set of
60256	** attached databases that will be use. A mask of these databases
60257	** is maintained in p->btreeMask. The p->lockMask value is the subset of
60258	** p->btreeMask of databases that will require a lock.
60259	*/
60260	SQLITE_PRIVATE void sqlite3VdbeUsesBtree(Vdbe *p, int i){
60261	assert( i>=0 && i<p->db->nDb && i<(int)sizeof(yDbMask)*8 );
60262	assert( i<(int)sizeof(p->btreeMask)*8 );
60263	p->btreeMask \|= ((yDbMask)1)<<i;
	@@ -60874,10 +60776,11 @@
60776	nVar = pParse->nVar;
60777	nMem = pParse->nMem;
60778	nCursor = pParse->nTab;
60779	nArg = pParse->nMaxArg;
60780	nOnce = pParse->nOnce;
60781	if( nOnce==0 ) nOnce = 1; /* Ensure at least one byte in p->aOnceFlag[] */
60782
60783	/* For each cursor required, also allocate a memory cell. Memory
60784	** cells (nMem+1-nCursor)..nMem, inclusive, will never be used by
60785	** the vdbe program. Instead they are used to allocate space for
60786	** VdbeCursor/BtCursor structures. The blob of memory associated with
	@@ -61225,11 +61128,11 @@
61128	int retryCount = 0;
61129	int nMainFile;
61130
61131	/* Select a master journal file name */
61132	nMainFile = sqlite3Strlen30(zMainFile);
61133	zMaster = sqlite3MPrintf(db, "%s-mjXXXXXX9XXz", zMainFile);
61134	if( zMaster==0 ) return SQLITE_NOMEM;
61135	do {
61136	u32 iRandom;
61137	if( retryCount ){
61138	if( retryCount>100 ){
	@@ -63013,11 +62916,11 @@
62916	** since any application that receives an SQLITE_MISUSE is broken by
62917	** definition.
62918	**
62919	** Nevertheless, some published applications that were originally written
62920	** for version 3.6.23 or earlier do in fact depend on SQLITE_MISUSE
62921	** returns, and those were broken by the automatic-reset change. As a
62922	** a work-around, the SQLITE_OMIT_AUTORESET compile-time restores the
62923	** legacy behavior of returning SQLITE_MISUSE for cases where the
62924	** previous sqlite3_step() returned something other than a SQLITE_LOCKED
62925	** or SQLITE_BUSY error.
62926	*/
	@@ -63359,17 +63262,17 @@
63262	pOut = &pVm->pResultSet[i];
63263	}else{
63264	/* If the value passed as the second argument is out of range, return
63265	** a pointer to the following static Mem object which contains the
63266	** value SQL NULL. Even though the Mem structure contains an element
63267	** of type i64, on certain architectures (x86) with certain compiler
63268	** switches (-Os), gcc may align this Mem object on a 4-byte boundary
63269	** instead of an 8-byte one. This all works fine, except that when
63270	** running with SQLITE_DEBUG defined the SQLite code sometimes assert()s
63271	** that a Mem structure is located on an 8-byte boundary. To prevent
63272	** these assert()s from failing, when building with SQLITE_DEBUG defined
63273	** using gcc, we force nullMem to be 8-byte aligned using the magical
63274	** __attribute__((aligned(8))) macro. */
63275	static const Mem nullMem
63276	#if defined(SQLITE_DEBUG) && defined(__GNUC__)
63277	__attribute__((aligned(8)))
63278	#endif
	@@ -64298,11 +64201,11 @@
64201	** Invoke this macro on memory cells just prior to changing the
64202	** value of the cell. This macro verifies that shallow copies are
64203	** not misused.
64204	*/
64205	#ifdef SQLITE_DEBUG
64206	# define memAboutToChange(P,M) sqlite3VdbeMemAboutToChange(P,M)
64207	#else
64208	# define memAboutToChange(P,M)
64209	#endif
64210
64211	/*
	@@ -64316,12 +64219,12 @@
64219	SQLITE_API int sqlite3_search_count = 0;
64220	#endif
64221
64222	/*
64223	** When this global variable is positive, it gets decremented once before
64224	** each instruction in the VDBE. When it reaches zero, the u1.isInterrupted
64225	** field of the sqlite3 structure is set in order to simulate an interrupt.
64226	**
64227	** This facility is used for testing purposes only. It does not function
64228	** in an ordinary build.
64229	*/
64230	#ifdef SQLITE_TEST
	@@ -64442,11 +64345,11 @@
64345	*/
64346	static VdbeCursor *allocateCursor(
64347	Vdbe p, / The virtual machine */
64348	int iCur, /* Index of the new VdbeCursor */
64349	int nField, /* Number of fields in the table or index */
64350	int iDb, /* Database the cursor belongs to, or -1 */
64351	int isBtreeCursor /* True for B-Tree. False for pseudo-table or vtab */
64352	){
64353	/* Find the memory cell that will be used to store the blob of memory
64354	** required for this VdbeCursor structure. It is convenient to use a
64355	** vdbe memory cell to manage the memory allocation required for a
	@@ -64813,11 +64716,11 @@
64716	** sqlite3_interrupt() routine has been called. If it has been, then
64717	** processing of the VDBE program is interrupted.
64718	**
64719	** This macro added to every instruction that does a jump in order to
64720	** implement a loop. This test used to be on every single instruction,
64721	** but that meant we more testing than we needed. By only testing the
64722	** flag on jump instructions, we get a (small) speed improvement.
64723	*/
64724	#define CHECK_FOR_INTERRUPT \
64725	if( db->u1.isInterrupted ) goto abort_due_to_interrupt;
64726
	@@ -65430,11 +65333,11 @@
65333	if( pOp->opflags & OPFLG_OUT2_PRERELEASE ){
65334	assert( pOp->p2>0 );
65335	assert( pOp->p2<=p->nMem );
65336	pOut = &aMem[pOp->p2];
65337	memAboutToChange(p, pOut);
65338	VdbeMemRelease(pOut);
65339	pOut->flags = MEM_Int;
65340	}
65341
65342	/* Sanity checking on other operands */
65343	#ifdef SQLITE_DEBUG
	@@ -65738,11 +65641,11 @@
65641	assert( pOp->p3<=p->nMem );
65642	pOut->flags = MEM_Null;
65643	while( u.ab.cnt>0 ){
65644	pOut++;
65645	memAboutToChange(p, pOut);
65646	VdbeMemRelease(pOut);
65647	pOut->flags = MEM_Null;
65648	u.ab.cnt--;
65649	}
65650	break;
65651	}
	@@ -65919,11 +65822,11 @@
65822	/* Invalidate all ephemeral cursor row caches */
65823	p->cacheCtr = (p->cacheCtr + 2)\|1;
65824
65825	/* Make sure the results of the current row are \000 terminated
65826	** and have an assigned type. The results are de-ephemeralized as
65827	** a side effect.
65828	*/
65829	u.ae.pMem = p->pResultSet = &aMem[pOp->p1];
65830	for(u.ae.i=0; u.ae.i<pOp->p2; u.ae.i++){
65831	assert( memIsValid(&u.ae.pMem[u.ae.i]) );
65832	Deephemeralize(&u.ae.pMem[u.ae.i]);
	@@ -67160,11 +67063,11 @@
67063	** a pointer to a Mem object.
67064	*/
67065	if( u.an.aOffset[u.an.p2] ){
67066	assert( rc==SQLITE_OK );
67067	if( u.an.zRec ){
67068	VdbeMemRelease(u.an.pDest);
67069	sqlite3VdbeSerialGet((u8 *)&u.an.zRec[u.an.aOffset[u.an.p2]], u.an.aType[u.an.p2], u.an.pDest);
67070	}else{
67071	u.an.len = sqlite3VdbeSerialTypeLen(u.an.aType[u.an.p2]);
67072	sqlite3VdbeMemMove(&u.an.sMem, u.an.pDest);
67073	rc = sqlite3VdbeMemFromBtree(u.an.pCrsr, u.an.aOffset[u.an.p2], u.an.len, u.an.pC->isIndex, &u.an.sMem);
	@@ -83758,11 +83661,11 @@
83661	** Allocate the index structure.
83662	*/
83663	nName = sqlite3Strlen30(zName);
83664	nCol = pList->nExpr;
83665	pIndex = sqlite3DbMallocZero(db,
83666	ROUND8(sizeof(Index)) + /* Index structure */
83667	ROUND8(sizeof(tRowcnt)(nCol+1)) + / Index.aiRowEst */
83668	sizeof(char )nCol + /* Index.azColl */
83669	sizeof(int)nCol + / Index.aiColumn */
83670	sizeof(u8)nCol + / Index.aSortOrder */
83671	nName + 1 + /* Index.zName */
	@@ -83769,11 +83672,12 @@
83672	nExtra /* Collation sequence names */
83673	);
83674	if( db->mallocFailed ){
83675	goto exit_create_index;
83676	}
83677	zExtra = (char*)pIndex;
83678	pIndex->aiRowEst = (tRowcnt*)&zExtra[ROUND8(sizeof(Index))];
83679	pIndex->azColl = (char**)
83680	((char)pIndex->aiRowEst + ROUND8(sizeof(tRowcnt)nCol+1));
83681	assert( EIGHT_BYTE_ALIGNMENT(pIndex->aiRowEst) );
83682	assert( EIGHT_BYTE_ALIGNMENT(pIndex->azColl) );
83683	pIndex->aiColumn = (int *)(&pIndex->azColl[nCol]);
	@@ -89966,11 +89870,11 @@
89870	** any ABORT Back out changes from the current command
89871	** only (do not do a complete rollback) then
89872	** cause sqlite3_exec() to return immediately
89873	** with SQLITE_CONSTRAINT.
89874	**
89875	** any FAIL Sqlite3_exec() returns immediately with a
89876	** return code of SQLITE_CONSTRAINT. The
89877	** transaction is not rolled back and any
89878	** prior changes are retained.
89879	**
89880	** any IGNORE The record number and data is popped from
	@@ -111450,11 +111354,11 @@
111354	*tokenType = TK_SPACE;
111355	return i;
111356	}
111357	case '-': {
111358	if( z[1]=='-' ){
111359	/* IMP: R-50417-27976 -- syntax diagram for comments */
111360	for(i=2; (c=z[i])!=0 && c!='\n'; i++){}
111361	tokenType = TK_SPACE; / IMP: R-22934-25134 */
111362	return i;
111363	}
111364	*tokenType = TK_MINUS;
	@@ -111483,11 +111387,11 @@
111387	case '/': {
111388	if( z[1]!='*' \|\| z[2]==0 ){
111389	*tokenType = TK_SLASH;
111390	return 1;
111391	}
111392	/* IMP: R-50417-27976 -- syntax diagram for comments */
111393	for(i=3, c=z[2]; (c!='*' \|\| z[i]!='/') && (c=z[i])!=0; i++){}
111394	if( c ) i++;
111395	tokenType = TK_SPACE; / IMP: R-22934-25134 */
111396	return i;
111397	}
	@@ -112277,12 +112181,12 @@
112181	/* IMPLEMENTATION-OF: R-35210-63508 The sqlite3_libversion_number() function
112182	** returns an integer equal to SQLITE_VERSION_NUMBER.
112183	*/
112184	SQLITE_API int sqlite3_libversion_number(void){ return SQLITE_VERSION_NUMBER; }
112185
112186	/* IMPLEMENTATION-OF: R-20790-14025 The sqlite3_threadsafe() function returns
112187	** zero if and only if SQLite was compiled with mutexing code omitted due to
112188	** the SQLITE_THREADSAFE compile-time option being set to 0.
112189	*/
112190	SQLITE_API int sqlite3_threadsafe(void){ return SQLITE_THREADSAFE; }
112191
112192	#if !defined(SQLITE_OMIT_TRACE) && defined(SQLITE_ENABLE_IOTRACE)
	@@ -112768,19 +112672,21 @@
112672	** Free up as much memory as we can from the given database
112673	** connection.
112674	*/
112675	SQLITE_API int sqlite3_db_release_memory(sqlite3 *db){
112676	int i;
112677	sqlite3_mutex_enter(db->mutex);
112678	sqlite3BtreeEnterAll(db);
112679	for(i=0; i<db->nDb; i++){
112680	Btree *pBt = db->aDb[i].pBt;
112681	if( pBt ){
112682	Pager *pPager = sqlite3BtreePager(pBt);
112683	sqlite3PagerShrink(pPager);
112684	}
112685	}
112686	sqlite3BtreeLeaveAll(db);
112687	sqlite3_mutex_leave(db->mutex);
112688	return SQLITE_OK;
112689	}
112690
112691	/*
112692	** Configuration settings for an individual database connection
112693

M src/sqlite3.h

+19 -22

		--- src/sqlite3.h
		+++ src/sqlite3.h
		@@ -107,11 +107,11 @@
107	107	** [sqlite3_libversion_number()], [sqlite3_sourceid()],
108	108	** [sqlite_version()] and [sqlite_source_id()].
109	109	*/
110	110	#define SQLITE_VERSION "3.7.10"
111	111	#define SQLITE_VERSION_NUMBER 3007010
112		-#define SQLITE_SOURCE_ID "2011-12-23 13:32:07 1f24ae716df6232de768e245ea990049deee3c22"
	112	+#define SQLITE_SOURCE_ID "2012-01-05 12:38:02 1378f905d37544701776d38987fe7a312b255983"
113	113
114	114	/*
115	115	** CAPI3REF: Run-Time Library Version Numbers
116	116	** KEYWORDS: sqlite3_version, sqlite3_sourceid
117	117	**
		@@ -175,11 +175,11 @@
175	175
176	176	/*
177	177	** CAPI3REF: Test To See If The Library Is Threadsafe
178	178	**
179	179	** ^The sqlite3_threadsafe() function returns zero if and only if
180		-** SQLite was compiled mutexing code omitted due to the
	180	+** SQLite was compiled with mutexing code omitted due to the
181	181	** [SQLITE_THREADSAFE] compile-time option being set to 0.
182	182	**
183	183	** SQLite can be compiled with or without mutexes. When
184	184	** the [SQLITE_THREADSAFE] C preprocessor macro is 1 or 2, mutexes
185	185	** are enabled and SQLite is threadsafe. When the
		@@ -369,11 +369,11 @@
369	369	** CAPI3REF: Result Codes
370	370	** KEYWORDS: SQLITE_OK {error code} {error codes}
371	371	** KEYWORDS: {result code} {result codes}
372	372	**
373	373	** Many SQLite functions return an integer result code from the set shown
374		-** here in order to indicates success or failure.
	374	+** here in order to indicate success or failure.
375	375	**
376	376	** New error codes may be added in future versions of SQLite.
377	377	**
378	378	** See also: [SQLITE_IOERR_READ \| extended result codes],
379	379	** [sqlite3_vtab_on_conflict()] [SQLITE_ROLLBACK \| result codes].
		@@ -508,17 +508,14 @@
508	508	** that when data is appended to a file, the data is appended
509	509	** first then the size of the file is extended, never the other
510	510	** way around. The SQLITE_IOCAP_SEQUENTIAL property means that
511	511	** information is written to disk in the same order as calls
512	512	** to xWrite(). The SQLITE_IOCAP_POWERSAFE_OVERWRITE property means that
513		-** after reboot following a crash or power loss, the value of
514		-** each byte in a file is a value that was actually written
515		-** into that byte at some point. In other words, a crash will
516		-** not cause unwritten bytes of the file to change nor introduce
517		-** randomness into a file nor zero out parts of the file, and any byte of
518		-** a file that are never written will not change values due to
519		-** writes to nearby bytes.
	513	+** after reboot following a crash or power loss, the only bytes in a
	514	+** file that were written at the application level might have changed
	515	+** and that adjacent bytes, even bytes within the same sector are
	516	+** guaranteed to be unchanged.
520	517	*/
521	518	#define SQLITE_IOCAP_ATOMIC 0x00000001
522	519	#define SQLITE_IOCAP_ATOMIC512 0x00000002
523	520	#define SQLITE_IOCAP_ATOMIC1K 0x00000004
524	521	#define SQLITE_IOCAP_ATOMIC2K 0x00000008
		@@ -750,16 +747,16 @@
750	747	** opcode as doing so may disrupt the operation of the specialized VFSes
751	748	** that do require it.
752	749	**
753	750	** ^The [SQLITE_FCNTL_WIN32_AV_RETRY] opcode is used to configure automatic
754	751	** retry counts and intervals for certain disk I/O operations for the
755		-** windows [VFS] in order to work to provide robustness against
	752	+** windows [VFS] in order to provide robustness in the presence of
756	753	** anti-virus programs. By default, the windows VFS will retry file read,
757	754	** file write, and file delete operations up to 10 times, with a delay
758	755	** of 25 milliseconds before the first retry and with the delay increasing
759	756	** by an additional 25 milliseconds with each subsequent retry. This
760		-** opcode allows those to values (10 retries and 25 milliseconds of delay)
	757	+** opcode allows these two values (10 retries and 25 milliseconds of delay)
761	758	** to be adjusted. The values are changed for all database connections
762	759	** within the same process. The argument is a pointer to an array of two
763	760	** integers where the first integer i the new retry count and the second
764	761	** integer is the delay. If either integer is negative, then the setting
765	762	** is not changed but instead the prior value of that setting is written
		@@ -870,11 +867,11 @@
870	867	** ^SQLite guarantees that the zFilename parameter to xOpen
871	868	** is either a NULL pointer or string obtained
872	869	** from xFullPathname() with an optional suffix added.
873	870	** ^If a suffix is added to the zFilename parameter, it will
874	871	** consist of a single "-" character followed by no more than
875		-** 10 alphanumeric and/or "-" characters.
	872	+** 11 alphanumeric and/or "-" characters.
876	873	** ^SQLite further guarantees that
877	874	** the string will be valid and unchanged until xClose() is
878	875	** called. Because of the previous sentence,
879	876	** the [sqlite3_file] can safely store a pointer to the
880	877	** filename if it needs to remember the filename for some reason.
		@@ -2021,11 +2018,11 @@
2021	2018	** These routines all implement some additional formatting
2022	2019	** options that are useful for constructing SQL statements.
2023	2020	** All of the usual printf() formatting options apply. In addition, there
2024	2021	** is are "%q", "%Q", and "%z" options.
2025	2022	**
2026		-** ^(The %q option works like %s in that it substitutes a null-terminated
	2023	+** ^(The %q option works like %s in that it substitutes a nul-terminated
2027	2024	** string from the argument list. But %q also doubles every '\'' character.
2028	2025	** %q is designed for use inside a string literal.)^ By doubling each '\''
2029	2026	** character it escapes that character and allows it to be inserted into
2030	2027	** the string.
2031	2028	**
		@@ -3541,11 +3538,11 @@
3541	3538	** of the string. ^For clarity: the values returned by
3542	3539	** [sqlite3_column_bytes()] and [sqlite3_column_bytes16()] are the number of
3543	3540	** bytes in the string, not the number of characters.
3544	3541	**
3545	3542	** ^Strings returned by sqlite3_column_text() and sqlite3_column_text16(),
3546		-** even empty strings, are always zero terminated. ^The return
	3543	+** even empty strings, are always zero-terminated. ^The return
3547	3544	** value from sqlite3_column_blob() for a zero-length BLOB is a NULL pointer.
3548	3545	**
3549	3546	** ^The object returned by [sqlite3_column_value()] is an
3550	3547	** [unprotected sqlite3_value] object. An unprotected sqlite3_value object
3551	3548	** may only be used with [sqlite3_bind_value()] and [sqlite3_result_value()].
		@@ -4621,11 +4618,11 @@
4621	4618	SQLITE_API int sqlite3_release_memory(int);
4622	4619
4623	4620	/*
4624	4621	** CAPI3REF: Free Memory Used By A Database Connection
4625	4622	**
4626		-** ^The sqlite3_db_shrink(D) interface attempts to free as much heap
	4623	+** ^The sqlite3_db_release_memory(D) interface attempts to free as much heap
4627	4624	** memory as possible from database connection D. Unlike the
4628	4625	** [sqlite3_release_memory()] interface, this interface is effect even
4629	4626	** when then [SQLITE_ENABLE_MEMORY_MANAGEMENT] compile-time option is
4630	4627	** omitted.
4631	4628	**
		@@ -5404,19 +5401,19 @@
5404	5401	** is selected automatically at compile-time. ^(The following
5405	5402	** implementations are available in the SQLite core:
5406	5403	**
5407	5404	** <ul>
5408	5405	** <li> SQLITE_MUTEX_OS2
5409		-** <li> SQLITE_MUTEX_PTHREAD
	5406	+** <li> SQLITE_MUTEX_PTHREADS
5410	5407	** <li> SQLITE_MUTEX_W32
5411	5408	** <li> SQLITE_MUTEX_NOOP
5412	5409	** </ul>)^
5413	5410	**
5414	5411	** ^The SQLITE_MUTEX_NOOP implementation is a set of routines
5415	5412	** that does no real locking and is appropriate for use in
5416	5413	** a single-threaded application. ^The SQLITE_MUTEX_OS2,
5417		-** SQLITE_MUTEX_PTHREAD, and SQLITE_MUTEX_W32 implementations
	5414	+** SQLITE_MUTEX_PTHREADS, and SQLITE_MUTEX_W32 implementations
5418	5415	** are appropriate for use on OS/2, Unix, and Windows.
5419	5416	**
5420	5417	** ^(If SQLite is compiled with the SQLITE_MUTEX_APPDEF preprocessor
5421	5418	** macro defined (with "-DSQLITE_MUTEX_APPDEF=1"), then no mutex
5422	5419	** implementation is included with the library. In this case the
		@@ -5602,11 +5599,11 @@
5602	5599	** defined and if NDEBUG is not defined.
5603	5600	**
5604	5601	** ^These routines should return true if the mutex in their argument
5605	5602	** is held or not held, respectively, by the calling thread.
5606	5603	**
5607		-** ^The implementation is not required to provided versions of these
	5604	+** ^The implementation is not required to provide versions of these
5608	5605	** routines that actually work. If the implementation does not provide working
5609	5606	** versions of these routines, it should at least provide stubs that always
5610	5607	** return true so that one does not get spurious assertion failures.
5611	5608	**
5612	5609	** ^If the argument to sqlite3_mutex_held() is a NULL pointer then
		@@ -6084,11 +6081,11 @@
6084	6081	**
6085	6082	** [[the xInit() page cache method]]
6086	6083	** ^(The xInit() method is called once for each effective
6087	6084	** call to [sqlite3_initialize()])^
6088	6085	** (usually only once during the lifetime of the process). ^(The xInit()
6089		-** method is passed a copy of the sqlite3_pcache_methods.pArg value.)^
	6086	+** method is passed a copy of the sqlite3_pcache_methods2.pArg value.)^
6090	6087	** The intent of the xInit() method is to set up global data structures
6091	6088	** required by the custom page cache implementation.
6092	6089	** ^(If the xInit() method is NULL, then the
6093	6090	** built-in default page cache is used instead of the application defined
6094	6091	** page cache.)^
		@@ -6110,11 +6107,11 @@
6110	6107	**
6111	6108	** [[the xCreate() page cache methods]]
6112	6109	** ^SQLite invokes the xCreate() method to construct a new cache instance.
6113	6110	** SQLite will typically create one cache instance for each open database file,
6114	6111	** though this is not guaranteed. ^The
6115		-** parameter parameter, szPage, is the size in bytes of the pages that must
	6112	+** first parameter, szPage, is the size in bytes of the pages that must
6116	6113	** be allocated by the cache. ^szPage will always a power of two. ^The
6117	6114	** second parameter szExtra is a number of bytes of extra storage
6118	6115	** associated with each page cache entry. ^The szExtra parameter will
6119	6116	** a number less than 250. SQLite will use the
6120	6117	** extra szExtra bytes on each page to store metadata about the underlying
		@@ -6205,11 +6202,11 @@
6205	6202	**
6206	6203	** [[the xDestroy() page cache method]]
6207	6204	** ^The xDestroy() method is used to delete a cache allocated by xCreate().
6208	6205	** All resources associated with the specified cache should be freed. ^After
6209	6206	** calling the xDestroy() method, SQLite considers the [sqlite3_pcache*]
6210		-** handle invalid, and will not use it with any other sqlite3_pcache_methods
	6207	+** handle invalid, and will not use it with any other sqlite3_pcache_methods2
6211	6208	** functions.
6212	6209	**
6213	6210	** [[the xShrink() page cache method]]
6214	6211	** ^SQLite invokes the xShrink() method when it wants the page cache to
6215	6212	** free up as much of heap memory as possible. The page cache implementation
6216	6213

	--- src/sqlite3.h
	+++ src/sqlite3.h
	@@ -107,11 +107,11 @@
107	** [sqlite3_libversion_number()], [sqlite3_sourceid()],
108	** [sqlite_version()] and [sqlite_source_id()].
109	*/
110	#define SQLITE_VERSION "3.7.10"
111	#define SQLITE_VERSION_NUMBER 3007010
112	#define SQLITE_SOURCE_ID "2011-12-23 13:32:07 1f24ae716df6232de768e245ea990049deee3c22"
113
114	/*
115	** CAPI3REF: Run-Time Library Version Numbers
116	** KEYWORDS: sqlite3_version, sqlite3_sourceid
117	**
	@@ -175,11 +175,11 @@
175
176	/*
177	** CAPI3REF: Test To See If The Library Is Threadsafe
178	**
179	** ^The sqlite3_threadsafe() function returns zero if and only if
180	** SQLite was compiled mutexing code omitted due to the
181	** [SQLITE_THREADSAFE] compile-time option being set to 0.
182	**
183	** SQLite can be compiled with or without mutexes. When
184	** the [SQLITE_THREADSAFE] C preprocessor macro is 1 or 2, mutexes
185	** are enabled and SQLite is threadsafe. When the
	@@ -369,11 +369,11 @@
369	** CAPI3REF: Result Codes
370	** KEYWORDS: SQLITE_OK {error code} {error codes}
371	** KEYWORDS: {result code} {result codes}
372	**
373	** Many SQLite functions return an integer result code from the set shown
374	** here in order to indicates success or failure.
375	**
376	** New error codes may be added in future versions of SQLite.
377	**
378	** See also: [SQLITE_IOERR_READ \| extended result codes],
379	** [sqlite3_vtab_on_conflict()] [SQLITE_ROLLBACK \| result codes].
	@@ -508,17 +508,14 @@
508	** that when data is appended to a file, the data is appended
509	** first then the size of the file is extended, never the other
510	** way around. The SQLITE_IOCAP_SEQUENTIAL property means that
511	** information is written to disk in the same order as calls
512	** to xWrite(). The SQLITE_IOCAP_POWERSAFE_OVERWRITE property means that
513	** after reboot following a crash or power loss, the value of
514	** each byte in a file is a value that was actually written
515	** into that byte at some point. In other words, a crash will
516	** not cause unwritten bytes of the file to change nor introduce
517	** randomness into a file nor zero out parts of the file, and any byte of
518	** a file that are never written will not change values due to
519	** writes to nearby bytes.
520	*/
521	#define SQLITE_IOCAP_ATOMIC 0x00000001
522	#define SQLITE_IOCAP_ATOMIC512 0x00000002
523	#define SQLITE_IOCAP_ATOMIC1K 0x00000004
524	#define SQLITE_IOCAP_ATOMIC2K 0x00000008
	@@ -750,16 +747,16 @@
750	** opcode as doing so may disrupt the operation of the specialized VFSes
751	** that do require it.
752	**
753	** ^The [SQLITE_FCNTL_WIN32_AV_RETRY] opcode is used to configure automatic
754	** retry counts and intervals for certain disk I/O operations for the
755	** windows [VFS] in order to work to provide robustness against
756	** anti-virus programs. By default, the windows VFS will retry file read,
757	** file write, and file delete operations up to 10 times, with a delay
758	** of 25 milliseconds before the first retry and with the delay increasing
759	** by an additional 25 milliseconds with each subsequent retry. This
760	** opcode allows those to values (10 retries and 25 milliseconds of delay)
761	** to be adjusted. The values are changed for all database connections
762	** within the same process. The argument is a pointer to an array of two
763	** integers where the first integer i the new retry count and the second
764	** integer is the delay. If either integer is negative, then the setting
765	** is not changed but instead the prior value of that setting is written
	@@ -870,11 +867,11 @@
870	** ^SQLite guarantees that the zFilename parameter to xOpen
871	** is either a NULL pointer or string obtained
872	** from xFullPathname() with an optional suffix added.
873	** ^If a suffix is added to the zFilename parameter, it will
874	** consist of a single "-" character followed by no more than
875	** 10 alphanumeric and/or "-" characters.
876	** ^SQLite further guarantees that
877	** the string will be valid and unchanged until xClose() is
878	** called. Because of the previous sentence,
879	** the [sqlite3_file] can safely store a pointer to the
880	** filename if it needs to remember the filename for some reason.
	@@ -2021,11 +2018,11 @@
2021	** These routines all implement some additional formatting
2022	** options that are useful for constructing SQL statements.
2023	** All of the usual printf() formatting options apply. In addition, there
2024	** is are "%q", "%Q", and "%z" options.
2025	**
2026	** ^(The %q option works like %s in that it substitutes a null-terminated
2027	** string from the argument list. But %q also doubles every '\'' character.
2028	** %q is designed for use inside a string literal.)^ By doubling each '\''
2029	** character it escapes that character and allows it to be inserted into
2030	** the string.
2031	**
	@@ -3541,11 +3538,11 @@
3541	** of the string. ^For clarity: the values returned by
3542	** [sqlite3_column_bytes()] and [sqlite3_column_bytes16()] are the number of
3543	** bytes in the string, not the number of characters.
3544	**
3545	** ^Strings returned by sqlite3_column_text() and sqlite3_column_text16(),
3546	** even empty strings, are always zero terminated. ^The return
3547	** value from sqlite3_column_blob() for a zero-length BLOB is a NULL pointer.
3548	**
3549	** ^The object returned by [sqlite3_column_value()] is an
3550	** [unprotected sqlite3_value] object. An unprotected sqlite3_value object
3551	** may only be used with [sqlite3_bind_value()] and [sqlite3_result_value()].
	@@ -4621,11 +4618,11 @@
4621	SQLITE_API int sqlite3_release_memory(int);
4622
4623	/*
4624	** CAPI3REF: Free Memory Used By A Database Connection
4625	**
4626	** ^The sqlite3_db_shrink(D) interface attempts to free as much heap
4627	** memory as possible from database connection D. Unlike the
4628	** [sqlite3_release_memory()] interface, this interface is effect even
4629	** when then [SQLITE_ENABLE_MEMORY_MANAGEMENT] compile-time option is
4630	** omitted.
4631	**
	@@ -5404,19 +5401,19 @@
5404	** is selected automatically at compile-time. ^(The following
5405	** implementations are available in the SQLite core:
5406	**
5407	** <ul>
5408	** <li> SQLITE_MUTEX_OS2
5409	** <li> SQLITE_MUTEX_PTHREAD
5410	** <li> SQLITE_MUTEX_W32
5411	** <li> SQLITE_MUTEX_NOOP
5412	** </ul>)^
5413	**
5414	** ^The SQLITE_MUTEX_NOOP implementation is a set of routines
5415	** that does no real locking and is appropriate for use in
5416	** a single-threaded application. ^The SQLITE_MUTEX_OS2,
5417	** SQLITE_MUTEX_PTHREAD, and SQLITE_MUTEX_W32 implementations
5418	** are appropriate for use on OS/2, Unix, and Windows.
5419	**
5420	** ^(If SQLite is compiled with the SQLITE_MUTEX_APPDEF preprocessor
5421	** macro defined (with "-DSQLITE_MUTEX_APPDEF=1"), then no mutex
5422	** implementation is included with the library. In this case the
	@@ -5602,11 +5599,11 @@
5602	** defined and if NDEBUG is not defined.
5603	**
5604	** ^These routines should return true if the mutex in their argument
5605	** is held or not held, respectively, by the calling thread.
5606	**
5607	** ^The implementation is not required to provided versions of these
5608	** routines that actually work. If the implementation does not provide working
5609	** versions of these routines, it should at least provide stubs that always
5610	** return true so that one does not get spurious assertion failures.
5611	**
5612	** ^If the argument to sqlite3_mutex_held() is a NULL pointer then
	@@ -6084,11 +6081,11 @@
6084	**
6085	** [[the xInit() page cache method]]
6086	** ^(The xInit() method is called once for each effective
6087	** call to [sqlite3_initialize()])^
6088	** (usually only once during the lifetime of the process). ^(The xInit()
6089	** method is passed a copy of the sqlite3_pcache_methods.pArg value.)^
6090	** The intent of the xInit() method is to set up global data structures
6091	** required by the custom page cache implementation.
6092	** ^(If the xInit() method is NULL, then the
6093	** built-in default page cache is used instead of the application defined
6094	** page cache.)^
	@@ -6110,11 +6107,11 @@
6110	**
6111	** [[the xCreate() page cache methods]]
6112	** ^SQLite invokes the xCreate() method to construct a new cache instance.
6113	** SQLite will typically create one cache instance for each open database file,
6114	** though this is not guaranteed. ^The
6115	** parameter parameter, szPage, is the size in bytes of the pages that must
6116	** be allocated by the cache. ^szPage will always a power of two. ^The
6117	** second parameter szExtra is a number of bytes of extra storage
6118	** associated with each page cache entry. ^The szExtra parameter will
6119	** a number less than 250. SQLite will use the
6120	** extra szExtra bytes on each page to store metadata about the underlying
	@@ -6205,11 +6202,11 @@
6205	**
6206	** [[the xDestroy() page cache method]]
6207	** ^The xDestroy() method is used to delete a cache allocated by xCreate().
6208	** All resources associated with the specified cache should be freed. ^After
6209	** calling the xDestroy() method, SQLite considers the [sqlite3_pcache*]
6210	** handle invalid, and will not use it with any other sqlite3_pcache_methods
6211	** functions.
6212	**
6213	** [[the xShrink() page cache method]]
6214	** ^SQLite invokes the xShrink() method when it wants the page cache to
6215	** free up as much of heap memory as possible. The page cache implementation
6216

	--- src/sqlite3.h
	+++ src/sqlite3.h
	@@ -107,11 +107,11 @@
107	** [sqlite3_libversion_number()], [sqlite3_sourceid()],
108	** [sqlite_version()] and [sqlite_source_id()].
109	*/
110	#define SQLITE_VERSION "3.7.10"
111	#define SQLITE_VERSION_NUMBER 3007010
112	#define SQLITE_SOURCE_ID "2012-01-05 12:38:02 1378f905d37544701776d38987fe7a312b255983"
113
114	/*
115	** CAPI3REF: Run-Time Library Version Numbers
116	** KEYWORDS: sqlite3_version, sqlite3_sourceid
117	**
	@@ -175,11 +175,11 @@
175
176	/*
177	** CAPI3REF: Test To See If The Library Is Threadsafe
178	**
179	** ^The sqlite3_threadsafe() function returns zero if and only if
180	** SQLite was compiled with mutexing code omitted due to the
181	** [SQLITE_THREADSAFE] compile-time option being set to 0.
182	**
183	** SQLite can be compiled with or without mutexes. When
184	** the [SQLITE_THREADSAFE] C preprocessor macro is 1 or 2, mutexes
185	** are enabled and SQLite is threadsafe. When the
	@@ -369,11 +369,11 @@
369	** CAPI3REF: Result Codes
370	** KEYWORDS: SQLITE_OK {error code} {error codes}
371	** KEYWORDS: {result code} {result codes}
372	**
373	** Many SQLite functions return an integer result code from the set shown
374	** here in order to indicate success or failure.
375	**
376	** New error codes may be added in future versions of SQLite.
377	**
378	** See also: [SQLITE_IOERR_READ \| extended result codes],
379	** [sqlite3_vtab_on_conflict()] [SQLITE_ROLLBACK \| result codes].
	@@ -508,17 +508,14 @@
508	** that when data is appended to a file, the data is appended
509	** first then the size of the file is extended, never the other
510	** way around. The SQLITE_IOCAP_SEQUENTIAL property means that
511	** information is written to disk in the same order as calls
512	** to xWrite(). The SQLITE_IOCAP_POWERSAFE_OVERWRITE property means that
513	** after reboot following a crash or power loss, the only bytes in a
514	** file that were written at the application level might have changed
515	** and that adjacent bytes, even bytes within the same sector are
516	** guaranteed to be unchanged.



517	*/
518	#define SQLITE_IOCAP_ATOMIC 0x00000001
519	#define SQLITE_IOCAP_ATOMIC512 0x00000002
520	#define SQLITE_IOCAP_ATOMIC1K 0x00000004
521	#define SQLITE_IOCAP_ATOMIC2K 0x00000008
	@@ -750,16 +747,16 @@
747	** opcode as doing so may disrupt the operation of the specialized VFSes
748	** that do require it.
749	**
750	** ^The [SQLITE_FCNTL_WIN32_AV_RETRY] opcode is used to configure automatic
751	** retry counts and intervals for certain disk I/O operations for the
752	** windows [VFS] in order to provide robustness in the presence of
753	** anti-virus programs. By default, the windows VFS will retry file read,
754	** file write, and file delete operations up to 10 times, with a delay
755	** of 25 milliseconds before the first retry and with the delay increasing
756	** by an additional 25 milliseconds with each subsequent retry. This
757	** opcode allows these two values (10 retries and 25 milliseconds of delay)
758	** to be adjusted. The values are changed for all database connections
759	** within the same process. The argument is a pointer to an array of two
760	** integers where the first integer i the new retry count and the second
761	** integer is the delay. If either integer is negative, then the setting
762	** is not changed but instead the prior value of that setting is written
	@@ -870,11 +867,11 @@
867	** ^SQLite guarantees that the zFilename parameter to xOpen
868	** is either a NULL pointer or string obtained
869	** from xFullPathname() with an optional suffix added.
870	** ^If a suffix is added to the zFilename parameter, it will
871	** consist of a single "-" character followed by no more than
872	** 11 alphanumeric and/or "-" characters.
873	** ^SQLite further guarantees that
874	** the string will be valid and unchanged until xClose() is
875	** called. Because of the previous sentence,
876	** the [sqlite3_file] can safely store a pointer to the
877	** filename if it needs to remember the filename for some reason.
	@@ -2021,11 +2018,11 @@
2018	** These routines all implement some additional formatting
2019	** options that are useful for constructing SQL statements.
2020	** All of the usual printf() formatting options apply. In addition, there
2021	** is are "%q", "%Q", and "%z" options.
2022	**
2023	** ^(The %q option works like %s in that it substitutes a nul-terminated
2024	** string from the argument list. But %q also doubles every '\'' character.
2025	** %q is designed for use inside a string literal.)^ By doubling each '\''
2026	** character it escapes that character and allows it to be inserted into
2027	** the string.
2028	**
	@@ -3541,11 +3538,11 @@
3538	** of the string. ^For clarity: the values returned by
3539	** [sqlite3_column_bytes()] and [sqlite3_column_bytes16()] are the number of
3540	** bytes in the string, not the number of characters.
3541	**
3542	** ^Strings returned by sqlite3_column_text() and sqlite3_column_text16(),
3543	** even empty strings, are always zero-terminated. ^The return
3544	** value from sqlite3_column_blob() for a zero-length BLOB is a NULL pointer.
3545	**
3546	** ^The object returned by [sqlite3_column_value()] is an
3547	** [unprotected sqlite3_value] object. An unprotected sqlite3_value object
3548	** may only be used with [sqlite3_bind_value()] and [sqlite3_result_value()].
	@@ -4621,11 +4618,11 @@
4618	SQLITE_API int sqlite3_release_memory(int);
4619
4620	/*
4621	** CAPI3REF: Free Memory Used By A Database Connection
4622	**
4623	** ^The sqlite3_db_release_memory(D) interface attempts to free as much heap
4624	** memory as possible from database connection D. Unlike the
4625	** [sqlite3_release_memory()] interface, this interface is effect even
4626	** when then [SQLITE_ENABLE_MEMORY_MANAGEMENT] compile-time option is
4627	** omitted.
4628	**
	@@ -5404,19 +5401,19 @@
5401	** is selected automatically at compile-time. ^(The following
5402	** implementations are available in the SQLite core:
5403	**
5404	** <ul>
5405	** <li> SQLITE_MUTEX_OS2
5406	** <li> SQLITE_MUTEX_PTHREADS
5407	** <li> SQLITE_MUTEX_W32
5408	** <li> SQLITE_MUTEX_NOOP
5409	** </ul>)^
5410	**
5411	** ^The SQLITE_MUTEX_NOOP implementation is a set of routines
5412	** that does no real locking and is appropriate for use in
5413	** a single-threaded application. ^The SQLITE_MUTEX_OS2,
5414	** SQLITE_MUTEX_PTHREADS, and SQLITE_MUTEX_W32 implementations
5415	** are appropriate for use on OS/2, Unix, and Windows.
5416	**
5417	** ^(If SQLite is compiled with the SQLITE_MUTEX_APPDEF preprocessor
5418	** macro defined (with "-DSQLITE_MUTEX_APPDEF=1"), then no mutex
5419	** implementation is included with the library. In this case the
	@@ -5602,11 +5599,11 @@
5599	** defined and if NDEBUG is not defined.
5600	**
5601	** ^These routines should return true if the mutex in their argument
5602	** is held or not held, respectively, by the calling thread.
5603	**
5604	** ^The implementation is not required to provide versions of these
5605	** routines that actually work. If the implementation does not provide working
5606	** versions of these routines, it should at least provide stubs that always
5607	** return true so that one does not get spurious assertion failures.
5608	**
5609	** ^If the argument to sqlite3_mutex_held() is a NULL pointer then
	@@ -6084,11 +6081,11 @@
6081	**
6082	** [[the xInit() page cache method]]
6083	** ^(The xInit() method is called once for each effective
6084	** call to [sqlite3_initialize()])^
6085	** (usually only once during the lifetime of the process). ^(The xInit()
6086	** method is passed a copy of the sqlite3_pcache_methods2.pArg value.)^
6087	** The intent of the xInit() method is to set up global data structures
6088	** required by the custom page cache implementation.
6089	** ^(If the xInit() method is NULL, then the
6090	** built-in default page cache is used instead of the application defined
6091	** page cache.)^
	@@ -6110,11 +6107,11 @@
6107	**
6108	** [[the xCreate() page cache methods]]
6109	** ^SQLite invokes the xCreate() method to construct a new cache instance.
6110	** SQLite will typically create one cache instance for each open database file,
6111	** though this is not guaranteed. ^The
6112	** first parameter, szPage, is the size in bytes of the pages that must
6113	** be allocated by the cache. ^szPage will always a power of two. ^The
6114	** second parameter szExtra is a number of bytes of extra storage
6115	** associated with each page cache entry. ^The szExtra parameter will
6116	** a number less than 250. SQLite will use the
6117	** extra szExtra bytes on each page to store metadata about the underlying
	@@ -6205,11 +6202,11 @@
6202	**
6203	** [[the xDestroy() page cache method]]
6204	** ^The xDestroy() method is used to delete a cache allocated by xCreate().
6205	** All resources associated with the specified cache should be freed. ^After
6206	** calling the xDestroy() method, SQLite considers the [sqlite3_pcache*]
6207	** handle invalid, and will not use it with any other sqlite3_pcache_methods2
6208	** functions.
6209	**
6210	** [[the xShrink() page cache method]]
6211	** ^SQLite invokes the xShrink() method when it wants the page cache to
6212	** free up as much of heap memory as possible. The page cache implementation
6213

Fossil SCM

Keyboard Shortcuts