System.Data.SQLite

/******************************************************************************
** This file is an amalgamation of many separate C source files from SQLite
** version 3.7.6.  By combining all the individual C code files into this 
** single large file, the entire code can be compiled as a single translation
** unit.  This allows many compilers to do optimizations that would not be
** possible if the files were compiled separately.  Performance improvements
** of 5% or more are commonly seen when SQLite is compiled as a single
** translation unit.
**
** This file is all you need to compile SQLite.  To use SQLite in other
** programs, you need this file and the "sqlite3.h" header file that defines

*/
#ifndef SQLITE_DEFAULT_WAL_AUTOCHECKPOINT
# define SQLITE_DEFAULT_WAL_AUTOCHECKPOINT  1000
#endif

/*
** The maximum number of attached databases.  This must be between 0
** and 62.  The upper bound on 62 is because a 64-bit integer bitmap
** is used internally to track attached databases.
*/
#ifndef SQLITE_MAX_ATTACHED
# define SQLITE_MAX_ATTACHED 10
#endif

** string contains the date and time of the check-in (UTC) and an SHA1
** hash of the entire source tree.
**
** See also: [sqlite3_libversion()],
** [sqlite3_libversion_number()], [sqlite3_sourceid()],
** [sqlite_version()] and [sqlite_source_id()].
*/
#define SQLITE_VERSION        "3.7.6"
#define SQLITE_VERSION_NUMBER 3007006
#define SQLITE_SOURCE_ID      "2011-04-11 18:35:09 51029d8430d2dbc782f161577d47e3dd11c4e4d7"

/*
** CAPI3REF: Run-Time Library Version Numbers
** KEYWORDS: sqlite3_version, sqlite3_sourceid
**
** These interfaces provide the same information as the [SQLITE_VERSION],
** [SQLITE_VERSION_NUMBER], and [SQLITE_SOURCE_ID] C preprocessor macros

#define SQLITE_OPEN_MASTER_JOURNAL   0x00004000  /* VFS only */
#define SQLITE_OPEN_NOMUTEX          0x00008000  /* Ok for sqlite3_open_v2() */
#define SQLITE_OPEN_FULLMUTEX        0x00010000  /* Ok for sqlite3_open_v2() */
#define SQLITE_OPEN_SHAREDCACHE      0x00020000  /* Ok for sqlite3_open_v2() */
#define SQLITE_OPEN_PRIVATECACHE     0x00040000  /* Ok for sqlite3_open_v2() */
#define SQLITE_OPEN_WAL              0x00080000  /* VFS only */

/* Reserved:                         0x00F00000 */

/*
** CAPI3REF: Device Characteristics
**
** The xDeviceCharacteristics method of the [sqlite3_io_methods]
** object returns an integer which is a vector of the these
** bit values expressing I/O characteristics of the mass storage
** device that holds the file that the [sqlite3_io_methods]

** ^(The [SQLITE_FCNTL_SYNC_OMITTED] opcode is generated internally by
** SQLite and sent to all VFSes in place of a call to the xSync method
** when the database connection has [PRAGMA synchronous] set to OFF.)^
** Some specialized VFSes need this signal in order to operate correctly
** when [PRAGMA synchronous | PRAGMA synchronous=OFF] is set, but most 
** VFSes do not need this signal and should silently ignore this opcode.
** Applications should not call [sqlite3_file_control()] with this
** opcode as doing so may disrupt the operation of the specialized VFSes
** that do require it.  
*/
#define SQLITE_FCNTL_LOCKSTATE        1
#define SQLITE_GET_LOCKPROXYFILE      2
#define SQLITE_SET_LOCKPROXYFILE      3
#define SQLITE_LAST_ERRNO             4
#define SQLITE_FCNTL_SIZE_HINT        5

** ^The xCurrentTimeInt64() method returns, as an integer, the Julian
** Day Number multipled by 86400000 (the number of milliseconds in 
** a 24-hour day).  
** ^SQLite will use the xCurrentTimeInt64() method to get the current
** date and time if that method is available (if iVersion is 2 or 
** greater and the function pointer is not NULL) and will fall back
** to xCurrentTime() if xCurrentTimeInt64() is unavailable.
**
** ^The xSetSystemCall(), xGetSystemCall(), and xNestSystemCall() interfaces
** are not used by the SQLite core.  These optional interfaces are provided
** by some VFSes to facilitate testing of the VFS code. By overriding 
** system calls with functions under its control, a test program can
** simulate faults and error conditions that would otherwise be difficult
** or impossible to induce.  The set of system calls that can be overridden
** varies from one VFS to another, and from one version of the same VFS to the
** next.  Applications that use these interfaces must be prepared for any
** or all of these interfaces to be NULL or for their behavior to change
** from one release to the next.  Applications must not attempt to access
** any of these methods if the iVersion of the VFS is less than 3.
*/
typedef struct sqlite3_vfs sqlite3_vfs;
typedef void (*sqlite3_syscall_ptr)(void);
struct sqlite3_vfs {
  int iVersion;            /* Structure version number (currently 3) */
  int szOsFile;            /* Size of subclassed sqlite3_file */
  int mxPathname;          /* Maximum file pathname length */
  sqlite3_vfs *pNext;      /* Next registered VFS */
  const char *zName;       /* Name of this virtual file system */
  void *pAppData;          /* Pointer to application-specific data */
  int (*xOpen)(sqlite3_vfs*, const char *zName, sqlite3_file*,
               int flags, int *pOutFlags);

  /*
  ** The methods above are in version 1 of the sqlite_vfs object
  ** definition.  Those that follow are added in version 2 or later
  */
  int (*xCurrentTimeInt64)(sqlite3_vfs*, sqlite3_int64*);
  /*
  ** The methods above are in versions 1 and 2 of the sqlite_vfs object.
  ** Those below are for version 3 and greater.
  */
  int (*xSetSystemCall)(sqlite3_vfs*, const char *zName, sqlite3_syscall_ptr);
  sqlite3_syscall_ptr (*xGetSystemCall)(sqlite3_vfs*, const char *zName);
  const char *(*xNextSystemCall)(sqlite3_vfs*, const char *zName);
  /*
  ** The methods above are in versions 1 through 3 of the sqlite_vfs object.
  ** New fields may be appended in figure versions.  The iVersion
  ** value will increment whenever this happens. 
  */
};

/*
** CAPI3REF: Flags for the xAccess VFS method

/*
** CAPI3REF: Configure database connections
**
** The sqlite3_db_config() interface is used to make configuration
** changes to a [database connection].  The interface is similar to
** [sqlite3_config()] except that the changes apply to a single
** [database connection] (specified in the first argument).



**
** The second argument to sqlite3_db_config(D,V,...)  is the
** [SQLITE_DBCONFIG_LOOKASIDE | configuration verb] - an integer code 
** that indicates what aspect of the [database connection] is being configured.


** Subsequent arguments vary depending on the configuration verb.
**
** ^Calls to sqlite3_db_config() return SQLITE_OK if and only if
** the call is considered successful.
*/
SQLITE_API int sqlite3_db_config(sqlite3*, int op, ...);

/*

** compiled with [SQLITE_DEFAULT_MEMSTATUS]=0 in which case memory
** allocation statistics are disabled by default.
** </dd>
**
** <dt>SQLITE_CONFIG_SCRATCH</dt>
** <dd> ^This option specifies a static memory buffer that SQLite can use for
** scratch memory.  There are three arguments:  A pointer an 8-byte
** aligned memory buffer from which the scratch allocations will be
** drawn, the size of each scratch allocation (sz),
** and the maximum number of scratch allocations (N).  The sz
** argument must be a multiple of 16.
** The first argument must be a pointer to an 8-byte aligned buffer
** of at least sz*N bytes of memory.
** ^SQLite will use no more than two scratch buffers per thread.  So
** N should be set to twice the expected maximum number of threads.

** ^If the first pointer (the memory pointer) is NULL, then SQLite reverts
** to using its default memory allocator (the system malloc() implementation),
** undoing any prior invocation of [SQLITE_CONFIG_MALLOC].  ^If the
** memory pointer is not NULL and either [SQLITE_ENABLE_MEMSYS3] or
** [SQLITE_ENABLE_MEMSYS5] are defined, then the alternative memory
** allocator is engaged to handle all of SQLites memory allocation needs.
** The first pointer (the memory pointer) must be aligned to an 8-byte
** boundary or subsequent behavior of SQLite will be undefined.
** The minimum allocation size is capped at 2^12. Reasonable values
** for the minimum allocation size are 2^5 through 2^8.</dd>
**
** <dt>SQLITE_CONFIG_MUTEX</dt>
** <dd> ^(This option takes a single argument which is a pointer to an
** instance of the [sqlite3_mutex_methods] structure.  The argument specifies
** alternative low-level mutex routines to be used in place
** the mutex routines built into SQLite.)^  ^SQLite makes a copy of the
** content of the [sqlite3_mutex_methods] structure before the call to

** is invoked.
**
** <dl>
** <dt>SQLITE_DBCONFIG_LOOKASIDE</dt>
** <dd> ^This option takes three additional arguments that determine the 
** [lookaside memory allocator] configuration for the [database connection].
** ^The first argument (the third parameter to [sqlite3_db_config()] is a
** pointer to a memory buffer to use for lookaside memory.
** ^The first argument after the SQLITE_DBCONFIG_LOOKASIDE verb
** may be NULL in which case SQLite will allocate the
** lookaside buffer itself using [sqlite3_malloc()]. ^The second argument is the
** size of each lookaside buffer slot.  ^The third argument is the number of
** slots.  The size of the buffer in the first argument must be greater than
** or equal to the product of the second and third arguments.  The buffer
** must be aligned to an 8-byte boundary.  ^If the second argument to
** SQLITE_DBCONFIG_LOOKASIDE is not a multiple of 8, it is internally
** rounded down to the next smaller multiple of 8.  ^(The lookaside memory
** configuration for a database connection can only be changed when that
** connection is not currently using lookaside memory, or in other words
** when the "current value" returned by
** [sqlite3_db_status](D,[SQLITE_CONFIG_LOOKASIDE],...) is zero.
** Any attempt to change the lookaside memory configuration when lookaside
** memory is in use leaves the configuration unchanged and returns 
** [SQLITE_BUSY].)^</dd>
**
** <dt>SQLITE_DBCONFIG_ENABLE_FKEY</dt>
** <dd> ^This option is used to enable or disable the enforcement of
** [foreign key constraints].  There should be two additional arguments.
** The first argument is an integer which is 0 to disable FK enforcement,
** positive to enable FK enforcement or negative to leave FK enforcement
** unchanged.  The second parameter is a pointer to an integer into which
** is written 0 or 1 to indicate whether FK enforcement is off or on
** following this call.  The second parameter may be a NULL pointer, in
** which case the FK enforcement setting is not reported back. </dd>
**
** <dt>SQLITE_DBCONFIG_ENABLE_TRIGGER</dt>
** <dd> ^This option is used to enable or disable [CREATE TRIGGER | triggers].
** There should be two additional arguments.
** The first argument is an integer which is 0 to disable triggers,
** positive to enable triggers or negative to leave the setting unchanged.
** The second parameter is a pointer to an integer into which
** is written 0 or 1 to indicate whether triggers are disabled or enabled
** following this call.  The second parameter may be a NULL pointer, in
** which case the trigger setting is not reported back. </dd>
**
** </dl>
*/
#define SQLITE_DBCONFIG_LOOKASIDE       1001  /* void* int int */
#define SQLITE_DBCONFIG_ENABLE_FKEY     1002  /* int int* */
#define SQLITE_DBCONFIG_ENABLE_TRIGGER  1003  /* int int* */


/*
** CAPI3REF: Enable Or Disable Extended Result Codes
**
** ^The sqlite3_extended_result_codes() routine enables or disables the
** [extended result codes] feature of SQLite. ^The extended result

** method.
*/
SQLITE_API void sqlite3_randomness(int N, void *P);

/*
** CAPI3REF: Compile-Time Authorization Callbacks
**
** ^This routine registers an authorizer callback with a particular
** [database connection], supplied in the first argument.
** ^The authorizer callback is invoked as SQL statements are being compiled
** by [sqlite3_prepare()] or its variants [sqlite3_prepare_v2()],
** [sqlite3_prepare16()] and [sqlite3_prepare16_v2()].  ^At various
** points during the compilation process, as logic is being created
** to perform various actions, the authorizer callback is invoked to
** see if those actions are allowed.  ^The authorizer callback should

** compiled using either [sqlite3_prepare_v2()] or [sqlite3_prepare16_v2()].
*/
SQLITE_API const char *sqlite3_sql(sqlite3_stmt *pStmt);

/*
** CAPI3REF: Determine If An SQL Statement Writes The Database
**
** ^The sqlite3_stmt_readonly(X) interface returns true (non-zero) if
** and only if the [prepared statement] X makes no direct changes to
** the content of the database file.
**
** Note that [application-defined SQL functions] or
** [virtual tables] might change the database indirectly as a side effect.  
** ^(For example, if an application defines a function "eval()" that 
** calls [sqlite3_exec()], then the following SQL statement would

** An sqlite3_value object may be either "protected" or "unprotected".
** Some interfaces require a protected sqlite3_value.  Other interfaces
** will accept either a protected or an unprotected sqlite3_value.
** Every interface that accepts sqlite3_value arguments specifies
** whether or not it requires a protected sqlite3_value.
**
** The terms "protected" and "unprotected" refer to whether or not
** a mutex is held.  An internal mutex is held for a protected
** sqlite3_value object but no mutex is held for an unprotected
** sqlite3_value object.  If SQLite is compiled to be single-threaded
** (with [SQLITE_THREADSAFE=0] and with [sqlite3_threadsafe()] returning 0)
** or if SQLite is run in one of reduced mutex modes 
** [SQLITE_CONFIG_SINGLETHREAD] or [SQLITE_CONFIG_MULTITHREAD]
** then there is no distinction between protected and unprotected
** sqlite3_value objects and they can be used interchangeably.  However,

** interface returns a pointer to a zero-terminated UTF-8 string
** and sqlite3_column_name16() returns a pointer to a zero-terminated
** UTF-16 string.  ^The first parameter is the [prepared statement]
** that implements the [SELECT] statement. ^The second parameter is the
** column number.  ^The leftmost column is number 0.
**
** ^The returned string pointer is valid until either the [prepared statement]
** is destroyed by [sqlite3_finalize()] or until the statement is automatically
** reprepared by the first call to [sqlite3_step()] for a particular run
** or until the next call to
** sqlite3_column_name() or sqlite3_column_name16() on the same column.
**
** ^If sqlite3_malloc() fails during the processing of either routine
** (for example during a conversion from UTF-8 to UTF-16) then a
** NULL pointer is returned.
**
** ^The name of a result column is the value of the "AS" clause for

** table column that is the origin of a particular result column in
** [SELECT] statement.
** ^The name of the database or table or column can be returned as
** either a UTF-8 or UTF-16 string.  ^The _database_ routines return
** the database name, the _table_ routines return the table name, and
** the origin_ routines return the column name.
** ^The returned string is valid until the [prepared statement] is destroyed
** using [sqlite3_finalize()] or until the statement is automatically
** reprepared by the first call to [sqlite3_step()] for a particular run
** or until the same information is requested
** again in a different encoding.
**
** ^The names returned are the original un-aliased names of the
** database, table, and column.
**
** ^The first argument to these interfaces is a [prepared statement].
** ^These functions return information about the Nth result column returned by

** KEYWORDS: {application-defined SQL function}
** KEYWORDS: {application-defined SQL functions}
**
** ^These functions (collectively known as "function creation routines")
** are used to add SQL functions or aggregates or to redefine the behavior
** of existing SQL functions or aggregates.  The only differences between
** these routines are the text encoding expected for
** the second parameter (the name of the function being created)
** and the presence or absence of a destructor callback for
** the application data pointer.
**
** ^The first parameter is the [database connection] to which the SQL
** function is to be added.  ^If an application uses more than one database
** connection then application-defined SQL functions must be added
** to each database connection separately.

**
** ^The sixth, seventh and eighth parameters, xFunc, xStep and xFinal, are
** pointers to C-language functions that implement the SQL function or
** aggregate. ^A scalar SQL function requires an implementation of the xFunc
** callback only; NULL pointers must be passed as the xStep and xFinal
** parameters. ^An aggregate SQL function requires an implementation of xStep
** and xFinal and NULL pointer must be passed for xFunc. ^To delete an existing
** SQL function or aggregate, pass NULL pointers for all three function
** callbacks.
**
** ^(If the ninth parameter to sqlite3_create_function_v2() is not NULL,
** then it is destructor for the application data pointer. 
** The destructor is invoked when the function is deleted, either by being
** overloaded or when the database connection closes.)^
** ^The destructor is also invoked if the call to

** ^The eTextRep argument determines the encoding of strings passed
** to the collating function callback, xCallback.
** ^The [SQLITE_UTF16] and [SQLITE_UTF16_ALIGNED] values for eTextRep
** force strings to be UTF16 with native byte order.
** ^The [SQLITE_UTF16_ALIGNED] value for eTextRep forces strings to begin
** on an even byte address.
**
** ^The fourth argument, pArg, is an application data pointer that is passed
** through as the first argument to the collating function callback.
**
** ^The fifth argument, xCallback, is a pointer to the collating function.
** ^Multiple collating functions can be registered using the same name but
** with different eTextRep parameters and SQLite will use whichever
** function requires the least amount of data transformation.
** ^If the xCallback argument is NULL then the collating function is
** deleted.  ^When all collating functions having the same name are deleted,
** that collation is no longer usable.
**
** ^The collating function callback is invoked with a copy of the pArg 
** application data pointer and with two strings in the encoding specified
** by the eTextRep argument.  The collating function must return an
** integer that is negative, zero, or positive
** if the first string is less than, equal to, or greater than the second,
** respectively.  A collating function must always return the same answer
** given the same inputs.  If two or more collating functions are registered
** to the same collation name (using different eTextRep values) then all
** must give an equivalent answer when invoked with equivalent strings.
** The collating function must obey the following properties for all
** strings A, B, and C:
**
** <ol>

** if one or more of following conditions are true:
**
** <ul>
** <li> The soft heap limit is set to zero.
** <li> Memory accounting is disabled using a combination of the
**      [sqlite3_config]([SQLITE_CONFIG_MEMSTATUS],...) start-time option and
**      the [SQLITE_DEFAULT_MEMSTATUS] compile-time option.
** <li> An alternative page cache implementation is specified using
**      [sqlite3_config]([SQLITE_CONFIG_PCACHE],...).
** <li> The page cache allocates from its own memory pool supplied
**      by [sqlite3_config]([SQLITE_CONFIG_PAGECACHE],...) rather than
**      from the heap.
** </ul>)^
**
** Beginning with SQLite version 3.7.3, the soft heap limit is enforced

typedef struct sqlite3_vtab_cursor sqlite3_vtab_cursor;
typedef struct sqlite3_module sqlite3_module;

/*
** CAPI3REF: Virtual Table Object
** KEYWORDS: sqlite3_module {virtual table module}
**
** This structure, sometimes called a "virtual table module", 
** defines the implementation of a [virtual tables].  
** This structure consists mostly of methods for the module.
**
** ^A virtual table module is created by filling in a persistent
** instance of this structure and passing a pointer to that instance
** to [sqlite3_create_module()] or [sqlite3_create_module_v2()].
** ^The registration remains valid until it is replaced by a different

**
** ^(If the row that a BLOB handle points to is modified by an
** [UPDATE], [DELETE], or by [ON CONFLICT] side-effects
** then the BLOB handle is marked as "expired".
** This is true if any column of the row is changed, even a column
** other than the one the BLOB handle is open on.)^
** ^Calls to [sqlite3_blob_read()] and [sqlite3_blob_write()] for
** an expired BLOB handle fail with a return code of [SQLITE_ABORT].
** ^(Changes written into a BLOB prior to the BLOB expiring are not
** rolled back by the expiration of the BLOB.  Such changes will eventually
** commit if the transaction continues to completion.)^
**
** ^Use the [sqlite3_blob_bytes()] interface to determine the size of
** the opened blob.  ^The size of a blob may not be changed by this
** interface.  Use the [UPDATE] SQL command to change the size of a

** ^(<dt>SQLITE_DBSTATUS_LOOKASIDE_USED</dt>
** <dd>This parameter returns the number of lookaside memory slots currently
** checked out.</dd>)^
**
** ^(<dt>SQLITE_DBSTATUS_LOOKASIDE_HIT</dt>
** <dd>This parameter returns the number malloc attempts that were 
** satisfied using lookaside memory. Only the high-water value is meaningful;
** the current value is always zero.)^

**
** ^(<dt>SQLITE_DBSTATUS_LOOKASIDE_MISS_SIZE</dt>
** <dd>This parameter returns the number malloc attempts that might have
** been satisfied using lookaside memory but failed due to the amount of
** memory requested being larger than the lookaside slot size.
** Only the high-water value is meaningful;
** the current value is always zero.)^

**
** ^(<dt>SQLITE_DBSTATUS_LOOKASIDE_MISS_FULL</dt>
** <dd>This parameter returns the number malloc attempts that might have
** been satisfied using lookaside memory but failed due to all lookaside
** memory already being in use.
** Only the high-water value is meaningful;
** the current value is always zero.)^

**
** ^(<dt>SQLITE_DBSTATUS_CACHE_USED</dt>
** <dd>This parameter returns the approximate number of of bytes of heap
** memory used by all pager caches associated with the database connection.)^
** ^The highwater mark associated with SQLITE_DBSTATUS_CACHE_USED is always 0.
**
** ^(<dt>SQLITE_DBSTATUS_SCHEMA_USED</dt>

** connection D.  ^If the database connection D is not in
** [WAL | write-ahead log mode] then this interface is a harmless no-op.
**
** ^The [wal_checkpoint pragma] can be used to invoke this interface
** from SQL.  ^The [sqlite3_wal_autocheckpoint()] interface and the
** [wal_autocheckpoint pragma] can be used to cause this interface to be
** run whenever the WAL reaches a certain size threshold.
**
** See also: [sqlite3_wal_checkpoint_v2()]
*/
SQLITE_API int sqlite3_wal_checkpoint(sqlite3 *db, const char *zDb);

/*
** CAPI3REF: Checkpoint a database
**
** Run a checkpoint operation on WAL database zDb attached to database 
** handle db. The specific operation is determined by the value of the 
** eMode parameter:
**
** <dl>
** <dt>SQLITE_CHECKPOINT_PASSIVE<dd>
**   Checkpoint as many frames as possible without waiting for any database 
**   readers or writers to finish. Sync the db file if all frames in the log
**   are checkpointed. This mode is the same as calling 
**   sqlite3_wal_checkpoint(). The busy-handler callback is never invoked.
**
** <dt>SQLITE_CHECKPOINT_FULL<dd>
**   This mode blocks (calls the busy-handler callback) until there is no
**   database writer and all readers are reading from the most recent database
**   snapshot. It then checkpoints all frames in the log file and syncs the
**   database file. This call blocks database writers while it is running,
**   but not database readers.
**
** <dt>SQLITE_CHECKPOINT_RESTART<dd>
**   This mode works the same way as SQLITE_CHECKPOINT_FULL, except after 
**   checkpointing the log file it blocks (calls the busy-handler callback)
**   until all readers are reading from the database file only. This ensures 
**   that the next client to write to the database file restarts the log file 
**   from the beginning. This call blocks database writers while it is running,
**   but not database readers.
** </dl>
**
** If pnLog is not NULL, then *pnLog is set to the total number of frames in
** the log file before returning. If pnCkpt is not NULL, then *pnCkpt is set to
** the total number of checkpointed frames (including any that were already
** checkpointed when this function is called). *pnLog and *pnCkpt may be
** populated even if sqlite3_wal_checkpoint_v2() returns other than SQLITE_OK.
** If no values are available because of an error, they are both set to -1
** before returning to communicate this to the caller.
**
** All calls obtain an exclusive "checkpoint" lock on the database file. If
** any other process is running a checkpoint operation at the same time, the 
** lock cannot be obtained and SQLITE_BUSY is returned. Even if there is a 
** busy-handler configured, it will not be invoked in this case.
**
** The SQLITE_CHECKPOINT_FULL and RESTART modes also obtain the exclusive 
** "writer" lock on the database file. If the writer lock cannot be obtained
** immediately, and a busy-handler is configured, it is invoked and the writer
** lock retried until either the busy-handler returns 0 or the lock is
** successfully obtained. The busy-handler is also invoked while waiting for
** database readers as described above. If the busy-handler returns 0 before
** the writer lock is obtained or while waiting for database readers, the
** checkpoint operation proceeds from that point in the same way as 
** SQLITE_CHECKPOINT_PASSIVE - checkpointing as many frames as possible 
** without blocking any further. SQLITE_BUSY is returned in this case.
**
** If parameter zDb is NULL or points to a zero length string, then the
** specified operation is attempted on all WAL databases. In this case the
** values written to output parameters *pnLog and *pnCkpt are undefined. If 
** an SQLITE_BUSY error is encountered when processing one or more of the 
** attached WAL databases, the operation is still attempted on any remaining 
** attached databases and SQLITE_BUSY is returned to the caller. If any other 
** error occurs while processing an attached database, processing is abandoned 
** and the error code returned to the caller immediately. If no error 
** (SQLITE_BUSY or otherwise) is encountered while processing the attached 
** databases, SQLITE_OK is returned.
**
** If database zDb is the name of an attached database that is not in WAL
** mode, SQLITE_OK is returned and both *pnLog and *pnCkpt set to -1. If
** zDb is not NULL (or a zero length string) and is not the name of any
** attached database, SQLITE_ERROR is returned to the caller.
*/
SQLITE_API int sqlite3_wal_checkpoint_v2(
  sqlite3 *db,                    /* Database handle */
  const char *zDb,                /* Name of attached database (or NULL) */
  int eMode,                      /* SQLITE_CHECKPOINT_* value */
  int *pnLog,                     /* OUT: Size of WAL log in frames */
  int *pnCkpt                     /* OUT: Total number of frames checkpointed */
);

/*
** CAPI3REF: Checkpoint operation parameters
**
** These constants can be used as the 3rd parameter to
** [sqlite3_wal_checkpoint_v2()].  See the [sqlite3_wal_checkpoint_v2()]
** documentation for additional information about the meaning and use of
** each of these values.
*/
#define SQLITE_CHECKPOINT_PASSIVE 0
#define SQLITE_CHECKPOINT_FULL    1
#define SQLITE_CHECKPOINT_RESTART 2


/*
** Undo the hack that converts floating point types to integer for
** builds on processors without floating point support.
*/
#ifdef SQLITE_OMIT_FLOATING_POINT
# undef double

/*
** Forward declarations of structure
*/
typedef struct Btree Btree;
typedef struct BtCursor BtCursor;
typedef struct BtShared BtShared;














SQLITE_PRIVATE int sqlite3BtreeOpen(
  const char *zFilename,   /* Name of database file to open */
  sqlite3 *db,             /* Associated database connection */
  Btree **ppBtree,         /* Return open Btree* here */
  int flags,               /* Flags */

SQLITE_PRIVATE u32 sqlite3BtreeLastPage(Btree*);
SQLITE_PRIVATE int sqlite3BtreeSecureDelete(Btree*,int);
SQLITE_PRIVATE int sqlite3BtreeGetReserve(Btree*);
SQLITE_PRIVATE int sqlite3BtreeSetAutoVacuum(Btree *, int);
SQLITE_PRIVATE int sqlite3BtreeGetAutoVacuum(Btree *);
SQLITE_PRIVATE int sqlite3BtreeBeginTrans(Btree*,int);
SQLITE_PRIVATE int sqlite3BtreeCommitPhaseOne(Btree*, const char *zMaster);
SQLITE_PRIVATE int sqlite3BtreeCommitPhaseTwo(Btree*, int);
SQLITE_PRIVATE int sqlite3BtreeCommit(Btree*);
SQLITE_PRIVATE int sqlite3BtreeRollback(Btree*);
SQLITE_PRIVATE int sqlite3BtreeBeginStmt(Btree*,int);
SQLITE_PRIVATE int sqlite3BtreeCreateTable(Btree*, int*, int flags);
SQLITE_PRIVATE int sqlite3BtreeIsInTrans(Btree*);
SQLITE_PRIVATE int sqlite3BtreeIsInReadTrans(Btree*);
SQLITE_PRIVATE int sqlite3BtreeIsInBackup(Btree*);

#ifdef SQLITE_TEST
SQLITE_PRIVATE int sqlite3BtreeCursorInfo(BtCursor*, int*, int);
SQLITE_PRIVATE void sqlite3BtreeCursorList(Btree*);
#endif

#ifndef SQLITE_OMIT_WAL
SQLITE_PRIVATE   int sqlite3BtreeCheckpoint(Btree*, int, int *, int *);
#endif

/*
** If we are not using shared cache, then there is no need to
** use mutexes to access the BtShared structures.  So make the
** Enter and Leave procedures no-ops.
*/
#ifndef SQLITE_OMIT_SHARED_CACHE
SQLITE_PRIVATE   void sqlite3BtreeEnter(Btree*);
SQLITE_PRIVATE   void sqlite3BtreeEnterAll(sqlite3*);
#else
# define sqlite3BtreeEnter(X) 
# define sqlite3BtreeEnterAll(X)
#endif

#if !defined(SQLITE_OMIT_SHARED_CACHE) && SQLITE_THREADSAFE
SQLITE_PRIVATE   int sqlite3BtreeSharable(Btree*);
SQLITE_PRIVATE   void sqlite3BtreeLeave(Btree*);
SQLITE_PRIVATE   void sqlite3BtreeEnterCursor(BtCursor*);
SQLITE_PRIVATE   void sqlite3BtreeLeaveCursor(BtCursor*);
SQLITE_PRIVATE   void sqlite3BtreeLeaveAll(sqlite3*);



#ifndef NDEBUG
  /* These routines are used inside assert() statements only. */
SQLITE_PRIVATE   int sqlite3BtreeHoldsMutex(Btree*);
SQLITE_PRIVATE   int sqlite3BtreeHoldsAllMutexes(sqlite3*);
SQLITE_PRIVATE   int sqlite3SchemaMutexHeld(sqlite3*,int,Schema*);
#endif
#else

# define sqlite3BtreeSharable(X) 0
# define sqlite3BtreeLeave(X)
# define sqlite3BtreeEnterCursor(X)
# define sqlite3BtreeLeaveCursor(X)
# define sqlite3BtreeLeaveAll(X)




# define sqlite3BtreeHoldsMutex(X) 1
# define sqlite3BtreeHoldsAllMutexes(X) 1
# define sqlite3SchemaMutexHeld(X,Y,Z) 1
#endif


#endif /* _BTREE_H_ */

/************** End of btree.h ***********************************************/
/************** Continuing where we left off in sqliteInt.h ******************/

#define P4_DYNAMIC  (-1)  /* Pointer to a string obtained from sqliteMalloc() */
#define P4_STATIC   (-2)  /* Pointer to a static string */
#define P4_COLLSEQ  (-4)  /* P4 is a pointer to a CollSeq structure */
#define P4_FUNCDEF  (-5)  /* P4 is a pointer to a FuncDef structure */
#define P4_KEYINFO  (-6)  /* P4 is a pointer to a KeyInfo structure */
#define P4_VDBEFUNC (-7)  /* P4 is a pointer to a VdbeFunc structure */
#define P4_MEM      (-8)  /* P4 is a pointer to a Mem*    structure */
#define P4_TRANSIENT  0   /* P4 is a pointer to a transient string */
#define P4_VTAB     (-10) /* P4 is a pointer to an sqlite3_vtab structure */
#define P4_MPRINTF  (-11) /* P4 is a string obtained from sqlite3_mprintf() */
#define P4_REAL     (-12) /* P4 is a 64-bit floating point value */
#define P4_INT64    (-13) /* P4 is a 64-bit signed integer */
#define P4_INT32    (-14) /* P4 is a 32-bit signed integer */
#define P4_INTARRAY (-15) /* P4 is a vector of 32-bit integers */
#define P4_SUBPROGRAM  (-18) /* P4 is a pointer to a SubProgram structure */

SQLITE_PRIVATE int sqlite3PagerSync(Pager *pPager);
SQLITE_PRIVATE int sqlite3PagerCommitPhaseTwo(Pager*);
SQLITE_PRIVATE int sqlite3PagerRollback(Pager*);
SQLITE_PRIVATE int sqlite3PagerOpenSavepoint(Pager *pPager, int n);
SQLITE_PRIVATE int sqlite3PagerSavepoint(Pager *pPager, int op, int iSavepoint);
SQLITE_PRIVATE int sqlite3PagerSharedLock(Pager *pPager);

SQLITE_PRIVATE int sqlite3PagerCheckpoint(Pager *pPager, int, int*, int*);
SQLITE_PRIVATE int sqlite3PagerWalSupported(Pager *pPager);
SQLITE_PRIVATE int sqlite3PagerWalCallback(Pager *pPager);
SQLITE_PRIVATE int sqlite3PagerOpenWal(Pager *pPager, int *pisOpen);
SQLITE_PRIVATE int sqlite3PagerCloseWal(Pager *pPager);

/* Functions used to query pager state and configuration. */
SQLITE_PRIVATE u8 sqlite3PagerIsreadonly(Pager*);

  u8 inTrans;          /* 0: not writable.  1: Transaction.  2: Checkpoint */
  u8 safety_level;     /* How aggressive at syncing data to disk */
  Schema *pSchema;     /* Pointer to database schema (possibly shared) */
};

/*
** An instance of the following structure stores a database schema.
**
** Most Schema objects are associated with a Btree.  The exception is
** the Schema for the TEMP databaes (sqlite3.aDb[1]) which is free-standing.
** In shared cache mode, a single Schema object can be shared by multiple
** Btrees that refer to the same underlying BtShared object.
** 
** Schema objects are automatically deallocated when the last Btree that
** references them is destroyed.   The TEMP Schema is manually freed by
** sqlite3_close().
*
** A thread must be holding a mutex on the corresponding Btree in order
** to access Schema content.  This implies that the thread must also be
** holding a mutex on the sqlite3 connection pointer that owns the Btree.
** For a TEMP Schema, on the connection mutex is required.
*/
struct Schema {
  int schema_cookie;   /* Database schema version number for this file */
  int iGeneration;     /* Generation counter.  Incremented with each change */
  Hash tblHash;        /* All tables indexed by name */
  Hash idxHash;        /* All (named) indices indexed by name */
  Hash trigHash;       /* All triggers indexed by name */
  Hash fkeyHash;       /* All foreign keys by referenced table name */
  Table *pSeqTab;      /* The sqlite_sequence table used by AUTOINCREMENT */
  u8 file_format;      /* Schema format version for this file */
  u8 enc;              /* Text encoding used by this database */

#define SQLITE_RecoveryMode   0x00800000  /* Ignore schema errors */
#define SQLITE_ReverseOrder   0x01000000  /* Reverse unordered SELECTs */
#define SQLITE_RecTriggers    0x02000000  /* Enable recursive triggers */
#define SQLITE_ForeignKeys    0x04000000  /* Enforce foreign key constraints  */
#define SQLITE_AutoIndex      0x08000000  /* Enable automatic indexes */
#define SQLITE_PreferBuiltin  0x10000000  /* Preference to built-in funcs */
#define SQLITE_LoadExtension  0x20000000  /* Enable load_extension */
#define SQLITE_EnableTrigger  0x40000000  /* True to enable triggers */

/*
** Bits of the sqlite3.flags field that are used by the
** sqlite3_test_control(SQLITE_TESTCTRL_OPTIMIZATIONS,...) interface.
** These must be the low-order bits of the flags field.
*/
#define SQLITE_QueryFlattener 0x01        /* Disable query flattening */

** schema. This is because each database connection requires its own unique
** instance of the sqlite3_vtab* handle used to access the virtual table 
** implementation. sqlite3_vtab* handles can not be shared between 
** database connections, even when the rest of the in-memory database 
** schema is shared, as the implementation often stores the database
** connection handle passed to it via the xConnect() or xCreate() method
** during initialization internally. This database connection handle may
** then be used by the virtual table implementation to access real tables 
** within the database. So that they appear as part of the callers 
** transaction, these accesses need to be made via the same database 
** connection as that used to execute SQL operations on the virtual table.
**
** All VTable objects that correspond to a single table in a shared
** database schema are initially stored in a linked-list pointed to by
** the Table.pVTable member variable of the corresponding Table object.

  int nColumn;     /* Number of columns in the table used by this index */
  int *aiColumn;   /* Which columns are used by this index.  1st is 0 */
  unsigned *aiRowEst; /* Result of ANALYZE: Est. rows selected by each column */
  Table *pTable;   /* The SQL table being indexed */
  int tnum;        /* Page containing root of this index in database file */
  u8 onError;      /* OE_Abort, OE_Ignore, OE_Replace, or OE_None */
  u8 autoIndex;    /* True if is automatically created (ex: by UNIQUE) */
  u8 bUnordered;   /* Use this index for == or IN queries only */
  char *zColAff;   /* String defining the affinity of each column */
  Index *pNext;    /* The next index associated with the same table */
  Schema *pSchema; /* Schema containing this index */
  u8 *aSortOrder;  /* Array of size Index.nColumn. True==DESC, False==ASC */
  char **azColl;   /* Array of collation sequence names for index */
  IndexSample *aSample;    /* Array of SQLITE_INDEX_SAMPLES samples */
};

*/
struct Expr {
  u8 op;                 /* Operation performed by this node */
  char affinity;         /* The affinity of the column or 0 if not a column */
  u16 flags;             /* Various flags.  EP_* See below */
  union {
    char *zToken;          /* Token value. Zero terminated and dequoted */
    int iValue;            /* Non-negative integer value if EP_IntValue */
  } u;

  /* If the EP_TokenOnly flag is set in the Expr.flags mask, then no
  ** space is allocated for the fields below this point. An attempt to
  ** access them will result in a segfault or malfunction. 
  *********************************************************************/

  Trigger *pTrigger;      /* Trigger this program was coded from */
  int orconf;             /* Default ON CONFLICT policy */
  SubProgram *pProgram;   /* Program implementing pTrigger/orconf */
  u32 aColmask[2];        /* Masks of old.*, new.* columns accessed */
  TriggerPrg *pNext;      /* Next entry in Parse.pTriggerPrg list */
};

/*
** The yDbMask datatype for the bitmask of all attached databases.
*/
#if SQLITE_MAX_ATTACHED>30
  typedef sqlite3_uint64 yDbMask;
#else
  typedef unsigned int yDbMask;
#endif

/*
** An SQL parser context.  A copy of this structure is passed through
** the parser and down into all the parser action routine in order to
** carry around information that is global to the entire parse.
**
** The structure is divided into two parts.  When the parser and code
** generate call themselves recursively, the first part of the structure

    int iTable;           /* Table cursor number */
    int iColumn;          /* Table column number */
    u8 tempReg;           /* iReg is a temp register that needs to be freed */
    int iLevel;           /* Nesting level */
    int iReg;             /* Reg with value of this column. 0 means none. */
    int lru;              /* Least recently used entry has the smallest value */
  } aColCache[SQLITE_N_COLCACHE];  /* One for each column cache entry */
  yDbMask writeMask;   /* Start a write transaction on these databases */
  yDbMask cookieMask;  /* Bitmask of schema verified databases */
  u8 isMultiWrite;     /* True if statement may affect/insert multiple rows */
  u8 mayAbort;         /* True if statement may throw an ABORT exception */
  int cookieGoto;      /* Address of OP_Goto to cookie verifier subroutine */
  int cookieValue[SQLITE_MAX_ATTACHED+2];  /* Values of cookies to verify */
#ifndef SQLITE_OMIT_SHARED_CACHE
  int nTableLock;        /* Number of locks in aTableLock */
  TableLock *aTableLock; /* Required table locks for shared-cache mode */

SQLITE_PRIVATE void sqlite3ExprAnalyzeAggList(NameContext*,ExprList*);
SQLITE_PRIVATE Vdbe *sqlite3GetVdbe(Parse*);
SQLITE_PRIVATE void sqlite3PrngSaveState(void);
SQLITE_PRIVATE void sqlite3PrngRestoreState(void);
SQLITE_PRIVATE void sqlite3PrngResetState(void);
SQLITE_PRIVATE void sqlite3RollbackAll(sqlite3*);
SQLITE_PRIVATE void sqlite3CodeVerifySchema(Parse*, int);
SQLITE_PRIVATE void sqlite3CodeVerifyNamedSchema(Parse*, const char *zDb);
SQLITE_PRIVATE void sqlite3BeginTransaction(Parse*, int);
SQLITE_PRIVATE void sqlite3CommitTransaction(Parse*);
SQLITE_PRIVATE void sqlite3RollbackTransaction(Parse*);
SQLITE_PRIVATE void sqlite3Savepoint(Parse*, int, Token*);
SQLITE_PRIVATE void sqlite3CloseSavepoints(sqlite3 *);
SQLITE_PRIVATE int sqlite3ExprIsConstant(Expr*);
SQLITE_PRIVATE int sqlite3ExprIsConstantNotJoin(Expr*);

SQLITE_PRIVATE CollSeq *sqlite3LocateCollSeq(Parse *pParse, const char*zName);
SQLITE_PRIVATE CollSeq *sqlite3ExprCollSeq(Parse *pParse, Expr *pExpr);
SQLITE_PRIVATE Expr *sqlite3ExprSetColl(Expr*, CollSeq*);
SQLITE_PRIVATE Expr *sqlite3ExprSetCollByToken(Parse *pParse, Expr*, Token*);
SQLITE_PRIVATE int sqlite3CheckCollSeq(Parse *, CollSeq *);
SQLITE_PRIVATE int sqlite3CheckObjectName(Parse *, const char *);
SQLITE_PRIVATE void sqlite3VdbeSetChanges(sqlite3 *, int);
SQLITE_PRIVATE int sqlite3AddInt64(i64*,i64);
SQLITE_PRIVATE int sqlite3SubInt64(i64*,i64);
SQLITE_PRIVATE int sqlite3MulInt64(i64*,i64);
SQLITE_PRIVATE int sqlite3AbsInt32(int);

SQLITE_PRIVATE const void *sqlite3ValueText(sqlite3_value*, u8);
SQLITE_PRIVATE int sqlite3ValueBytes(sqlite3_value*, u8);
SQLITE_PRIVATE void sqlite3ValueSetStr(sqlite3_value*, int, const void *,u8, 
                        void(*)(void*));
SQLITE_PRIVATE void sqlite3ValueFree(sqlite3_value*);
SQLITE_PRIVATE sqlite3_value *sqlite3ValueNew(sqlite3 *);

SQLITE_PRIVATE const Token sqlite3IntTokens[];
SQLITE_PRIVATE SQLITE_WSD struct Sqlite3Config sqlite3Config;
SQLITE_PRIVATE SQLITE_WSD FuncDefHash sqlite3GlobalFunctions;
#ifndef SQLITE_OMIT_WSD
SQLITE_PRIVATE int sqlite3PendingByte;
#endif
#endif
SQLITE_PRIVATE void sqlite3RootPageMoved(sqlite3*, int, int, int);
SQLITE_PRIVATE void sqlite3Reindex(Parse*, Token*, Token*);
SQLITE_PRIVATE void sqlite3AlterFunctions(void);
SQLITE_PRIVATE void sqlite3AlterRenameTable(Parse*, SrcList*, Token*);
SQLITE_PRIVATE int sqlite3GetToken(const unsigned char *, int *);
SQLITE_PRIVATE void sqlite3NestedParse(Parse*, const char*, ...);
SQLITE_PRIVATE void sqlite3ExpirePreparedStatements(sqlite3*);
SQLITE_PRIVATE int sqlite3CodeSubselect(Parse *, Expr *, int, int);

SQLITE_PRIVATE int sqlite3FindDbName(sqlite3 *, const char *);
SQLITE_PRIVATE int sqlite3AnalysisLoad(sqlite3*,int iDB);
SQLITE_PRIVATE void sqlite3DeleteIndexSamples(sqlite3*,Index*);
SQLITE_PRIVATE void sqlite3DefaultRowEst(Index*);
SQLITE_PRIVATE void sqlite3RegisterLikeFunctions(sqlite3*, int);
SQLITE_PRIVATE int sqlite3IsLikeFunction(sqlite3*,Expr*,int*,char*);
SQLITE_PRIVATE void sqlite3MinimumFileFormat(Parse*, int, int);
SQLITE_PRIVATE void sqlite3SchemaClear(void *);
SQLITE_PRIVATE Schema *sqlite3SchemaGet(sqlite3 *, Btree *);
SQLITE_PRIVATE int sqlite3SchemaToIndex(sqlite3 *db, Schema *);
SQLITE_PRIVATE KeyInfo *sqlite3IndexKeyinfo(Parse *, Index *);
SQLITE_PRIVATE int sqlite3CreateFunc(sqlite3 *, const char *, int, int, void *, 
  void (*)(sqlite3_context*,int,sqlite3_value **),
  void (*)(sqlite3_context*,int,sqlite3_value **), void (*)(sqlite3_context*),
  FuncDestructor *pDestructor

SQLITE_PRIVATE int sqlite3TransferBindings(sqlite3_stmt *, sqlite3_stmt *);
SQLITE_PRIVATE int sqlite3Reprepare(Vdbe*);
SQLITE_PRIVATE void sqlite3ExprListCheckLength(Parse*, ExprList*, const char*);
SQLITE_PRIVATE CollSeq *sqlite3BinaryCompareCollSeq(Parse *, Expr *, Expr *);
SQLITE_PRIVATE int sqlite3TempInMemory(const sqlite3*);
SQLITE_PRIVATE VTable *sqlite3GetVTable(sqlite3*, Table*);
SQLITE_PRIVATE const char *sqlite3JournalModename(int);
SQLITE_PRIVATE int sqlite3Checkpoint(sqlite3*, int, int, int*, int*);
SQLITE_PRIVATE int sqlite3WalDefaultHook(void*,sqlite3*,const char*,int);

/* Declarations for functions in fkey.c. All of these are replaced by
** no-op macros if OMIT_FOREIGN_KEY is defined. In this case no foreign
** key functionality is available. If OMIT_TRIGGER is defined but
** OMIT_FOREIGN_KEY is not, only some of the functions are no-oped. In
** this case foreign keys are parsed, but no other functionality is 

  u8 runOnlyOnce;         /* Automatically expire on reset */
  u8 minWriteFileFormat;  /* Minimum file format for writable database files */
  u8 inVtabMethod;        /* See comments above */
  u8 usesStmtJournal;     /* True if uses a statement journal */
  u8 readOnly;            /* True for read-only statements */
  u8 isPrepareV2;         /* True if prepared with prepare_v2() */
  int nChange;            /* Number of db changes made since last reset */
  yDbMask btreeMask;      /* Bitmask of db->aDb[] entries referenced */
  yDbMask lockMask;       /* Subset of btreeMask that requires a lock */
  int iStatement;         /* Statement number (or 0 if has not opened stmt) */
  int aCounter[3];        /* Counters used by sqlite3_stmt_status() */

#ifndef SQLITE_OMIT_TRACE
  i64 startTime;          /* Time when query started - used for profiling */
#endif
  i64 nFkConstraint;      /* Number of imm. FK constraints this VM */
  i64 nStmtDefCons;       /* Number of def. constraints when stmt started */
  char *zSql;             /* Text of the SQL statement that generated this */
  void *pFree;            /* Free this when deleting the vdbe */

SQLITE_PRIVATE int sqlite3VdbeMemFinalize(Mem*, FuncDef*);
SQLITE_PRIVATE const char *sqlite3OpcodeName(int);
SQLITE_PRIVATE int sqlite3VdbeMemGrow(Mem *pMem, int n, int preserve);
SQLITE_PRIVATE int sqlite3VdbeCloseStatement(Vdbe *, int);
SQLITE_PRIVATE void sqlite3VdbeFrameDelete(VdbeFrame*);
SQLITE_PRIVATE int sqlite3VdbeFrameRestore(VdbeFrame *);
SQLITE_PRIVATE void sqlite3VdbeMemStoreType(Mem *pMem);

#if !defined(SQLITE_OMIT_SHARED_CACHE) && SQLITE_THREADSAFE>0
SQLITE_PRIVATE   void sqlite3VdbeEnter(Vdbe*);
SQLITE_PRIVATE   void sqlite3VdbeLeave(Vdbe*);
#else
# define sqlite3VdbeEnter(X)
# define sqlite3VdbeLeave(X)
#endif

#ifdef SQLITE_DEBUG
SQLITE_PRIVATE void sqlite3VdbeMemPrepareToChange(Vdbe*,Mem*);
#endif

#ifndef SQLITE_OMIT_FOREIGN_KEY
SQLITE_PRIVATE int sqlite3VdbeCheckFk(Vdbe *, int);
#else
# define sqlite3VdbeCheckFk(p,i) 0
#endif







SQLITE_PRIVATE int sqlite3VdbeMemTranslate(Mem*, u8);
#ifdef SQLITE_DEBUG
SQLITE_PRIVATE   void sqlite3VdbePrintSql(Vdbe*);
SQLITE_PRIVATE   void sqlite3VdbeMemPrettyPrint(Mem *pMem, char *zBuf);
#endif
SQLITE_PRIVATE int sqlite3VdbeMemHandleBom(Mem *pMem);

    ** to store the schema for all databases (main, temp, and any ATTACHed
    ** databases.  *pHighwater is set to zero.
    */
    case SQLITE_DBSTATUS_SCHEMA_USED: {
      int i;                      /* Used to iterate through schemas */
      int nByte = 0;              /* Used to accumulate return value */

      sqlite3BtreeEnterAll(db);
      db->pnBytesFreed = &nByte;
      for(i=0; i<db->nDb; i++){
        Schema *pSchema = db->aDb[i].pSchema;
        if( ALWAYS(pSchema!=0) ){
          HashElem *p;

          nByte += sqlite3GlobalConfig.m.xRoundup(sizeof(HashElem)) * (

          }
          for(p=sqliteHashFirst(&pSchema->tblHash); p; p=sqliteHashNext(p)){
            sqlite3DeleteTable(db, (Table *)sqliteHashData(p));
          }
        }
      }
      db->pnBytesFreed = 0;
      sqlite3BtreeLeaveAll(db);

      *pHighwater = 0;
      *pCurrent = nByte;
      break;
    }

    /*

  /*
  ** Space for tracking which blocks are checked out and the size
  ** of each block.  One byte per block.
  */
  u8 *aCtrl;

} mem5;

/*
** Access the static variable through a macro for SQLITE_OMIT_WSD
*/
#define mem5 GLOBAL(struct Mem5Global, mem5)

/*

**             memsys5Log(4) -> 2
**             memsys5Log(5) -> 3
**             memsys5Log(8) -> 3
**             memsys5Log(9) -> 4
*/
static int memsys5Log(int iValue){
  int iLog;
  for(iLog=0; (iLog<(int)((sizeof(int)*8)-1)) && (1<<iLog)<iValue; iLog++);
  return iLog;
}

/*
** Initialize the memory allocator.
**
** This routine is not threadsafe.  The caller must be holding a mutex

  */
  assert( (sizeof(Mem5Link)&(sizeof(Mem5Link)-1))==0 );

  nByte = sqlite3GlobalConfig.nHeap;
  zByte = (u8*)sqlite3GlobalConfig.pHeap;
  assert( zByte!=0 );  /* sqlite3_config() does not allow otherwise */

  /* boundaries on sqlite3GlobalConfig.mnReq are enforced in sqlite3_config() */
  nMinLog = memsys5Log(sqlite3GlobalConfig.mnReq);
  mem5.szAtom = (1<<nMinLog);
  while( (int)sizeof(Mem5Link)>mem5.szAtom ){
    mem5.szAtom = mem5.szAtom << 1;
  }

  mem5.nBlock = (nByte / (mem5.szAtom+sizeof(u8)));

/*
** The mutex object
** Each recursive mutex is an instance of the following structure.
*/
struct sqlite3_mutex {
  HMTX mutex;       /* Mutex controlling the lock */
  int  id;          /* Mutex type */
#ifdef SQLITE_DEBUG
 int   trace;       /* True to trace changes */
#endif
};

#ifdef SQLITE_DEBUG
#define SQLITE3_MUTEX_INITIALIZER { 0, 0, 0 }
#else
#define SQLITE3_MUTEX_INITIALIZER { 0, 0 }
#endif

/*
** Initialize and deinitialize the mutex subsystem.
*/
static int os2MutexInit(void){ return SQLITE_OK; }
static int os2MutexEnd(void){ return SQLITE_OK; }

/*
** The sqlite3_mutex_alloc() routine allocates a new
** mutex and returns a pointer to it.  If it returns NULL
** that means that a mutex could not be allocated. 
** SQLite will unwind its stack and return an error.  The argument
** to sqlite3_mutex_alloc() is one of these integer constants:
**
** <ul>
** <li>  SQLITE_MUTEX_FAST
** <li>  SQLITE_MUTEX_RECURSIVE
** <li>  SQLITE_MUTEX_STATIC_MASTER
** <li>  SQLITE_MUTEX_STATIC_MEM
** <li>  SQLITE_MUTEX_STATIC_MEM2
** <li>  SQLITE_MUTEX_STATIC_PRNG
** <li>  SQLITE_MUTEX_STATIC_LRU
** <li>  SQLITE_MUTEX_STATIC_LRU2
** </ul>
**
** The first two constants cause sqlite3_mutex_alloc() to create
** a new mutex.  The new mutex is recursive when SQLITE_MUTEX_RECURSIVE
** is used but not necessarily so when SQLITE_MUTEX_FAST is used.
** The mutex implementation does not need to make a distinction
** between SQLITE_MUTEX_RECURSIVE and SQLITE_MUTEX_FAST if it does
** not want to.  But SQLite will only request a recursive mutex in
** cases where it really needs one.  If a faster non-recursive mutex
** implementation is available on the host platform, the mutex subsystem
** might return such a mutex in response to SQLITE_MUTEX_FAST.
**
** The other allowed parameters to sqlite3_mutex_alloc() each return
** a pointer to a static preexisting mutex.  Six static mutexes are
** used by the current version of SQLite.  Future versions of SQLite
** may add additional static mutexes.  Static mutexes are for internal
** use by SQLite only.  Applications that use SQLite mutexes should
** use only the dynamic mutexes returned by SQLITE_MUTEX_FAST or
** SQLITE_MUTEX_RECURSIVE.
**
** Note that if one of the dynamic mutex parameters (SQLITE_MUTEX_FAST

          p = NULL;
        }
      }
      break;
    }
    default: {
      static volatile int isInit = 0;
      static sqlite3_mutex staticMutexes[6] = {
        SQLITE3_MUTEX_INITIALIZER,
        SQLITE3_MUTEX_INITIALIZER,
        SQLITE3_MUTEX_INITIALIZER,
        SQLITE3_MUTEX_INITIALIZER,
        SQLITE3_MUTEX_INITIALIZER,
        SQLITE3_MUTEX_INITIALIZER,
      };
      if ( !isInit ){
        APIRET rc;
        PTIB ptib;
        PPIB ppib;
        HMTX mutex;
        char name[32];

/*
** This routine deallocates a previously allocated mutex.
** SQLite is careful to deallocate every mutex that it allocates.
*/
static void os2MutexFree(sqlite3_mutex *p){
#ifdef SQLITE_DEBUG
  TID tid;
  PID pid;
  ULONG ulCount;
  DosQueryMutexSem(p->mutex, &pid, &tid, &ulCount);
  assert( ulCount==0 );
  assert( p->id==SQLITE_MUTEX_FAST || p->id==SQLITE_MUTEX_RECURSIVE );
#endif
  DosCloseMutexSem( p->mutex );
  sqlite3_free( p );
}

#ifdef SQLITE_DEBUG
/*
** The sqlite3_mutex_held() and sqlite3_mutex_notheld() routine are
** intended for use inside assert() statements.
*/
static int os2MutexHeld(sqlite3_mutex *p){
  TID tid;
  PID pid;
  ULONG ulCount;
  PTIB ptib;

  DosQueryMutexSem(p->mutex, &pid, &tid, &ulCount);
  if( ulCount==0 || ( ulCount>1 && p->id!=SQLITE_MUTEX_RECURSIVE ) )
    return 0;
  DosGetInfoBlocks(&ptib, NULL);
  return tid==ptib->tib_ptib2->tib2_ultid;


}
static int os2MutexNotheld(sqlite3_mutex *p){
  TID tid;
  PID pid;
  ULONG ulCount;
  PTIB ptib;

  DosQueryMutexSem(p->mutex, &pid, &tid, &ulCount);
  if( ulCount==0 )
    return 1;
  DosGetInfoBlocks(&ptib, NULL);
  return tid!=ptib->tib_ptib2->tib2_ultid;
}
static void os2MutexTrace(sqlite3_mutex *p, char *pAction){
  TID   tid;
  PID   pid;
  ULONG ulCount;
  DosQueryMutexSem(p->mutex, &pid, &tid, &ulCount);
  printf("%s mutex %p (%d) with nRef=%ld\n", pAction, (void*)p, p->trace, ulCount);
}
#endif

/*
** The sqlite3_mutex_enter() and sqlite3_mutex_try() routines attempt
** to enter a mutex.  If another thread is already within the mutex,
** sqlite3_mutex_enter() will block and sqlite3_mutex_try() will return
** SQLITE_BUSY.  The sqlite3_mutex_try() interface returns SQLITE_OK
** upon successful entry.  Mutexes created using SQLITE_MUTEX_RECURSIVE can
** be entered multiple times by the same thread.  In such cases the,
** mutex must be exited an equal number of times before another thread
** can enter.  If the same thread tries to enter any other kind of mutex
** more than once, the behavior is undefined.
*/
static void os2MutexEnter(sqlite3_mutex *p){




  assert( p->id==SQLITE_MUTEX_RECURSIVE || os2MutexNotheld(p) );
  DosRequestMutexSem(p->mutex, SEM_INDEFINITE_WAIT);
#ifdef SQLITE_DEBUG
  if( p->trace ) os2MutexTrace(p, "enter");

#endif
}
static int os2MutexTry(sqlite3_mutex *p){
  int rc = SQLITE_BUSY;




  assert( p->id==SQLITE_MUTEX_RECURSIVE || os2MutexNotheld(p) );
  if( DosRequestMutexSem(p->mutex, SEM_IMMEDIATE_RETURN) == NO_ERROR ) {



    rc = SQLITE_OK;

#ifdef SQLITE_DEBUG
    if( p->trace ) os2MutexTrace(p, "try");
#endif
  }

  return rc;
}

/*
** The sqlite3_mutex_leave() routine exits a mutex that was
** previously entered by the same thread.  The behavior
** is undefined if the mutex is not currently entered or
** is not currently allocated.  SQLite will never do either.
*/
static void os2MutexLeave(sqlite3_mutex *p){




  assert( os2MutexHeld(p) );




  DosReleaseMutexSem(p->mutex);
#ifdef SQLITE_DEBUG
  if( p->trace ) os2MutexTrace(p, "leave");
#endif
}

SQLITE_PRIVATE sqlite3_mutex_methods const *sqlite3DefaultMutex(void){
  static const sqlite3_mutex_methods sMutex = {
    os2MutexInit,
    os2MutexEnd,
    os2MutexAlloc,
    os2MutexFree,
    os2MutexEnter,
    os2MutexTry,
    os2MutexLeave,
#ifdef SQLITE_DEBUG
    os2MutexHeld,
    os2MutexNotheld
#else
    0,
    0
#endif
  };

  return &sMutex;
}
#endif /* SQLITE_MUTEX_OS2 */

    rc = SQLITE_OK;
  }
#else
  UNUSED_PARAMETER(p);
#endif
#ifdef SQLITE_DEBUG
  if( rc==SQLITE_OK && p->trace ){
    printf("try mutex %p (%d) with nRef=%d\n", p, p->trace, p->nRef);
  }
#endif
  return rc;
}

/*
** The sqlite3_mutex_leave() routine exits a mutex that was

      /* Release memory from the SQLITE_CONFIG_SCRATCH allocation */
      ScratchFreeslot *pSlot;
      pSlot = (ScratchFreeslot*)p;
      sqlite3_mutex_enter(mem0.mutex);
      pSlot->pNext = mem0.pScratchFree;
      mem0.pScratchFree = pSlot;
      mem0.nScratchFree++;
      assert( mem0.nScratchFree <= (u32)sqlite3GlobalConfig.nScratch );
      sqlite3StatusAdd(SQLITE_STATUS_SCRATCH_USED, -1);
      sqlite3_mutex_leave(mem0.mutex);
    }else{
      /* Release memory back to the heap */
      assert( sqlite3MemdebugHasType(p, MEMTYPE_SCRATCH) );
      assert( sqlite3MemdebugNoType(p, ~MEMTYPE_SCRATCH) );
      sqlite3MemdebugSetType(p, MEMTYPE_HEAP);

            v = va_arg(ap,i64);
          }else if( flag_long ){
            v = va_arg(ap,long int);
          }else{
            v = va_arg(ap,int);
          }
          if( v<0 ){
            if( v==SMALLEST_INT64 ){
              longvalue = ((u64)1)<<63;
            }else{
              longvalue = -v;
            }
            prefix = '-';
          }else{
            longvalue = v;
            if( flag_plussign )        prefix = '+';
            else if( flag_blanksign )  prefix = ' ';
            else                       prefix = 0;
          }

        || (c&0xFFFFF800)==0xD800                          \
        || (c&0xFFFFFFFE)==0xFFFE ){  c = 0xFFFD; }        \
  }
SQLITE_PRIVATE int sqlite3Utf8Read(
  const unsigned char *zIn,       /* First byte of UTF-8 character */
  const unsigned char **pzNext    /* Write first byte past UTF-8 char here */
){
  unsigned int c;

  /* Same as READ_UTF8() above but without the zTerm parameter.
  ** For this routine, we assume the UTF8 string is always zero-terminated.
  */
  c = *(zIn++);
  if( c>=0xc0 ){
    c = sqlite3Utf8Trans1[c-0xc0];

#if defined(SQLITE_TEST) && defined(SQLITE_DEBUG)
/*
** Translate UTF-8 to UTF-8.
**
** This has the effect of making sure that the string is well-formed
** UTF-8.  Miscoded characters are removed.
**
** The translation is done in-place and aborted if the output
** overruns the input.
*/
SQLITE_PRIVATE int sqlite3Utf8To8(unsigned char *zIn){
  unsigned char *zOut = zIn;
  unsigned char *zStart = zIn;
  u32 c;

  while( zIn[0] && zOut<=zIn ){
    c = sqlite3Utf8Read(zIn, (const u8**)&zIn);
    if( c!=0xfffd ){
      WRITE_UTF8(zOut, c);
    }
  }
  *zOut = 0;
  return (int)(zOut - zStart);

#endif

/*
** Routine needed to support the testcase() macro.
*/
#ifdef SQLITE_COVERAGE_TEST
SQLITE_PRIVATE void sqlite3Coverage(int x){
  static unsigned dummy = 0;
  dummy += (unsigned)x;
}
#endif

#ifndef SQLITE_OMIT_FLOATING_POINT
/*
** Return true if the floating point value is Not a Number (NaN).
**

    testcase( c==(+1) );
  }
  return c;
}


/*
** Convert zNum to a 64-bit signed integer.
**
** If the zNum value is representable as a 64-bit twos-complement 
** integer, then write that value into *pNum and return 0.
**
** If zNum is exactly 9223372036854665808, return 2.  This special
** case is broken out because while 9223372036854665808 cannot be a 
** signed 64-bit integer, its negative -9223372036854665808 can be.
**
** If zNum is too big for a 64-bit integer and is not

** 9223372036854665808 then return 1.
**
** length is the number of bytes in the string (bytes, not characters).
** The string is not necessarily zero-terminated.  The encoding is
** given by enc.
*/
SQLITE_PRIVATE int sqlite3Atoi64(const char *zNum, i64 *pNum, int length, u8 enc){
  int incr = (enc==SQLITE_UTF8?1:2);
  u64 u = 0;
  int neg = 0; /* assume positive */
  int i;
  int c = 0;
  const char *zStart;
  const char *zEnd = zNum + length;
  if( enc==SQLITE_UTF16BE ) zNum++;
  while( zNum<zEnd && sqlite3Isspace(*zNum) ) zNum+=incr;
  if( zNum<zEnd ){
    if( *zNum=='-' ){
      neg = 1;
      zNum+=incr;
    }else if( *zNum=='+' ){
      zNum+=incr;
    }

  }
  zStart = zNum;
  while( zNum<zEnd && zNum[0]=='0' ){ zNum+=incr; } /* Skip leading zeros. */
  for(i=0; &zNum[i]<zEnd && (c=zNum[i])>='0' && c<='9'; i+=incr){
    u = u*10 + c - '0';
  }
  if( u>LARGEST_INT64 ){
    *pNum = SMALLEST_INT64;
  }else if( neg ){
    *pNum = -(i64)u;
  }else{
    *pNum = (i64)u;
  }
  testcase( i==18 );
  testcase( i==19 );
  testcase( i==20 );
  if( (c!=0 && &zNum[i]<zEnd) || (i==0 && zStart==zNum) || i>19*incr ){
    /* zNum is empty or contains non-numeric text or is longer
    ** than 19 digits (thus guaranteeing that it is too large) */
    return 1;
  }else if( i<19*incr ){
    /* Less than 19 digits, so we know that it fits in 64 bits */
    assert( u<=LARGEST_INT64 );
    return 0;
  }else{

    /* zNum is a 19-digit numbers.  Compare it against 9223372036854775808. */

    c = compare2pow63(zNum, incr);
    if( c<0 ){
      /* zNum is less than 9223372036854775808 so it fits */
      assert( u<=LARGEST_INT64 );
      return 0;
    }else if( c>0 ){
      /* zNum is greater than 9223372036854775808 so it overflows */
      return 1;
    }else{
      /* zNum is exactly 9223372036854775808.  Fits if negative.  The
      ** special case 2 overflow if positive */
      assert( u-1==LARGEST_INT64 );
      assert( (*pNum)==SMALLEST_INT64 );
      return neg ? 0 : 2;
    }
  }
}

/*
** If zNum represents an integer that will fit in 32-bits, then set
** *pValue to that integer and return true.  Otherwise return false.
**

    testcase( sqlite3GlobalConfig.xLog!=0 );
    logBadConnection("invalid");
    return 0;
  }else{
    return 1;
  }
}

/*
** Attempt to add, substract, or multiply the 64-bit signed value iB against
** the other 64-bit signed integer at *pA and store the result in *pA.
** Return 0 on success.  Or if the operation would have resulted in an
** overflow, leave *pA unchanged and return 1.
*/
SQLITE_PRIVATE int sqlite3AddInt64(i64 *pA, i64 iB){
  i64 iA = *pA;
  testcase( iA==0 ); testcase( iA==1 );
  testcase( iB==-1 ); testcase( iB==0 );
  if( iB>=0 ){
    testcase( iA>0 && LARGEST_INT64 - iA == iB );
    testcase( iA>0 && LARGEST_INT64 - iA == iB - 1 );
    if( iA>0 && LARGEST_INT64 - iA < iB ) return 1;
    *pA += iB;
  }else{
    testcase( iA<0 && -(iA + LARGEST_INT64) == iB + 1 );
    testcase( iA<0 && -(iA + LARGEST_INT64) == iB + 2 );
    if( iA<0 && -(iA + LARGEST_INT64) > iB + 1 ) return 1;
    *pA += iB;
  }
  return 0; 
}
SQLITE_PRIVATE int sqlite3SubInt64(i64 *pA, i64 iB){
  testcase( iB==SMALLEST_INT64+1 );
  if( iB==SMALLEST_INT64 ){
    testcase( (*pA)==(-1) ); testcase( (*pA)==0 );
    if( (*pA)>=0 ) return 1;
    *pA -= iB;
    return 0;
  }else{
    return sqlite3AddInt64(pA, -iB);
  }
}
#define TWOPOWER32 (((i64)1)<<32)
#define TWOPOWER31 (((i64)1)<<31)
SQLITE_PRIVATE int sqlite3MulInt64(i64 *pA, i64 iB){
  i64 iA = *pA;
  i64 iA1, iA0, iB1, iB0, r;

  iA1 = iA/TWOPOWER32;
  iA0 = iA % TWOPOWER32;
  iB1 = iB/TWOPOWER32;
  iB0 = iB % TWOPOWER32;
  if( iA1*iB1 != 0 ) return 1;
  assert( iA1*iB0==0 || iA0*iB1==0 );
  r = iA1*iB0 + iA0*iB1;
  testcase( r==(-TWOPOWER31)-1 );
  testcase( r==(-TWOPOWER31) );
  testcase( r==TWOPOWER31 );
  testcase( r==TWOPOWER31-1 );
  if( r<(-TWOPOWER31) || r>=TWOPOWER31 ) return 1;
  r *= TWOPOWER32;
  if( sqlite3AddInt64(&r, iA0*iB0) ) return 1;
  *pA = r;
  return 0;
}

/*
** Compute the absolute value of a 32-bit signed integer, of possible.  Or 
** if the integer has a value of -2147483648, return +2147483647
*/
SQLITE_PRIVATE int sqlite3AbsInt32(int x){
  if( x>=0 ) return x;
  if( x==(int)0x80000000 ) return 0x7fffffff;
  return -x;
}

/************** End of util.c ************************************************/
/************** Begin file hash.c ********************************************/
/*
** 2001 September 22
**
** The author disclaims copyright to this source code.  In place of

#endif

#endif /* !defined(_OS_COMMON_H_) */

/************** End of os_common.h *******************************************/
/************** Continuing where we left off in os_os2.c *********************/

/* Forward references */
typedef struct os2File os2File;         /* The file structure */
typedef struct os2ShmNode os2ShmNode;   /* A shared descritive memory node */
typedef struct os2ShmLink os2ShmLink;   /* A connection to shared-memory */

/*
** The os2File structure is subclass of sqlite3_file specific for the OS/2
** protability layer.
*/

struct os2File {
  const sqlite3_io_methods *pMethod;  /* Always the first entry */
  HFILE h;                  /* Handle for accessing the file */
  int flags;                /* Flags provided to os2Open() */
  int locktype;             /* Type of lock currently held on this file */
  int szChunk;              /* Chunk size configured by FCNTL_CHUNK_SIZE */
  char *zFullPathCp;        /* Full path name of this file */
  os2ShmLink *pShmLink;     /* Instance of shared memory on this file */
};

#define LOCK_TIMEOUT 10L /* the default locking timeout */

/*
** Missing from some versions of the OS/2 toolkit -
** used to allocate from high memory if possible
*/
#ifndef OBJ_ANY
# define OBJ_ANY 0x00000400
#endif

/*****************************************************************************
** The next group of routines implement the I/O methods specified
** by the sqlite3_io_methods object.
******************************************************************************/

/*
** Close a file.
*/
static int os2Close( sqlite3_file *id ){
  APIRET rc;
  os2File *pFile = (os2File*)id;

  assert( id!=0 );
  OSTRACE(( "CLOSE %d (%s)\n", pFile->h, pFile->zFullPathCp ));

  rc = DosClose( pFile->h );

  if( pFile->flags & SQLITE_OPEN_DELETEONCLOSE )
    DosForceDelete( (PSZ)pFile->zFullPathCp );

  free( pFile->zFullPathCp );
  pFile->zFullPathCp = NULL;
  pFile->locktype = NO_LOCK;
  pFile->h = (HFILE)-1;
  pFile->flags = 0;


  OpenCounter( -1 );


  return rc == NO_ERROR ? SQLITE_OK : SQLITE_IOERR;
}

/*
** Read data from a file into a buffer.  Return SQLITE_OK if all
** bytes were read successfully and SQLITE_IOERR if anything goes
** wrong.

  return ( rc != NO_ERROR || amt > (int)wrote ) ? SQLITE_FULL : SQLITE_OK;
}

/*
** Truncate an open file to a specified size
*/
static int os2Truncate( sqlite3_file *id, i64 nByte ){
  APIRET rc;
  os2File *pFile = (os2File*)id;
  assert( id!=0 );
  OSTRACE(( "TRUNCATE %d %lld\n", pFile->h, nByte ));
  SimulateIOError( return SQLITE_IOERR_TRUNCATE );

  /* If the user has configured a chunk-size for this file, truncate the
  ** file so that it consists of an integer number of chunks (i.e. the
  ** actual file size after the operation may be larger than the requested
  ** size).
  */
  if( pFile->szChunk ){
    nByte = ((nByte + pFile->szChunk - 1)/pFile->szChunk) * pFile->szChunk;
  }
  
  rc = DosSetFileSize( pFile->h, nByte );
  return rc == NO_ERROR ? SQLITE_OK : SQLITE_IOERR_TRUNCATE;
}

#ifdef SQLITE_TEST
/*
** Count the number of fullsyncs and normal syncs.  This is used to test

  switch( op ){
    case SQLITE_FCNTL_LOCKSTATE: {
      *(int*)pArg = ((os2File*)id)->locktype;
      OSTRACE(( "FCNTL_LOCKSTATE %d lock=%d\n",
                ((os2File*)id)->h, ((os2File*)id)->locktype ));
      return SQLITE_OK;
    }
    case SQLITE_FCNTL_CHUNK_SIZE: {
      ((os2File*)id)->szChunk = *(int*)pArg;
      return SQLITE_OK;
    }
    case SQLITE_FCNTL_SIZE_HINT: {
      sqlite3_int64 sz = *(sqlite3_int64*)pArg;
      SimulateIOErrorBenign(1);
      os2Truncate(id, sz);
      SimulateIOErrorBenign(0);
      return SQLITE_OK;
    }
    case SQLITE_FCNTL_SYNC_OMITTED: {
      return SQLITE_OK;
    }
  }
  return SQLITE_NOTFOUND;
}

/*
** Return the sector size in bytes of the underlying block device for
** the specified file. This is almost always 512 bytes, but may be
** larger for some devices.
**
** SQLite code assumes this function cannot fail. It also assumes that
** if two files are created in the same file-system directory (i.e.
** a database and its journal file) that the sector size will be the
** same for both.
*/
static int os2SectorSize(sqlite3_file *id){
  UNUSED_PARAMETER(id);
  return SQLITE_DEFAULT_SECTOR_SIZE;
}

/*
** Return a vector of device characteristics.
*/
static int os2DeviceCharacteristics(sqlite3_file *id){
  UNUSED_PARAMETER(id);
  return SQLITE_IOCAP_UNDELETABLE_WHEN_OPEN;
}


/*
** Character set conversion objects used by conversion routines.
*/
static UconvObject ucUtf8 = NULL; /* convert between UTF-8 and UCS-2 */

    return out; /* if conversion fails, return the empty string */

  /* determine string for the conversion of UTF-8 which is CP1208 */
  UniStrFromUcs( ucUtf8, out, tempPath, CCHMAXPATH );

  return out;
}


#ifndef SQLITE_OMIT_WAL

/*
** Use main database file for interprocess locking. If un-defined
** a separate file is created for this purpose. The file will be
** used only to set file locks. There will be no data written to it.
*/
#define SQLITE_OS2_NO_WAL_LOCK_FILE     

#if 0
static void _ERR_TRACE( const char *fmt, ... ) {
  va_list  ap;
  va_start(ap, fmt);
  vfprintf(stderr, fmt, ap);
  fflush(stderr);
}
#define ERR_TRACE(rc, msg)        \
        if( (rc) != SQLITE_OK ) _ERR_TRACE msg;
#else
#define ERR_TRACE(rc, msg)
#endif

/*
** Helper functions to obtain and relinquish the global mutex. The
** global mutex is used to protect os2ShmNodeList.
**
** Function os2ShmMutexHeld() is used to assert() that the global mutex 
** is held when required. This function is only used as part of assert() 
** statements. e.g.
**
**   os2ShmEnterMutex()
**     assert( os2ShmMutexHeld() );
**   os2ShmLeaveMutex()
*/
static void os2ShmEnterMutex(void){
  sqlite3_mutex_enter(sqlite3MutexAlloc(SQLITE_MUTEX_STATIC_MASTER));
}
static void os2ShmLeaveMutex(void){
  sqlite3_mutex_leave(sqlite3MutexAlloc(SQLITE_MUTEX_STATIC_MASTER));
}
#ifdef SQLITE_DEBUG
static int os2ShmMutexHeld(void) {
  return sqlite3_mutex_held(sqlite3MutexAlloc(SQLITE_MUTEX_STATIC_MASTER));
}
int GetCurrentProcessId(void) {
  PPIB pib;
  DosGetInfoBlocks(NULL, &pib);
  return (int)pib->pib_ulpid;
}
#endif

/*
** Object used to represent a the shared memory area for a single log file.
** When multiple threads all reference the same log-summary, each thread has
** its own os2File object, but they all point to a single instance of this 
** object.  In other words, each log-summary is opened only once per process.
**
** os2ShmMutexHeld() must be true when creating or destroying
** this object or while reading or writing the following fields:
**
**      nRef
**      pNext 
**
** The following fields are read-only after the object is created:
** 
**      szRegion
**      hLockFile
**      shmBaseName
**
** Either os2ShmNode.mutex must be held or os2ShmNode.nRef==0 and
** os2ShmMutexHeld() is true when reading or writing any other field
** in this structure.
**
*/
struct os2ShmNode {
  sqlite3_mutex *mutex;      /* Mutex to access this object */
  os2ShmNode *pNext;         /* Next in list of all os2ShmNode objects */

  int szRegion;              /* Size of shared-memory regions */

  int nRegion;               /* Size of array apRegion */
  void **apRegion;           /* Array of pointers to shared-memory regions */

  int nRef;                  /* Number of os2ShmLink objects pointing to this */
  os2ShmLink *pFirst;        /* First os2ShmLink object pointing to this */

  HFILE hLockFile;           /* File used for inter-process memory locking */
  char shmBaseName[1];       /* Name of the memory object !!! must last !!! */
};


/*
** Structure used internally by this VFS to record the state of an
** open shared memory connection.
**
** The following fields are initialized when this object is created and
** are read-only thereafter:
**
**    os2Shm.pShmNode
**    os2Shm.id
**
** All other fields are read/write.  The os2Shm.pShmNode->mutex must be held
** while accessing any read/write fields.
*/
struct os2ShmLink {
  os2ShmNode *pShmNode;      /* The underlying os2ShmNode object */
  os2ShmLink *pNext;         /* Next os2Shm with the same os2ShmNode */
  u32 sharedMask;            /* Mask of shared locks held */
  u32 exclMask;              /* Mask of exclusive locks held */
#ifdef SQLITE_DEBUG
  u8 id;                     /* Id of this connection with its os2ShmNode */
#endif
};


/*
** A global list of all os2ShmNode objects.
**
** The os2ShmMutexHeld() must be true while reading or writing this list.
*/
static os2ShmNode *os2ShmNodeList = NULL;

/*
** Constants used for locking
*/
#ifdef  SQLITE_OS2_NO_WAL_LOCK_FILE
#define OS2_SHM_BASE   (PENDING_BYTE + 0x10000)         /* first lock byte */
#else
#define OS2_SHM_BASE   ((22+SQLITE_SHM_NLOCK)*4)        /* first lock byte */
#endif

#define OS2_SHM_DMS    (OS2_SHM_BASE+SQLITE_SHM_NLOCK)  /* deadman switch */

/*
** Apply advisory locks for all n bytes beginning at ofst.
*/
#define _SHM_UNLCK  1   /* no lock */
#define _SHM_RDLCK  2   /* shared lock, no wait */
#define _SHM_WRLCK  3   /* exlusive lock, no wait */
#define _SHM_WRLCK_WAIT 4 /* exclusive lock, wait */
static int os2ShmSystemLock(
  os2ShmNode *pNode,    /* Apply locks to this open shared-memory segment */
  int lockType,         /* _SHM_UNLCK, _SHM_RDLCK, _SHM_WRLCK or _SHM_WRLCK_WAIT */
  int ofst,             /* Offset to first byte to be locked/unlocked */
  int nByte             /* Number of bytes to lock or unlock */
){
  APIRET rc;
  FILELOCK area;
  ULONG mode, timeout;

  /* Access to the os2ShmNode object is serialized by the caller */
  assert( sqlite3_mutex_held(pNode->mutex) || pNode->nRef==0 );

  mode = 1;     /* shared lock */
  timeout = 0;  /* no wait */
  area.lOffset = ofst;
  area.lRange = nByte;

  switch( lockType ) {
    case _SHM_WRLCK_WAIT:
      timeout = (ULONG)-1;      /* wait forever */
    case _SHM_WRLCK:
      mode = 0;                 /* exclusive lock */
    case _SHM_RDLCK:
      rc = DosSetFileLocks(pNode->hLockFile, 
                           NULL, &area, timeout, mode);
      break;
    /* case _SHM_UNLCK: */
    default:
      rc = DosSetFileLocks(pNode->hLockFile, 
                           &area, NULL, 0, 0);
      break;
  }
                          
  OSTRACE(("SHM-LOCK %d %s %s 0x%08lx\n", 
           pNode->hLockFile,
           rc==SQLITE_OK ? "ok" : "failed",
           lockType==_SHM_UNLCK ? "Unlock" : "Lock",
           rc));

  ERR_TRACE(rc, ("os2ShmSystemLock: %d %s\n", rc, pNode->shmBaseName))

  return ( rc == 0 ) ?  SQLITE_OK : SQLITE_BUSY;
}

/*
** Find an os2ShmNode in global list or allocate a new one, if not found.
**
** This is not a VFS shared-memory method; it is a utility function called
** by VFS shared-memory methods.
*/
static int os2OpenSharedMemory( os2File *fd, int szRegion ) {
  os2ShmLink *pLink;
  os2ShmNode *pNode;
  int cbShmName, rc = SQLITE_OK;
  char shmName[CCHMAXPATH + 30];
#ifndef SQLITE_OS2_NO_WAL_LOCK_FILE
  ULONG action;
#endif
  
  /* We need some additional space at the end to append the region number */
  cbShmName = sprintf(shmName, "\\SHAREMEM\\%s", fd->zFullPathCp );
  if( cbShmName >= CCHMAXPATH-8 )
    return SQLITE_IOERR_SHMOPEN; 

  /* Replace colon in file name to form a valid shared memory name */
  shmName[10+1] = '!';

  /* Allocate link object (we free it later in case of failure) */
  pLink = sqlite3_malloc( sizeof(*pLink) );
  if( !pLink )
    return SQLITE_NOMEM;

  /* Access node list */
  os2ShmEnterMutex();

  /* Find node by it's shared memory base name */
  for( pNode = os2ShmNodeList; 
       pNode && stricmp(shmName, pNode->shmBaseName) != 0; 
       pNode = pNode->pNext )   ;

  /* Not found: allocate a new node */
  if( !pNode ) {
    pNode = sqlite3_malloc( sizeof(*pNode) + cbShmName );
    if( pNode ) {
      memset(pNode, 0, sizeof(*pNode) );
      pNode->szRegion = szRegion;
      pNode->hLockFile = (HFILE)-1;      
      strcpy(pNode->shmBaseName, shmName);

#ifdef SQLITE_OS2_NO_WAL_LOCK_FILE
      if( DosDupHandle(fd->h, &pNode->hLockFile) != 0 ) {
#else
      sprintf(shmName, "%s-lck", fd->zFullPathCp);
      if( DosOpen((PSZ)shmName, &pNode->hLockFile, &action, 0, FILE_NORMAL, 
                  OPEN_ACTION_OPEN_IF_EXISTS | OPEN_ACTION_CREATE_IF_NEW,
                  OPEN_ACCESS_READWRITE | OPEN_SHARE_DENYNONE | 
                  OPEN_FLAGS_NOINHERIT | OPEN_FLAGS_FAIL_ON_ERROR,
                  NULL) != 0 ) {
#endif
        sqlite3_free(pNode);  
        rc = SQLITE_IOERR;
      } else {
        pNode->mutex = sqlite3_mutex_alloc(SQLITE_MUTEX_FAST);
        if( !pNode->mutex ) {
          sqlite3_free(pNode);  
          rc = SQLITE_NOMEM;
        }
      }   
    } else {
      rc = SQLITE_NOMEM;
    }
    
    if( rc == SQLITE_OK ) {
      pNode->pNext = os2ShmNodeList;
      os2ShmNodeList = pNode;
    } else {
      pNode = NULL;
    }
  } else if( pNode->szRegion != szRegion ) {
    rc = SQLITE_IOERR_SHMSIZE;
    pNode = NULL;
  }

  if( pNode ) {
    sqlite3_mutex_enter(pNode->mutex);

    memset(pLink, 0, sizeof(*pLink));

    pLink->pShmNode = pNode;
    pLink->pNext = pNode->pFirst;
    pNode->pFirst = pLink;
    pNode->nRef++;

    fd->pShmLink = pLink;

    sqlite3_mutex_leave(pNode->mutex);
    
  } else {
    /* Error occured. Free our link object. */
    sqlite3_free(pLink);  
  }

  os2ShmLeaveMutex();

  ERR_TRACE(rc, ("os2OpenSharedMemory: %d  %s\n", rc, fd->zFullPathCp))  
  
  return rc;
}

/*
** Purge the os2ShmNodeList list of all entries with nRef==0.
**
** This is not a VFS shared-memory method; it is a utility function called
** by VFS shared-memory methods.
*/
static void os2PurgeShmNodes( int deleteFlag ) {
  os2ShmNode *pNode;
  os2ShmNode **ppNode;

  os2ShmEnterMutex();
  
  ppNode = &os2ShmNodeList;

  while( *ppNode ) {
    pNode = *ppNode;

    if( pNode->nRef == 0 ) {
      *ppNode = pNode->pNext;   
     
      if( pNode->apRegion ) {
        /* Prevent other processes from resizing the shared memory */
        os2ShmSystemLock(pNode, _SHM_WRLCK_WAIT, OS2_SHM_DMS, 1);

        while( pNode->nRegion-- ) {
#ifdef SQLITE_DEBUG
          int rc = 
#endif          
          DosFreeMem(pNode->apRegion[pNode->nRegion]);

          OSTRACE(("SHM-PURGE pid-%d unmap region=%d %s\n",
                  (int)GetCurrentProcessId(), pNode->nRegion,
                  rc == 0 ? "ok" : "failed"));
        }

        /* Allow other processes to resize the shared memory */
        os2ShmSystemLock(pNode, _SHM_UNLCK, OS2_SHM_DMS, 1);

        sqlite3_free(pNode->apRegion);
      }  

      DosClose(pNode->hLockFile);
      
#ifndef SQLITE_OS2_NO_WAL_LOCK_FILE
      if( deleteFlag ) {
         char fileName[CCHMAXPATH];
         /* Skip "\\SHAREMEM\\" */
         sprintf(fileName, "%s-lck", pNode->shmBaseName + 10);
         /* restore colon */
         fileName[1] = ':';
         
         DosForceDelete(fileName); 
      }
#endif

      sqlite3_mutex_free(pNode->mutex);

      sqlite3_free(pNode);
      
    } else {
      ppNode = &pNode->pNext;
    }
  } 

  os2ShmLeaveMutex();
}

/*
** This function is called to obtain a pointer to region iRegion of the
** shared-memory associated with the database file id. Shared-memory regions
** are numbered starting from zero. Each shared-memory region is szRegion
** bytes in size.
**
** If an error occurs, an error code is returned and *pp is set to NULL.
**
** Otherwise, if the bExtend parameter is 0 and the requested shared-memory
** region has not been allocated (by any client, including one running in a
** separate process), then *pp is set to NULL and SQLITE_OK returned. If
** bExtend is non-zero and the requested shared-memory region has not yet
** been allocated, it is allocated by this function.
**
** If the shared-memory region has already been allocated or is allocated by
** this call as described above, then it is mapped into this processes
** address space (if it is not already), *pp is set to point to the mapped
** memory and SQLITE_OK returned.
*/
static int os2ShmMap(
  sqlite3_file *id,               /* Handle open on database file */
  int iRegion,                    /* Region to retrieve */
  int szRegion,                   /* Size of regions */
  int bExtend,                    /* True to extend block if necessary */
  void volatile **pp              /* OUT: Mapped memory */
){
  PVOID pvTemp;
  void **apRegion;
  os2ShmNode *pNode;
  int n, rc = SQLITE_OK;
  char shmName[CCHMAXPATH];
  os2File *pFile = (os2File*)id;
  
  *pp = NULL;

  if( !pFile->pShmLink )
    rc = os2OpenSharedMemory( pFile, szRegion );
  
  if( rc == SQLITE_OK ) {
    pNode = pFile->pShmLink->pShmNode ;
    
    sqlite3_mutex_enter(pNode->mutex);
    
    assert( szRegion==pNode->szRegion );

    /* Unmapped region ? */
    if( iRegion >= pNode->nRegion ) {
      /* Prevent other processes from resizing the shared memory */
      os2ShmSystemLock(pNode, _SHM_WRLCK_WAIT, OS2_SHM_DMS, 1);

      apRegion = sqlite3_realloc(
        pNode->apRegion, (iRegion + 1) * sizeof(apRegion[0]));

      if( apRegion ) {
        pNode->apRegion = apRegion;

        while( pNode->nRegion <= iRegion ) {
          sprintf(shmName, "%s-%u", 
                  pNode->shmBaseName, pNode->nRegion);

          if( DosGetNamedSharedMem(&pvTemp, (PSZ)shmName, 
                PAG_READ | PAG_WRITE) != NO_ERROR ) {
            if( !bExtend )
              break;

            if( DosAllocSharedMem(&pvTemp, (PSZ)shmName, szRegion,
                  PAG_READ | PAG_WRITE | PAG_COMMIT | OBJ_ANY) != NO_ERROR && 
                DosAllocSharedMem(&pvTemp, (PSZ)shmName, szRegion,
                  PAG_READ | PAG_WRITE | PAG_COMMIT) != NO_ERROR ) { 
              rc = SQLITE_NOMEM;
              break;
            }
          }

          apRegion[pNode->nRegion++] = pvTemp;
        }

        /* zero out remaining entries */ 
        for( n = pNode->nRegion; n <= iRegion; n++ )
          pNode->apRegion[n] = NULL;

        /* Return this region (maybe zero) */
        *pp = pNode->apRegion[iRegion];
      } else {
        rc = SQLITE_NOMEM;
      }

      /* Allow other processes to resize the shared memory */
      os2ShmSystemLock(pNode, _SHM_UNLCK, OS2_SHM_DMS, 1);
      
    } else {
      /* Region has been mapped previously */
      *pp = pNode->apRegion[iRegion];
    }

    sqlite3_mutex_leave(pNode->mutex);
  } 

  ERR_TRACE(rc, ("os2ShmMap: %s iRgn = %d, szRgn = %d, bExt = %d : %d\n", 
                 pFile->zFullPathCp, iRegion, szRegion, bExtend, rc))
          
  return rc;
}

/*
** Close a connection to shared-memory.  Delete the underlying
** storage if deleteFlag is true.
**
** If there is no shared memory associated with the connection then this
** routine is a harmless no-op.
*/
static int os2ShmUnmap(
  sqlite3_file *id,               /* The underlying database file */
  int deleteFlag                  /* Delete shared-memory if true */
){
  os2File *pFile = (os2File*)id;
  os2ShmLink *pLink = pFile->pShmLink;
  
  if( pLink ) {
    int nRef = -1;
    os2ShmLink **ppLink;
    os2ShmNode *pNode = pLink->pShmNode;

    sqlite3_mutex_enter(pNode->mutex);
    
    for( ppLink = &pNode->pFirst;
         *ppLink && *ppLink != pLink;
         ppLink = &(*ppLink)->pNext )   ;
         
    assert(*ppLink);

    if( *ppLink ) {
      *ppLink = pLink->pNext;
      nRef = --pNode->nRef;
    } else {
      ERR_TRACE(1, ("os2ShmUnmap: link not found ! %s\n", 
                    pNode->shmBaseName))
    }
    
    pFile->pShmLink = NULL;
    sqlite3_free(pLink);

    sqlite3_mutex_leave(pNode->mutex);
    
    if( nRef == 0 )
      os2PurgeShmNodes( deleteFlag );
  }

  return SQLITE_OK;
}

/*
** Change the lock state for a shared-memory segment.
**
** Note that the relationship between SHAREd and EXCLUSIVE locks is a little
** different here than in posix.  In xShmLock(), one can go from unlocked
** to shared and back or from unlocked to exclusive and back.  But one may
** not go from shared to exclusive or from exclusive to shared.
*/
static int os2ShmLock(
  sqlite3_file *id,          /* Database file holding the shared memory */
  int ofst,                  /* First lock to acquire or release */
  int n,                     /* Number of locks to acquire or release */
  int flags                  /* What to do with the lock */
){
  u32 mask;                             /* Mask of locks to take or release */
  int rc = SQLITE_OK;                   /* Result code */
  os2File *pFile = (os2File*)id;
  os2ShmLink *p = pFile->pShmLink;      /* The shared memory being locked */
  os2ShmLink *pX;                       /* For looping over all siblings */
  os2ShmNode *pShmNode = p->pShmNode;   /* Our node */
  
  assert( ofst>=0 && ofst+n<=SQLITE_SHM_NLOCK );
  assert( n>=1 );
  assert( flags==(SQLITE_SHM_LOCK | SQLITE_SHM_SHARED)
       || flags==(SQLITE_SHM_LOCK | SQLITE_SHM_EXCLUSIVE)
       || flags==(SQLITE_SHM_UNLOCK | SQLITE_SHM_SHARED)
       || flags==(SQLITE_SHM_UNLOCK | SQLITE_SHM_EXCLUSIVE) );
  assert( n==1 || (flags & SQLITE_SHM_EXCLUSIVE)!=0 );

  mask = (u32)((1U<<(ofst+n)) - (1U<<ofst));
  assert( n>1 || mask==(1<<ofst) );


  sqlite3_mutex_enter(pShmNode->mutex);

  if( flags & SQLITE_SHM_UNLOCK ){
    u32 allMask = 0; /* Mask of locks held by siblings */

    /* See if any siblings hold this same lock */
    for(pX=pShmNode->pFirst; pX; pX=pX->pNext){
      if( pX==p ) continue;
      assert( (pX->exclMask & (p->exclMask|p->sharedMask))==0 );
      allMask |= pX->sharedMask;
    }

    /* Unlock the system-level locks */
    if( (mask & allMask)==0 ){
      rc = os2ShmSystemLock(pShmNode, _SHM_UNLCK, ofst+OS2_SHM_BASE, n);
    }else{
      rc = SQLITE_OK;
    }

    /* Undo the local locks */
    if( rc==SQLITE_OK ){
      p->exclMask &= ~mask;
      p->sharedMask &= ~mask;
    } 
  }else if( flags & SQLITE_SHM_SHARED ){
    u32 allShared = 0;  /* Union of locks held by connections other than "p" */

    /* Find out which shared locks are already held by sibling connections.
    ** If any sibling already holds an exclusive lock, go ahead and return
    ** SQLITE_BUSY.
    */
    for(pX=pShmNode->pFirst; pX; pX=pX->pNext){
      if( (pX->exclMask & mask)!=0 ){
        rc = SQLITE_BUSY;
        break;
      }
      allShared |= pX->sharedMask;
    }

    /* Get shared locks at the system level, if necessary */
    if( rc==SQLITE_OK ){
      if( (allShared & mask)==0 ){
        rc = os2ShmSystemLock(pShmNode, _SHM_RDLCK, ofst+OS2_SHM_BASE, n);
      }else{
        rc = SQLITE_OK;
      }
    }

    /* Get the local shared locks */
    if( rc==SQLITE_OK ){
      p->sharedMask |= mask;
    }
  }else{
    /* Make sure no sibling connections hold locks that will block this
    ** lock.  If any do, return SQLITE_BUSY right away.
    */
    for(pX=pShmNode->pFirst; pX; pX=pX->pNext){
      if( (pX->exclMask & mask)!=0 || (pX->sharedMask & mask)!=0 ){
        rc = SQLITE_BUSY;
        break;
      }
    }
  
    /* Get the exclusive locks at the system level.  Then if successful
    ** also mark the local connection as being locked.
    */
    if( rc==SQLITE_OK ){
      rc = os2ShmSystemLock(pShmNode, _SHM_WRLCK, ofst+OS2_SHM_BASE, n);
      if( rc==SQLITE_OK ){
        assert( (p->sharedMask & mask)==0 );
        p->exclMask |= mask;
      }
    }
  }

  sqlite3_mutex_leave(pShmNode->mutex);
  
  OSTRACE(("SHM-LOCK shmid-%d, pid-%d got %03x,%03x %s\n",
           p->id, (int)GetCurrentProcessId(), p->sharedMask, p->exclMask,
           rc ? "failed" : "ok"));

  ERR_TRACE(rc, ("os2ShmLock: ofst = %d, n = %d, flags = 0x%x -> %d \n", 
                 ofst, n, flags, rc))
                  
  return rc; 
}

/*
** Implement a memory barrier or memory fence on shared memory.
**
** All loads and stores begun before the barrier must complete before
** any load or store begun after the barrier.
*/
static void os2ShmBarrier(
  sqlite3_file *id                /* Database file holding the shared memory */
){
  UNUSED_PARAMETER(id);
  os2ShmEnterMutex();
  os2ShmLeaveMutex();
}

#else
# define os2ShmMap     0
# define os2ShmLock    0
# define os2ShmBarrier 0
# define os2ShmUnmap   0
#endif /* #ifndef SQLITE_OMIT_WAL */


/*
** This vector defines all the methods that can operate on an
** sqlite3_file for os2.
*/
static const sqlite3_io_methods os2IoMethod = {
  2,                              /* iVersion */
  os2Close,                       /* xClose */
  os2Read,                        /* xRead */
  os2Write,                       /* xWrite */
  os2Truncate,                    /* xTruncate */
  os2Sync,                        /* xSync */
  os2FileSize,                    /* xFileSize */
  os2Lock,                        /* xLock */
  os2Unlock,                      /* xUnlock */
  os2CheckReservedLock,           /* xCheckReservedLock */
  os2FileControl,                 /* xFileControl */
  os2SectorSize,                  /* xSectorSize */
  os2DeviceCharacteristics,       /* xDeviceCharacteristics */
  os2ShmMap,                      /* xShmMap */
  os2ShmLock,                     /* xShmLock */
  os2ShmBarrier,                  /* xShmBarrier */
  os2ShmUnmap                     /* xShmUnmap */
};


/***************************************************************************
** Here ends the I/O methods that form the sqlite3_io_methods object.
**
** The next block of code implements the VFS methods.
****************************************************************************/

/*
** Create a temporary file name in zBuf.  zBuf must be big enough to
** hold at pVfs->mxPathname characters.
*/
static int getTempname(int nBuf, char *zBuf ){
  static const char zChars[] =
    "abcdefghijklmnopqrstuvwxyz"
    "ABCDEFGHIJKLMNOPQRSTUVWXYZ"
    "0123456789";
  int i, j;
  PSZ zTempPathCp;      
  char zTempPath[CCHMAXPATH];
  ULONG ulDriveNum, ulDriveMap;
  
  /* It's odd to simulate an io-error here, but really this is just
  ** using the io-error infrastructure to test that SQLite handles this
  ** function failing. 
  */
  SimulateIOError( return SQLITE_IOERR );

  if( sqlite3_temp_directory ) {
    sqlite3_snprintf(CCHMAXPATH-30, zTempPath, "%s", sqlite3_temp_directory);

  } else if( DosScanEnv( (PSZ)"TEMP",   &zTempPathCp ) == NO_ERROR ||
             DosScanEnv( (PSZ)"TMP",    &zTempPathCp ) == NO_ERROR ||
             DosScanEnv( (PSZ)"TMPDIR", &zTempPathCp ) == NO_ERROR ) {

    char *zTempPathUTF = convertCpPathToUtf8( (char *)zTempPathCp );
    sqlite3_snprintf(CCHMAXPATH-30, zTempPath, "%s", zTempPathUTF);
    free( zTempPathUTF );
  } else if( DosQueryCurrentDisk( &ulDriveNum, &ulDriveMap ) == NO_ERROR ) {
    zTempPath[0] = (char)('A' + ulDriveNum - 1);
    zTempPath[1] = ':'; 
    zTempPath[2] = '\0'; 
  } else {
    zTempPath[0] = '\0'; 
  }
  

  /* Strip off a trailing slashes or backslashes, otherwise we would get *
   * multiple (back)slashes which causes DosOpen() to fail.              *
   * Trailing spaces are not allowed, either.                            */
  j = sqlite3Strlen30(zTempPath);
  while( j > 0 && ( zTempPath[j-1] == '\\' || zTempPath[j-1] == '/' || 
                    zTempPath[j-1] == ' ' ) ){
    j--;
  }
  zTempPath[j] = '\0';
  
  /* We use 20 bytes to randomize the name */
  sqlite3_snprintf(nBuf-22, zBuf,
                   "%s\\"SQLITE_TEMP_FILE_PREFIX, zTempPath);





  j = sqlite3Strlen30(zBuf);
  sqlite3_randomness( 20, &zBuf[j] );
  for( i = 0; i < 20; i++, j++ ){
    zBuf[j] = zChars[ ((unsigned char)zBuf[j])%(sizeof(zChars)-1) ];
  }
  zBuf[j] = 0;

  OSTRACE(( "TEMP FILENAME: %s\n", zBuf ));
  return SQLITE_OK;
}


/*
** Turn a relative pathname into a full pathname.  Write the full
** pathname into zFull[].  zFull[] will be at least pVfs->mxPathname
** bytes in size.
*/
static int os2FullPathname(
  sqlite3_vfs *pVfs,          /* Pointer to vfs object */
  const char *zRelative,      /* Possibly relative input path */
  int nFull,                  /* Size of output buffer in bytes */
  char *zFull                 /* Output buffer */
){
  char *zRelativeCp = convertUtf8PathToCp( zRelative );
  char zFullCp[CCHMAXPATH] = "\0";
  char *zFullUTF;
  APIRET rc = DosQueryPathInfo( (PSZ)zRelativeCp, FIL_QUERYFULLNAME, 
                                zFullCp, CCHMAXPATH );
  free( zRelativeCp );
  zFullUTF = convertCpPathToUtf8( zFullCp );
  sqlite3_snprintf( nFull, zFull, zFullUTF );
  free( zFullUTF );
  return rc == NO_ERROR ? SQLITE_OK : SQLITE_IOERR;
}


/*
** Open a file.
*/
static int os2Open(
  sqlite3_vfs *pVfs,            /* Not used */
  const char *zName,            /* Name of the file (UTF-8) */
  sqlite3_file *id,             /* Write the SQLite file handle here */
  int flags,                    /* Open mode flags */
  int *pOutFlags                /* Status return flags */
){
  HFILE h;

  ULONG ulOpenFlags = 0;
  ULONG ulOpenMode = 0;
  ULONG ulAction = 0;
  ULONG rc;
  os2File *pFile = (os2File*)id;
  const char *zUtf8Name = zName;

  char *zNameCp;
  char  zTmpname[CCHMAXPATH];

  int isExclusive  = (flags & SQLITE_OPEN_EXCLUSIVE);
  int isCreate     = (flags & SQLITE_OPEN_CREATE);
  int isReadWrite  = (flags & SQLITE_OPEN_READWRITE);
#ifndef NDEBUG
  int isDelete     = (flags & SQLITE_OPEN_DELETEONCLOSE);
  int isReadonly   = (flags & SQLITE_OPEN_READONLY);
  int eType        = (flags & 0xFFFFFF00);
  int isOpenJournal = (isCreate && (
        eType==SQLITE_OPEN_MASTER_JOURNAL 
     || eType==SQLITE_OPEN_MAIN_JOURNAL 
     || eType==SQLITE_OPEN_WAL
  ));
#endif

  UNUSED_PARAMETER(pVfs);
  assert( id!=0 );

  /* Check the following statements are true: 
  **
  **   (a) Exactly one of the READWRITE and READONLY flags must be set, and 
  **   (b) if CREATE is set, then READWRITE must also be set, and
  **   (c) if EXCLUSIVE is set, then CREATE must also be set.
  **   (d) if DELETEONCLOSE is set, then CREATE must also be set.
  */
  assert((isReadonly==0 || isReadWrite==0) && (isReadWrite || isReadonly));
  assert(isCreate==0 || isReadWrite);
  assert(isExclusive==0 || isCreate);
  assert(isDelete==0 || isCreate);

  /* The main DB, main journal, WAL file and master journal are never 
  ** automatically deleted. Nor are they ever temporary files.  */
  assert( (!isDelete && zName) || eType!=SQLITE_OPEN_MAIN_DB );
  assert( (!isDelete && zName) || eType!=SQLITE_OPEN_MAIN_JOURNAL );
  assert( (!isDelete && zName) || eType!=SQLITE_OPEN_MASTER_JOURNAL );
  assert( (!isDelete && zName) || eType!=SQLITE_OPEN_WAL );

  /* Assert that the upper layer has set one of the "file-type" flags. */
  assert( eType==SQLITE_OPEN_MAIN_DB      || eType==SQLITE_OPEN_TEMP_DB 
       || eType==SQLITE_OPEN_MAIN_JOURNAL || eType==SQLITE_OPEN_TEMP_JOURNAL 
       || eType==SQLITE_OPEN_SUBJOURNAL   || eType==SQLITE_OPEN_MASTER_JOURNAL 
       || eType==SQLITE_OPEN_TRANSIENT_DB || eType==SQLITE_OPEN_WAL
  );

  memset( pFile, 0, sizeof(*pFile) );
  pFile->h = (HFILE)-1;

  /* If the second argument to this function is NULL, generate a 
  ** temporary file name to use 
  */
  if( !zUtf8Name ){
    assert(isDelete && !isOpenJournal);
    rc = getTempname(CCHMAXPATH, zTmpname);
    if( rc!=SQLITE_OK ){
      return rc;
    }
    zUtf8Name = zTmpname;
  }

  if( isReadWrite ){





    ulOpenMode |= OPEN_ACCESS_READWRITE;

  }else{
    ulOpenMode |= OPEN_ACCESS_READONLY;

  }

  /* Open in random access mode for possibly better speed.  Allow full
  ** sharing because file locks will provide exclusive access when needed.
  ** The handle should not be inherited by child processes and we don't 
  ** want popups from the critical error handler.




  */

  ulOpenMode |= OPEN_FLAGS_RANDOM | OPEN_SHARE_DENYNONE | 


                OPEN_FLAGS_NOINHERIT | OPEN_FLAGS_FAIL_ON_ERROR;


  /* SQLITE_OPEN_EXCLUSIVE is used to make sure that a new file is 


  ** created. SQLite doesn't use it to indicate "exclusive access" 
  ** as it is usually understood.

  */
  if( isExclusive ){


    /* Creates a new file, only if it does not already exist. */
    /* If the file exists, it fails. */
    ulOpenFlags |= OPEN_ACTION_CREATE_IF_NEW | OPEN_ACTION_FAIL_IF_EXISTS;
  }else if( isCreate ){
    /* Open existing file, or create if it doesn't exist */
    ulOpenFlags |= OPEN_ACTION_CREATE_IF_NEW | OPEN_ACTION_OPEN_IF_EXISTS;
  }else{

    /* Opens a file, only if it exists. */
    ulOpenFlags |= OPEN_ACTION_FAIL_IF_NEW | OPEN_ACTION_OPEN_IF_EXISTS;
  }






  zNameCp = convertUtf8PathToCp( zUtf8Name );
  rc = DosOpen( (PSZ)zNameCp,
                &h,
                &ulAction,
                0L,
                FILE_NORMAL,
                ulOpenFlags,
                ulOpenMode,
                (PEAOP2)NULL );
  free( zNameCp );

  if( rc != NO_ERROR ){
    OSTRACE(( "OPEN Invalid handle rc=%d: zName=%s, ulAction=%#lx, ulFlags=%#lx, ulMode=%#lx\n",
              rc, zUtf8Name, ulAction, ulOpenFlags, ulOpenMode ));



    if( isReadWrite ){


      return os2Open( pVfs, zName, id,
                      ((flags|SQLITE_OPEN_READONLY)&~(SQLITE_OPEN_CREATE|SQLITE_OPEN_READWRITE)),
                      pOutFlags );
    }else{
      return SQLITE_CANTOPEN;
    }
  }

  if( pOutFlags ){
    *pOutFlags = isReadWrite ? SQLITE_OPEN_READWRITE : SQLITE_OPEN_READONLY;
  }

  os2FullPathname( pVfs, zUtf8Name, sizeof( zTmpname ), zTmpname );
  pFile->zFullPathCp = convertUtf8PathToCp( zTmpname );
  pFile->pMethod = &os2IoMethod;
  pFile->flags = flags;
  pFile->h = h;

  OpenCounter(+1);
  OSTRACE(( "OPEN %d pOutFlags=%d\n", pFile->h, pOutFlags ));
  return SQLITE_OK;
}

/*
** Delete the named file.
*/
static int os2Delete(
  sqlite3_vfs *pVfs,                     /* Not used on os2 */
  const char *zFilename,                 /* Name of file to delete */
  int syncDir                            /* Not used on os2 */
){
  APIRET rc;
  char *zFilenameCp;
  SimulateIOError( return SQLITE_IOERR_DELETE );
  zFilenameCp = convertUtf8PathToCp( zFilename );
  rc = DosDelete( (PSZ)zFilenameCp );
  free( zFilenameCp );
  OSTRACE(( "DELETE \"%s\"\n", zFilename ));
  return (rc == NO_ERROR ||
          rc == ERROR_FILE_NOT_FOUND ||
          rc == ERROR_PATH_NOT_FOUND ) ? SQLITE_OK : SQLITE_IOERR_DELETE;
}

/*
** Check the existance and status of a file.
*/
static int os2Access(
  sqlite3_vfs *pVfs,        /* Not used on os2 */
  const char *zFilename,    /* Name of file to check */
  int flags,                /* Type of test to make on this file */
  int *pOut                 /* Write results here */
){
  APIRET rc;
  FILESTATUS3 fsts3ConfigInfo;

  char *zFilenameCp;

  UNUSED_PARAMETER(pVfs);
  SimulateIOError( return SQLITE_IOERR_ACCESS; );
  
  zFilenameCp = convertUtf8PathToCp( zFilename );
  rc = DosQueryPathInfo( (PSZ)zFilenameCp, FIL_STANDARD,
                         &fsts3ConfigInfo, sizeof(FILESTATUS3) );
  free( zFilenameCp );
  OSTRACE(( "ACCESS fsts3ConfigInfo.attrFile=%d flags=%d rc=%d\n",
            fsts3ConfigInfo.attrFile, flags, rc ));

  switch( flags ){
    case SQLITE_ACCESS_EXISTS:
      /* For an SQLITE_ACCESS_EXISTS query, treat a zero-length file
      ** as if it does not exist.
      */
      if( fsts3ConfigInfo.cbFile == 0 ) 
        rc = ERROR_FILE_NOT_FOUND;
      break;
    case SQLITE_ACCESS_READ:


      break;
    case SQLITE_ACCESS_READWRITE:
      if( fsts3ConfigInfo.attrFile & FILE_READONLY )
        rc = ERROR_ACCESS_DENIED;
      break;
    default:
      rc = ERROR_FILE_NOT_FOUND;
      assert( !"Invalid flags argument" );
  }

  *pOut = (rc == NO_ERROR);
  OSTRACE(( "ACCESS %s flags %d: rc=%d\n", zFilename, flags, *pOut ));

  return SQLITE_OK;
}


#ifndef SQLITE_OMIT_LOAD_EXTENSION
/*
** Interfaces for opening a shared library, finding entry points
** within the shared library, and closing the shared library.
*/
/*
** Interfaces for opening a shared library, finding entry points
** within the shared library, and closing the shared library.
*/
static void *os2DlOpen(sqlite3_vfs *pVfs, const char *zFilename){

  HMODULE hmod;
  APIRET rc;
  char *zFilenameCp = convertUtf8PathToCp(zFilename);
  rc = DosLoadModule(NULL, 0, (PSZ)zFilenameCp, &hmod);
  free(zFilenameCp);
  return rc != NO_ERROR ? 0 : (void*)hmod;
}
/*
** A no-op since the error code is returned on the DosLoadModule call.
** os2Dlopen returns zero if DosLoadModule is not successful.
*/
static void os2DlError(sqlite3_vfs *pVfs, int nBuf, char *zBufOut){
/* no-op */
}
static void (*os2DlSym(sqlite3_vfs *pVfs, void *pHandle, const char *zSymbol))(void){
  PFN pfn;
  APIRET rc;
  rc = DosQueryProcAddr((HMODULE)pHandle, 0L, (PSZ)zSymbol, &pfn);
  if( rc != NO_ERROR ){
    /* if the symbol itself was not found, search again for the same
     * symbol with an extra underscore, that might be needed depending
     * on the calling convention */
    char _zSymbol[256] = "_";
    strncat(_zSymbol, zSymbol, 254);
    rc = DosQueryProcAddr((HMODULE)pHandle, 0L, (PSZ)_zSymbol, &pfn);
  }
  return rc != NO_ERROR ? 0 : (void(*)(void))pfn;
}
static void os2DlClose(sqlite3_vfs *pVfs, void *pHandle){
  DosFreeModule((HMODULE)pHandle);
}
#else /* if SQLITE_OMIT_LOAD_EXTENSION is defined: */
  #define os2DlOpen 0
  #define os2DlError 0

*/
static int os2Randomness(sqlite3_vfs *pVfs, int nBuf, char *zBuf ){
  int n = 0;
#if defined(SQLITE_TEST)
  n = nBuf;
  memset(zBuf, 0, nBuf);
#else
  int i;                           
  PPIB ppib;
  PTIB ptib;
  DATETIME dt; 
  static unsigned c = 0;

  /* Ordered by variation probability */
  static ULONG svIdx[6] = { QSV_MS_COUNT, QSV_TIME_LOW,
                            QSV_MAXPRMEM, QSV_MAXSHMEM,
                            QSV_TOTAVAILMEM, QSV_TOTRESMEM };

  /* 8 bytes; timezone and weekday don't increase the randomness much */
  if( (int)sizeof(dt)-3 <= nBuf - n ){



    c += 0x0100;

    DosGetDateTime(&dt);



    dt.year = (USHORT)((dt.year - 1900) | c);
    memcpy(&zBuf[n], &dt, sizeof(dt)-3);
    n += sizeof(dt)-3;
  }



  /* 4 bytes; PIDs and TIDs are 16 bit internally, so combine them */
  if( (int)sizeof(ULONG) <= nBuf - n ){


    DosGetInfoBlocks(&ptib, &ppib);
    *(PULONG)&zBuf[n] = MAKELONG(ppib->pib_ulpid,

                                 ptib->tib_ptib2->tib2_ultid);


    n += sizeof(ULONG);
  }




  /* Up to 6 * 4 bytes; variables depend on the system state */
  for( i = 0; i < 6 && (int)sizeof(ULONG) <= nBuf - n; i++ ){
    DosQuerySysInfo(svIdx[i], svIdx[i], 
                    (PULONG)&zBuf[n], sizeof(ULONG));



    n += sizeof(ULONG);
  } 

#endif

  return n;
}

/*
** Sleep for a little while.  Return the amount of time slept.

** returned from sqlite3OsCurrentTime().  This is used for testing.
*/
#ifdef SQLITE_TEST
SQLITE_API int sqlite3_current_time = 0;
#endif

/*
** Find the current time (in Universal Coordinated Time).  Write into *piNow
** the current time and date as a Julian Day number times 86_400_000.  In
** other words, write into *piNow the number of milliseconds since the Julian
** epoch of noon in Greenwich on November 24, 4714 B.C according to the
** proleptic Gregorian calendar.
**
** On success, return 0.  Return 1 if the time and date cannot be found.
*/
static int os2CurrentTimeInt64(sqlite3_vfs *pVfs, sqlite3_int64 *piNow){


#ifdef SQLITE_TEST
  static const sqlite3_int64 unixEpoch = 24405875*(sqlite3_int64)8640000;
#endif
  int year, month, datepart, timepart;
 
  DATETIME dt;
  DosGetDateTime( &dt );


  year = dt.year;

  month = dt.month;


  /* Calculations from http://www.astro.keele.ac.uk/~rno/Astronomy/hjd.html
  ** http://www.astro.keele.ac.uk/~rno/Astronomy/hjd-0.1.c
  ** Calculate the Julian days
  */
  datepart = (int)dt.day - 32076 +
    1461*(year + 4800 + (month - 14)/12)/4 +
    367*(month - 2 - (month - 14)/12*12)/12 -
    3*((year + 4900 + (month - 14)/12)/100)/4;

  /* Time in milliseconds, hours to noon added */
  timepart = 12*3600*1000 + dt.hundredths*10 + dt.seconds*1000 +
    ((int)dt.minutes + dt.timezone)*60*1000 + dt.hours*3600*1000;

  *piNow = (sqlite3_int64)datepart*86400*1000 + timepart;
   
#ifdef SQLITE_TEST
  if( sqlite3_current_time ){
    *piNow = 1000*(sqlite3_int64)sqlite3_current_time + unixEpoch;
  }
#endif

  UNUSED_PARAMETER(pVfs);
  return 0;
}

/*
** Find the current time (in Universal Coordinated Time).  Write the
** current time and date as a Julian Day number into *prNow and
** return 0.  Return 1 if the time and date cannot be found.
*/
static int os2CurrentTime( sqlite3_vfs *pVfs, double *prNow ){
  int rc;
  sqlite3_int64 i;
  rc = os2CurrentTimeInt64(pVfs, &i);
  if( !rc ){
    *prNow = i/86400000.0;
  }
  return rc;
}

/*
** The idea is that this function works like a combination of
** GetLastError() and FormatMessage() on windows (or errno and
** strerror_r() on unix). After an error is returned by an OS
** function, SQLite calls this function with zBuf pointing to
** a buffer of nBuf bytes. The OS layer should populate the
** buffer with a nul-terminated UTF-8 encoded error message
** describing the last IO error to have occurred within the calling
** thread.
**
** If the error message is too large for the supplied buffer,
** it should be truncated. The return value of xGetLastError
** is zero if the error message fits in the buffer, or non-zero
** otherwise (if the message was truncated). If non-zero is returned,
** then it is not necessary to include the nul-terminator character
** in the output buffer.
**
** Not supplying an error message will have no adverse effect
** on SQLite. It is fine to have an implementation that never
** returns an error message:
**
**   int xGetLastError(sqlite3_vfs *pVfs, int nBuf, char *zBuf){
**     assert(zBuf[0]=='\0');
**     return 0;
**   }
**
** However if an error message is supplied, it will be incorporated
** by sqlite into the error message available to the user using
** sqlite3_errmsg(), possibly making IO errors easier to debug.
*/
static int os2GetLastError(sqlite3_vfs *pVfs, int nBuf, char *zBuf){
  assert(zBuf[0]=='\0');
  return 0;
}

/*
** Initialize and deinitialize the operating system interface.
*/
SQLITE_API int sqlite3_os_init(void){
  static sqlite3_vfs os2Vfs = {
    3,                 /* iVersion */
    sizeof(os2File),   /* szOsFile */
    CCHMAXPATH,        /* mxPathname */
    0,                 /* pNext */
    "os2",             /* zName */
    0,                 /* pAppData */

    os2Open,           /* xOpen */
    os2Delete,         /* xDelete */
    os2Access,         /* xAccess */
    os2FullPathname,   /* xFullPathname */
    os2DlOpen,         /* xDlOpen */
    os2DlError,        /* xDlError */
    os2DlSym,          /* xDlSym */
    os2DlClose,        /* xDlClose */
    os2Randomness,     /* xRandomness */
    os2Sleep,          /* xSleep */
    os2CurrentTime,    /* xCurrentTime */
    os2GetLastError,   /* xGetLastError */
    os2CurrentTimeInt64, /* xCurrentTimeInt64 */
    0,                 /* xSetSystemCall */
    0,                 /* xGetSystemCall */
    0                  /* xNextSystemCall */
  };
  sqlite3_vfs_register(&os2Vfs, 1);
  initUconvObjects();
/*  sqlite3OSTrace = 1; */
  return SQLITE_OK;
}
SQLITE_API int sqlite3_os_end(void){
  freeUconvObjects();
  return SQLITE_OK;
}

typedef struct unixFile unixFile;
struct unixFile {
  sqlite3_io_methods const *pMethod;  /* Always the first entry */
  unixInodeInfo *pInode;              /* Info about locks on this inode */
  int h;                              /* The file descriptor */
  int dirfd;                          /* File descriptor for the directory */
  unsigned char eFileLock;            /* The type of lock held on this fd */
  unsigned char ctrlFlags;            /* Behavioral bits.  UNIXFILE_* flags */
  int lastErrno;                      /* The unix errno from last I/O error */
  void *lockingContext;               /* Locking style specific state */
  UnixUnusedFd *pUnused;              /* Pre-allocated UnixUnusedFd */

  const char *zPath;                  /* Name of the file */
  unixShm *pShm;                      /* Shared memory segment information */
  int szChunk;                        /* Configured by FCNTL_CHUNK_SIZE */
#if SQLITE_ENABLE_LOCKING_STYLE
  int openFlags;                      /* The flags specified at open() */
#endif
#if SQLITE_ENABLE_LOCKING_STYLE || defined(__APPLE__)

  ** it is larger than the struct CrashFile defined in test6.c.
  */
  char aPadding[32];
#endif
};

/*
** Allowed values for the unixFile.ctrlFlags bitmask:
*/
#define UNIXFILE_EXCL   0x01     /* Connections from one process only */
#define UNIXFILE_RDONLY 0x02     /* Connection is read only */

/*
** Include code that is common to all os_*.c files
*/
/************** Include os_common.h in the middle of os_unix.c ***************/
/************** Begin file os_common.h ***************************************/
/*

#ifndef O_NOFOLLOW
# define O_NOFOLLOW 0
#endif
#ifndef O_BINARY
# define O_BINARY 0
#endif











/*
** The threadid macro resolves to the thread-id or to 0.  Used for
** testing and debugging only.
*/
#if SQLITE_THREADSAFE
#define threadid pthread_self()
#else
#define threadid 0
#endif

/*
** Many system calls are accessed through pointer-to-functions so that
** they may be overridden at runtime to facilitate fault injection during
** testing and sandboxing.  The following array holds the names and pointers
** to all overrideable system calls.
*/
static struct unix_syscall {
  const char *zName;            /* Name of the sytem call */
  sqlite3_syscall_ptr pCurrent; /* Current value of the system call */
  sqlite3_syscall_ptr pDefault; /* Default value */
} aSyscall[] = {
  { "open",         (sqlite3_syscall_ptr)open,       0  },
#define osOpen      ((int(*)(const char*,int,int))aSyscall[0].pCurrent)

  { "close",        (sqlite3_syscall_ptr)close,      0  },
#define osClose     ((int(*)(int))aSyscall[1].pCurrent)

  { "access",       (sqlite3_syscall_ptr)access,     0  },
#define osAccess    ((int(*)(const char*,int))aSyscall[2].pCurrent)

  { "getcwd",       (sqlite3_syscall_ptr)getcwd,     0  },
#define osGetcwd    ((char*(*)(char*,size_t))aSyscall[3].pCurrent)

  { "stat",         (sqlite3_syscall_ptr)stat,       0  },
#define osStat      ((int(*)(const char*,struct stat*))aSyscall[4].pCurrent)

/*
** The DJGPP compiler environment looks mostly like Unix, but it
** lacks the fcntl() system call.  So redefine fcntl() to be something
** that always succeeds.  This means that locking does not occur under
** DJGPP.  But it is DOS - what did you expect?
*/
#ifdef __DJGPP__
  { "fstat",        0,                 0  },
#define osFstat(a,b,c)    0
#else     
  { "fstat",        (sqlite3_syscall_ptr)fstat,      0  },
#define osFstat     ((int(*)(int,struct stat*))aSyscall[5].pCurrent)
#endif

  { "ftruncate",    (sqlite3_syscall_ptr)ftruncate,  0  },
#define osFtruncate ((int(*)(int,off_t))aSyscall[6].pCurrent)

  { "fcntl",        (sqlite3_syscall_ptr)fcntl,      0  },
#define osFcntl     ((int(*)(int,int,...))aSyscall[7].pCurrent)

  { "read",         (sqlite3_syscall_ptr)read,       0  },
#define osRead      ((ssize_t(*)(int,void*,size_t))aSyscall[8].pCurrent)

#if defined(USE_PREAD) || defined(SQLITE_ENABLE_LOCKING_STYLE)
  { "pread",        (sqlite3_syscall_ptr)pread,      0  },
#else
  { "pread",        (sqlite3_syscall_ptr)0,          0  },
#endif
#define osPread     ((ssize_t(*)(int,void*,size_t,off_t))aSyscall[9].pCurrent)

#if defined(USE_PREAD64)
  { "pread64",      (sqlite3_syscall_ptr)pread64,    0  },
#else
  { "pread64",      (sqlite3_syscall_ptr)0,          0  },
#endif
#define osPread64   ((ssize_t(*)(int,void*,size_t,off_t))aSyscall[10].pCurrent)

  { "write",        (sqlite3_syscall_ptr)write,      0  },
#define osWrite     ((ssize_t(*)(int,const void*,size_t))aSyscall[11].pCurrent)

#if defined(USE_PREAD) || defined(SQLITE_ENABLE_LOCKING_STYLE)
  { "pwrite",       (sqlite3_syscall_ptr)pwrite,     0  },
#else
  { "pwrite",       (sqlite3_syscall_ptr)0,          0  },
#endif
#define osPwrite    ((ssize_t(*)(int,const void*,size_t,off_t))\
                    aSyscall[12].pCurrent)

#if defined(USE_PREAD64)
  { "pwrite64",     (sqlite3_syscall_ptr)pwrite64,   0  },
#else
  { "pwrite64",     (sqlite3_syscall_ptr)0,          0  },
#endif
#define osPwrite64  ((ssize_t(*)(int,const void*,size_t,off_t))\
                    aSyscall[13].pCurrent)

#if SQLITE_ENABLE_LOCKING_STYLE
  { "fchmod",       (sqlite3_syscall_ptr)fchmod,     0  },
#define osFchmod    ((int(*)(int,mode_t))aSyscall[14].pCurrent)
#endif

#if defined(HAVE_POSIX_FALLOCATE) && HAVE_POSIX_FALLOCATE
  { "fallocate",    (sqlite3_syscall_ptr)posix_fallocate,  0 },
#else
  { "fallocate",    (sqlite3_syscall_ptr)0,                0 },
#endif
#define osFallocate ((int(*)(int,off_t,off_t))aSyscall[15].pCurrent)

}; /* End of the overrideable system calls */

/*
** This is the xSetSystemCall() method of sqlite3_vfs for all of the
** "unix" VFSes.  Return SQLITE_OK opon successfully updating the
** system call pointer, or SQLITE_NOTFOUND if there is no configurable
** system call named zName.
*/
static int unixSetSystemCall(
  sqlite3_vfs *pNotUsed,        /* The VFS pointer.  Not used */
  const char *zName,            /* Name of system call to override */
  sqlite3_syscall_ptr pNewFunc  /* Pointer to new system call value */
){
  unsigned int i;
  int rc = SQLITE_NOTFOUND;

  UNUSED_PARAMETER(pNotUsed);
  if( zName==0 ){
    /* If no zName is given, restore all system calls to their default
    ** settings and return NULL
    */
    rc = SQLITE_OK;
    for(i=0; i<sizeof(aSyscall)/sizeof(aSyscall[0]); i++){
      if( aSyscall[i].pDefault ){
        aSyscall[i].pCurrent = aSyscall[i].pDefault;
      }
    }
  }else{
    /* If zName is specified, operate on only the one system call
    ** specified.
    */
    for(i=0; i<sizeof(aSyscall)/sizeof(aSyscall[0]); i++){
      if( strcmp(zName, aSyscall[i].zName)==0 ){
        if( aSyscall[i].pDefault==0 ){
          aSyscall[i].pDefault = aSyscall[i].pCurrent;
        }
        rc = SQLITE_OK;
        if( pNewFunc==0 ) pNewFunc = aSyscall[i].pDefault;
        aSyscall[i].pCurrent = pNewFunc;
        break;
      }
    }
  }
  return rc;
}

/*
** Return the value of a system call.  Return NULL if zName is not a
** recognized system call name.  NULL is also returned if the system call
** is currently undefined.
*/
static sqlite3_syscall_ptr unixGetSystemCall(
  sqlite3_vfs *pNotUsed,
  const char *zName
){
  unsigned int i;

  UNUSED_PARAMETER(pNotUsed);
  for(i=0; i<sizeof(aSyscall)/sizeof(aSyscall[0]); i++){
    if( strcmp(zName, aSyscall[i].zName)==0 ) return aSyscall[i].pCurrent;
  }
  return 0;
}

/*
** Return the name of the first system call after zName.  If zName==NULL
** then return the name of the first system call.  Return NULL if zName
** is the last system call or if zName is not the name of a valid
** system call.
*/
static const char *unixNextSystemCall(sqlite3_vfs *p, const char *zName){
  int i = -1;

  UNUSED_PARAMETER(p);
  if( zName ){
    for(i=0; i<ArraySize(aSyscall)-1; i++){
      if( strcmp(zName, aSyscall[i].zName)==0 ) break;
    }
  }
  for(i++; i<ArraySize(aSyscall); i++){
    if( aSyscall[i].pCurrent!=0 ) return aSyscall[i].zName;
  }
  return 0;
}

/*
** Retry open() calls that fail due to EINTR
*/
static int robust_open(const char *z, int f, int m){
  int rc;
  do{ rc = osOpen(z,f,m); }while( rc<0 && errno==EINTR );
  return rc;
}

/*
** Helper functions to obtain and relinquish the global mutex. The
** global mutex is used to protect the unixInodeInfo and
** vxworksFileId objects used by this file, all of which may be 
** shared by multiple threads.
**

  int s;
  int savedErrno;
  if( op==F_GETLK ){
    zOpName = "GETLK";
  }else if( op==F_SETLK ){
    zOpName = "SETLK";
  }else{
    s = osFcntl(fd, op, p);
    sqlite3DebugPrintf("fcntl unknown %d %d %d\n", fd, op, s);
    return s;
  }
  if( p->l_type==F_RDLCK ){
    zType = "RDLCK";
  }else if( p->l_type==F_WRLCK ){
    zType = "WRLCK";
  }else if( p->l_type==F_UNLCK ){
    zType = "UNLCK";
  }else{
    assert( 0 );
  }
  assert( p->l_whence==SEEK_SET );
  s = osFcntl(fd, op, p);
  savedErrno = errno;
  sqlite3DebugPrintf("fcntl %d %d %s %s %d %d %d %d\n",
     threadid, fd, zOpName, zType, (int)p->l_start, (int)p->l_len,
     (int)p->l_pid, s);
  if( s==(-1) && op==F_SETLK && (p->l_type==F_RDLCK || p->l_type==F_WRLCK) ){
    struct flock l2;
    l2 = *p;
    osFcntl(fd, F_GETLK, &l2);
    if( l2.l_type==F_RDLCK ){
      zType = "RDLCK";
    }else if( l2.l_type==F_WRLCK ){
      zType = "WRLCK";
    }else if( l2.l_type==F_UNLCK ){
      zType = "UNLCK";
    }else{
      assert( 0 );
    }
    sqlite3DebugPrintf("fcntl-failure-reason: %s %d %d %d\n",
       zType, (int)l2.l_start, (int)l2.l_len, (int)l2.l_pid);
  }
  errno = savedErrno;
  return s;
}
#undef osFcntl
#define osFcntl lockTrace
#endif /* SQLITE_LOCK_TRACE */

/*
** Retry ftruncate() calls that fail due to EINTR
*/
static int robust_ftruncate(int h, sqlite3_int64 sz){
  int rc;
  do{ rc = osFtruncate(h,sz); }while( rc<0 && errno==EINTR );
  return rc;
}

/*
** This routine translates a standard POSIX errno code into something
** useful to the clients of the sqlite3 functions.  Specifically, it is
** intended to translate a variety of "try again" errors into SQLITE_BUSY
** and a variety of "please close the file descriptor NOW" errors into 
** SQLITE_IOERR
** 
** Errors during initialization of locks, or file system support for locks,
** should handle ENOLCK, ENOTSUP, EOPNOTSUPP separately.
*/
static int sqliteErrorFromPosixError(int posixError, int sqliteIOErr) {
  switch (posixError) {
#if 0
  /* At one point this code was not commented out. In theory, this branch
  ** should never be hit, as this function should only be called after
  ** a locking-related function (i.e. fcntl()) has returned non-zero with
  ** the value of errno as the first argument. Since a system call has failed,
  ** errno should be non-zero.
  **
  ** Despite this, if errno really is zero, we still don't want to return
  ** SQLITE_OK. The system call failed, and *some* SQLite error should be
  ** propagated back to the caller. Commenting this branch out means errno==0
  ** will be handled by the "default:" case below.
  */
  case 0: 
    return SQLITE_OK;
#endif

  case EAGAIN:
  case ETIMEDOUT:
  case EBUSY:
  case EINTR:
  case ENOLCK:  
    /* random NFS retry error, unless during file system support 
     * introspection, in which it actually means what it says */

	(sqliteIOErr == SQLITE_IOERR_CHECKRESERVEDLOCK) ){
      return SQLITE_BUSY;
    }
    /* else fall through */
  case EPERM: 
    return SQLITE_PERM;
    
  /* EDEADLK is only possible if a call to fcntl(F_SETLKW) is made. And
  ** this module never makes such a call. And the code in SQLite itself 
  ** asserts that SQLITE_IOERR_BLOCKED is never returned. For these reasons
  ** this case is also commented out. If the system does set errno to EDEADLK,
  ** the default SQLITE_IOERR_XXX code will be returned. */
#if 0
  case EDEADLK:
    return SQLITE_IOERR_BLOCKED;
#endif
    
#if EOPNOTSUPP!=ENOTSUP
  case EOPNOTSUPP: 
    /* something went terribly awry, unless during file system support 
     * introspection, in which it actually means what it says */
#endif
#ifdef ENOTSUP

** A single inode can have multiple file descriptors, so each unixFile
** structure contains a pointer to an instance of this object and this
** object keeps a count of the number of unixFile pointing to it.
*/
struct unixInodeInfo {
  struct unixFileId fileId;       /* The lookup key */
  int nShared;                    /* Number of SHARED locks held */
  unsigned char eFileLock;        /* One of SHARED_LOCK, RESERVED_LOCK etc. */
  unsigned char bProcessLock;     /* An exclusive process lock is held */
  int nRef;                       /* Number of pointers to this structure */
  unixShmNode *pShmNode;          /* Shared memory associated with this inode */
  int nLock;                      /* Number of outstanding file locks */
  UnixUnusedFd *pUnused;          /* Unused file descriptors to close */
  unixInodeInfo *pNext;           /* List of all unixInodeInfo objects */
  unixInodeInfo *pPrev;           /*    .... doubly linked */
#if defined(SQLITE_ENABLE_LOCKING_STYLE)
  unsigned long long sharedByte;  /* for AFP simulated shared lock */
#endif
#if OS_VXWORKS
  sem_t *pSem;                    /* Named POSIX semaphore */
  char aSemName[MAX_PATHNAME+2];  /* Name of that semaphore */
#endif
};

/*
** A lists of all unixInodeInfo objects.
*/
static unixInodeInfo *inodeList = 0;

/*
**
** This function - unixLogError_x(), is only ever called via the macro
** unixLogError().
**
** It is invoked after an error occurs in an OS function and errno has been
** set. It logs a message using sqlite3_log() containing the current value of
** errno and, if possible, the human-readable equivalent from strerror() or
** strerror_r().
**
** The first argument passed to the macro should be the error code that
** will be returned to SQLite (e.g. SQLITE_IOERR_DELETE, SQLITE_CANTOPEN). 
** The two subsequent arguments should be the name of the OS function that
** failed (e.g. "unlink", "open") and the the associated file-system path,
** if any.
*/
#define unixLogError(a,b,c)     unixLogErrorAtLine(a,b,c,__LINE__)
static int unixLogErrorAtLine(
  int errcode,                    /* SQLite error code */
  const char *zFunc,              /* Name of OS function that failed */
  const char *zPath,              /* File path associated with error */
  int iLine                       /* Source line number where error occurred */
){
  char *zErr;                     /* Message from strerror() or equivalent */
  int iErrno = errno;             /* Saved syscall error number */

  /* If this is not a threadsafe build (SQLITE_THREADSAFE==0), then use
  ** the strerror() function to obtain the human-readable error message
  ** equivalent to errno. Otherwise, use strerror_r().
  */ 
#if SQLITE_THREADSAFE && defined(HAVE_STRERROR_R)
  char aErr[80];
  memset(aErr, 0, sizeof(aErr));
  zErr = aErr;

  /* If STRERROR_R_CHAR_P (set by autoconf scripts) or __USE_GNU is defined,
  ** assume that the system provides the the GNU version of strerror_r() that 
  ** returns a pointer to a buffer containing the error message. That pointer 
  ** may point to aErr[], or it may point to some static storage somewhere. 
  ** Otherwise, assume that the system provides the POSIX version of 
  ** strerror_r(), which always writes an error message into aErr[].
  **
  ** If the code incorrectly assumes that it is the POSIX version that is
  ** available, the error message will often be an empty string. Not a
  ** huge problem. Incorrectly concluding that the GNU version is available 
  ** could lead to a segfault though.
  */
#if defined(STRERROR_R_CHAR_P) || defined(__USE_GNU)
  zErr = 
# endif
  strerror_r(iErrno, aErr, sizeof(aErr)-1);

#elif SQLITE_THREADSAFE
  /* This is a threadsafe build, but strerror_r() is not available. */
  zErr = "";
#else
  /* Non-threadsafe build, use strerror(). */
  zErr = strerror(iErrno);
#endif

  assert( errcode!=SQLITE_OK );
  if( zPath==0 ) zPath = "";
  sqlite3_log(errcode,
      "os_unix.c:%d: (%d) %s(%s) - %s",
      iLine, iErrno, zFunc, zPath, zErr
  );

  return errcode;
}

/*
** Close a file descriptor.
**
** We assume that close() almost always works, since it is only in a
** very sick application or on a very sick platform that it might fail.
** If it does fail, simply leak the file descriptor, but do log the
** error.
**
** Note that it is not safe to retry close() after EINTR since the
** file descriptor might have already been reused by another thread.
** So we don't even try to recover from an EINTR.  Just log the error
** and move on.
*/
static void robust_close(unixFile *pFile, int h, int lineno){
  if( osClose(h) ){
    unixLogErrorAtLine(SQLITE_IOERR_CLOSE, "close",
                       pFile ? pFile->zPath : 0, lineno);
  }
}

/*
** Close all file descriptors accumuated in the unixInodeInfo->pUnused list.
*/ 
static void closePendingFds(unixFile *pFile){

  unixInodeInfo *pInode = pFile->pInode;

  UnixUnusedFd *p;
  UnixUnusedFd *pNext;
  for(p=pInode->pUnused; p; p=pNext){
    pNext = p->pNext;
    robust_close(pFile, p->fd, __LINE__);





    sqlite3_free(p);
  }

  pInode->pUnused = 0;

}

/*
** Release a unixInodeInfo structure previously allocated by findInodeInfo().
**
** The mutex entered using the unixEnterMutex() function must be held
** when this function is called.
*/
static void releaseInodeInfo(unixFile *pFile){
  unixInodeInfo *pInode = pFile->pInode;
  assert( unixMutexHeld() );
  if( ALWAYS(pInode) ){
    pInode->nRef--;
    if( pInode->nRef==0 ){
      assert( pInode->pShmNode==0 );
      closePendingFds(pFile);
      if( pInode->pPrev ){
        assert( pInode->pPrev->pNext==pInode );
        pInode->pPrev->pNext = pInode->pNext;

  assert( unixMutexHeld() );

  /* Get low-level information about the file that we can used to
  ** create a unique name for the file.
  */
  fd = pFile->h;
  rc = osFstat(fd, &statbuf);
  if( rc!=0 ){
    pFile->lastErrno = errno;
#ifdef EOVERFLOW
    if( pFile->lastErrno==EOVERFLOW ) return SQLITE_NOLFS;
#endif
    return SQLITE_IOERR;
  }

#ifdef __APPLE__
  /* On OS X on an msdos filesystem, the inode number is reported
  ** incorrectly for zero-size files.  See ticket #3260.  To work
  ** around this problem (we consider it a bug in OS X, not SQLite)
  ** we always increase the file size to 1 by writing a single byte
  ** prior to accessing the inode number.  The one byte written is
  ** an ASCII 'S' character which also happens to be the first byte
  ** in the header of every SQLite database.  In this way, if there
  ** is a race condition such that another thread has already populated
  ** the first page of the database, no damage is done.
  */
  if( statbuf.st_size==0 && (pFile->fsFlags & SQLITE_FSFLAGS_IS_MSDOS)!=0 ){
    do{ rc = osWrite(fd, "S", 1); }while( rc<0 && errno==EINTR );
    if( rc!=1 ){
      pFile->lastErrno = errno;
      return SQLITE_IOERR;
    }
    rc = osFstat(fd, &statbuf);
    if( rc!=0 ){
      pFile->lastErrno = errno;
      return SQLITE_IOERR;
    }
  }
#endif

  if( pFile->pInode->eFileLock>SHARED_LOCK ){
    reserved = 1;
  }

  /* Otherwise see if some other process holds it.
  */
#ifndef __DJGPP__
  if( !reserved && !pFile->pInode->bProcessLock ){
    struct flock lock;
    lock.l_whence = SEEK_SET;
    lock.l_start = RESERVED_BYTE;
    lock.l_len = 1;
    lock.l_type = F_WRLCK;
    if( osFcntl(pFile->h, F_GETLK, &lock) ){

      rc = SQLITE_IOERR_CHECKRESERVEDLOCK;
      pFile->lastErrno = errno;
    } else if( lock.l_type!=F_UNLCK ){
      reserved = 1;
    }
  }
#endif
  
  unixLeaveMutex();
  OSTRACE(("TEST WR-LOCK %d %d %d (unix)\n", pFile->h, rc, reserved));

  *pResOut = reserved;
  return rc;
}

/*
** Attempt to set a system-lock on the file pFile.  The lock is 
** described by pLock.
**
** If the pFile was opened read/write from unix-excl, then the only lock
** ever obtained is an exclusive lock, and it is obtained exactly once
** the first time any lock is attempted.  All subsequent system locking
** operations become no-ops.  Locking operations still happen internally,
** in order to coordinate access between separate database connections
** within this process, but all of that is handled in memory and the
** operating system does not participate.
**
** This function is a pass-through to fcntl(F_SETLK) if pFile is using
** any VFS other than "unix-excl" or if pFile is opened on "unix-excl"
** and is read-only.
**
** Zero is returned if the call completes successfully, or -1 if a call
** to fcntl() fails. In this case, errno is set appropriately (by fcntl()).
*/
static int unixFileLock(unixFile *pFile, struct flock *pLock){
  int rc;
  unixInodeInfo *pInode = pFile->pInode;
  assert( unixMutexHeld() );
  assert( pInode!=0 );
  if( ((pFile->ctrlFlags & UNIXFILE_EXCL)!=0 || pInode->bProcessLock)
   && ((pFile->ctrlFlags & UNIXFILE_RDONLY)==0)
  ){
    if( pInode->bProcessLock==0 ){
      struct flock lock;
      assert( pInode->nLock==0 );
      lock.l_whence = SEEK_SET;
      lock.l_start = SHARED_FIRST;
      lock.l_len = SHARED_SIZE;
      lock.l_type = F_WRLCK;
      rc = osFcntl(pFile->h, F_SETLK, &lock);
      if( rc<0 ) return rc;
      pInode->bProcessLock = 1;
      pInode->nLock++;
    }else{
      rc = 0;
    }
  }else{
    rc = osFcntl(pFile->h, F_SETLK, pLock);
  }
  return rc;
}

/*
** Lock the file with the lock specified by parameter eFileLock - one
** of the following:
**
**     (1) SHARED_LOCK
**     (2) RESERVED_LOCK

  ** locking a random byte from a range, concurrent SHARED locks may exist
  ** even if the locking primitive used is always a write-lock.
  */
  int rc = SQLITE_OK;
  unixFile *pFile = (unixFile*)id;
  unixInodeInfo *pInode = pFile->pInode;
  struct flock lock;

  int tErrno = 0;

  assert( pFile );
  OSTRACE(("LOCK    %d %s was %s(%s,%d) pid=%d (unix)\n", pFile->h,
      azFileLock(eFileLock), azFileLock(pFile->eFileLock),
      azFileLock(pInode->eFileLock), pInode->nShared , getpid()));

  lock.l_len = 1L;
  lock.l_whence = SEEK_SET;
  if( eFileLock==SHARED_LOCK 
      || (eFileLock==EXCLUSIVE_LOCK && pFile->eFileLock<PENDING_LOCK)
  ){
    lock.l_type = (eFileLock==SHARED_LOCK?F_RDLCK:F_WRLCK);
    lock.l_start = PENDING_BYTE;
    if( unixFileLock(pFile, &lock) ){

      tErrno = errno;
      rc = sqliteErrorFromPosixError(tErrno, SQLITE_IOERR_LOCK);
      if( rc!=SQLITE_BUSY ){
        pFile->lastErrno = tErrno;
      }
      goto end_lock;
    }
  }


  /* If control gets to this point, then actually go ahead and make
  ** operating system calls for the specified lock.
  */
  if( eFileLock==SHARED_LOCK ){
    assert( pInode->nShared==0 );
    assert( pInode->eFileLock==0 );
    assert( rc==SQLITE_OK );

    /* Now get the read-lock */
    lock.l_start = SHARED_FIRST;
    lock.l_len = SHARED_SIZE;
    if( unixFileLock(pFile, &lock) ){
      tErrno = errno;
      rc = sqliteErrorFromPosixError(tErrno, SQLITE_IOERR_LOCK);
    }

    /* Drop the temporary PENDING lock */
    lock.l_start = PENDING_BYTE;
    lock.l_len = 1L;
    lock.l_type = F_UNLCK;
    if( unixFileLock(pFile, &lock) && rc==SQLITE_OK ){

      /* This could happen with a network mount */
      tErrno = errno;
      rc = SQLITE_IOERR_UNLOCK; 


    }



    if( rc ){
      if( rc!=SQLITE_BUSY ){

        pFile->lastErrno = tErrno;
      }
      goto end_lock;
    }else{
      pFile->eFileLock = SHARED_LOCK;
      pInode->nLock++;
      pInode->nShared = 1;
    }
  }else if( eFileLock==EXCLUSIVE_LOCK && pInode->nShared>1 ){
    /* We are trying for an exclusive lock but another thread in this
    ** same process is still holding a shared lock. */
    rc = SQLITE_BUSY;
  }else{
    /* The request was for a RESERVED or EXCLUSIVE lock.  It is
    ** assumed that there is a SHARED or greater lock on the file
    ** already.
    */
    assert( 0!=pFile->eFileLock );
    lock.l_type = F_WRLCK;

    assert( eFileLock==RESERVED_LOCK || eFileLock==EXCLUSIVE_LOCK );
    if( eFileLock==RESERVED_LOCK ){
      lock.l_start = RESERVED_BYTE;
      lock.l_len = 1L;
    }else{

      lock.l_start = SHARED_FIRST;
      lock.l_len = SHARED_SIZE;



    }

    if( unixFileLock(pFile, &lock) ){
      tErrno = errno;
      rc = sqliteErrorFromPosixError(tErrno, SQLITE_IOERR_LOCK);
      if( rc!=SQLITE_BUSY ){
        pFile->lastErrno = tErrno;
      }
    }
  }
  

#ifndef NDEBUG

** 
** If handleNFSUnlock is true, then on downgrading an EXCLUSIVE_LOCK to SHARED
** the byte range is divided into 2 parts and the first part is unlocked then
** set to a read lock, then the other part is simply unlocked.  This works 
** around a bug in BSD NFS lockd (also seen on MacOSX 10.3+) that fails to 
** remove the write lock on a region when a read lock is set.
*/
static int posixUnlock(sqlite3_file *id, int eFileLock, int handleNFSUnlock){
  unixFile *pFile = (unixFile*)id;
  unixInodeInfo *pInode;
  struct flock lock;
  int rc = SQLITE_OK;
  int h;


  assert( pFile );
  OSTRACE(("UNLOCK  %d %d was %d(%d,%d) pid=%d (unix)\n", pFile->h, eFileLock,
      pFile->eFileLock, pFile->pInode->eFileLock, pFile->pInode->nShared,
      getpid()));

  assert( eFileLock<=SHARED_LOCK );

    ** write lock until the rest is covered by a read lock:
    **  1:   [WWWWW]
    **  2:   [....W]
    **  3:   [RRRRW]
    **  4:   [RRRR.]
    */
    if( eFileLock==SHARED_LOCK ){

#if !defined(__APPLE__) || !SQLITE_ENABLE_LOCKING_STYLE
      (void)handleNFSUnlock;
      assert( handleNFSUnlock==0 );
#endif
#if defined(__APPLE__) && SQLITE_ENABLE_LOCKING_STYLE
      if( handleNFSUnlock ){
        int tErrno;               /* Error code from system call errors */
        off_t divSize = SHARED_SIZE - 1;
        
        lock.l_type = F_UNLCK;
        lock.l_whence = SEEK_SET;
        lock.l_start = SHARED_FIRST;
        lock.l_len = divSize;
        if( unixFileLock(pFile, &lock)==(-1) ){
          tErrno = errno;
          rc = SQLITE_IOERR_UNLOCK;
          if( IS_LOCK_ERROR(rc) ){
            pFile->lastErrno = tErrno;
          }
          goto end_unlock;
        }
        lock.l_type = F_RDLCK;
        lock.l_whence = SEEK_SET;
        lock.l_start = SHARED_FIRST;
        lock.l_len = divSize;
        if( unixFileLock(pFile, &lock)==(-1) ){
          tErrno = errno;
          rc = sqliteErrorFromPosixError(tErrno, SQLITE_IOERR_RDLOCK);
          if( IS_LOCK_ERROR(rc) ){
            pFile->lastErrno = tErrno;
          }
          goto end_unlock;
        }
        lock.l_type = F_UNLCK;
        lock.l_whence = SEEK_SET;
        lock.l_start = SHARED_FIRST+divSize;
        lock.l_len = SHARED_SIZE-divSize;
        if( unixFileLock(pFile, &lock)==(-1) ){
          tErrno = errno;
          rc = SQLITE_IOERR_UNLOCK;
          if( IS_LOCK_ERROR(rc) ){
            pFile->lastErrno = tErrno;
          }
          goto end_unlock;
        }
      }else
#endif /* defined(__APPLE__) && SQLITE_ENABLE_LOCKING_STYLE */
      {
        lock.l_type = F_RDLCK;
        lock.l_whence = SEEK_SET;
        lock.l_start = SHARED_FIRST;
        lock.l_len = SHARED_SIZE;
        if( unixFileLock(pFile, &lock) ){
          /* In theory, the call to unixFileLock() cannot fail because another
          ** process is holding an incompatible lock. If it does, this 
          ** indicates that the other process is not following the locking
          ** protocol. If this happens, return SQLITE_IOERR_RDLOCK. Returning
          ** SQLITE_BUSY would confuse the upper layer (in practice it causes 
          ** an assert to fail). */ 
          rc = SQLITE_IOERR_RDLOCK;

          pFile->lastErrno = errno;

          goto end_unlock;
        }
      }
    }
    lock.l_type = F_UNLCK;
    lock.l_whence = SEEK_SET;
    lock.l_start = PENDING_BYTE;
    lock.l_len = 2L;  assert( PENDING_BYTE+1==RESERVED_BYTE );
    if( unixFileLock(pFile, &lock)==0 ){
      pInode->eFileLock = SHARED_LOCK;
    }else{

      rc = SQLITE_IOERR_UNLOCK;

      pFile->lastErrno = errno;

      goto end_unlock;
    }
  }
  if( eFileLock==NO_LOCK ){
    /* Decrement the shared lock counter.  Release the lock using an
    ** OS call only when all threads in this same process have released
    ** the lock.
    */
    pInode->nShared--;
    if( pInode->nShared==0 ){
      lock.l_type = F_UNLCK;
      lock.l_whence = SEEK_SET;
      lock.l_start = lock.l_len = 0L;
      SimulateIOErrorBenign(1);
      SimulateIOError( h=(-1) )
      SimulateIOErrorBenign(0);
      if( unixFileLock(pFile, &lock)==0 ){
        pInode->eFileLock = NO_LOCK;
      }else{

        rc = SQLITE_IOERR_UNLOCK;

	pFile->lastErrno = errno;

        pInode->eFileLock = NO_LOCK;
        pFile->eFileLock = NO_LOCK;
      }
    }

    /* Decrement the count of locks against this same file.  When the
    ** count reaches zero, close any other file descriptors whose close
    ** was deferred because of outstanding locks.
    */
    pInode->nLock--;
    assert( pInode->nLock>=0 );
    if( pInode->nLock==0 ){
      closePendingFds(pFile);



    }
  }
	
end_unlock:
  unixLeaveMutex();
  if( rc==SQLITE_OK ) pFile->eFileLock = eFileLock;
  return rc;
}

/*
** Lower the locking level on file descriptor pFile to eFileLock.  eFileLock
** must be either NO_LOCK or SHARED_LOCK.
**
** If the locking level of the file descriptor is already at or below
** the requested locking level, this routine is a no-op.
*/
static int unixUnlock(sqlite3_file *id, int eFileLock){
  return posixUnlock(id, eFileLock, 0);
}

/*
** This function performs the parts of the "close file" operation 
** common to all locking schemes. It closes the directory and file
** handles, if they are valid, and sets all fields of the unixFile
** structure to 0.
**
** It is *not* necessary to hold the mutex when this routine is called,
** even on VxWorks.  A mutex will be acquired on VxWorks by the
** vxworksReleaseFileId() routine.
*/
static int closeUnixFile(sqlite3_file *id){
  unixFile *pFile = (unixFile*)id;

  if( pFile->dirfd>=0 ){
    robust_close(pFile, pFile->dirfd, __LINE__);




    pFile->dirfd=-1;
  }

  if( pFile->h>=0 ){
    robust_close(pFile, pFile->h, __LINE__);

    pFile->h = -1;

  }

#if OS_VXWORKS
  if( pFile->pId ){
    if( pFile->isDelete ){
      unlink(pFile->pId->zCanonicalName);
    }
    vxworksReleaseFileId(pFile->pId);
    pFile->pId = 0;
  }
#endif
  OSTRACE(("CLOSE   %-3d\n", pFile->h));
  OpenCounter(-1);
  sqlite3_free(pFile->pUnused);
  memset(pFile, 0, sizeof(unixFile));

  return SQLITE_OK;
}

/*
** Close a file.
*/
static int unixClose(sqlite3_file *id){
  int rc = SQLITE_OK;

  unixFile *pFile = (unixFile *)id;
  unixUnlock(id, NO_LOCK);
  unixEnterMutex();

  /* unixFile.pInode is always valid here. Otherwise, a different close
  ** routine (e.g. nolockClose()) would be called instead.
  */
  assert( pFile->pInode->nLock>0 || pFile->pInode->bProcessLock==0 );
  if( ALWAYS(pFile->pInode) && pFile->pInode->nLock ){
    /* If there are outstanding locks, do not actually close the file just
    ** yet because that would clear those locks.  Instead, add the file
    ** descriptor to pInode->pUnused list.  It will be automatically closed 
    ** when the last lock is cleared.
    */
    setPendingFd(pFile);
  }
  releaseInodeInfo(pFile);
  rc = closeUnixFile(id);
  unixLeaveMutex();

  return rc;
}

/************** End of the posix advisory lock implementation *****************
******************************************************************************/

/******************************************************************************

  if( pFile->eFileLock>SHARED_LOCK ){
    /* Either this connection or some other connection in the same process
    ** holds a lock on the file.  No need to check further. */
    reserved = 1;
  }else{
    /* The lock is held if and only if the lockfile exists */
    const char *zLockFile = (const char*)pFile->lockingContext;
    reserved = osAccess(zLockFile, 0)==0;
  }
  OSTRACE(("TEST WR-LOCK %d %d %d (dotlock)\n", pFile->h, rc, reserved));
  *pResOut = reserved;
  return rc;
}

/*

    /* Always update the timestamp on the old file */
    utimes(zLockFile, NULL);
#endif
    return SQLITE_OK;
  }
  
  /* grab an exclusive lock */
  fd = robust_open(zLockFile,O_RDONLY|O_CREAT|O_EXCL,0600);
  if( fd<0 ){
    /* failed to open/create the file, someone else may have stolen the lock */
    int tErrno = errno;
    if( EEXIST == tErrno ){
      rc = SQLITE_BUSY;
    } else {
      rc = sqliteErrorFromPosixError(tErrno, SQLITE_IOERR_LOCK);
      if( IS_LOCK_ERROR(rc) ){
        pFile->lastErrno = tErrno;
      }
    }
    return rc;
  } 
  robust_close(pFile, fd, __LINE__);



  
  /* got it, set the type and return ok */
  pFile->eFileLock = eFileLock;
  return rc;
}

/*

  
  /* To fully unlock the database, delete the lock file */
  assert( eFileLock==NO_LOCK );
  if( unlink(zLockFile) ){
    int rc = 0;
    int tErrno = errno;
    if( ENOENT != tErrno ){
      rc = SQLITE_IOERR_UNLOCK;
    }
    if( IS_LOCK_ERROR(rc) ){
      pFile->lastErrno = tErrno;
    }
    return rc; 
  }
  pFile->eFileLock = NO_LOCK;

** only a single process can be reading the database at a time.
**
** Omit this section if SQLITE_ENABLE_LOCKING_STYLE is turned off or if
** compiling for VXWORKS.
*/
#if SQLITE_ENABLE_LOCKING_STYLE && !OS_VXWORKS

/*
** Retry flock() calls that fail with EINTR
*/
#ifdef EINTR
static int robust_flock(int fd, int op){
  int rc;
  do{ rc = flock(fd,op); }while( rc<0 && errno==EINTR );
  return rc;
}
#else
# define robust_flock(a,b) flock(a,b)
#endif
     

/*
** This routine checks if there is a RESERVED lock held on the specified
** file by this or any other process. If such a lock is held, set *pResOut
** to a non-zero value otherwise *pResOut is set to zero.  The return value
** is set to SQLITE_OK unless an I/O error occurs during lock checking.
*/
static int flockCheckReservedLock(sqlite3_file *id, int *pResOut){

  if( pFile->eFileLock>SHARED_LOCK ){
    reserved = 1;
  }
  
  /* Otherwise see if some other process holds it. */
  if( !reserved ){
    /* attempt to get the lock */
    int lrc = robust_flock(pFile->h, LOCK_EX | LOCK_NB);
    if( !lrc ){
      /* got the lock, unlock it */
      lrc = robust_flock(pFile->h, LOCK_UN);
      if ( lrc ) {
        int tErrno = errno;
        /* unlock failed with an error */
        lrc = SQLITE_IOERR_UNLOCK; 
        if( IS_LOCK_ERROR(lrc) ){
          pFile->lastErrno = tErrno;
          rc = lrc;
        }
      }
    } else {
      int tErrno = errno;

  if (pFile->eFileLock > NO_LOCK) {
    pFile->eFileLock = eFileLock;
    return SQLITE_OK;
  }
  
  /* grab an exclusive lock */
  
  if (robust_flock(pFile->h, LOCK_EX | LOCK_NB)) {
    int tErrno = errno;
    /* didn't get, must be busy */
    rc = sqliteErrorFromPosixError(tErrno, SQLITE_IOERR_LOCK);
    if( IS_LOCK_ERROR(rc) ){
      pFile->lastErrno = tErrno;
    }
  } else {

  /* shared can just be set because we always have an exclusive */
  if (eFileLock==SHARED_LOCK) {
    pFile->eFileLock = eFileLock;
    return SQLITE_OK;
  }
  
  /* no, really, unlock. */
  if( robust_flock(pFile->h, LOCK_UN) ){






#ifdef SQLITE_IGNORE_FLOCK_LOCK_ERRORS

    return SQLITE_OK;

#endif /* SQLITE_IGNORE_FLOCK_LOCK_ERRORS */

    return SQLITE_IOERR_UNLOCK;
  }else{
    pFile->eFileLock = NO_LOCK;
    return SQLITE_OK;
  }
}

/*
** Close a file.

        pFile->eFileLock = NO_LOCK;
      }
    }
    if( rc==SQLITE_OK ){
      pInode->nLock--;
      assert( pInode->nLock>=0 );
      if( pInode->nLock==0 ){
        closePendingFds(pFile);
      }
    }
  }
  
  unixLeaveMutex();
  if( rc==SQLITE_OK ) pFile->eFileLock = eFileLock;
  return rc;

 ** Lower the locking level on file descriptor pFile to eFileLock.  eFileLock
 ** must be either NO_LOCK or SHARED_LOCK.
 **
 ** If the locking level of the file descriptor is already at or below
 ** the requested locking level, this routine is a no-op.
 */
static int nfsUnlock(sqlite3_file *id, int eFileLock){
  return posixUnlock(id, eFileLock, 1);
}

#endif /* defined(__APPLE__) && SQLITE_ENABLE_LOCKING_STYLE */
/*
** The code above is the NFS lock implementation.  The code is specific
** to MacOSX and does not work on other unix platforms.  No alternative
** is available.  

static int seekAndRead(unixFile *id, sqlite3_int64 offset, void *pBuf, int cnt){
  int got;
#if (!defined(USE_PREAD) && !defined(USE_PREAD64))
  i64 newOffset;
#endif
  TIMER_START;
#if defined(USE_PREAD)
  do{ got = osPread(id->h, pBuf, cnt, offset); }while( got<0 && errno==EINTR );
  SimulateIOError( got = -1 );
#elif defined(USE_PREAD64)
  do{ got = osPread64(id->h, pBuf, cnt, offset); }while( got<0 && errno==EINTR);
  SimulateIOError( got = -1 );
#else
  newOffset = lseek(id->h, offset, SEEK_SET);
  SimulateIOError( newOffset-- );
  if( newOffset!=offset ){
    if( newOffset == -1 ){
      ((unixFile*)id)->lastErrno = errno;
    }else{
      ((unixFile*)id)->lastErrno = 0;			
    }
    return -1;
  }
  do{ got = osRead(id->h, pBuf, cnt); }while( got<0 && errno==EINTR );
#endif
  TIMER_END;
  if( got<0 ){
    ((unixFile*)id)->lastErrno = errno;
  }
  OSTRACE(("READ    %-3d %5d %7lld %llu\n", id->h, got, offset, TIMER_ELAPSED));
  return got;

static int seekAndWrite(unixFile *id, i64 offset, const void *pBuf, int cnt){
  int got;
#if (!defined(USE_PREAD) && !defined(USE_PREAD64))
  i64 newOffset;
#endif
  TIMER_START;
#if defined(USE_PREAD)
  do{ got = osPwrite(id->h, pBuf, cnt, offset); }while( got<0 && errno==EINTR );
#elif defined(USE_PREAD64)
  do{ got = osPwrite64(id->h, pBuf, cnt, offset);}while( got<0 && errno==EINTR);
#else
  newOffset = lseek(id->h, offset, SEEK_SET);
  SimulateIOError( newOffset-- );
  if( newOffset!=offset ){
    if( newOffset == -1 ){
      ((unixFile*)id)->lastErrno = errno;
    }else{
      ((unixFile*)id)->lastErrno = 0;			
    }
    return -1;
  }
  do{ got = osWrite(id->h, pBuf, cnt); }while( got<0 && errno==EINTR );
#endif
  TIMER_END;
  if( got<0 ){
    ((unixFile*)id)->lastErrno = errno;
  }

  OSTRACE(("WRITE   %-3d %5d %7lld %llu\n", id->h, got, offset, TIMER_ELAPSED));

  /* If we compiled with the SQLITE_NO_SYNC flag, then syncing is a
  ** no-op
  */
#ifdef SQLITE_NO_SYNC
  rc = SQLITE_OK;
#elif HAVE_FULLFSYNC
  if( fullSync ){
    rc = osFcntl(fd, F_FULLFSYNC, 0);
  }else{
    rc = 1;
  }
  /* If the FULLFSYNC failed, fall back to attempting an fsync().
  ** It shouldn't be possible for fullfsync to fail on the local 
  ** file system (on OSX), so failure indicates that FULLFSYNC
  ** isn't supported for this file system. So, attempt an fsync 

  assert( pFile );
  OSTRACE(("SYNC    %-3d\n", pFile->h));
  rc = full_fsync(pFile->h, isFullsync, isDataOnly);
  SimulateIOError( rc=1 );
  if( rc ){
    pFile->lastErrno = errno;
    return unixLogError(SQLITE_IOERR_FSYNC, "full_fsync", pFile->zPath);
  }
  if( pFile->dirfd>=0 ){

    OSTRACE(("DIRSYNC %-3d (have_fullfsync=%d fullsync=%d)\n", pFile->dirfd,
            HAVE_FULLFSYNC, isFullsync));
#ifndef SQLITE_DISABLE_DIRSYNC
    /* The directory sync is only attempted if full_fsync is
    ** turned off or unavailable.  If a full_fsync occurred above,
    ** then the directory sync is superfluous.
    */
    if( (!HAVE_FULLFSYNC || !isFullsync) && full_fsync(pFile->dirfd,0,0) ){
       /*
       ** We have received multiple reports of fsync() returning
       ** errors when applied to directories on certain file systems.
       ** A failed directory sync is not a big deal.  So it seems
       ** better to ignore the error.  Ticket #1657
       */
       /* pFile->lastErrno = errno; */
       /* return SQLITE_IOERR; */
    }
#endif
    /* Only need to sync once, so close the  directory when we are done */
    robust_close(pFile, pFile->dirfd, __LINE__);
    pFile->dirfd = -1;




  }
  return rc;
}

/*
** Truncate an open file to a specified size
*/

  ** actual file size after the operation may be larger than the requested
  ** size).
  */
  if( pFile->szChunk ){
    nByte = ((nByte + pFile->szChunk - 1)/pFile->szChunk) * pFile->szChunk;
  }

  rc = robust_ftruncate(pFile->h, (off_t)nByte);
  if( rc ){
    pFile->lastErrno = errno;
    return unixLogError(SQLITE_IOERR_TRUNCATE, "ftruncate", pFile->zPath);
  }else{
#ifndef NDEBUG
    /* If we are doing a normal write to a database file (as opposed to
    ** doing a hot-journal rollback or a write to some file other than a
    ** normal database file) and we truncate the file to zero length,
    ** that effectively updates the change counter.  This might happen
    ** when restoring a database using the backup API from a zero-length

/*
** Determine the current size of a file in bytes
*/
static int unixFileSize(sqlite3_file *id, i64 *pSize){
  int rc;
  struct stat buf;
  assert( id );
  rc = osFstat(((unixFile*)id)->h, &buf);
  SimulateIOError( rc=1 );
  if( rc!=0 ){
    ((unixFile*)id)->lastErrno = errno;
    return SQLITE_IOERR_FSTAT;
  }
  *pSize = buf.st_size;

** SQLITE_FCNTL_SIZE_HINT operation is a no-op for Unix.
*/
static int fcntlSizeHint(unixFile *pFile, i64 nByte){
  if( pFile->szChunk ){
    i64 nSize;                    /* Required file size */
    struct stat buf;              /* Used to hold return values of fstat() */
   
    if( osFstat(pFile->h, &buf) ) return SQLITE_IOERR_FSTAT;

    nSize = ((nByte+pFile->szChunk-1) / pFile->szChunk) * pFile->szChunk;
    if( nSize>(i64)buf.st_size ){

#if defined(HAVE_POSIX_FALLOCATE) && HAVE_POSIX_FALLOCATE
      /* The code below is handling the return value of osFallocate() 
      ** correctly. posix_fallocate() is defined to "returns zero on success, 
      ** or an error number on  failure". See the manpage for details. */
      int err;
      do{
        err = osFallocate(pFile->h, buf.st_size, nSize-buf.st_size);
      }while( err==EINTR );
      if( err ) return SQLITE_IOERR_WRITE;

#else
      /* If the OS does not have posix_fallocate(), fake it. First use
      ** ftruncate() to set the file size, then write a single byte to
      ** the last byte in each block within the extended region. This
      ** is the same technique used by glibc to implement posix_fallocate()
      ** on systems that do not have a real fallocate() system call.
      */
      int nBlk = buf.st_blksize;  /* File-system block size */
      i64 iWrite;                 /* Next offset to write to */


      if( robust_ftruncate(pFile->h, nSize) ){
        pFile->lastErrno = errno;
        return unixLogError(SQLITE_IOERR_TRUNCATE, "ftruncate", pFile->zPath);
      }
      iWrite = ((buf.st_size + 2*nBlk - 1)/nBlk)*nBlk-1;

      while( iWrite<nSize ){
        int nWrite = seekAndWrite(pFile, iWrite, "", 1);


        if( nWrite!=1 ) return SQLITE_IOERR_WRITE;
        iWrite += nBlk;
      }
#endif
    }
  }

  return SQLITE_OK;
}

  /* Shared locks never span more than one byte */
  assert( n==1 || lockType!=F_RDLCK );

  /* Locks are within range */
  assert( n>=1 && n<SQLITE_SHM_NLOCK );

  if( pShmNode->h>=0 ){
    /* Initialize the locking parameters */
    memset(&f, 0, sizeof(f));
    f.l_type = lockType;
    f.l_whence = SEEK_SET;
    f.l_start = ofst;
    f.l_len = n;

    rc = osFcntl(pShmNode->h, F_SETLK, &f);
    rc = (rc!=(-1)) ? SQLITE_OK : SQLITE_BUSY;
  }

  /* Update the global lock state and do debug tracing */
#ifdef SQLITE_DEBUG
  { u16 mask;
  OSTRACE(("SHM-LOCK "));
  mask = (1<<(ofst+n)) - (1<<ofst);
  if( rc==SQLITE_OK ){

  unixShmNode *p = pFd->pInode->pShmNode;
  assert( unixMutexHeld() );
  if( p && p->nRef==0 ){
    int i;
    assert( p->pInode==pFd->pInode );
    if( p->mutex ) sqlite3_mutex_free(p->mutex);
    for(i=0; i<p->nRegion; i++){
      if( p->h>=0 ){
        munmap(p->apRegion[i], p->szRegion);
      }else{
        sqlite3_free(p->apRegion[i]);
      }
    }
    sqlite3_free(p->apRegion);
    if( p->h>=0 ){
      robust_close(pFd, p->h, __LINE__);
      p->h = -1;
    }
    p->pInode->pShmNode = 0;
    sqlite3_free(p);
  }
}

/*
** Open a shared-memory area associated with open database file pDbFd.  

** same database file at the same time, database corruption will likely
** result. The SQLITE_SHM_DIRECTORY compile-time option is considered
** "unsupported" and may go away in a future SQLite release.
**
** When opening a new shared-memory file, if no other instances of that
** file are currently open, in this process or in other processes, then
** the file must be truncated to zero length or have its header cleared.
**
** If the original database file (pDbFd) is using the "unix-excl" VFS
** that means that an exclusive lock is held on the database file and
** that no other processes are able to read or write the database.  In
** that case, we do not really need shared memory.  No shared memory
** file is created.  The shared memory will be simulated with heap memory.
*/
static int unixOpenSharedMemory(unixFile *pDbFd){
  struct unixShm *p = 0;          /* The connection to be opened */
  struct unixShmNode *pShmNode;   /* The underlying mmapped file */
  int rc;                         /* Result code */
  unixInodeInfo *pInode;          /* The inode of fd */
  char *zShmFilename;             /* Name of the file used for SHM */

    struct stat sStat;                 /* fstat() info for database file */

    /* Call fstat() to figure out the permissions on the database file. If
    ** a new *-shm file is created, an attempt will be made to create it
    ** with the same permissions. The actual permissions the file is created
    ** with are subject to the current umask setting.
    */
    if( osFstat(pDbFd->h, &sStat) && pInode->bProcessLock==0 ){
      rc = SQLITE_IOERR_FSTAT;
      goto shm_open_err;
    }

#ifdef SQLITE_SHM_DIRECTORY
    nShmFilename = sizeof(SQLITE_SHM_DIRECTORY) + 30;
#else

    pShmNode->pInode = pDbFd->pInode;
    pShmNode->mutex = sqlite3_mutex_alloc(SQLITE_MUTEX_FAST);
    if( pShmNode->mutex==0 ){
      rc = SQLITE_NOMEM;
      goto shm_open_err;
    }

    if( pInode->bProcessLock==0 ){
      pShmNode->h = robust_open(zShmFilename, O_RDWR|O_CREAT,
                               (sStat.st_mode & 0777));
      if( pShmNode->h<0 ){
        rc = unixLogError(SQLITE_CANTOPEN_BKPT, "open", zShmFilename);
        goto shm_open_err;
      }
  
      /* Check to see if another process is holding the dead-man switch.
      ** If not, truncate the file to zero length. 
      */
      rc = SQLITE_OK;
      if( unixShmSystemLock(pShmNode, F_WRLCK, UNIX_SHM_DMS, 1)==SQLITE_OK ){
        if( robust_ftruncate(pShmNode->h, 0) ){
          rc = unixLogError(SQLITE_IOERR_SHMOPEN, "ftruncate", zShmFilename);
        }
      }
      if( rc==SQLITE_OK ){
        rc = unixShmSystemLock(pShmNode, F_RDLCK, UNIX_SHM_DMS, 1);
      }
      if( rc ) goto shm_open_err;
    }
  }

  /* Make the new connection a child of the unixShmNode */
  p->pShmNode = pShmNode;
#ifdef SQLITE_DEBUG
  p->id = pShmNode->nextShmId++;
#endif

    if( rc!=SQLITE_OK ) return rc;
  }

  p = pDbFd->pShm;
  pShmNode = p->pShmNode;
  sqlite3_mutex_enter(pShmNode->mutex);
  assert( szRegion==pShmNode->szRegion || pShmNode->nRegion==0 );
  assert( pShmNode->pInode==pDbFd->pInode );
  assert( pShmNode->h>=0 || pDbFd->pInode->bProcessLock==1 );
  assert( pShmNode->h<0 || pDbFd->pInode->bProcessLock==0 );

  if( pShmNode->nRegion<=iRegion ){
    char **apNew;                      /* New apRegion[] array */
    int nByte = (iRegion+1)*szRegion;  /* Minimum required file size */
    struct stat sStat;                 /* Used by fstat() */

    pShmNode->szRegion = szRegion;

    if( pShmNode->h>=0 ){
      /* The requested region is not mapped into this processes address space.
      ** Check to see if it has been allocated (i.e. if the wal-index file is
      ** large enough to contain the requested region).
      */
      if( osFstat(pShmNode->h, &sStat) ){
        rc = SQLITE_IOERR_SHMSIZE;
        goto shmpage_out;
      }
  
      if( sStat.st_size<nByte ){
        /* The requested memory region does not exist. If bExtend is set to
        ** false, exit early. *pp will be set to NULL and SQLITE_OK returned.
        **
        ** Alternatively, if bExtend is true, use ftruncate() to allocate
        ** the requested memory region.
        */
        if( !bExtend ) goto shmpage_out;
        if( robust_ftruncate(pShmNode->h, nByte) ){
          rc = unixLogError(SQLITE_IOERR_SHMSIZE, "ftruncate",
                            pShmNode->zFilename);
          goto shmpage_out;
        }
      }
    }

    /* Map the requested memory region into this processes address space. */
    apNew = (char **)sqlite3_realloc(
        pShmNode->apRegion, (iRegion+1)*sizeof(char *)
    );
    if( !apNew ){
      rc = SQLITE_IOERR_NOMEM;
      goto shmpage_out;
    }
    pShmNode->apRegion = apNew;
    while(pShmNode->nRegion<=iRegion){
      void *pMem;
      if( pShmNode->h>=0 ){
        pMem = mmap(0, szRegion, PROT_READ|PROT_WRITE, 
            MAP_SHARED, pShmNode->h, pShmNode->nRegion*szRegion
        );
        if( pMem==MAP_FAILED ){
          rc = SQLITE_IOERR;
          goto shmpage_out;
        }
      }else{
        pMem = sqlite3_malloc(szRegion);
        if( pMem==0 ){
          rc = SQLITE_NOMEM;
          goto shmpage_out;
        }
        memset(pMem, 0, szRegion);
      }
      pShmNode->apRegion[pShmNode->nRegion] = pMem;
      pShmNode->nRegion++;
    }
  }

shmpage_out:

  assert( ofst>=0 && ofst+n<=SQLITE_SHM_NLOCK );
  assert( n>=1 );
  assert( flags==(SQLITE_SHM_LOCK | SQLITE_SHM_SHARED)
       || flags==(SQLITE_SHM_LOCK | SQLITE_SHM_EXCLUSIVE)
       || flags==(SQLITE_SHM_UNLOCK | SQLITE_SHM_SHARED)
       || flags==(SQLITE_SHM_UNLOCK | SQLITE_SHM_EXCLUSIVE) );
  assert( n==1 || (flags & SQLITE_SHM_EXCLUSIVE)!=0 );
  assert( pShmNode->h>=0 || pDbFd->pInode->bProcessLock==1 );
  assert( pShmNode->h<0 || pDbFd->pInode->bProcessLock==0 );

  mask = (1<<(ofst+n)) - (1<<ofst);
  assert( n>1 || mask==(1<<ofst) );
  sqlite3_mutex_enter(pShmNode->mutex);
  if( flags & SQLITE_SHM_UNLOCK ){
    u16 allMask = 0; /* Mask of locks held by siblings */

  /* If pShmNode->nRef has reached 0, then close the underlying
  ** shared-memory file, too */
  unixEnterMutex();
  assert( pShmNode->nRef>0 );
  pShmNode->nRef--;
  if( pShmNode->nRef==0 ){
    if( deleteFlag && pShmNode->h>=0 ) unlink(pShmNode->zFilename);
    unixShmPurge(pDbFd);
  }
  unixLeaveMutex();

  return SQLITE_OK;
}

  ** Test byte-range lock using fcntl(). If the call succeeds, 
  ** assume that the file-system supports POSIX style locks. 
  */
  lockInfo.l_len = 1;
  lockInfo.l_start = 0;
  lockInfo.l_whence = SEEK_SET;
  lockInfo.l_type = F_RDLCK;
  if( osFcntl(pNew->h, F_GETLK, &lockInfo)!=-1 ) {
    if( strcmp(fsInfo.f_fstypename, "nfs")==0 ){
      return &nfsIoMethods;
    } else {
      return &posixIoMethods;
    }
  }else{
    return &dotlockIoMethods;

  /* Test if fcntl() is supported and use POSIX style locks.
  ** Otherwise fall back to the named semaphore method.
  */
  lockInfo.l_len = 1;
  lockInfo.l_start = 0;
  lockInfo.l_whence = SEEK_SET;
  lockInfo.l_type = F_RDLCK;
  if( osFcntl(pNew->h, F_GETLK, &lockInfo)!=-1 ) {
    return &posixIoMethods;
  }else{
    return &semIoMethods;
  }
}
static const sqlite3_io_methods 
  *(*const autolockIoFinder)(const char*,unixFile*) = autolockIoFinderImpl;

static int fillInUnixFile(
  sqlite3_vfs *pVfs,      /* Pointer to vfs object */
  int h,                  /* Open file descriptor of file being opened */
  int dirfd,              /* Directory file descriptor */
  sqlite3_file *pId,      /* Write to the unixFile structure here */
  const char *zFilename,  /* Name of the file being opened */
  int noLock,             /* Omit locking if true */
  int isDelete,           /* Delete on close if true */
  int isReadOnly          /* True if the file is opened read-only */
){
  const sqlite3_io_methods *pLockingStyle;
  unixFile *pNew = (unixFile *)pId;
  int rc = SQLITE_OK;

  assert( pNew->pInode==NULL );

#else
  assert( zFilename==0 || zFilename[0]=='/' );
#endif

  OSTRACE(("OPEN    %-3d %s\n", h, zFilename));
  pNew->h = h;
  pNew->dirfd = dirfd;

  pNew->zPath = zFilename;
  if( memcmp(pVfs->zName,"unix-excl",10)==0 ){
    pNew->ctrlFlags = UNIXFILE_EXCL;
  }else{
    pNew->ctrlFlags = 0;
  }
  if( isReadOnly ){
    pNew->ctrlFlags |= UNIXFILE_RDONLY;
  }

#if OS_VXWORKS
  pNew->pId = vxworksFindFileId(zFilename);
  if( pNew->pId==0 ){
    noLock = 1;
    rc = SQLITE_NOMEM;
  }

      ** handle h - as it is guaranteed that no posix locks will be released
      ** by doing so.
      **
      ** If scenario (a) caused the error then things are not so safe. The
      ** implicit assumption here is that if fstat() fails, things are in
      ** such bad shape that dropping a lock or two doesn't matter much.
      */
      robust_close(pNew, h, __LINE__);
      h = -1;
    }
    unixLeaveMutex();
  }

#if SQLITE_ENABLE_LOCKING_STYLE && defined(__APPLE__)
  else if( pLockingStyle == &afpIoMethods ){

      pCtx->dbPath = zFilename;
      pCtx->reserved = 0;
      srandomdev();
      unixEnterMutex();
      rc = findInodeInfo(pNew, &pNew->pInode);
      if( rc!=SQLITE_OK ){
        sqlite3_free(pNew->lockingContext);
        robust_close(pNew, h, __LINE__);
        h = -1;
      }
      unixLeaveMutex();        
    }
  }
#endif

    unixLeaveMutex();
  }
#endif
  
  pNew->lastErrno = 0;
#if OS_VXWORKS
  if( rc!=SQLITE_OK ){
    if( h>=0 ) robust_close(pNew, h, __LINE__);
    h = -1;
    unlink(zFilename);
    isDelete = 0;
  }
  pNew->isDelete = isDelete;
#endif
  if( rc!=SQLITE_OK ){
    if( dirfd>=0 ) robust_close(pNew, dirfd, __LINE__);
    if( h>=0 ) robust_close(pNew, h, __LINE__);
  }else{
    pNew->pMethod = pLockingStyle;
    OpenCounter(+1);
  }
  return rc;
}

  int fd = -1;
  char zDirname[MAX_PATHNAME+1];

  sqlite3_snprintf(MAX_PATHNAME, zDirname, "%s", zFilename);
  for(ii=(int)strlen(zDirname); ii>1 && zDirname[ii]!='/'; ii--);
  if( ii>0 ){
    zDirname[ii] = '\0';
    fd = robust_open(zDirname, O_RDONLY|O_BINARY, 0);
    if( fd>=0 ){
#ifdef FD_CLOEXEC
      osFcntl(fd, F_SETFD, osFcntl(fd, F_GETFD, 0) | FD_CLOEXEC);
#endif
      OSTRACE(("OPENDIR %-3d %s\n", fd, zDirname));
    }
  }
  *pFd = fd;
  return (fd>=0?SQLITE_OK:unixLogError(SQLITE_CANTOPEN_BKPT, "open", zDirname));
}

/*
** Return the name of a directory in which to put temporary files.
** If no suitable temporary file directory can be found, return NULL.
*/
static const char *unixTempFileDir(void){

  struct stat buf;
  const char *zDir = 0;

  azDirs[0] = sqlite3_temp_directory;
  if( !azDirs[1] ) azDirs[1] = getenv("TMPDIR");
  for(i=0; i<sizeof(azDirs)/sizeof(azDirs[0]); zDir=azDirs[i++]){
    if( zDir==0 ) continue;
    if( osStat(zDir, &buf) ) continue;
    if( !S_ISDIR(buf.st_mode) ) continue;
    if( osAccess(zDir, 07) ) continue;
    break;
  }
  return zDir;
}

/*
** Create a temporary file name in zBuf.  zBuf must be allocated

    sqlite3_snprintf(nBuf-17, zBuf, "%s/"SQLITE_TEMP_FILE_PREFIX, zDir);
    j = (int)strlen(zBuf);
    sqlite3_randomness(15, &zBuf[j]);
    for(i=0; i<15; i++, j++){
      zBuf[j] = (char)zChars[ ((unsigned char)zBuf[j])%(sizeof(zChars)-1) ];
    }
    zBuf[j] = 0;
  }while( osAccess(zBuf,0)==0 );
  return SQLITE_OK;
}

#if SQLITE_ENABLE_LOCKING_STYLE && defined(__APPLE__)
/*
** Routine to transform a unixFile into a proxy-locking unixFile.
** Implementation in the proxy-lock division, but used by unixOpen()

    mode_t openMode;              /* Permissions to create file with */
    rc = findCreateFileMode(zName, flags, &openMode);
    if( rc!=SQLITE_OK ){
      assert( !p->pUnused );
      assert( eType==SQLITE_OPEN_WAL || eType==SQLITE_OPEN_MAIN_JOURNAL );
      return rc;
    }
    fd = robust_open(zName, openFlags, openMode);
    OSTRACE(("OPENX   %-3d %s 0%o\n", fd, zName, openFlags));
    if( fd<0 && errno!=EISDIR && isReadWrite && !isExclusive ){
      /* Failed to open the file for read/write access. Try read-only. */
      flags &= ~(SQLITE_OPEN_READWRITE|SQLITE_OPEN_CREATE);
      openFlags &= ~(O_RDWR|O_CREAT);
      flags |= SQLITE_OPEN_READONLY;
      openFlags |= O_RDONLY;
      isReadonly = 1;
      fd = robust_open(zName, openFlags, openMode);
    }
    if( fd<0 ){
      rc = unixLogError(SQLITE_CANTOPEN_BKPT, "open", zName);
      goto open_finished;
    }
  }
  assert( fd>=0 );
  if( pOutFlags ){
    *pOutFlags = flags;
  }

    rc = openDirectory(zPath, &dirfd);
    if( rc!=SQLITE_OK ){
      /* It is safe to close fd at this point, because it is guaranteed not
      ** to be open on a database file. If it were open on a database file,
      ** it would not be safe to close as this would release any locks held
      ** on the file by this process.  */
      assert( eType!=SQLITE_OPEN_MAIN_DB );
      robust_close(p, fd, __LINE__);
      goto open_finished;
    }
  }

#ifdef FD_CLOEXEC
  osFcntl(fd, F_SETFD, osFcntl(fd, F_GETFD, 0) | FD_CLOEXEC);
#endif

  noLock = eType!=SQLITE_OPEN_MAIN_DB;

  
#if defined(__APPLE__) || SQLITE_ENABLE_LOCKING_STYLE
  struct statfs fsInfo;
  if( fstatfs(fd, &fsInfo) == -1 ){
    ((unixFile*)pFile)->lastErrno = errno;
    if( dirfd>=0 ) robust_close(p, dirfd, __LINE__);
    robust_close(p, fd, __LINE__);
    return SQLITE_IOERR_ACCESS;
  }
  if (0 == strncmp("msdos", fsInfo.f_fstypename, 5)) {
    ((unixFile*)pFile)->fsFlags |= SQLITE_FSFLAGS_IS_MSDOS;
  }
#endif
  

        ** on that file that are currently holding advisory locks on it,
        ** then the call to close() will cancel those locks. In practice,
        ** we're assuming that statfs() doesn't fail very often. At least
        ** not while other file descriptors opened by the same process on
        ** the same file are working.  */
        p->lastErrno = errno;
        if( dirfd>=0 ){
          robust_close(p, dirfd, __LINE__);
        }
        robust_close(p, fd, __LINE__);
        rc = SQLITE_IOERR_ACCESS;
        goto open_finished;
      }
      useProxy = !(fsInfo.f_flags&MNT_LOCAL);
    }
    if( useProxy ){
      rc = fillInUnixFile(pVfs, fd, dirfd, pFile, zPath, noLock,
                          isDelete, isReadonly);
      if( rc==SQLITE_OK ){
        rc = proxyTransformUnixFile((unixFile*)pFile, ":auto:");
        if( rc!=SQLITE_OK ){
          /* Use unixClose to clean up the resources added in fillInUnixFile 
          ** and clear all the structure's references.  Specifically, 
          ** pFile->pMethods will be NULL so sqlite3OsClose will be a no-op 
          */
          unixClose(pFile);
          return rc;
        }
      }
      goto open_finished;
    }
  }
#endif
  
  rc = fillInUnixFile(pVfs, fd, dirfd, pFile, zPath, noLock,
                      isDelete, isReadonly);
open_finished:
  if( rc!=SQLITE_OK ){
    sqlite3_free(p->pUnused);
  }
  return rc;
}

  const char *zPath,        /* Name of file to be deleted */
  int dirSync               /* If true, fsync() directory after deleting file */
){
  int rc = SQLITE_OK;
  UNUSED_PARAMETER(NotUsed);
  SimulateIOError(return SQLITE_IOERR_DELETE);
  if( unlink(zPath)==(-1) && errno!=ENOENT ){
    return unixLogError(SQLITE_IOERR_DELETE, "unlink", zPath);
  }
#ifndef SQLITE_DISABLE_DIRSYNC
  if( dirSync ){
    int fd;
    rc = openDirectory(zPath, &fd);
    if( rc==SQLITE_OK ){
#if OS_VXWORKS
      if( fsync(fd)==-1 )
#else
      if( fsync(fd) )
#endif
      {
        rc = unixLogError(SQLITE_IOERR_DIR_FSYNC, "fsync", zPath);
      }
      robust_close(0, fd, __LINE__);


    }
  }
#endif
  return rc;
}

/*

    case SQLITE_ACCESS_READ:
      amode = R_OK;
      break;

    default:
      assert(!"Invalid flags argument");
  }
  *pResOut = (osAccess(zPath, amode)==0);
  if( flags==SQLITE_ACCESS_EXISTS && *pResOut ){
    struct stat buf;
    if( 0==stat(zPath, &buf) && buf.st_size==0 ){
      *pResOut = 0;
    }
  }
  return SQLITE_OK;

  UNUSED_PARAMETER(pVfs);

  zOut[nOut-1] = '\0';
  if( zPath[0]=='/' ){
    sqlite3_snprintf(nOut, zOut, "%s", zPath);
  }else{
    int nCwd;
    if( osGetcwd(zOut, nOut-1)==0 ){
      return unixLogError(SQLITE_CANTOPEN_BKPT, "getcwd", zPath);
    }
    nCwd = (int)strlen(zOut);
    sqlite3_snprintf(nOut-nCwd, &zOut[nCwd], "/%s", zPath);
  }
  return SQLITE_OK;
}

  ** that we always use the same random number sequence.  This makes the
  ** tests repeatable.
  */
  memset(zBuf, 0, nBuf);
#if !defined(SQLITE_TEST)
  {
    int pid, fd;
    fd = robust_open("/dev/urandom", O_RDONLY, 0);
    if( fd<0 ){
      time_t t;
      time(&t);
      memcpy(zBuf, &t, sizeof(t));
      pid = getpid();
      memcpy(&zBuf[sizeof(t)], &pid, sizeof(pid));
      assert( sizeof(t)+sizeof(pid)<=(size_t)nBuf );
      nBuf = sizeof(t) + sizeof(pid);
    }else{
      do{ nBuf = osRead(fd, zBuf, nBuf); }while( nBuf<0 && errno==EINTR );
      robust_close(0, fd, __LINE__);
    }
  }
#endif
  return nBuf;
}

  }else{
    pUnused = sqlite3_malloc(sizeof(*pUnused));
    if( !pUnused ){
      return SQLITE_NOMEM;
    }
  }
  if( fd<0 ){
    fd = robust_open(path, openFlags, SQLITE_DEFAULT_FILE_PERMISSIONS);
    terrno = errno;
    if( fd<0 && errno==ENOENT && islockfile ){
      if( proxyCreateLockPath(path) == SQLITE_OK ){
        fd = robust_open(path, openFlags, SQLITE_DEFAULT_FILE_PERMISSIONS);
      }
    }
  }
  if( fd<0 ){
    openFlags = O_RDONLY;
    fd = robust_open(path, openFlags, SQLITE_DEFAULT_FILE_PERMISSIONS);
    terrno = errno;
  }
  if( fd<0 ){
    if( islockfile ){
      return SQLITE_BUSY;
    }
    switch (terrno) {

  pNew = (unixFile *)sqlite3_malloc(sizeof(*pNew));
  if( pNew==NULL ){
    rc = SQLITE_NOMEM;
    goto end_create_proxy;
  }
  memset(pNew, 0, sizeof(unixFile));
  pNew->openFlags = openFlags;
  memset(&dummyVfs, 0, sizeof(dummyVfs));
  dummyVfs.pAppData = (void*)&autolockIoFinder;
  dummyVfs.zName = "dummy";
  pUnused->fd = fd;
  pUnused->flags = openFlags;
  pNew->pUnused = pUnused;
  
  rc = fillInUnixFile(&dummyVfs, fd, dirfd, (sqlite3_file*)pNew, path, 0, 0, 0);
  if( rc==SQLITE_OK ){
    *ppFile = pNew;
    return SQLITE_OK;
  }
end_create_proxy:    
  robust_close(pNew, fd, __LINE__);
  sqlite3_free(pNew);
  sqlite3_free(pUnused);
  return rc;
}

#ifdef SQLITE_TEST
/* simulate multiple hosts by creating unique hostid file paths */

  pathLen = strlcpy(tPath, cPath, MAXPATHLEN);
  if( pathLen>MAXPATHLEN || pathLen<6 || 
     (strlcpy(&tPath[pathLen-5], "break", 6) != 5) ){
    sqlite3_snprintf(sizeof(errmsg),errmsg,"path error (len %d)",(int)pathLen);
    goto end_breaklock;
  }
  /* read the conch content */
  readLen = osPread(conchFile->h, buf, PROXY_MAXCONCHLEN, 0);
  if( readLen<PROXY_PATHINDEX ){
    sqlite3_snprintf(sizeof(errmsg),errmsg,"read error (len %d)",(int)readLen);
    goto end_breaklock;
  }
  /* write it out to the temporary break file */
  fd = robust_open(tPath, (O_RDWR|O_CREAT|O_EXCL),
                   SQLITE_DEFAULT_FILE_PERMISSIONS);
  if( fd<0 ){
    sqlite3_snprintf(sizeof(errmsg), errmsg, "create failed (%d)", errno);
    goto end_breaklock;
  }
  if( osPwrite(fd, buf, readLen, 0) != (ssize_t)readLen ){
    sqlite3_snprintf(sizeof(errmsg), errmsg, "write failed (%d)", errno);
    goto end_breaklock;
  }
  if( rename(tPath, cPath) ){
    sqlite3_snprintf(sizeof(errmsg), errmsg, "rename failed (%d)", errno);
    goto end_breaklock;
  }
  rc = 0;
  fprintf(stderr, "broke stale lock on %s\n", cPath);
  robust_close(pFile, conchFile->h, __LINE__);
  conchFile->h = fd;
  conchFile->openFlags = O_RDWR | O_CREAT;

end_breaklock:
  if( rc ){
    if( fd>=0 ){
      unlink(tPath);
      robust_close(pFile, fd, __LINE__);
    }
    fprintf(stderr, "failed to break stale lock on %s, %s\n", cPath, errmsg);
  }
  return rc;
}

/* Take the requested lock on the conch file and break a stale lock if the 

      /* If the lock failed (busy):
       * 1st try: get the mod time of the conch, wait 0.5s and try again. 
       * 2nd try: fail if the mod time changed or host id is different, wait 
       *           10 sec and try again
       * 3rd try: break the lock unless the mod time has changed.
       */
      struct stat buf;
      if( osFstat(conchFile->h, &buf) ){
        pFile->lastErrno = errno;
        return SQLITE_IOERR_LOCK;
      }
      
      if( nTries==1 ){
        conchModTime = buf.st_mtimespec;
        usleep(500000); /* wait 0.5 sec and try the lock again*/
        continue;  
      }

      assert( nTries>1 );
      if( conchModTime.tv_sec != buf.st_mtimespec.tv_sec || 
         conchModTime.tv_nsec != buf.st_mtimespec.tv_nsec ){
        return SQLITE_BUSY;
      }
      
      if( nTries==2 ){  
        char tBuf[PROXY_MAXCONCHLEN];
        int len = osPread(conchFile->h, tBuf, PROXY_MAXCONCHLEN, 0);
        if( len<0 ){
          pFile->lastErrno = errno;
          return SQLITE_IOERR_LOCK;
        }
        if( len>PROXY_PATHINDEX && tBuf[0]==(char)PROXY_CONCHVERSION){
          /* don't break the lock if the host id doesn't match */
          if( 0!=memcmp(&tBuf[PROXY_HEADERLEN], myHostID, PROXY_HOSTIDLEN) ){

        memcpy(&writeBuffer[PROXY_HEADERLEN], myHostID, PROXY_HOSTIDLEN);
        if( pCtx->lockProxyPath!=NULL ){
          strlcpy(&writeBuffer[PROXY_PATHINDEX], pCtx->lockProxyPath, MAXPATHLEN);
        }else{
          strlcpy(&writeBuffer[PROXY_PATHINDEX], tempLockPath, MAXPATHLEN);
        }
        writeSize = PROXY_PATHINDEX + strlen(&writeBuffer[PROXY_PATHINDEX]);
        robust_ftruncate(conchFile->h, writeSize);
        rc = unixWrite((sqlite3_file *)conchFile, writeBuffer, writeSize, 0);
        fsync(conchFile->h);
        /* If we created a new conch file (not just updated the contents of a 
         ** valid conch file), try to match the permissions of the database 
         */
        if( rc==SQLITE_OK && createConch ){
          struct stat buf;
          int err = osFstat(pFile->h, &buf);
          if( err==0 ){
            mode_t cmode = buf.st_mode&(S_IRUSR|S_IWUSR | S_IRGRP|S_IWGRP |
                                        S_IROTH|S_IWOTH);
            /* try to match the database file R/W permissions, ignore failure */
#ifndef SQLITE_PROXY_DEBUG
            osFchmod(conchFile->h, cmode);
#else
            do{
              rc = osFchmod(conchFile->h, cmode);
            }while( rc==(-1) && errno==EINTR );
            if( rc!=0 ){
              int code = errno;
              fprintf(stderr, "fchmod %o FAILED with %d %s\n",
                      cmode, code, strerror(code));
            } else {
              fprintf(stderr, "fchmod %o SUCCEDED\n",cmode);
            }
          }else{
            int code = errno;
            fprintf(stderr, "STAT FAILED[%d] with %d %s\n", 
                    err, code, strerror(code));
#endif
          }
        }
      }
      conchFile->pMethod->xUnlock((sqlite3_file*)conchFile, SHARED_LOCK);
      
    end_takeconch:
      OSTRACE(("TRANSPROXY: CLOSE  %d\n", pFile->h));
      if( rc==SQLITE_OK && pFile->openFlags ){
        if( pFile->h>=0 ){

          robust_close(pFile, pFile->h, __LINE__);


        }




        pFile->h = -1;
        int fd = robust_open(pCtx->dbPath, pFile->openFlags,
                      SQLITE_DEFAULT_FILE_PERMISSIONS);
        OSTRACE(("TRANSPROXY: OPEN  %d\n", fd));
        if( fd>=0 ){
          pFile->h = fd;
        }else{
          rc=SQLITE_CANTOPEN_BKPT; /* SQLITE_BUSY? proxyTakeConch called
           during locking */

      ** Ugh, since O_RDONLY==0x0000 we test for !O_RDWR since unixOpen asserts
      ** that openFlags will have only one of O_RDONLY or O_RDWR.
      */
      struct statfs fsInfo;
      struct stat conchInfo;
      int goLockless = 0;

      if( osStat(pCtx->conchFilePath, &conchInfo) == -1 ) {
        int err = errno;
        if( (err==ENOENT) && (statfs(dbPath, &fsInfo) != -1) ){
          goLockless = (fsInfo.f_flags&MNT_RDONLY) == MNT_RDONLY;
        }
      }
      if( goLockless ){
        pCtx->conchHeld = -1; /* read only FS/ lockless */

  ** Most finders simply return a pointer to a fixed sqlite3_io_methods
  ** object.  But the "autolockIoFinder" available on MacOSX does a little
  ** more than that; it looks at the filesystem type that hosts the 
  ** database file and tries to choose an locking method appropriate for
  ** that filesystem time.
  */
  #define UNIXVFS(VFSNAME, FINDER) {                        \
    3,                    /* iVersion */                    \
    sizeof(unixFile),     /* szOsFile */                    \
    MAX_PATHNAME,         /* mxPathname */                  \
    0,                    /* pNext */                       \
    VFSNAME,              /* zName */                       \
    (void*)&FINDER,       /* pAppData */                    \
    unixOpen,             /* xOpen */                       \
    unixDelete,           /* xDelete */                     \
    unixAccess,           /* xAccess */                     \
    unixFullPathname,     /* xFullPathname */               \
    unixDlOpen,           /* xDlOpen */                     \
    unixDlError,          /* xDlError */                    \
    unixDlSym,            /* xDlSym */                      \
    unixDlClose,          /* xDlClose */                    \
    unixRandomness,       /* xRandomness */                 \
    unixSleep,            /* xSleep */                      \
    unixCurrentTime,      /* xCurrentTime */                \
    unixGetLastError,     /* xGetLastError */               \
    unixCurrentTimeInt64, /* xCurrentTimeInt64 */           \
    unixSetSystemCall,    /* xSetSystemCall */              \
    unixGetSystemCall,    /* xGetSystemCall */              \
    unixNextSystemCall,   /* xNextSystemCall */             \
  }

  /*
  ** All default VFSes for unix are contained in the following array.
  **
  ** Note that the sqlite3_vfs.pNext field of the VFS object is modified
  ** by the SQLite core when the VFS is registered.  So the following
  ** array cannot be const.
  */
  static sqlite3_vfs aVfs[] = {
#if SQLITE_ENABLE_LOCKING_STYLE && (OS_VXWORKS || defined(__APPLE__))
    UNIXVFS("unix",          autolockIoFinder ),
#else
    UNIXVFS("unix",          posixIoFinder ),
#endif
    UNIXVFS("unix-none",     nolockIoFinder ),
    UNIXVFS("unix-dotfile",  dotlockIoFinder ),
    UNIXVFS("unix-excl",     posixIoFinder ),
#if OS_VXWORKS
    UNIXVFS("unix-namedsem", semIoFinder ),
#endif
#if SQLITE_ENABLE_LOCKING_STYLE
    UNIXVFS("unix-posix",    posixIoFinder ),
#if !OS_VXWORKS
    UNIXVFS("unix-flock",    flockIoFinder ),

/*
** Initialize and deinitialize the operating system interface.
*/
SQLITE_API int sqlite3_os_init(void){
  static sqlite3_vfs winVfs = {
    3,                   /* iVersion */
    sizeof(winFile),     /* szOsFile */
    MAX_PATH,            /* mxPathname */
    0,                   /* pNext */
    "win32",             /* zName */
    0,                   /* pAppData */
    winOpen,             /* xOpen */
    winDelete,           /* xDelete */
    winAccess,           /* xAccess */
    winFullPathname,     /* xFullPathname */
    winDlOpen,           /* xDlOpen */
    winDlError,          /* xDlError */
    winDlSym,            /* xDlSym */
    winDlClose,          /* xDlClose */
    winRandomness,       /* xRandomness */
    winSleep,            /* xSleep */
    winCurrentTime,      /* xCurrentTime */
    winGetLastError,     /* xGetLastError */
    winCurrentTimeInt64, /* xCurrentTimeInt64 */
    0,                   /* xSetSystemCall */
    0,                   /* xGetSystemCall */
    0,                   /* xNextSystemCall */
  };

#ifndef SQLITE_OMIT_WAL
  /* get memory map allocation granularity */
  memset(&winSysInfo, 0, sizeof(SYSTEM_INFO));
  GetSystemInfo(&winSysInfo);
  assert(winSysInfo.dwAllocationGranularity > 0);

*/

#ifndef _WAL_H_
#define _WAL_H_


#ifdef SQLITE_OMIT_WAL
# define sqlite3WalOpen(x,y,z)                   0
# define sqlite3WalClose(w,x,y,z)                0
# define sqlite3WalBeginReadTransaction(y,z)     0
# define sqlite3WalEndReadTransaction(z)
# define sqlite3WalRead(v,w,x,y,z)               0
# define sqlite3WalDbsize(y)                     0
# define sqlite3WalBeginWriteTransaction(y)      0
# define sqlite3WalEndWriteTransaction(x)        0
# define sqlite3WalUndo(x,y,z)                   0
# define sqlite3WalSavepoint(y,z)
# define sqlite3WalSavepointUndo(y,z)            0
# define sqlite3WalFrames(u,v,w,x,y,z)           0
# define sqlite3WalCheckpoint(r,s,t,u,v,w,x,y,z) 0
# define sqlite3WalCallback(z)                   0
# define sqlite3WalExclusiveMode(y,z)            0
# define sqlite3WalHeapMemory(z)                 0
#else

#define WAL_SAVEPOINT_NDATA 4

/* Connection to a write-ahead log (WAL) file. 
** There is one object of this type for each pager. 
*/

/* Write a frame or frames to the log. */
SQLITE_PRIVATE int sqlite3WalFrames(Wal *pWal, int, PgHdr *, Pgno, int, int);

/* Copy pages from the log to the database file */ 
SQLITE_PRIVATE int sqlite3WalCheckpoint(
  Wal *pWal,                      /* Write-ahead log connection */
  int eMode,                      /* One of PASSIVE, FULL and RESTART */
  int (*xBusy)(void*),            /* Function to call when busy */
  void *pBusyArg,                 /* Context argument for xBusyHandler */
  int sync_flags,                 /* Flags to sync db file with (or 0) */
  int nBuf,                       /* Size of buffer nBuf */
  u8 *zBuf,                       /* Temporary buffer to use */
  int *pnLog,                     /* OUT: Number of frames in WAL */
  int *pnCkpt                     /* OUT: Number of backfilled frames in WAL */
);

/* Return the value to pass to a sqlite3_wal_hook callback, the
** number of frames in the WAL at the point of the last commit since
** sqlite3WalCallback() was called.  If no commits have occurred since
** the last call, then return 0.
*/

  PAGER_INCR(pPager->nRead);
  IOTRACE(("PGIN %p %d\n", pPager, pgno));
  PAGERTRACE(("FETCH %d page %d hash(%08x)\n",
               PAGERID(pPager), pgno, pager_pagehash(pPg)));

  return rc;
}

/*
** Update the value of the change-counter at offsets 24 and 92 in
** the header and the sqlite version number at offset 96.
**
** This is an unconditional update.  See also the pager_incr_changecounter()
** routine which only updates the change-counter if the update is actually
** needed, as determined by the pPager->changeCountDone state variable.
*/
static void pager_write_changecounter(PgHdr *pPg){
  u32 change_counter;

  /* Increment the value just read and write it back to byte 24. */
  change_counter = sqlite3Get4byte((u8*)pPg->pPager->dbFileVers)+1;
  put32bits(((char*)pPg->pData)+24, change_counter);

  /* Also store the SQLite version number in bytes 96..99 and in
  ** bytes 92..95 store the change counter for which the version number
  ** is valid. */
  put32bits(((char*)pPg->pData)+92, change_counter);
  put32bits(((char*)pPg->pData)+96, SQLITE_VERSION_NUMBER);
}

#ifndef SQLITE_OMIT_WAL
/*
** This function is invoked once for each page that has already been 
** written into the log file when a WAL transaction is rolled back.
** Parameter iPg is the page number of said page. The pCtx argument 
** is actually a pointer to the Pager structure.

    rc = pagerUndoCallback((void *)pPager, pList->pgno);
    pList = pNext;
  }

  return rc;
}
























/*
** This function is a wrapper around sqlite3WalFrames(). As well as logging
** the contents of the list of pages headed by pList (connected by pDirty),
** this function notifies any active backup processes that the pages have
** changed. 
**
** The list of pages passed into this routine is always sorted by page number.
** Hence, if page 1 appears anywhere on the list, it will be the first page.
*/ 
static int pagerWalFrames(
  Pager *pPager,                  /* Pager object */
  PgHdr *pList,                   /* List of frames to log */

  assert( pPager->pWal );
#ifdef SQLITE_DEBUG
  /* Verify that the page list is in accending order */
  for(p=pList; p && p->pDirty; p=p->pDirty){
    assert( p->pgno < p->pDirty->pgno );
  }
#endif

  if( isCommit ){
    /* If a WAL transaction is being committed, there is no point in writing
    ** any pages with page numbers greater than nTruncate into the WAL file.
    ** They will never be read by any client. So remove them from the pDirty
    ** list here. */
    PgHdr *p;
    PgHdr **ppNext = &pList;
    for(p=pList; (*ppNext = p); p=p->pDirty){
      if( p->pgno<=nTruncate ) ppNext = &p->pDirty;
    }
    assert( pList );
  }

  if( pList->pgno==1 ) pager_write_changecounter(pList);
  rc = sqlite3WalFrames(pPager->pWal, 
      pPager->pageSize, pList, nTruncate, isCommit, syncFlags
  );
  if( rc==SQLITE_OK && pPager->pBackup ){
    PgHdr *p;
    for(p=pList; p; p=p->pDirty){
      sqlite3BackupUpdate(pPager->pBackup, p->pgno, (u8 *)p->pData);
    }
  }

#ifdef SQLITE_CHECK_PAGES
  pList = sqlite3PcacheDirtyList(pPager->pPCache);
  for(p=pList; p; p=p->pDirty){
    pager_set_pagehash(p);
  }
#endif

  return rc;
}

*/
SQLITE_PRIVATE sqlite3_backup **sqlite3PagerBackupPtr(Pager *pPager){
  return &pPager->pBackup;
}

#ifndef SQLITE_OMIT_WAL
/*
** This function is called when the user invokes "PRAGMA wal_checkpoint",
** "PRAGMA wal_blocking_checkpoint" or calls the sqlite3_wal_checkpoint()
** or wal_blocking_checkpoint() API functions.
**
** Parameter eMode is one of SQLITE_CHECKPOINT_PASSIVE, FULL or RESTART.
*/
SQLITE_PRIVATE int sqlite3PagerCheckpoint(Pager *pPager, int eMode, int *pnLog, int *pnCkpt){
  int rc = SQLITE_OK;
  if( pPager->pWal ){

    rc = sqlite3WalCheckpoint(pPager->pWal, eMode,
        pPager->xBusyHandler, pPager->pBusyHandlerArg,
        pPager->ckptSyncFlags, pPager->pageSize, (u8 *)pPager->pTmpSpace,
        pnLog, pnCkpt
    );
  }
  return rc;
}

SQLITE_PRIVATE int sqlite3PagerWalCallback(Pager *pPager){
  return sqlite3WalCallback(pPager->pWal);
}

*/
static int pagerExclusiveLock(Pager *pPager){
  int rc;                         /* Return code */

  assert( pPager->eLock==SHARED_LOCK || pPager->eLock==EXCLUSIVE_LOCK );
  rc = pagerLockDb(pPager, EXCLUSIVE_LOCK);
  if( rc!=SQLITE_OK ){
    /* If the attempt to grab the exclusive lock failed, release the 
    ** pending lock that may have been obtained instead.  */
    pagerUnlockDb(pPager, SHARED_LOCK);
  }

  return rc;
}

/*

  if( rc!=SQLITE_OK ){
    walIteratorFree(p);
  }
  *pp = p;
  return rc;
}

/*
** Attempt to obtain the exclusive WAL lock defined by parameters lockIdx and
** n. If the attempt fails and parameter xBusy is not NULL, then it is a
** busy-handler function. Invoke it and retry the lock until either the
** lock is successfully obtained or the busy-handler returns 0.
*/
static int walBusyLock(
  Wal *pWal,                      /* WAL connection */
  int (*xBusy)(void*),            /* Function to call when busy */
  void *pBusyArg,                 /* Context argument for xBusyHandler */
  int lockIdx,                    /* Offset of first byte to lock */
  int n                           /* Number of bytes to lock */
){
  int rc;
  do {
    rc = walLockExclusive(pWal, lockIdx, n);
  }while( xBusy && rc==SQLITE_BUSY && xBusy(pBusyArg) );
  return rc;
}

/*
** The cache of the wal-index header must be valid to call this function.
** Return the page-size in bytes used by the database.
*/
static int walPagesize(Wal *pWal){
  return (pWal->hdr.szPage&0xfe00) + ((pWal->hdr.szPage&0x0001)<<16);
}

/*
** Copy as much content as we can from the WAL back into the database file
** in response to an sqlite3_wal_checkpoint() request or the equivalent.
**
** The amount of information copies from WAL to database might be limited
** by active readers.  This routine will never overwrite a database page

**
** The caller must be holding sufficient locks to ensure that no other
** checkpoint is running (in any other thread or process) at the same
** time.
*/
static int walCheckpoint(
  Wal *pWal,                      /* Wal connection */
  int eMode,                      /* One of PASSIVE, FULL or RESTART */
  int (*xBusyCall)(void*),        /* Function to call when busy */
  void *pBusyArg,                 /* Context argument for xBusyHandler */
  int sync_flags,                 /* Flags for OsSync() (or 0) */

  u8 *zBuf                        /* Temporary buffer to use */
){
  int rc;                         /* Return code */
  int szPage;                     /* Database page-size */
  WalIterator *pIter = 0;         /* Wal iterator context */
  u32 iDbpage = 0;                /* Next database page to write */
  u32 iFrame = 0;                 /* Wal frame containing data for iDbpage */
  u32 mxSafeFrame;                /* Max frame that can be backfilled */
  u32 mxPage;                     /* Max database page to write */
  int i;                          /* Loop counter */
  volatile WalCkptInfo *pInfo;    /* The checkpoint status information */
  int (*xBusy)(void*) = 0;        /* Function to call when waiting for locks */

  szPage = walPagesize(pWal);
  testcase( szPage<=32768 );
  testcase( szPage>=65536 );
  pInfo = walCkptInfo(pWal);
  if( pInfo->nBackfill>=pWal->hdr.mxFrame ) return SQLITE_OK;

  /* Allocate the iterator */
  rc = walIteratorInit(pWal, &pIter);
  if( rc!=SQLITE_OK ){
    return rc;
  }
  assert( pIter );



  if( eMode!=SQLITE_CHECKPOINT_PASSIVE ) xBusy = xBusyCall;



  /* Compute in mxSafeFrame the index of the last frame of the WAL that is
  ** safe to write into the database.  Frames beyond mxSafeFrame might
  ** overwrite database pages that are in use by active readers and thus
  ** cannot be backfilled from the WAL.
  */
  mxSafeFrame = pWal->hdr.mxFrame;
  mxPage = pWal->hdr.nPage;
  for(i=1; i<WAL_NREADER; i++){
    u32 y = pInfo->aReadMark[i];
    if( mxSafeFrame>y ){
      assert( y<=pWal->hdr.mxFrame );
      rc = walBusyLock(pWal, xBusy, pBusyArg, WAL_READ_LOCK(i), 1);
      if( rc==SQLITE_OK ){
        pInfo->aReadMark[i] = READMARK_NOT_USED;
        walUnlockExclusive(pWal, WAL_READ_LOCK(i), 1);
      }else if( rc==SQLITE_BUSY ){
        mxSafeFrame = y;
        xBusy = 0;
      }else{
        goto walcheckpoint_out;
      }
    }
  }

  if( pInfo->nBackfill<mxSafeFrame
   && (rc = walBusyLock(pWal, xBusy, pBusyArg, WAL_READ_LOCK(0), 1))==SQLITE_OK
  ){
    i64 nSize;                    /* Current size of database file */
    u32 nBackfill = pInfo->nBackfill;

    /* Sync the WAL to disk */
    if( sync_flags ){
      rc = sqlite3OsSync(pWal->pWalFd, sync_flags);

      if( rc==SQLITE_OK ){
        pInfo->nBackfill = mxSafeFrame;
      }
    }

    /* Release the reader lock held while backfilling */
    walUnlockExclusive(pWal, WAL_READ_LOCK(0), 1);
  }

  if( rc==SQLITE_BUSY ){
    /* Reset the return code so as not to report a checkpoint failure
    ** just because there are active readers.  */
    rc = SQLITE_OK;
  }

  /* If this is an SQLITE_CHECKPOINT_RESTART operation, and the entire wal
  ** file has been copied into the database file, then block until all
  ** readers have finished using the wal file. This ensures that the next
  ** process to write to the database restarts the wal file.
  */
  if( rc==SQLITE_OK && eMode!=SQLITE_CHECKPOINT_PASSIVE ){
    assert( pWal->writeLock );
    if( pInfo->nBackfill<pWal->hdr.mxFrame ){
      rc = SQLITE_BUSY;
    }else if( eMode==SQLITE_CHECKPOINT_RESTART ){
      assert( mxSafeFrame==pWal->hdr.mxFrame );
      rc = walBusyLock(pWal, xBusy, pBusyArg, WAL_READ_LOCK(1), WAL_NREADER-1);
      if( rc==SQLITE_OK ){
        walUnlockExclusive(pWal, WAL_READ_LOCK(1), WAL_NREADER-1);
      }
    }
  }

 walcheckpoint_out:
  walIteratorFree(pIter);
  return rc;
}

/*

    ** The EXCLUSIVE lock is not released before returning.
    */
    rc = sqlite3OsLock(pWal->pDbFd, SQLITE_LOCK_EXCLUSIVE);
    if( rc==SQLITE_OK ){
      if( pWal->exclusiveMode==WAL_NORMAL_MODE ){
        pWal->exclusiveMode = WAL_EXCLUSIVE_MODE;
      }
      rc = sqlite3WalCheckpoint(
          pWal, SQLITE_CHECKPOINT_PASSIVE, 0, 0, sync_flags, nBuf, zBuf, 0, 0
      );
      if( rc==SQLITE_OK ){
        isDelete = 1;
      }
    }

    walIndexClose(pWal, isDelete);
    sqlite3OsClose(pWal->pWalFd);

  u32 mxReadMark;                 /* Largest aReadMark[] value */
  int mxI;                        /* Index of largest aReadMark[] value */
  int i;                          /* Loop counter */
  int rc = SQLITE_OK;             /* Return code  */

  assert( pWal->readLock<0 );     /* Not currently locked */

  /* Take steps to avoid spinning forever if there is a protocol error.
  **
  ** Circumstances that cause a RETRY should only last for the briefest
  ** instances of time.  No I/O or other system calls are done while the
  ** locks are held, so the locks should not be held for very long. But 
  ** if we are unlucky, another process that is holding a lock might get
  ** paged out or take a page-fault that is time-consuming to resolve, 
  ** during the few nanoseconds that it is holding the lock.  In that case,
  ** it might take longer than normal for the lock to free.
  **
  ** After 5 RETRYs, we begin calling sqlite3OsSleep().  The first few
  ** calls to sqlite3OsSleep() have a delay of 1 microsecond.  Really this
  ** is more of a scheduler yield than an actual delay.  But on the 10th
  ** an subsequent retries, the delays start becoming longer and longer, 
  ** so that on the 100th (and last) RETRY we delay for 21 milliseconds.
  ** The total delay time before giving up is less than 1 second.
  */
  if( cnt>5 ){
    int nDelay = 1;                      /* Pause time in microseconds */
    if( cnt>100 ){
      VVA_ONLY( pWal->lockError = 1; )
      return SQLITE_PROTOCOL;
    }
    if( cnt>=10 ) nDelay = (cnt-9)*238;  /* Max delay 21ms. Total delay 996ms */
    sqlite3OsSleep(pWal->pVfs, nDelay);
  }

  if( !useWal ){
    rc = walIndexReadHdr(pWal, pChanged);
    if( rc==SQLITE_BUSY ){
      /* If there is not a recovery running in another thread or process
      ** then convert BUSY errors to WAL_RETRY.  If recovery is known to

    u32 thisMark = pInfo->aReadMark[i];
    if( mxReadMark<=thisMark && thisMark<=pWal->hdr.mxFrame ){
      assert( thisMark!=READMARK_NOT_USED );
      mxReadMark = thisMark;
      mxI = i;
    }
  }
  /* There was once an "if" here. The extra "{" is to preserve indentation. */














  {
    if( mxReadMark < pWal->hdr.mxFrame || mxI==0 ){
      for(i=1; i<WAL_NREADER; i++){
        rc = walLockExclusive(pWal, WAL_READ_LOCK(i), 1);
        if( rc==SQLITE_OK ){
          mxReadMark = pInfo->aReadMark[i] = pWal->hdr.mxFrame;
          mxI = i;
          walUnlockExclusive(pWal, WAL_READ_LOCK(i), 1);
          break;
        }else if( rc!=SQLITE_BUSY ){
          return rc;
        }
      }
    }
    if( mxI==0 ){
      assert( rc==SQLITE_BUSY );
      return WAL_RETRY;
    }

    rc = walLockShared(pWal, WAL_READ_LOCK(mxI));
    if( rc ){
      return rc==SQLITE_BUSY ? WAL_RETRY : rc;
    }
    /* Now that the read-lock has been obtained, check that neither the
    ** value in the aReadMark[] array or the contents of the wal-index

SQLITE_PRIVATE int sqlite3WalBeginReadTransaction(Wal *pWal, int *pChanged){
  int rc;                         /* Return code */
  int cnt = 0;                    /* Number of TryBeginRead attempts */

  do{
    rc = walTryBeginRead(pWal, pChanged, 0, ++cnt);
  }while( rc==WAL_RETRY );
  testcase( (rc&0xff)==SQLITE_BUSY );
  testcase( (rc&0xff)==SQLITE_IOERR );
  testcase( rc==SQLITE_PROTOCOL );
  testcase( rc==SQLITE_OK );
  return rc;
}

/*
** Finish with a read transaction.  All this does is release the
** read-lock.
*/

  int rc = SQLITE_OK;
  int cnt;

  if( pWal->readLock==0 ){
    volatile WalCkptInfo *pInfo = walCkptInfo(pWal);
    assert( pInfo->nBackfill==pWal->hdr.mxFrame );
    if( pInfo->nBackfill>0 ){
      u32 salt1;
      sqlite3_randomness(4, &salt1);
      rc = walLockExclusive(pWal, WAL_READ_LOCK(1), WAL_NREADER-1);
      if( rc==SQLITE_OK ){
        /* If all readers are using WAL_READ_LOCK(0) (in other words if no
        ** readers are currently using the WAL), then the transactions
        ** frames will overwrite the start of the existing log. Update the
        ** wal-index header to reflect this.
        **
        ** In theory it would be Ok to update the cache of the header only
        ** at this point. But updating the actual wal-index header is also
        ** safe and means there is no special case for sqlite3WalUndo()
        ** to handle if this transaction is rolled back.
        */
        int i;                    /* Loop counter */
        u32 *aSalt = pWal->hdr.aSalt;       /* Big-endian salt values */
        pWal->nCkpt++;
        pWal->hdr.mxFrame = 0;
        sqlite3Put4byte((u8*)&aSalt[0], 1 + sqlite3Get4byte((u8*)&aSalt[0]));
        aSalt[1] = salt1;
        walIndexWriteHdr(pWal);
        pInfo->nBackfill = 0;
        for(i=1; i<WAL_NREADER; i++) pInfo->aReadMark[i] = READMARK_NOT_USED;
        assert( pInfo->aReadMark[0]==0 );
        walUnlockExclusive(pWal, WAL_READ_LOCK(1), WAL_NREADER-1);
      }else if( rc!=SQLITE_BUSY ){
        return rc;
      }
    }
    walUnlockShared(pWal, WAL_READ_LOCK(0));
    pWal->readLock = -1;
    cnt = 0;
    do{
      int notUsed;
      rc = walTryBeginRead(pWal, &notUsed, 1, ++cnt);
    }while( rc==WAL_RETRY );
    assert( (rc&0xff)!=SQLITE_BUSY ); /* BUSY not possible when useWal==1 */
    testcase( (rc&0xff)==SQLITE_IOERR );
    testcase( rc==SQLITE_PROTOCOL );
    testcase( rc==SQLITE_OK );
  }
  return rc;
}

/* 
** Write a set of frames to the log. The caller must hold the write-lock
** on the log file (obtained using sqlite3WalBeginWriteTransaction()).

/* 
** This routine is called to implement sqlite3_wal_checkpoint() and
** related interfaces.
**
** Obtain a CHECKPOINT lock and then backfill as much information as
** we can from WAL into the database.
**
** If parameter xBusy is not NULL, it is a pointer to a busy-handler
** callback. In this case this function runs a blocking checkpoint.
*/
SQLITE_PRIVATE int sqlite3WalCheckpoint(
  Wal *pWal,                      /* Wal connection */
  int eMode,                      /* PASSIVE, FULL or RESTART */
  int (*xBusy)(void*),            /* Function to call when busy */
  void *pBusyArg,                 /* Context argument for xBusyHandler */
  int sync_flags,                 /* Flags to sync db file with (or 0) */
  int nBuf,                       /* Size of temporary buffer */
  u8 *zBuf,                       /* Temporary buffer to use */
  int *pnLog,                     /* OUT: Number of frames in WAL */
  int *pnCkpt                     /* OUT: Number of backfilled frames in WAL */
){
  int rc;                         /* Return code */
  int isChanged = 0;              /* True if a new wal-index header is loaded */
  int eMode2 = eMode;             /* Mode to pass to walCheckpoint() */

  assert( pWal->ckptLock==0 );
  assert( pWal->writeLock==0 );

  WALTRACE(("WAL%p: checkpoint begins\n", pWal));
  rc = walLockExclusive(pWal, WAL_CKPT_LOCK, 1);
  if( rc ){
    /* Usually this is SQLITE_BUSY meaning that another thread or process
    ** is already running a checkpoint, or maybe a recovery.  But it might
    ** also be SQLITE_IOERR. */
    return rc;
  }
  pWal->ckptLock = 1;

  /* If this is a blocking-checkpoint, then obtain the write-lock as well
  ** to prevent any writers from running while the checkpoint is underway.
  ** This has to be done before the call to walIndexReadHdr() below.
  **
  ** If the writer lock cannot be obtained, then a passive checkpoint is
  ** run instead. Since the checkpointer is not holding the writer lock,
  ** there is no point in blocking waiting for any readers. Assuming no 
  ** other error occurs, this function will return SQLITE_BUSY to the caller.
  */
  if( eMode!=SQLITE_CHECKPOINT_PASSIVE ){
    rc = walBusyLock(pWal, xBusy, pBusyArg, WAL_WRITE_LOCK, 1);
    if( rc==SQLITE_OK ){
      pWal->writeLock = 1;
    }else if( rc==SQLITE_BUSY ){
      eMode2 = SQLITE_CHECKPOINT_PASSIVE;
      rc = SQLITE_OK;
    }
  }

  /* Read the wal-index header. */
  if( rc==SQLITE_OK ){
    rc = walIndexReadHdr(pWal, &isChanged);
  }

  /* Copy data from the log to the database file. */
  if( rc==SQLITE_OK ){
    if( pWal->hdr.mxFrame && walPagesize(pWal)!=nBuf ){
      rc = SQLITE_CORRUPT_BKPT;
    }else{
      rc = walCheckpoint(pWal, eMode2, xBusy, pBusyArg, sync_flags, zBuf);
    }

    /* If no error occurred, set the output variables. */
    if( rc==SQLITE_OK || rc==SQLITE_BUSY ){
      if( pnLog ) *pnLog = (int)pWal->hdr.mxFrame;
      if( pnCkpt ) *pnCkpt = (int)(walCkptInfo(pWal)->nBackfill);
    }
  }

  if( isChanged ){
    /* If a new wal-index header was loaded before the checkpoint was 
    ** performed, then the pager-cache associated with pWal is now
    ** out of date. So zero the cached wal-index header to ensure that
    ** next time the pager opens a snapshot on this database it knows that
    ** the cache needs to be reset.
    */
    memset(&pWal->hdr, 0, sizeof(WalIndexHdr));
  }

  /* Release the locks. */
  sqlite3WalEndWriteTransaction(pWal);
  walUnlockExclusive(pWal, WAL_CKPT_LOCK, 1);
  pWal->ckptLock = 0;
  WALTRACE(("WAL%p: checkpoint %s\n", pWal, rc ? "failed" : "ok"));
  return (rc==SQLITE_OK && eMode!=eMode2 ? SQLITE_BUSY : rc);
}

/* Return the value to pass to a sqlite3_wal_hook callback, the
** number of frames in the WAL at the point of the last commit since
** sqlite3WalCallback() was called.  If no commits have occurred since
** the last call, then return 0.
*/

**      *     zero or more pages numbers of leaves
*/


/* The following value is the maximum cell size assuming a maximum page
** size give above.
*/
#define MX_CELL_SIZE(pBt)  ((int)(pBt->pageSize-8))

/* The maximum number of cells on a single page of the database.  This
** assumes a minimum cell size of 6 bytes  (4 bytes for the cell itself
** plus 2 bytes for the index to the cell in the page header).  Such
** small cells will be rare, but they are possible.
*/
#define MX_CELL(pBt) ((pBt->pageSize-8)/6)

** points to the same BtShared object.  The database cache and the
** schema associated with the database file are all contained within
** the BtShared object.
**
** All fields in this structure are accessed under sqlite3.mutex.
** The pBt pointer itself may not be changed while there exists cursors 
** in the referenced BtShared that point back to this Btree since those
** cursors have to go through this Btree to find their BtShared and
** they often do so without holding sqlite3.mutex.
*/
struct Btree {
  sqlite3 *db;       /* The database connection holding this btree */
  BtShared *pBt;     /* Sharable content of this btree */
  u8 inTrans;        /* TRANS_NONE, TRANS_READ or TRANS_WRITE */
  u8 sharable;       /* True if we can share pBt with another db */

  u16 minLeaf;          /* Minimum local payload in a LEAFDATA table */
  u32 pageSize;         /* Total number of bytes on a page */
  u32 usableSize;       /* Number of usable bytes on each page */
  int nTransaction;     /* Number of open transactions (read + write) */
  u32 nPage;            /* Number of pages in the database */
  void *pSchema;        /* Pointer to space allocated by sqlite3BtreeSchema() */
  void (*xFreeSchema)(void*);  /* Destructor for BtShared.pSchema */
  sqlite3_mutex *mutex; /* Non-recursive mutex required to access this object */
  Bitvec *pHasContent;  /* Set of pages moved to free-list this transaction */
#ifndef SQLITE_OMIT_SHARED_CACHE
  int nRef;             /* Number of references to this structure */
  BtShared *pNext;      /* Next on a list of sharable BtShared structs */
  BtLock *pLock;        /* List of locks held on this shared-btree struct */
  Btree *pWriter;       /* Btree with currently open write transaction */
  u8 isExclusive;       /* True if pWriter has an EXCLUSIVE lock on the db */

}

/*
** Release the BtShared mutex associated with B-Tree handle p and
** clear the p->locked boolean.
*/
static void unlockBtreeMutex(Btree *p){
  BtShared *pBt = p->pBt;
  assert( p->locked==1 );
  assert( sqlite3_mutex_held(pBt->mutex) );
  assert( sqlite3_mutex_held(p->db->mutex) );
  assert( p->db==pBt->db );

  sqlite3_mutex_leave(pBt->mutex);
  p->locked = 0;
}

/*
** Enter a mutex on the given BTree object.
**
** If the object is not sharable, then no mutex is ever required

** Enter the mutexes in accending order by BtShared pointer address
** to avoid the possibility of deadlock when two threads with
** two or more btrees in common both try to lock all their btrees
** at the same instant.
*/
SQLITE_PRIVATE void sqlite3BtreeEnterAll(sqlite3 *db){
  int i;
  Btree *p;
  assert( sqlite3_mutex_held(db->mutex) );
  for(i=0; i<db->nDb; i++){
    p = db->aDb[i].pBt;














    if( p ) sqlite3BtreeEnter(p);





  }
}
SQLITE_PRIVATE void sqlite3BtreeLeaveAll(sqlite3 *db){
  int i;
  Btree *p;
  assert( sqlite3_mutex_held(db->mutex) );
  for(i=0; i<db->nDb; i++){
    p = db->aDb[i].pBt;
    if( p ) sqlite3BtreeLeave(p);




  }
}

/*
** Return true if a particular Btree requires a lock.  Return FALSE if
** no lock is ever required since it is not sharable.
*/
SQLITE_PRIVATE int sqlite3BtreeSharable(Btree *p){
  return p->sharable;
}

#ifndef NDEBUG
/*
** Return true if the current thread holds the database connection
** mutex and all required BtShared mutexes.
**

      return 0;
    }
  }
  return 1;
}
#endif /* NDEBUG */


















#ifndef NDEBUG

/*
** Return true if the correct mutexes are held for accessing the
** db->aDb[iDb].pSchema structure.  The mutexes required for schema





** access are:







**
**   (1) The mutex on db





**   (2) if iDb!=1, then the mutex on db->aDb[iDb].pBt.
**

** If pSchema is not NULL, then iDb is computed from pSchema and
** db using sqlite3SchemaToIndex().

*/
SQLITE_PRIVATE int sqlite3SchemaMutexHeld(sqlite3 *db, int iDb, Schema *pSchema){


  Btree *p;

  assert( db!=0 );
  if( pSchema ) iDb = sqlite3SchemaToIndex(db, pSchema);
  assert( iDb>=0 && iDb<db->nDb );
  if( !sqlite3_mutex_held(db->mutex) ) return 0;
  if( iDb==1 ) return 1;
  p = db->aDb[iDb].pBt;
  assert( p!=0 );
  return p->sharable==0 || p->locked==1;
}
#endif /* NDEBUG */






#else /* SQLITE_THREADSAFE>0 above.  SQLITE_THREADSAFE==0 below */







/*
** The following are special cases for mutex enter routines for use
** in single threaded applications that use shared cache.  Except for
** these two routines, all mutex operations are no-ops in that case and
** are null #defines in btree.h.
**



** If shared cache is disabled, then all btree mutex routines, including




** the ones below, are no-ops and are null #defines in btree.h.


*/








SQLITE_PRIVATE void sqlite3BtreeEnter(Btree *p){
  p->pBt->db = p->db;
}
SQLITE_PRIVATE void sqlite3BtreeEnterAll(sqlite3 *db){
  int i;
  for(i=0; i<db->nDb; i++){
    Btree *p = db->aDb[i].pBt;

  ** and the cell content area.  The btreeInitPage() call has already
  ** validated the freelist.  Given that the freelist is valid, there
  ** is no way that the allocation can extend off the end of the page.
  ** The assert() below verifies the previous sentence.
  */
  top -= nByte;
  put2byte(&data[hdr+5], top);
  assert( top+nByte <= (int)pPage->pBt->usableSize );
  *pIdx = top;
  return SQLITE_OK;
}

/*
** Return a section of the pPage->aData to the freelist.
** The first byte of the new free block is pPage->aDisk[start]
** and the size of the block is "size" bytes.
**
** Most of the effort here is involved in coalesing adjacent
** free blocks into a single big free block.
*/
static int freeSpace(MemPage *pPage, int start, int size){
  int addr, pbegin, hdr;
  int iLast;                        /* Largest possible freeblock offset */
  unsigned char *data = pPage->aData;

  assert( pPage->pBt!=0 );
  assert( sqlite3PagerIswriteable(pPage->pDbPage) );
  assert( start>=pPage->hdrOffset+6+pPage->childPtrSize );
  assert( (start + size) <= (int)pPage->pBt->usableSize );
  assert( sqlite3_mutex_held(pPage->pBt->mutex) );
  assert( size>=0 );   /* Minimum cell size is 4 */

  if( pPage->pBt->secureDelete ){
    /* Overwrite deleted information with zeros when the secure_delete
    ** option is enabled */
    memset(&data[start], 0, size);

  pPage->nFree = pPage->nFree + (u16)size;

  /* Coalesce adjacent free blocks */
  addr = hdr + 1;
  while( (pbegin = get2byte(&data[addr]))>0 ){
    int pnext, psize, x;
    assert( pbegin>addr );
    assert( pbegin <= (int)pPage->pBt->usableSize-4 );
    pnext = get2byte(&data[pbegin]);
    psize = get2byte(&data[pbegin+2]);
    if( pbegin + psize + 3 >= pnext && pnext>0 ){
      int frag = pnext - (pbegin+psize);
      if( (frag<0) || (frag>(int)data[hdr+7]) ){
        return SQLITE_CORRUPT_BKPT;
      }

  sqlite3BtreeEnter(p);
  assert( pBt && pBt->pPager );
  rc = sqlite3PagerNosync(pBt->pPager);
  sqlite3BtreeLeave(p);
  return rc;
}


/*
** Change the default pages size and the number of reserved bytes per page.
** Or, if the page size has already been fixed, return SQLITE_READONLY 
** without changing anything.
**
** The page size must be a power of 2 between 512 and 65536.  If the page
** size supplied does not meet this constraint then the page size is not

/*
** Return the currently defined page size
*/
SQLITE_PRIVATE int sqlite3BtreeGetPageSize(Btree *p){
  return p->pBt->pageSize;
}

#if !defined(SQLITE_OMIT_PAGER_PRAGMAS) || !defined(SQLITE_OMIT_VACUUM)
/*
** Return the number of bytes of space at the end of every page that
** are intentually left unused.  This is the "reserved" space that is
** sometimes used by extensions.
*/
SQLITE_PRIVATE int sqlite3BtreeGetReserve(Btree *p){
  int n;