SQLite

#define _SQLITE_OS_C_ 1
#include "sqliteInt.h"
#include "os.h"
#undef _SQLITE_OS_C_

/*
** The following routines are convenience wrappers around methods
** of the sqlite3_file object.  This is mostly just syntactic sugar. All
** of this would be completely automatic if SQLite were coded using
** C++ instead of plain old C.
*/
int sqlite3OsClose(sqlite3_file **pId){
  int rc = SQLITE_OK;
  sqlite3_file *id;
  if( pId!=0 && (id = *pId)!=0 ){
    rc = id->pMethods->xClose(id);

    if( rc==SQLITE_OK ){
      *pId = 0;
    }
  }

  return rc;
}
int sqlite3OsRead(sqlite3_file *id, void *pBuf, int amt, i64 offset){
  return id->pMethods->xRead(id, pBuf, amt, offset);
}
int sqlite3OsWrite(sqlite3_file *id, const void *pBuf, int amt, i64 offset){
  return id->pMethods->xWrite(id, pBuf, amt, offset);
}
int sqlite3OsTruncate(sqlite3_file *id, i64 size){
  return id->pMethods->xTruncate(id, size);
}
int sqlite3OsSync(sqlite3_file *id, int fullsync){
  return id->pMethods->xSync(id, fullsync);
}
int sqlite3OsFileSize(sqlite3_file *id, i64 *pSize){
  return id->pMethods->xFileSize(id, pSize);
}



int sqlite3OsLock(sqlite3_file *id, int lockType){
  return id->pMethods->xLock(id, lockType);
}
int sqlite3OsUnlock(sqlite3_file *id, int lockType){
  return id->pMethods->xUnlock(id, lockType);
}
int sqlite3OsBreakLock(sqlite3_file *id){
  return id->pMethods->xBreakLock(id);
}
int sqlite3OsCheckReservedLock(sqlite3_file *id){
  return id->pMethods->xCheckReservedLock(id);
}
int sqlite3OsSectorSize(sqlite3_file *id){
  int (*xSectorSize)(sqlite3_file*) = id->pMethods->xSectorSize;
  return xSectorSize ? xSectorSize(id) : SQLITE_DEFAULT_SECTOR_SIZE;
}
int sqlite3OsDeviceCharacteristics(sqlite3_file *id){
  return id->pMethods->xDeviceCharacteristics(id);
}

#if defined(SQLITE_TEST) || defined(SQLITE_DEBUG)
  /* These methods are currently only used for testing and debugging. */
  int sqlite3OsFileHandle(sqlite3_file *id){
    /* return id->pMethods->xFileHandle(id); */
    return 0;
  }
  int sqlite3OsLockState(sqlite3_file *id){
    /* return id->pMethods->xLockState(id); */
    return 0;
  }
#endif

#ifdef SQLITE_ENABLE_REDEF_IO
/*
** A function to return a pointer to the virtual function table.
** This routine really does not accomplish very much since the

#define RESERVED_BYTE     (PENDING_BYTE+1)
#define SHARED_FIRST      (PENDING_BYTE+2)
#define SHARED_SIZE       510

/*
** Prototypes for operating system interface routines.
*/
int sqlite3OsClose(sqlite3_file**);

int sqlite3OsRead(sqlite3_file*, void*, int amt, i64 offset);
int sqlite3OsWrite(sqlite3_file*, const void*, int amt, i64 offset);

int sqlite3OsTruncate(sqlite3_file*, i64 size);
int sqlite3OsSync(sqlite3_file*, int);

int sqlite3OsFileSize(sqlite3_file*, i64 *pSize);
int sqlite3OsLock(sqlite3_file*, int);
int sqlite3OsUnlock(sqlite3_file*, int);
int sqlite3OsBreakLock(sqlite3_file*);
int sqlite3OsCheckReservedLock(sqlite3_file *id);
int sqlite3OsSectorSize(sqlite3_file *id);
int sqlite3OsDeviceCharacteristics(sqlite3_file *id);

int sqlite3OsOpenReadWrite(const char*, sqlite3_file**, int*);
int sqlite3OsOpenExclusive(const char*, sqlite3_file**, int);
int sqlite3OsOpenReadOnly(const char*, sqlite3_file**);

int sqlite3OsDelete(const char*);
int sqlite3OsFileExists(const char*);
char *sqlite3OsFullPathname(const char*);
int sqlite3OsIsDirWritable(char*);
int sqlite3OsSyncDirectory(const char*);

int sqlite3OsTempFileName(char*);
int sqlite3OsRandomSeed(char*);
int sqlite3OsSleep(int ms);
int sqlite3OsCurrentTime(double*);
void sqlite3OsEnterMutex(void);
void sqlite3OsLeaveMutex(void);
int sqlite3OsInMutex(int);
ThreadData *sqlite3OsThreadSpecificData(int);
void *sqlite3OsMalloc(int);
void *sqlite3OsRealloc(void *, int);
void sqlite3OsFree(void *);
int sqlite3OsAllocationSize(void *);
void *sqlite3OsDlopen(const char*);
void *sqlite3OsDlsym(void*, const char*);
int sqlite3OsDlclose(void*);

#if defined(SQLITE_TEST) || defined(SQLITE_DEBUG)
  int sqlite3OsFileHandle(sqlite3_file *id);
  int sqlite3OsLockState(sqlite3_file *id);
#endif

/*
** If the SQLITE_ENABLE_REDEF_IO macro is defined, then the OS-layer
** interface routines are not called directly but are invoked using
** pointers to functions.  This allows the implementation of various
** OS-layer interface routines to be modified at run-time.  There are

** When redefinable I/O is enabled, a single global instance of the
** following structure holds pointers to the routines that SQLite 
** uses to talk with the underlying operating system.  Modify this
** structure (before using any SQLite API!) to accomodate perculiar
** operating system interfaces or behaviors.
*/
struct sqlite3OsVtbl {
  int (*xOpenReadWrite)(const char*, sqlite3_file**, int*);
  int (*xOpenExclusive)(const char*, sqlite3_file**, int);
  int (*xOpenReadOnly)(const char*, sqlite3_file**);

  int (*xDelete)(const char*);
  int (*xFileExists)(const char*);
  char *(*xFullPathname)(const char*);
  int (*xIsDirWritable)(char*);
  int (*xSyncDirectory)(const char*);
  int (*xTempFileName)(char*);

/*
** The unixFile structure is subclass of OsFile specific for the unix
** protability layer.
*/
typedef struct unixFile unixFile;
struct unixFile {
  sqlite3_io_methods const *pMethod;  /* Always the first entry */
  struct openCnt *pOpen;    /* Info about all open fd's on this inode */
  struct lockInfo *pLock;   /* Info about locks on this inode */
#ifdef SQLITE_ENABLE_LOCKING_STYLE
  void *lockingContext;     /* Locking style specific state */
#endif /* SQLITE_ENABLE_LOCKING_STYLE */
  int h;                    /* The file descriptor */
  unsigned char locktype;   /* The type of lock held on this fd */
  unsigned char isOpen;     /* True if needs to be closed */
  unsigned char fullSync;   /* Use F_FULLSYNC if available */
  int dirfd;                /* File descriptor for the directory */
  i64 offset;               /* Seek offset */
#ifdef SQLITE_UNIX_THREADS
  pthread_t tid;            /* The thread that "owns" this OsFile */
#endif
};


/*
** Provide the ability to override some OS-layer functions during
** testing.  This is used to simulate OS crashes to verify that 
** commits are atomic even in the event of an OS crash.
*/
#ifdef SQLITE_CRASH_TEST
  extern int sqlite3CrashTestEnable;
  int sqlite3CrashFileWrap(sqlite3_file *, const char *, sqlite3_file **);


  static int CRASH_TEST_OVERRIDE(const char *zFile, sqlite3_file **pId, int rc){
    if( rc==SQLITE_OK && sqlite3CrashTestEnable ){
      rc = sqlite3CrashFileWrap(*pId, zFile, pId); 
    }
    return rc;
  }
#else
# define CRASH_TEST_OVERRIDE(A,B,C) C
#endif


/*
** Include code that is common to all os_*.c files
*/
#include "os_common.h"

int sqlite3UnixFileExists(const char *zFilename){
  return access(zFilename, 0)==0;
}

/* Forward declaration */
static int allocateUnixFile(
  int h,                    /* File descriptor of the open file */
  sqlite3_file **pId,       /* Write the real file descriptor here */
  const char *zFilename,    /* Name of the file being opened */
  int delFlag               /* If true, make sure the file deletes on close */
);

/*
** Attempt to open a file for both reading and writing.  If that
** fails, try opening it read-only.  If the file does not exist,
** try to create it.
**
** On success, a handle for the open file is written to *id
** and *pReadonly is set to 0 if the file was opened for reading and
** writing or 1 if the file was opened read-only.  The function returns
** SQLITE_OK.
**
** On failure, the function returns SQLITE_CANTOPEN and leaves
** *id and *pReadonly unchanged.
*/
int sqlite3UnixOpenReadWrite(
  const char *zFilename,
  sqlite3_file **pId,
  int *pReadonly
){
  int h;
  

  assert( 0==*pId );
  h = open(zFilename, O_RDWR|O_CREAT|O_LARGEFILE|O_BINARY,
                        SQLITE_DEFAULT_FILE_PERMISSIONS);
  if( h<0 ){
#ifdef EISDIR
    if( errno==EISDIR ){
      return SQLITE_CANTOPEN;
    }
#endif
    h = open(zFilename, O_RDONLY|O_LARGEFILE|O_BINARY);
    if( h<0 ){
      return SQLITE_CANTOPEN; 
    }
    *pReadonly = 1;
  }else{
    *pReadonly = 0;
  }

  return CRASH_TEST_OVERRIDE(
    zFilename, pId, allocateUnixFile(h, pId, zFilename, 0)
  );
}


/*
** Attempt to open a new file for exclusive access by this process.
** The file will be opened for both reading and writing.  To avoid
** a potential security problem, we do not allow the file to have
** previously existed.  Nor do we allow the file to be a symbolic
** link.
**
** If delFlag is true, then make arrangements to automatically delete
** the file when it is closed.
**
** On success, write the file handle into *id and return SQLITE_OK.
**
** On failure, return SQLITE_CANTOPEN.
*/
int sqlite3UnixOpenExclusive(
  const char *zFilename, 
  sqlite3_file **pId, 
  int delFlag
){
  int h;


  assert( 0==*pId );
  h = open(zFilename,
                O_RDWR|O_CREAT|O_EXCL|O_NOFOLLOW|O_LARGEFILE|O_BINARY,
                delFlag ? 0600 : SQLITE_DEFAULT_FILE_PERMISSIONS);
  if( h<0 ){
    return SQLITE_CANTOPEN;
  }
  return CRASH_TEST_OVERRIDE(
    zFilename, pId, allocateUnixFile(h, pId, zFilename, delFlag)
  );
}

/*
** Attempt to open a new file for read-only access.
**
** On success, write the file handle into *id and return SQLITE_OK.
**
** On failure, return SQLITE_CANTOPEN.
*/
int sqlite3UnixOpenReadOnly(const char *zFilename, sqlite3_file **pId){
  int h;
  

  assert( 0==*pId );
  h = open(zFilename, O_RDONLY|O_LARGEFILE|O_BINARY);
  if( h<0 ){
    return SQLITE_CANTOPEN;
  }
  return CRASH_TEST_OVERRIDE(
    zFilename, pId, allocateUnixFile(h, pId, zFilename, 0)
  );
}

/*
** Attempt to open a file descriptor for the directory that contains a
** file.  This file descriptor can be used to fsync() the directory
** in order to make sure the creation of a new file is actually written
** to disk.

  return 1;
}

/*
** Seek to the offset in id->offset then read cnt bytes into pBuf.
** Return the number of bytes actually read.  Update the offset.
*/
static int seekAndRead(unixFile *id, sqlite3_int64 offset, void *pBuf, int cnt){
  int got;
  i64 newOffset;
  TIMER_START;
#if defined(USE_PREAD)
  got = pread(id->h, pBuf, cnt, offset);
  SimulateIOError( got = -1 );
#elif defined(USE_PREAD64)
  got = pread64(id->h, pBuf, cnt, offset);
  SimulateIOError( got = -1 );
#else
  newOffset = lseek(id->h, offset, SEEK_SET);
  SimulateIOError( newOffset-- );
  if( newOffset!=offset ){
    return -1;
  }
  got = read(id->h, pBuf, cnt);
#endif
  TIMER_END;
  OSTRACE5("READ    %-3d %5d %7lld %d\n", id->h, got, id->offset, TIMER_ELAPSED);



  return got;
}

/*
** Read data from a file into a buffer.  Return SQLITE_OK if all
** bytes were read successfully and SQLITE_IOERR if anything goes
** wrong.
*/
static int unixRead(
  sqlite3_file *id, 
  void *pBuf, 
  int amt,
  sqlite3_int64 offset
){
  int got;
  assert( id );
  got = seekAndRead((unixFile*)id, offset, pBuf, amt);
  if( got==amt ){
    return SQLITE_OK;
  }else if( got<0 ){
    return SQLITE_IOERR_READ;
  }else{
    memset(&((char*)pBuf)[got], 0, amt-got);
    return SQLITE_IOERR_SHORT_READ;
  }
}

/*
** Seek to the offset in id->offset then read cnt bytes into pBuf.
** Return the number of bytes actually read.  Update the offset.
*/
static int seekAndWrite(unixFile *id, i64 offset, const void *pBuf, int cnt){
  int got;
  i64 newOffset;
  TIMER_START;
#if defined(USE_PREAD)
  got = pwrite(id->h, pBuf, cnt, offset);
#elif defined(USE_PREAD64)
  got = pwrite64(id->h, pBuf, cnt, offset);
#else
  newOffset = lseek(id->h, offset, SEEK_SET);
  if( newOffset!=offset ){
    return -1;
  }
  got = write(id->h, pBuf, cnt);
#endif
  TIMER_END;
  OSTRACE5("WRITE   %-3d %5d %7lld %d\n", id->h, got, offset, TIMER_ELAPSED);



  return got;
}


/*
** Write data from a buffer into a file.  Return SQLITE_OK on success
** or some other error code on failure.
*/
static int unixWrite(
  sqlite3_file *id, 
  const void *pBuf, 
  int amt,
  sqlite3_int64 offset 
){
  int wrote = 0;
  assert( id );
  assert( amt>0 );
  while( amt>0 && (wrote = seekAndWrite((unixFile*)id, offset, pBuf, amt))>0 ){
    amt -= wrote;
    offset += wrote;
    pBuf = &((char*)pBuf)[wrote];
  }
  SimulateIOError(( wrote=(-1), amt=1 ));
  SimulateDiskfullError(( wrote=0, amt=1 ));
  if( amt>0 ){
    if( wrote<0 ){
      return SQLITE_IOERR_WRITE;

** has been created by fsync-ing the directory that contains the file.
** If we do not do this and we encounter a power failure, the directory
** entry for the journal might not exist after we reboot.  The next
** SQLite to access the file will not know that the journal exists (because
** the directory entry for the journal was never created) and the transaction
** will not roll back - possibly leading to database corruption.
*/
static int unixSync(sqlite3_file *id, int dataOnly){
  int rc;
  unixFile *pFile = (unixFile*)id;
  assert( pFile );
  OSTRACE2("SYNC    %-3d\n", pFile->h);
  rc = full_fsync(pFile->h, pFile->fullSync, dataOnly);
  SimulateIOError( rc=1 );
  if( rc ){

  }
#endif
}

/*
** Truncate an open file to a specified size
*/
static int unixTruncate(sqlite3_file *id, i64 nByte){
  int rc;
  assert( id );
  rc = ftruncate(((unixFile*)id)->h, (off_t)nByte);
  SimulateIOError( rc=1 );
  if( rc ){
    return SQLITE_IOERR_TRUNCATE;
  }else{
    return SQLITE_OK;
  }
}

/*
** 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 = fstat(((unixFile*)id)->h, &buf);
  SimulateIOError( rc=1 );
  if( rc!=0 ){
    return SQLITE_IOERR_FSTAT;
  }
  *pSize = buf.st_size;
  return SQLITE_OK;
}

/*
** 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, return
** non-zero.  If the file is unlocked or holds only SHARED locks, then
** return zero.
*/
static int unixCheckReservedLock(sqlite3_file *id){
  int r = 0;
  unixFile *pFile = (unixFile*)id;

  assert( pFile );
  sqlite3OsEnterMutex(); /* Because pFile->pLock is shared across threads */

  /* Check if a thread in this process holds such a lock */

**    SHARED -> (PENDING) -> EXCLUSIVE
**    RESERVED -> (PENDING) -> EXCLUSIVE
**    PENDING -> EXCLUSIVE
**
** This routine will only increase a lock.  Use the sqlite3OsUnlock()
** routine to lower a locking level.
*/
static int unixLock(sqlite3_file *id, int locktype){
  /* The following describes the implementation of the various locks and
  ** lock transitions in terms of the POSIX advisory shared and exclusive
  ** lock primitives (called read-locks and write-locks below, to avoid
  ** confusion with SQLite lock names). The algorithms are complicated
  ** slightly in order to be compatible with windows systems simultaneously
  ** accessing the same database file, in case that is ever required.
  **

/*
** Lower the locking level on file descriptor pFile to locktype.  locktype
** 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 locktype){
  struct lockInfo *pLock;
  struct flock lock;
  int rc = SQLITE_OK;
  unixFile *pFile = (unixFile*)id;

  assert( pFile );
  OSTRACE7("UNLOCK  %d %d was %d(%d,%d) pid=%d\n", pFile->h, locktype,

  pFile->locktype = locktype;
  return rc;
}

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

  if( !pFile ) return SQLITE_OK;
  unixUnlock(id, NO_LOCK);
  if( pFile->dirfd>=0 ) close(pFile->dirfd);
  pFile->dirfd = -1;
  sqlite3OsEnterMutex();

  if( pFile->pOpen->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 pOpen->aPending.  It will be automatically closed when
    ** the last lock is cleared.
    */
    int *aNew;
    struct openCnt *pOpen = pFile->pOpen;
    aNew = realloc( pOpen->aPending, (pOpen->nPending+1)*sizeof(int) );
    if( aNew==0 ){
      /* If a malloc fails, just leak the file descriptor */
    }else{
      pOpen->aPending = aNew;
      pOpen->aPending[pOpen->nPending] = pFile->h;
      pOpen->nPending++;
    }
  }else{
    /* There are no outstanding locks so we can close the file immediately */
    close(pFile->h);
  }
  releaseLockInfo(pFile->pLock);
  releaseOpenCnt(pFile->pOpen);

  sqlite3OsLeaveMutex();
  pFile->isOpen = 0;
  OSTRACE2("CLOSE   %-3d\n", pFile->h);
  OpenCounter(-1);
  sqlite3ThreadSafeFree(id);

  return SQLITE_OK;
}


#ifdef SQLITE_ENABLE_LOCKING_STYLE
#pragma mark AFP Support

** 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 it's journal file) that the sector size will be the
** same for both.
*/
static int unixSectorSize(sqlite3_file *id){
  return SQLITE_DEFAULT_SECTOR_SIZE;
}

static int unixDeviceCharacteristics(sqlite3_file *id){
  return 0;
}

static int unixBreakLock(sqlite3_file *id){
  assert(!"TODO: unixBreakLock()");
  return 0;
}

/*
** This vector defines all the methods that can operate on an OsFile
** for unix.
*/
static const sqlite3_io_methods sqlite3UnixIoMethod = {
  1,                        /* iVersion */
  unixClose,

  unixRead,
  unixWrite,

  unixTruncate,
  unixSync,


  unixFileSize,
  unixLock,
  unixUnlock,

  unixCheckReservedLock,
  unixBreakLock,
  unixSectorSize,
  unixDeviceCharacteristics
};

#ifdef SQLITE_ENABLE_LOCKING_STYLE
/*
 ** This vector defines all the methods that can operate on an OsFile
 ** for unix with AFP style file locking.
 */

    OpenCounter(+1);
    return SQLITE_OK;
  }
}
#else /* SQLITE_ENABLE_LOCKING_STYLE */
static int allocateUnixFile(
  int h,                 /* Open file descriptor on file being opened */
  sqlite3_file **pId,    /* Write the resul unixFile structure here */
  const char *zFilename, /* Name of the file being opened */
  int delFlag            /* If true, delete the file on or before closing */
){
  unixFile *pNew;
  unixFile f;
  int rc;

    releaseOpenCnt(f.pOpen);
    sqlite3OsLeaveMutex();
    *pId = 0;
    return SQLITE_NOMEM;
  }else{
    *pNew = f;
    pNew->pMethod = &sqlite3UnixIoMethod;
    *pId = (sqlite3_file*)pNew;
    OpenCounter(+1);
    return SQLITE_OK;
  }
}
#endif /* SQLITE_ENABLE_LOCKING_STYLE */

#endif /* SQLITE_OMIT_DISKIO */

** The pager is used to access a database disk file.  It implements
** atomic commit and rollback through the use of a journal file that
** is separate from the database file.  The pager also implements file
** locking to prevent two processes from writing the same database
** file simultaneously, or one process from reading the database while
** another is writing.
**
** @(#) $Id: pager.c,v 1.356 2007/08/15 17:08:46 danielk1977 Exp $
*/
#ifndef SQLITE_OMIT_DISKIO
#include "sqliteInt.h"
#include "os.h"
#include "pager.h"
#include <assert.h>
#include <string.h>

#endif

/*
** The following two macros are used within the PAGERTRACEX() macros above
** to print out file-descriptors. 
**
** PAGERID() takes a pointer to a Pager struct as it's argument. The
** associated file-descriptor is returned. FILEHANDLEID() takes an sqlite3_file
** struct as it's argument.
*/
#define PAGERID(p) ((int)(p->fd))
#define FILEHANDLEID(fd) ((int)fd)

/*
** The page cache as a whole is always in one of the following

  int mxPage;                 /* Maximum number of pages to hold in cache */
  Pgno mxPgno;                /* Maximum allowed size of the database */
  u8 *aInJournal;             /* One bit for each page in the database file */
  u8 *aInStmt;                /* One bit for each page in the database */
  char *zFilename;            /* Name of the database file */
  char *zJournal;             /* Name of the journal file */
  char *zDirectory;           /* Directory hold database and journal files */
  sqlite3_file *fd, *jfd;     /* File descriptors for database and journal */
  sqlite3_file *stfd;         /* File descriptor for the statement subjournal*/
  BusyHandler *pBusyHandler;  /* Pointer to sqlite.busyHandler */
  PgHdr *pFirst, *pLast;      /* List of free pages */
  PgHdr *pFirstSynced;        /* First free page with PgHdr.needSync==0 */
  PgHdr *pAll;                /* List of all pages */
  PgHdr *pStmt;               /* List of pages in the statement subjournal */
  PgHdr *pDirty;              /* List of all dirty pages */
  i64 journalOff;             /* Current byte offset in the journal file */

/*
** Read a 32-bit integer from the given file descriptor.  Store the integer
** that is read in *pRes.  Return SQLITE_OK if everything worked, or an
** error code is something goes wrong.
**
** All values are stored on disk as big-endian.
*/
static int read32bits(sqlite3_file *fd, i64 offset, u32 *pRes){
  unsigned char ac[4];
  int rc = sqlite3OsRead(fd, ac, sizeof(ac), offset);
  if( rc==SQLITE_OK ){
    *pRes = sqlite3Get4byte(ac);
  }
  return rc;
}

/*
** Write a 32-bit integer into a string buffer in big-endian byte order.
*/
#define put32bits(A,B)  sqlite3Put4byte((u8*)A,B)

/*
** Write a 32-bit integer into the given file descriptor.  Return SQLITE_OK
** on success or an error code is something goes wrong.
*/
static int write32bits(sqlite3_file *fd, i64 offset, u32 val){
  char ac[4];
  put32bits(ac, val);
  return sqlite3OsWrite(fd, ac, 4, offset);
}

/*
** This function should be called when an error occurs within the pager
** code. The first argument is a pointer to the pager structure, the
** second the error-code about to be returned by a pager API function. 
** The value returned is a copy of the second argument to this function. 

** written into memory obtained from sqliteMalloc(). *pzMaster is
** set to point at the memory and SQLITE_OK returned. The caller must
** sqliteFree() *pzMaster.
**
** If no master journal file name is present *pzMaster is set to 0 and
** SQLITE_OK returned.
*/
static int readMasterJournal(sqlite3_file *pJrnl, char **pzMaster){
  int rc;
  u32 len;
  i64 szJ;
  u32 cksum;
  int i;
  unsigned char aMagic[8]; /* A buffer to hold the magic header */

  *pzMaster = 0;

  rc = sqlite3OsFileSize(pJrnl, &szJ);
  if( rc!=SQLITE_OK || szJ<16 ) return rc;




  rc = read32bits(pJrnl, szJ-16, &len);
  if( rc!=SQLITE_OK ) return rc;

  rc = read32bits(pJrnl, szJ-12, &cksum);
  if( rc!=SQLITE_OK ) return rc;

  rc = sqlite3OsRead(pJrnl, aMagic, 8, szJ-8);
  if( rc!=SQLITE_OK || memcmp(aMagic, aJournalMagic, 8) ) return rc;




  *pzMaster = (char *)sqliteMalloc(len+1);
  if( !*pzMaster ){
    return SQLITE_NOMEM;
  }
  rc = sqlite3OsRead(pJrnl, *pzMaster, len, szJ-16-len);
  if( rc!=SQLITE_OK ){
    sqliteFree(*pzMaster);
    *pzMaster = 0;
    return rc;
  }

  /* See if the checksum matches the master journal name */

** ---------------------------------------
** 0                         0
** 512                       512
** 100                       512
** 2000                      2048
** 
*/
static void seekJournalHdr(Pager *pPager){
  i64 offset = 0;
  i64 c = pPager->journalOff;
  if( c ){
    offset = ((c-1)/JOURNAL_HDR_SZ(pPager) + 1) * JOURNAL_HDR_SZ(pPager);
  }
  assert( offset%JOURNAL_HDR_SZ(pPager)==0 );
  assert( offset>=c );
  assert( (offset-c)<JOURNAL_HDR_SZ(pPager) );
  pPager->journalOff = offset;

}

/*
** The journal file must be open when this routine is called. A journal
** header (JOURNAL_HDR_SZ bytes) is written into the journal file at the
** current location.
**

  char zHeader[sizeof(aJournalMagic)+16];
  int rc;

  if( pPager->stmtHdrOff==0 ){
    pPager->stmtHdrOff = pPager->journalOff;
  }

  seekJournalHdr(pPager);


  pPager->journalHdr = pPager->journalOff;


  /* FIX ME: 
  **
  ** Possibly for a pager not in no-sync mode, the journal magic should not
  ** be written until nRec is filled in as part of next syncJournal(). 
  **
  ** Actually maybe the whole journal header should be delayed until that
  ** point. Think about this.
  */
  memcpy(zHeader, aJournalMagic, sizeof(aJournalMagic));
  /* The nRec Field. 0xFFFFFFFF for no-sync journals. */
  put32bits(&zHeader[sizeof(aJournalMagic)], pPager->noSync ? 0xffffffff : 0);
  /* The random check-hash initialiser */ 
  sqlite3Randomness(sizeof(pPager->cksumInit), &pPager->cksumInit);
  put32bits(&zHeader[sizeof(aJournalMagic)+4], pPager->cksumInit);
  /* The initial database size */
  put32bits(&zHeader[sizeof(aJournalMagic)+8], pPager->dbSize);
  /* The assumed sector size for this process */
  put32bits(&zHeader[sizeof(aJournalMagic)+12], pPager->sectorSize);
  IOTRACE(("JHDR %p %lld %d\n", pPager, pPager->journalHdr, sizeof(zHeader)))
  rc = sqlite3OsWrite(pPager->jfd, zHeader, sizeof(zHeader),pPager->journalOff);
  pPager->journalOff += JOURNAL_HDR_SZ(pPager);

  /* The journal header has been written successfully. Seek the journal
  ** file descriptor to the end of the journal header sector.
  */
  if( rc==SQLITE_OK ){
    IOTRACE(("JTAIL %p %lld\n", pPager, pPager->journalOff-1))


    rc = sqlite3OsWrite(pPager->jfd, "\000", 1, pPager->journalOff-1);

  }
  return rc;
}

/*
** The journal file must be open when this is called. A journal header file
** (JOURNAL_HDR_SZ bytes) is read from the current location in the journal

  Pager *pPager, 
  i64 journalSize,
  u32 *pNRec, 
  u32 *pDbSize
){
  int rc;
  unsigned char aMagic[8]; /* A buffer to hold the magic header */
  i64 jrnlOff;

  seekJournalHdr(pPager);


  if( pPager->journalOff+JOURNAL_HDR_SZ(pPager) > journalSize ){
    return SQLITE_DONE;
  }
  jrnlOff = pPager->journalOff;

  rc = sqlite3OsRead(pPager->jfd, aMagic, sizeof(aMagic), jrnlOff);
  if( rc ) return rc;
  jrnlOff += sizeof(aMagic);

  if( memcmp(aMagic, aJournalMagic, sizeof(aMagic))!=0 ){
    return SQLITE_DONE;
  }

  rc = read32bits(pPager->jfd, jrnlOff, pNRec);
  if( rc ) return rc;

  rc = read32bits(pPager->jfd, jrnlOff+4, &pPager->cksumInit);
  if( rc ) return rc;

  rc = read32bits(pPager->jfd, jrnlOff+8, pDbSize);
  if( rc ) return rc;

  /* Update the assumed sector-size to match the value used by 
  ** the process that created this journal. If this journal was
  ** created by a process other than this one, then this routine
  ** is being called from within pager_playback(). The local value
  ** of Pager.sectorSize is restored at the end of that routine.
  */
  rc = read32bits(pPager->jfd, jrnlOff+12, (u32 *)&pPager->sectorSize);
  if( rc ) return rc;

  pPager->journalOff += JOURNAL_HDR_SZ(pPager);

  return SQLITE_OK;
}


/*
** Write the supplied master journal name into the journal file for pager
** pPager at the current location. The master journal name must be the last
** thing written to a journal file. If the pager is in full-sync mode, the

** If zMaster is a NULL pointer (occurs for a single database transaction), 
** this call is a no-op.
*/
static int writeMasterJournal(Pager *pPager, const char *zMaster){
  int rc;
  int len; 
  int i; 
  i64 jrnlOff;
  u32 cksum = 0;
  char zBuf[sizeof(aJournalMagic)+2*4];

  if( !zMaster || pPager->setMaster) return SQLITE_OK;
  pPager->setMaster = 1;

  len = strlen(zMaster);
  for(i=0; i<len; i++){
    cksum += zMaster[i];
  }

  /* If in full-sync mode, advance to the next disk sector before writing
  ** the master journal name. This is in case the previous page written to
  ** the journal has already been synced.
  */
  if( pPager->fullSync ){
    seekJournalHdr(pPager);

  }
  jrnlOff = pPager->journalOff;
  pPager->journalOff += (len+20);

  rc = write32bits(pPager->jfd, jrnlOff, PAGER_MJ_PGNO(pPager));
  if( rc!=SQLITE_OK ) return rc;
  jrnlOff += 4;

  rc = sqlite3OsWrite(pPager->jfd, zMaster, len, jrnlOff);
  if( rc!=SQLITE_OK ) return rc;
  jrnlOff += len;

  put32bits(zBuf, len);
  put32bits(&zBuf[4], cksum);
  memcpy(&zBuf[8], aJournalMagic, sizeof(aJournalMagic));
  rc = sqlite3OsWrite(pPager->jfd, zBuf, 8+sizeof(aJournalMagic), jrnlOff);
  pPager->needSync = !pPager->noSync;
  return rc;
}

/*
** Add or remove a page from the list of all pages that are in the
** statement journal.

  if( pPager->stmtOpen && !pPager->exclusiveMode ){
    sqlite3OsClose(&pPager->stfd);
    pPager->stmtOpen = 0;
  }
  if( pPager->journalOpen ){
    if( pPager->exclusiveMode 
          && (rc = sqlite3OsTruncate(pPager->jfd, 0))==SQLITE_OK ){;

      pPager->journalOff = 0;
      pPager->journalStarted = 0;
    }else{
      sqlite3OsClose(&pPager->jfd);
      pPager->journalOpen = 0;
      if( rc==SQLITE_OK ){
        rc = sqlite3OsDelete(pPager->zJournal);

** Read a single page from the journal file opened on file descriptor
** jfd.  Playback this one page.
**
** If useCksum==0 it means this journal does not use checksums.  Checksums
** are not used in statement journals because statement journals do not
** need to survive power failures.
*/
static int pager_playback_one_page(
  Pager *pPager, 
  sqlite3_file *jfd,
  i64 offset,
  int useCksum
){
  int rc;
  PgHdr *pPg;                   /* An existing page in the cache */
  Pgno pgno;                    /* The page number of a page in journal */
  u32 cksum;                    /* Checksum used for sanity checking */
  u8 *aData = (u8 *)pPager->pTmpSpace;   /* Temp storage for a page */

  /* useCksum should be true for the main journal and false for
  ** statement journals.  Verify that this is always the case
  */
  assert( jfd == (useCksum ? pPager->jfd : pPager->stfd) );
  assert( aData );

  rc = read32bits(jfd, offset, &pgno);
  if( rc!=SQLITE_OK ) return rc;
  rc = sqlite3OsRead(jfd, aData, pPager->pageSize, offset+4);
  if( rc!=SQLITE_OK ) return rc;
  pPager->journalOff += pPager->pageSize + 4;

  /* Sanity checking on the page.  This is more important that I originally
  ** thought.  If a power failure occurs while the journal is being written,
  ** it could cause invalid data to be written into the journal.  We need to
  ** detect this invalid data (with high probability) and ignore it.
  */
  if( pgno==0 || pgno==PAGER_MJ_PGNO(pPager) ){
    return SQLITE_DONE;
  }
  if( pgno>(unsigned)pPager->dbSize ){
    return SQLITE_OK;
  }
  if( useCksum ){
    rc = read32bits(jfd, offset+pPager->pageSize+4, &cksum);
    if( rc ) return rc;
    pPager->journalOff += 4;
    if( pager_cksum(pPager, aData)!=cksum ){
      return SQLITE_DONE;
    }
  }

  ** in the main journal either because the page is not in cache or else
  ** the page is marked as needSync==0.
  */
  pPg = pager_lookup(pPager, pgno);
  PAGERTRACE4("PLAYBACK %d page %d hash(%08x)\n",
               PAGERID(pPager), pgno, pager_datahash(pPager->pageSize, aData));
  if( pPager->state>=PAGER_EXCLUSIVE && (pPg==0 || pPg->needSync==0) ){
    i64 offset = (pgno-1)*(i64)pPager->pageSize;

    rc = sqlite3OsWrite(pPager->fd, aData, pPager->pageSize, offset);

    if( pPg ){
      makeClean(pPg);
    }
  }
  if( pPg ){
    /* No page should ever be explicitly rolled back that is in use, except
    ** for page 1 which is held in use in order to keep the lock on the

** To tell if a master journal can be deleted, check to each of the
** children.  If all children are either missing or do not refer to
** a different master journal, then this master journal can be deleted.
*/
static int pager_delmaster(const char *zMaster){
  int rc;
  int master_open = 0;
  sqlite3_file *master = 0;
  char *zMasterJournal = 0; /* Contents of master journal file */
  i64 nMasterJournal;       /* Size of master journal file */

  /* Open the master journal file exclusively in case some other process
  ** is running this routine also. Not that it makes too much difference.
  */
  rc = sqlite3OsOpenReadOnly(zMaster, &master);

    ** sqliteMalloc() and pointed to by zMasterJournal. 
    */
    zMasterJournal = (char *)sqliteMalloc(nMasterJournal);
    if( !zMasterJournal ){
      rc = SQLITE_NOMEM;
      goto delmaster_out;
    }
    rc = sqlite3OsRead(master, zMasterJournal, nMasterJournal, 0);
    if( rc!=SQLITE_OK ) goto delmaster_out;

    zJournal = zMasterJournal;
    while( (zJournal-zMasterJournal)<nMasterJournal ){
      if( sqlite3OsFileExists(zJournal) ){
        /* One of the journals pointed to by the master journal exists.
        ** Open it and check if it points at the master journal. If
        ** so, return without deleting the master journal file.
        */
        sqlite3_file *journal = 0;
        int c;

        rc = sqlite3OsOpenReadOnly(zJournal, &journal);
        assert( rc!=SQLITE_OK || journal );
        if( rc!=SQLITE_OK ){
          goto delmaster_out;
        }

  assert( rc!=SQLITE_DONE );
  if( rc!=SQLITE_OK || (zMaster && !sqlite3OsFileExists(zMaster)) ){
    sqliteFree(zMaster);
    zMaster = 0;
    if( rc==SQLITE_DONE ) rc = SQLITE_OK;
    goto end_playback;
  }

  pPager->journalOff = 0;

  /* This loop terminates either when the readJournalHdr() call returns
  ** SQLITE_DONE or an IO error occurs. */
  while( 1 ){

    /* Read the next journal header from the journal file.  If there are

        goto end_playback;
      }
    }

    /* Copy original pages out of the journal and back into the database file.
    */
    for(i=0; i<nRec; i++){
      rc = pager_playback_one_page(pPager, pPager->jfd, pPager->journalOff, 1);
      if( rc!=SQLITE_OK ){
        if( rc==SQLITE_DONE ){
          rc = SQLITE_OK;
          pPager->journalOff = szJ;
          break;
        }else{
          goto end_playback;

  */
  rc = pager_truncate(pPager, pPager->stmtSize);
  assert( pPager->state>=PAGER_SHARED );

  /* Figure out how many records are in the statement journal.
  */
  assert( pPager->stmtInUse && pPager->journalOpen );

  nRec = pPager->stmtNRec;
  
  /* Copy original pages out of the statement journal and back into the
  ** database file.  Note that the statement journal omits checksums from
  ** each record since power-failure recovery is not important to statement
  ** journals.
  */
  for(i=0; i<nRec; i++){
    i64 offset = i*(4+pPager->pageSize);
    rc = pager_playback_one_page(pPager, pPager->stfd, offset, 0);
    assert( rc!=SQLITE_DONE );
    if( rc!=SQLITE_OK ) goto end_stmt_playback;
  }

  /* Now roll some pages back from the transaction journal. Pager.stmtJSize
  ** was the size of the journal file when this statement was started, so
  ** everything after that needs to be rolled back, either into the
  ** database, the memory cache, or both.
  **
  ** If it is not zero, then Pager.stmtHdrOff is the offset to the start
  ** of the first journal header written during this statement transaction.
  */




  pPager->journalOff = pPager->stmtJSize;
  pPager->cksumInit = pPager->stmtCksum;
  while( pPager->journalOff < hdrOff ){
    rc = pager_playback_one_page(pPager, pPager->jfd, pPager->journalOff, 1);
    assert( rc!=SQLITE_DONE );
    if( rc!=SQLITE_OK ) goto end_stmt_playback;
  }

  while( pPager->journalOff < szJ ){
    u32 nJRec;         /* Number of Journal Records */
    u32 dummy;
    rc = readJournalHdr(pPager, szJ, &nJRec, &dummy);
    if( rc!=SQLITE_OK ){
      assert( rc!=SQLITE_DONE );
      goto end_stmt_playback;
    }
    if( nJRec==0 ){
      nJRec = (szJ - pPager->journalOff) / (pPager->pageSize+8);
    }
    for(i=nJRec-1; i>=0 && pPager->journalOff < szJ; i--){
      rc = pager_playback_one_page(pPager, pPager->jfd, pPager->journalOff, 1);
      assert( rc!=SQLITE_DONE );
      if( rc!=SQLITE_OK ) goto end_stmt_playback;
    }
  }

  pPager->journalOff = szJ;
  

**
** Write the file descriptor into *fd.  Return SQLITE_OK on success or some
** other error code if we fail.
**
** The OS will automatically delete the temporary file when it is
** closed.
*/
static int sqlite3PagerOpentemp(sqlite3_file **pFd){
  int cnt = 8;
  int rc;
  char zFile[SQLITE_TEMPNAME_SIZE];

#ifdef SQLITE_TEST
  sqlite3_opentemp_count++;  /* Used for testing and analysis only */
#endif

  const char *zFilename,   /* Name of the database file to open */
  int nExtra,              /* Extra bytes append to each in-memory page */
  int flags                /* flags controlling this file */
){
  Pager *pPager = 0;
  char *zFullPathname = 0;
  int nameLen;  /* Compiler is wrong. This is always initialized before use */
  sqlite3_file *fd = 0;
  int rc = SQLITE_OK;
  int i;
  int tempFile = 0;
  int memDb = 0;
  int readOnly = 0;
  int useJournal = (flags & PAGER_OMIT_JOURNAL)==0;
  int noReadlock = (flags & PAGER_NO_READLOCK)!=0;

** response is to zero the memory at pDest and continue.  A real IO error 
** will presumably recur and be picked up later (Todo: Think about this).
*/
int sqlite3PagerReadFileheader(Pager *pPager, int N, unsigned char *pDest){
  int rc = SQLITE_OK;
  memset(pDest, 0, N);
  if( MEMDB==0 ){



    IOTRACE(("DBHDR %p 0 %d\n", pPager, N))
    rc = sqlite3OsRead(pPager->fd, pDest, N, 0);
    if( rc==SQLITE_IOERR_SHORT_READ ){
      rc = SQLITE_OK;
    }
  }
  return rc;
}

        i64 jSz;
        rc = sqlite3OsFileSize(pPager->jfd, &jSz);
        if( rc!=0 ) return rc;
        assert( pPager->journalOff==jSz );
      }
#endif
      {
        i64 jrnlOff;

        /* Write the nRec value into the journal file header. If in
        ** full-synchronous mode, sync the journal first. This ensures that
        ** all data has really hit the disk before nRec is updated to mark
        ** it as a candidate for rollback. 
        */
        if( pPager->fullSync ){
          PAGERTRACE2("SYNC journal of %d\n", PAGERID(pPager));
          IOTRACE(("JSYNC %p\n", pPager))
          rc = sqlite3OsSync(pPager->jfd, 0);
          if( rc!=0 ) return rc;
        }

        jrnlOff = pPager->journalHdr + sizeof(aJournalMagic);

        IOTRACE(("JHDR %p %lld %d\n", pPager, jrnlOff, 4));

        rc = write32bits(pPager->jfd, jrnlOff, pPager->nRec);



        if( rc ) return rc;
      }
      PAGERTRACE2("SYNC journal of %d\n", PAGERID(pPager));
      IOTRACE(("JSYNC %p\n", pPager))
      rc = sqlite3OsSync(pPager->jfd, pPager->full_fsync);
      if( rc!=0 ) return rc;
      pPager->journalStarted = 1;

  if( rc!=SQLITE_OK ){
    return rc;
  }

  pList = sort_pagelist(pList);
  while( pList ){
    assert( pList->dirty );


    /* If there are dirty pages in the page cache with page numbers greater
    ** than Pager.dbSize, this means sqlite3PagerTruncate() was called to
    ** make the file smaller (presumably by auto-vacuum code). Do not write
    ** any such pages to the file.
    */
    if( pList->pgno<=pPager->dbSize ){
      i64 offset = (pList->pgno-1)*(i64)pPager->pageSize;
      char *pData = CODEC2(pPager, PGHDR_TO_DATA(pList), pList->pgno, 6);
      PAGERTRACE4("STORE %d page %d hash(%08x)\n",
                   PAGERID(pPager), pList->pgno, pager_pagehash(pList));
      IOTRACE(("PGOUT %p %d\n", pPager, pList->pgno));
      rc = sqlite3OsWrite(pPager->fd, pData, pPager->pageSize, offset);
      PAGER_INCR(sqlite3_pager_writedb_count);
      PAGER_INCR(pPager->nWrite);
      if( pList->pgno==1 ){
        memcpy(&pPager->dbFileVers, &pData[24], sizeof(pPager->dbFileVers));
      }
    }
#ifndef NDEBUG

#endif /* SQLITE_ENABLE_MEMORY_MANAGEMENT && !SQLITE_OMIT_DISKIO */

/*
** Read the content of page pPg out of the database file.
*/
static int readDbPage(Pager *pPager, PgHdr *pPg, Pgno pgno){
  int rc;
  i64 offset;
  assert( MEMDB==0 );
  offset = (pgno-1)*(i64)pPager->pageSize;

  rc = sqlite3OsRead(pPager->fd, PGHDR_TO_DATA(pPg), pPager->pageSize, offset);


  PAGER_INCR(sqlite3_pager_readdb_count);
  PAGER_INCR(pPager->nRead);
  IOTRACE(("PGIN %p %d\n", pPager, pgno));
  if( pgno==1 ){
    memcpy(&pPager->dbFileVers, &((u8*)PGHDR_TO_DATA(pPg))[24],
                                              sizeof(pPager->dbFileVers));
  }

        if( pPager->errCode ){
          return pPager->errCode;
        }

        if( pPager->dbSize>0 ){
          IOTRACE(("CKVERS %p %d\n", pPager, sizeof(dbFileVers)));




          rc = sqlite3OsRead(pPager->fd, &dbFileVers, sizeof(dbFileVers), 24);
          if( rc!=SQLITE_OK ){
            return rc;
          }
        }else{
          memset(dbFileVers, 0, sizeof(dbFileVers));
        }

  pPager->journalHdr = 0;
  if( rc!=SQLITE_OK ){
    if( rc==SQLITE_NOMEM ){
      sqlite3OsDelete(pPager->zJournal);
    }
    goto failed_to_open_journal;
  }
#if 0
  sqlite3OsSetFullSync(pPager->jfd, pPager->full_fsync);
  sqlite3OsSetFullSync(pPager->fd, pPager->full_fsync);
  sqlite3OsOpenDirectory(pPager->jfd, pPager->zDirectory);
#endif
  pPager->journalOpen = 1;
  pPager->journalStarted = 0;
  pPager->needSync = 0;
  pPager->alwaysRollback = 0;
  pPager->nRec = 0;
  if( pPager->errCode ){
    rc = pPager->errCode;

          cksum = pager_cksum(pPager, (u8*)pData2);
          pEnd = pData2 + pPager->pageSize;
          pData2 -= 4;
          saved = *(u32*)pEnd;
          put32bits(pEnd, cksum);
          szPg = pPager->pageSize+8;
          put32bits(pData2, pPg->pgno);
          rc = sqlite3OsWrite(pPager->jfd, pData2, szPg, pPager->journalOff);
          IOTRACE(("JOUT %p %d %lld %d\n", pPager, pPg->pgno,
                   pPager->journalOff, szPg));
          PAGER_INCR(sqlite3_pager_writej_count);
          pPager->journalOff += szPg;
          PAGERTRACE5("JOURNAL %d page %d needSync=%d hash(%08x)\n",
               PAGERID(pPager), pPg->pgno, pPg->needSync, pager_pagehash(pPg));
          *(u32*)pEnd = saved;

        pHist->pStmt = sqliteMallocRaw( pPager->pageSize );
        if( pHist->pStmt ){
          memcpy(pHist->pStmt, PGHDR_TO_DATA(pPg), pPager->pageSize);
        }
        PAGERTRACE3("STMT-JOURNAL %d page %d\n", PAGERID(pPager), pPg->pgno);
        page_add_to_stmt_list(pPg);
      }else{
        i64 offset = pPager->stmtNRec*(4+pPager->pageSize);
        char *pData2 = CODEC2(pPager, pData, pPg->pgno, 7)-4;
        put32bits(pData2, pPg->pgno);
        rc = sqlite3OsWrite(pPager->stfd, pData2, pPager->pageSize+4, offset);
        PAGERTRACE3("STMT-JOURNAL %d page %d\n", PAGERID(pPager), pPg->pgno);
        if( rc!=SQLITE_OK ){
          return rc;
        }
        pPager->stmtNRec++;
        assert( pPager->aInStmt!=0 );
        pPager->aInStmt[pPg->pgno/8] |= 1<<(pPg->pgno&7);

** Commit a statement.
*/
int sqlite3PagerStmtCommit(Pager *pPager){
  if( pPager->stmtInUse ){
    PgHdr *pPg, *pNext;
    PAGERTRACE2("STMT-COMMIT %d\n", PAGERID(pPager));
    if( !MEMDB ){

      /* sqlite3OsTruncate(pPager->stfd, 0); */
      sqliteFree( pPager->aInStmt );
      pPager->aInStmt = 0;
    }else{
      for(pPg=pPager->pStmt; pPg; pPg=pNext){
        PgHistory *pHist = PGHDR_TO_HIST(pPg, pPager);
        pNext = pHist->pNextStmt;

** on how SQLite interfaces are suppose to operate.
**
** The name of this file under configuration management is "sqlite.h.in".
** The makefile makes some minor changes to this file (such as inserting
** the version number) and changes its name to "sqlite3.h" as
** part of the build process.
**
** @(#) $Id: sqlite.h.in,v 1.224 2007/08/15 17:08:46 danielk1977 Exp $
*/
#ifndef _SQLITE3_H_
#define _SQLITE3_H_
#include <stdarg.h>     /* Needed for the definition of va_list */

/*
** Make sure we can call this stuff from C++.

** to xWrite().
*/
typedef struct sqlite3_io_methods sqlite3_io_methods;
struct sqlite3_io_methods {
  int iVersion;
  int (*xClose)(sqlite3_file*);
  int (*xRead)(sqlite3_file*, void*, int iAmt, sqlite_int64 iOfst);
  int (*xWrite)(sqlite3_file*, const void*, int iAmt, sqlite_int64 iOfst);
  int (*xTruncate)(sqlite3_file*, sqlite_int64 size);
  int (*xSync)(sqlite3_file*, int flags);
  int (*xFileSize)(sqlite3_file*, sqlite_int64 *pSize);
  int (*xLock)(sqlite3_file*, int);
  int (*xUnlock)(sqlite3_file*, int);
  int (*xCheckReservedLock)(sqlite3_file*);
  int (*xBreakLock)(sqlite3_file*);

**    May you share freely, never taking more than you give.
**
*************************************************************************
** Code for testing the pager.c module in SQLite.  This code
** is not included in the SQLite library.  It is used for automated
** testing of the SQLite library.
**
** $Id: test2.c,v 1.44 2007/08/15 17:08:46 danielk1977 Exp $
*/
#include "sqliteInt.h"
#include "os.h"
#include "pager.h"
#include "tcl.h"
#include <stdlib.h>
#include <string.h>

  rc = sqlite3OsOpenReadWrite(argv[2], &fd, &readOnly);
  if( rc ){
    Tcl_AppendResult(interp, "open failed: ", errorName(rc), 0);
    return TCL_ERROR;
  }
  offset = n;
  offset *= 1024*1024;





  rc = sqlite3OsWrite(fd, "Hello, World!", 14, offset);
  sqlite3OsClose(&fd);
  if( rc ){
    Tcl_AppendResult(interp, "write failed: ", errorName(rc), 0);
    return TCL_ERROR;
  }
  return TCL_OK;
}

#if SQLITE_TEST          /* This file is used for the testing only */
#include "sqliteInt.h"
#include "os.h"
#include "tcl.h"

#ifndef SQLITE_OMIT_DISKIO  /* This file is a no-op if disk I/O is disabled */

typedef struct CrashFile CrashFile;
typedef struct CrashGlobal CrashGlobal;
typedef struct WriteBuffer WriteBuffer;

/*
** Method:
**
**   This layer is implemented as a wrapper around the "real" 
**   sqlite3_file object for the host system. Each time data is 

**   that occur before at least one write() on the file-handle specified
**   as part of the xSync() are written to their associated real files.
**
**   If IOCAP_SAFEAPPEND is set and the first byte written by the write()
**   operation is one byte past the current end of the file, then option
**   (1) is always selected.
*/
















































/*
** Each write operation in the write-list is represented by an instance
** of the following structure.
**

** If zBuf is 0, then this structure represents a call to xTruncate(), 





** not xWrite(). In that case, iOffset is the size that the file is

** truncated to.
*/

struct WriteBuffer {




  i64 iOffset;                 /* Byte offset of the start of this write() */






  int nBuf;                    /* Number of bytes written */

















  u8 *zBuf;                    /* Pointer to copy of written data */
  CrashFile *pFile;            /* File this write() applies to */





















  WriteBuffer *pNext;          /* Next in CrashGlobal.pWriteList */











};






























































































struct CrashFile {













  const sqlite3_io_methods *pMethod;   /* Must be first */








  sqlite3_file *pRealFile;             /* Underlying "real" file handle */





  const char *zName;



};









struct CrashGlobal {



  WriteBuffer *pWriteList;     /* Head of write-list */























  int iSectorSize;             /* Value of simulated sector size */



  int iDeviceCharacteristics;  /* Value of simulated device characteristics */



  int iCrash;                  /* Crash on the iCrash'th call to xSync() */








  char zCrashFile[500];        /* Crash during an xSync() on this file */ 




















};














static CrashGlobal g = {0, SQLITE_DEFAULT_SECTOR_SIZE, 0, 0};
































/*
** Set this global variable to 1 to enable crash testing.
*/
int sqlite3CrashTestEnable = 0;

/*
** Flush the write-list as if xSync() had been called on file handle
** pFile. If isCrash is true, simulate a crash.
*/
static int writeListSync(CrashFile *pFile, int isCrash){
  int rc = SQLITE_OK;
  int iDc = g.iDeviceCharacteristics;

  WriteBuffer *pWrite;
  WriteBuffer **ppPtr;

  /* Set pFinal to point to the last element of the write-list that
  ** is associated with file handle pFile.
  */
  WriteBuffer *pFinal = 0;
  if( !isCrash ){
    for(pWrite=g.pWriteList; pWrite; pWrite=pWrite->pNext){
      if( pWrite->pFile==pFile ){
        pFinal = pWrite;
      }
    }
  }

  ppPtr = &g.pWriteList;
  for(pWrite=*ppPtr; rc==SQLITE_OK && pWrite; pWrite=*ppPtr){

    /* (eAction==1)      -> write block out normally,
    ** (eAction==2)      -> do nothing,
    ** (eAction==3)      -> trash sectors.
    */
    int eAction = 0;
    if( !isCrash ){
      eAction = 2;
      if( (pWrite->pFile==pFile || iDc&SQLITE_IOCAP_SEQUENTIAL) ){
        eAction = 1;
      }
    }else{
      char random;
      sqlite3Randomness(1, &random);

      if( iDc&SQLITE_IOCAP_ATOMIC || pWrite->zBuf==0 ){
        random &= 0x01;
      }

      if( (random&0x06)==0x06 ){
        eAction = 3;
      }else{
        eAction = ((random&0x01)?2:1);
      }
    }

    switch( eAction ){
      case 1: {               /* Write out correctly */
        if( pWrite->zBuf ){
          rc = sqlite3OsWrite(
              pFile->pRealFile, pWrite->zBuf, pWrite->nBuf, pWrite->iOffset
          );
        }else{
          rc = sqlite3OsTruncate(pFile->pRealFile, pWrite->iOffset);
        }
        *ppPtr = pWrite->pNext;
        sqliteFree(pWrite);
        break;
      }
      case 2: {               /* Do nothing */
        ppPtr = &pWrite->pNext;
        break;
      }
      case 3: {               /* Trash sectors */
        u8 *zGarbage;
        sqlite3_int64 iFirst = (pWrite->iOffset%g.iSectorSize);
        sqlite3_int64 iLast = (pWrite->iOffset+pWrite->nBuf-1)%g.iSectorSize;

        zGarbage = sqliteMalloc(g.iSectorSize);
        if( zGarbage ){
          sqlite3_int64 i;
          for(i=iFirst; rc==SQLITE_OK && i<=iLast; i++){
            sqlite3Randomness(g.iSectorSize, zGarbage); 
            rc = sqlite3OsWrite(
              pFile->pRealFile, i*g.iSectorSize, zGarbage, g.iSectorSize
            );
          }
          sqliteFree(zGarbage);
        }else{
          rc = SQLITE_NOMEM;
        }

        ppPtr = &pWrite->pNext;
        break;
      }

      default:
        assert(!"Cannot happen");
    }


    if( pWrite==pFinal ) break;
  }



  if( rc==SQLITE_OK && isCrash ){
    exit(-1);
  }

  return rc;
}

/*

** Add an entry to the end of the write-list.
*/
static int writeListAppend(
  sqlite3_file *pFile,
  sqlite3_int64 iOffset,
  const u8 *zBuf,
  int nBuf
){
  WriteBuffer *pNew;

  assert((zBuf && nBuf) || (!nBuf && !zBuf));

  pNew = (WriteBuffer *)sqliteMalloc(sizeof(WriteBuffer) + nBuf);
  pNew->iOffset = iOffset;
  pNew->nBuf = nBuf;
  pNew->pFile = (CrashFile *)pFile;
  if( zBuf ){
    pNew->zBuf = (u8 *)&pNew[1];
    memcpy(pNew->zBuf, zBuf, nBuf);
  }

  if( g.pWriteList ){
    WriteBuffer *pEnd;
    for(pEnd=g.pWriteList; pEnd->pNext; pEnd=pEnd->pNext);
    pEnd->pNext = pNew;
  }else{
    g.pWriteList = pNew;
  }
  
  return SQLITE_OK;
}

/*
** Close a crash-file.
*/
int cfClose(sqlite3_file *pFile){
  CrashFile *pCrash = (CrashFile *)pFile;
  writeListSync(pCrash, 0);
  sqlite3OsClose(&pCrash->pRealFile);
  return SQLITE_OK;
}

/*

** Read data from a crash-file.
*/
int cfRead(sqlite3_file *pFile, void *zBuf, int iAmt, sqlite_int64 iOfst){
  CrashFile *pCrash = (CrashFile *)pFile;
  sqlite3_int64 iSize;
  int rc;
  WriteBuffer *pWrite;

  /* Check the file-size to see if this is a short-read */
  rc = sqlite3OsFileSize(pFile, &iSize);
  if( rc!=SQLITE_OK ){
    return rc;
  }

  if( iSize<(iOfst+iAmt) ){
    return SQLITE_IOERR_SHORT_READ;
  }

  /* Zero the output buffer */
  memset(zBuf, 0, iAmt);

  /* Read some data from the real file */
  rc = sqlite3OsFileSize(pCrash->pRealFile, &iSize);
  if( rc==SQLITE_OK && iSize>iOfst ){
    int nRead = iAmt;
    if( iSize<(iOfst+iAmt) ){
      nRead = iSize - iOfst;
    }
    rc = sqlite3OsRead(pCrash->pRealFile, zBuf, nRead, iOfst);
  }

  /* Fill in the buffer by traversing the write-list */
  for(pWrite=g.pWriteList; rc==SQLITE_OK && pWrite; pWrite=pWrite->pNext){
    if( pWrite->pFile==pCrash ){
      int iWriteOffset;
      int nWriteBuf;
      u8 *zWriteBuf;

      iWriteOffset = pWrite->iOffset - iOfst;
      nWriteBuf = pWrite->nBuf;
      zWriteBuf = pWrite->zBuf;
      if( iWriteOffset<0 ){
        nWriteBuf += iWriteOffset;
        zWriteBuf -= iWriteOffset;
        iWriteOffset = 0;
      }
      if( (iWriteOffset+nWriteBuf)>iAmt ){
        nWriteBuf = iAmt - iWriteOffset;
      }
      
      if( pWrite->zBuf && nWriteBuf>0){
        /* Copy data to the buffer */
        memcpy(&((u8 *)zBuf)[iWriteOffset], zWriteBuf, nWriteBuf);
      }

      if( pWrite->zBuf==0 && iWriteOffset<iAmt ){
        /* Zero part of the buffer to simulate a truncate */
        memset(&((u8 *)zBuf)[iWriteOffset], 0, iAmt-iWriteOffset);
      }
    }
  }

  return rc;
}

/*
** Write data to a crash-file.
*/
int cfWrite(sqlite3_file *pFile, void *zBuf, int iAmt, sqlite_int64 iOfst){
  return writeListAppend(pFile, iOfst, zBuf, iAmt);
}

/*
** Truncate a crash-file.
*/
int cfTruncate(sqlite3_file *pFile, sqlite_int64 size){

  return writeListAppend(pFile, size, 0, 0);
}

/*
** Sync a crash-file.
*/
int cfSync(sqlite3_file *pFile, int flags){
  CrashFile *pCrash = (CrashFile *)pFile;
  int isCrash = 0;

  if( 0==strcmp(pCrash->zName, g.zCrashFile) ){
    if( (--g.iCrash==0) ){
      isCrash = 1;
    }
  }

  return writeListSync(pCrash, isCrash);
}

/*

** Return the current file-size of the crash-file.
*/

int cfFileSize(sqlite3_file *pFile, sqlite_int64 *pSize){
  CrashFile *pCrash = (CrashFile *)pFile;


  WriteBuffer *pWrite;
  int rc;
  sqlite_int64 iSize;

  rc = sqlite3OsFileSize(pCrash->pRealFile, &iSize);
  if( rc!=SQLITE_OK ){
    return rc;
  }

  for(pWrite=g.pWriteList; pWrite; pWrite=pWrite->pNext){
    sqlite_int64 iEnd = pWrite->nBuf+pWrite->iOffset;
    if( pWrite->pFile==pCrash && (pWrite->zBuf==0 || iEnd>iSize) ){
      iSize = iEnd;
    }
  }
  *pSize = iSize;

  return SQLITE_OK;
}

/*
** Calls related to file-locks are passed on to the real file handle.
*/
int cfLock(sqlite3_file *pFile, int eLock){
  return sqlite3OsLock(((CrashFile *)pFile)->pRealFile, eLock);
}
int cfUnlock(sqlite3_file *pFile, int eLock){
  return sqlite3OsUnlock(((CrashFile *)pFile)->pRealFile, eLock);
}
int cfCheckReservedLock(sqlite3_file *pFile){
  return sqlite3OsCheckReservedLock(((CrashFile *)pFile)->pRealFile);
}
int cfBreakLock(sqlite3_file *pFile){
  return sqlite3OsBreakLock(((CrashFile *)pFile)->pRealFile);
}

/*
** The xSectorSize() and xDeviceCharacteristics() functions return
** the global values configured by the [sqlite_crashparams] tcl
*  interface.
*/
int cfSectorSize(sqlite3_file *pFile){
  return g.iSectorSize;
}
int cfDeviceCharacteristics(sqlite3_file *pFile){
  return g.iDeviceCharacteristics;
}

static const sqlite3_io_methods CrashFileVtab = {
  1,                            /* iVersion */
  cfClose,                      /* xClose */
  cfRead,                       /* xRead */
  cfWrite,                      /* xWrite */
  cfTruncate,                   /* xTruncate */
  cfSync,                       /* xSync */
  cfFileSize,                   /* xFileSize */
  cfLock,                       /* xLock */
  cfUnlock,                     /* xUnlock */
  cfCheckReservedLock,          /* xCheckReservedLock */
  cfBreakLock,                  /* xBreakLock */
  cfSectorSize,                 /* xSectorSize */
  cfDeviceCharacteristics       /* xDeviceCharacteristics */
};

/*
** Open a crash-file file handle. The vfs pVfs is used to open
** the underlying real file.
*/
int sqlite3CrashFileOpen(
  sqlite3_vfs *pVfs,
  const char *zName,
  sqlite3_file *pFile,
  int flags,
  int *pOutFlags
){
  CrashFile *pWrapper = (CrashFile *)pFile;
  int rc = SQLITE_NOMEM;
  sqlite3_file *pReal;
  pReal = (sqlite3_file *)sqliteMalloc(pVfs->szOsFile);
  if( pReal ){
    pWrapper->pMethod = &CrashFileVtab;
    pWrapper->zName = zName;
    rc = pVfs->xOpen(pVfs->pAppData, zName, pReal, flags, pOutFlags);


    if( rc==SQLITE_OK ){
      pWrapper->pRealFile = pFile;

    }else{
      sqliteFree(pReal);
    }
  }
  return rc;
}

int sqlite3CrashFileWrap(
  sqlite3_file *pFile,
  const char *zName,
  sqlite3_file **ppWrapper
){
  CrashFile *pWrapper;
  pWrapper = (CrashFile *)sqliteMalloc(sizeof(CrashFile)+strlen(zName)+1);
  if( !pWrapper ){
    return SQLITE_NOMEM;
  }

  pWrapper->pMethod = &CrashFileVtab;
  pWrapper->pRealFile = pFile;
  pWrapper->zName = &pWrapper[1];
  memcpy(pWrapper->zName, zName, strlen(zName)+1);

  *ppWrapper = (sqlite3_file *)pWrapper;
  return SQLITE_OK;
}


int sqlite3CrashFileSize(){
  return (int)sizeof(CrashFile);
}

/*
** tclcmd:   sqlite_crashparams ?OPTIONS? DELAY CRASHFILE
**
** This procedure implements a TCL command that enables crash testing
** in testfixture.  Once enabled, crash testing cannot be disabled.
**
** Available options are "-characteristics" and "-sectorsize". Both require
** an argument. For -sectorsize, this is the simulated sector size in
** bytes. For -characteristics, the argument must be a list of io-capability
** flags to simulate. Valid flags are "atomic", "atomic512", "atomic1K",
** "atomic2K", "atomic4K", "atomic8K", "atomic16K", "atomic32K", 
** "atomic64K", "sequential" and "safe_append".
**
** Example:
**
**   sqlite_crashparams -sect 1024 -char {atomic sequential} ./test.db 1
**
*/
static int crashParamsObjCmd(
  void * clientData,
  Tcl_Interp *interp,
  int objc,
  Tcl_Obj *CONST objv[]
){
  int i;

  int iDelay;
  const char *zCrashFile;
  int nCrashFile;

  int iDc = 0;
  int iSectorSize = 0;
  int setSectorsize = 0;
  int setDeviceChar = 0;

  struct DeviceFlag {
    char *zName;
    int iValue;
  } aFlag[] = {
    { "atomic",      SQLITE_IOCAP_ATOMIC      },
    { "atomic512",   SQLITE_IOCAP_ATOMIC512   },
    { "atomic1k",    SQLITE_IOCAP_ATOMIC1K    },
    { "atomic2k",    SQLITE_IOCAP_ATOMIC2K    },
    { "atomic4k",    SQLITE_IOCAP_ATOMIC4K    },
    { "atomic8k",    SQLITE_IOCAP_ATOMIC8K    },
    { "atomic16k",   SQLITE_IOCAP_ATOMIC16K   },
    { "atomic32k",   SQLITE_IOCAP_ATOMIC32K   },
    { "atomic64k",   SQLITE_IOCAP_ATOMIC64K   },
    { "sequential",  SQLITE_IOCAP_SEQUENTIAL  },
    { "safe_append", SQLITE_IOCAP_SAFE_APPEND },
    { 0, 0 }
  };
  
  if( objc<3 ){
    Tcl_WrongNumArgs(interp, 1, objv, "?OPTIONS? DELAY CRASHFILE");
    return TCL_ERROR;
  }

  zCrashFile = sqlite3OsFullPathname(Tcl_GetString(objv[objc-1]));
  nCrashFile = strlen(zCrashFile);
  if( nCrashFile>=sizeof(g.zCrashFile) ){
    Tcl_AppendResult(interp, "Filename is too long: \"", zCrashFile, "\"", 0);
    return TCL_ERROR;
  }
  if( Tcl_GetIntFromObj(interp, objv[objc-2], &iDelay) ){
    return TCL_ERROR;
  }

  for(i=1; i<(objc-2); i+=2){
    int nOpt;
    char *zOpt = Tcl_GetStringFromObj(objv[i], &nOpt);

    if( (nOpt>11 || nOpt<2 || strncmp("-sectorsize", zOpt, nOpt)) 
     && (nOpt>16 || nOpt<2 || strncmp("-characteristics", zOpt, nOpt))
    ){
      Tcl_AppendResult(interp, 
        "Bad option: \"", zOpt, 
        "\" - must be \"-characteristics\" or \"-sectorsize\"", 0
      );
      return TCL_ERROR;
    }
    if( i==objc-3 ){
      Tcl_AppendResult(interp, "Option requires an argument: \"", zOpt, "\"",0);
      return TCL_ERROR;
    }

    if( zOpt[1]=='s' ){
      if( Tcl_GetIntFromObj(interp, objv[i+1], &iSectorSize) ){



        return TCL_ERROR;
      }
      setSectorsize = 1;
    }else{
      int j;
      Tcl_Obj **apObj;
      int nObj;
      if( Tcl_ListObjGetElements(interp, objv[i+1], &nObj, &apObj) ){
        return TCL_ERROR;
      }
      for(j=0; j<nObj; j++){
        int rc;
        int iChoice;
        Tcl_Obj *pFlag = Tcl_DuplicateObj(apObj[j]);
        Tcl_IncrRefCount(pFlag);
        Tcl_UtfToLower(Tcl_GetString(pFlag));
 

        rc = Tcl_GetIndexFromObjStruct(
            interp, pFlag, aFlag, sizeof(aFlag[0]), "no such flag", 0, &iChoice
        );
        Tcl_DecrRefCount(pFlag);
        if( rc ){
          return TCL_ERROR;
        }

        iDc |= aFlag[iChoice].iValue;
      }
      setDeviceChar = 1;
    }
  }

  if( setDeviceChar ){
    g.iDeviceCharacteristics = iDc;
  }
  if( setSectorsize ){
    g.iSectorSize = iSectorSize;
  }
  g.iCrash = iDelay;
  memcpy(g.zCrashFile, zCrashFile, nCrashFile+1);

  sqlite3CrashTestEnable = 1;
  return TCL_OK;
}

#endif /* SQLITE_OMIT_DISKIO */

/*

  ** committed atomicly.
  */
#ifndef SQLITE_OMIT_DISKIO
  else{
    int needSync = 0;
    char *zMaster = 0;   /* File-name for the master journal */
    char const *zMainFile = sqlite3BtreeGetFilename(db->aDb[0].pBt);
    sqlite3_file *master = 0;
    i64 offset = 0;

    /* Select a master journal file name */
    do {
      u32 random;
      sqliteFree(zMaster);
      sqlite3Randomness(sizeof(random), &random);
      zMaster = sqlite3MPrintf("%s-mj%08X", zMainFile, random&0x7fffffff);

      if( i==1 ) continue;   /* Ignore the TEMP database */
      if( pBt && sqlite3BtreeIsInTrans(pBt) ){
        char const *zFile = sqlite3BtreeGetJournalname(pBt);
        if( zFile[0]==0 ) continue;  /* Ignore :memory: databases */
        if( !needSync && !sqlite3BtreeSyncDisabled(pBt) ){
          needSync = 1;
        }
        rc = sqlite3OsWrite(master, zFile, strlen(zFile)+1, offset);
        offset += strlen(zFile)+1;
        if( rc!=SQLITE_OK ){
          sqlite3OsClose(&master);
          sqlite3OsDelete(zMaster);
          sqliteFree(zMaster);
          return rc;
        }
      }
    }


    /* Sync the master journal file. Before doing this, open the directory
    ** the master journal file is store in so that it gets synced too.
    */
    zMainFile = sqlite3BtreeGetDirname(db->aDb[0].pBt);
#if 0
    rc = sqlite3OsOpenDirectory(master, zMainFile);
    if( rc!=SQLITE_OK ||
          (needSync && (rc=sqlite3OsSync(master,0))!=SQLITE_OK) ){
      sqlite3OsClose(&master);
      sqlite3OsDelete(zMaster);
      sqliteFree(zMaster);
      return rc;
    }
#endif

    /* Sync all the db files involved in the transaction. The same call
    ** sets the master journal pointer in each individual journal. If
    ** an error occurs here, do not delete the master journal file.
    **
    ** If the error occurs during the first call to
    ** sqlite3BtreeCommitPhaseOne(), then there is a chance that the