SQLite

** Clear operations are exceedingly rare.  There are usually between
** 5 and 500 set operations per Bitvec object, though the number of sets can
** sometimes grow into tens of thousands or larger.  The size of the
** Bitvec object is the number of pages in the database file at the
** start of a transaction, and is thus usually less than a few thousand,
** but can be as large as 2 billion for a really big database.
**
** @(#) $Id: bitvec.c,v 1.11 2009/01/16 15:21:05 danielk1977 Exp $
*/
#include "sqliteInt.h"

/* Size of the Bitvec structure in bytes. */
#define BITVEC_SZ        512

/* Round the union size down to the nearest pointer boundary, since that's how 

    for(i=0; i<BITVEC_NPTR; i++){
      sqlite3BitvecDestroy(p->u.apSub[i]);
    }
  }
  sqlite3_free(p);
}

/*
** Return the value of the iSize parameter specified when Bitvec *p
** was created.
*/
u32 sqlite3BitvecSize(Bitvec *p){
  return p->iSize;
}

#ifndef SQLITE_OMIT_BUILTIN_TEST
/*
** Let V[] be an array of unsigned characters sufficient to hold
** up to N bits.  Let I be an integer between 0 and N.  0<=I<N.
** Then the following macros can be used to set, clear, or test
** individual bits within V.
*/

/*
** 2004 April 6
**
** The author disclaims copyright to this source code.  In place of
** a legal notice, here is a blessing:
**
**    May you do good and not evil.
**    May you find forgiveness for yourself and forgive others.
**    May you share freely, never taking more than you give.
**
*************************************************************************
** $Id: btree.c,v 1.559 2009/01/16 15:21:06 danielk1977 Exp $
**
** This file implements a external (disk-based) database using BTrees.
** See the header comment on "btreeInt.h" for additional information.
** Including a description of file format and an overview of operation.
*/
#include "btreeInt.h"

  }
}
#else
  #define invalidateOverflowCache(x)
  #define invalidateAllOverflowCache(x)
#endif

/*
** Set bit pgno of the BtShared.pHasContent bitvec. This is called 
** when a page that previously contained data becomes a free-list leaf 
** page.
**
** The BtShared.pHasContent bitvec exists to work around an obscure
** bug caused by the interaction of two useful IO optimizations surrounding
** free-list leaf pages:
**
**   1) When all data is deleted from a page and the page becomes
**      a free-list leaf page, the page is not written to the database
**      (as free-list leaf pages contain no meaningful data). Sometimes
**      such a page is not even journalled (as it will not be modified,
**      why bother journalling it?).
**
**   2) When a free-list leaf page is reused, its content is not read
**      from the database or written to the journal file (why should it
**      be, if it is not at all meaningful?).
**
** By themselves, these optimizations work fine and provide a handy
** performance boost to bulk delete or insert operations. However, if
** a page is moved to the free-list and then reused within the same
** transaction, a problem comes up. If the page is not journalled when
** it is moved to the free-list and it is also not journalled when it
** is extracted from the free-list and reused, then the original data
** may be lost. In the event of a rollback, it may not be possible
** to restore the database to its original configuration.
**
** The solution is the BtShared.pHasContent bitvec. Whenever a page is 
** moved to become a free-list leaf page, the corresponding bit is
** set in the bitvec. Whenever a leaf page is extracted from the free-list,
** optimization 2 above is ommitted if the corresponding bit is already
** set in BtShared.pHasContent. The contents of the bitvec are cleared
** at the end of every transaction.
*/
static int btreeSetHasContent(BtShared *pBt, Pgno pgno){
  int rc = SQLITE_OK;
  if( !pBt->pHasContent ){
    int nPage;
    rc = sqlite3PagerPagecount(pBt->pPager, &nPage);
    if( rc==SQLITE_OK ){
      pBt->pHasContent = sqlite3BitvecCreate((u32)nPage);
      if( !pBt->pHasContent ){
        rc = SQLITE_NOMEM;
      }
    }
  }
  if( rc==SQLITE_OK && pgno<=sqlite3BitvecSize(pBt->pHasContent) ){
    rc = sqlite3BitvecSet(pBt->pHasContent, pgno);
  }
  return rc;
}

/*
** Query the BtShared.pHasContent vector.
**
** This function is called when a free-list leaf page is removed from the
** free-list for reuse. It returns false if it is safe to retrieve the
** page from the pager layer with the 'no-content' flag set. True otherwise.
*/
static int btreeGetHasContent(BtShared *pBt, Pgno pgno){
  Bitvec *p = pBt->pHasContent;
  return (p && (pgno>sqlite3BitvecSize(p) || sqlite3BitvecTest(p, pgno)));
}

/*
** Clear (destroy) the BtShared.pHasContent bitvec. This should be
** invoked at the conclusion of each write-transaction.
*/
static void btreeClearHasContent(BtShared *pBt){
  sqlite3BitvecDestroy(pBt->pHasContent);
  pBt->pHasContent = 0;
}

/*
** Save the current cursor position in the variables BtCursor.nKey 
** and BtCursor.pKey. The cursor's state is set to CURSOR_REQUIRESEEK.
*/
static int saveCursorPosition(BtCursor *pCur){
  int rc;

  assert( sqlite3_mutex_held(pBt->mutex) );
  rc = sqlite3PagerAcquire(pBt->pPager, pgno, (DbPage**)&pDbPage, noContent);
  if( rc ) return rc;
  *ppPage = btreePageFromDbPage(pDbPage, pgno, pBt);
  return SQLITE_OK;
}

/*
** Retrieve a page from the pager cache. If the requested page is not
** already in the pager cache return NULL. Initialize the MemPage.pBt and
** MemPage.aData elements if needed.
*/
static MemPage *btreePageLookup(BtShared *pBt, Pgno pgno){
  DbPage *pDbPage;
  assert( sqlite3_mutex_held(pBt->mutex) );
  pDbPage = sqlite3PagerLookup(pBt->pPager, pgno);
  if( pDbPage ){
    return btreePageFromDbPage(pDbPage, pgno, pBt);
  }
  return 0;
}

/*
** Return the size of the database file in pages. If there is any kind of
** error, return ((unsigned int)-1).
*/
static Pgno pagerPagecount(BtShared *pBt){
  int nPage = -1;
  int rc;

*/
static int getAndInitPage(
  BtShared *pBt,          /* The database file */
  Pgno pgno,           /* Number of the page to get */
  MemPage **ppPage     /* Write the page pointer here */
){
  int rc;

  MemPage *pPage;

  assert( sqlite3_mutex_held(pBt->mutex) );
  if( pgno==0 ){
    return SQLITE_CORRUPT_BKPT; 
  }

  /* It is often the case that the page we want is already in cache.
  ** If so, get it directly.  This saves us from having to call
  ** pagerPagecount() to make sure pgno is within limits, which results
  ** in a measureable performance improvements.
  */
  *ppPage = pPage = btreePageLookup(pBt, pgno);
  if( pPage ){
    /* Page is already in cache */

    rc = SQLITE_OK;
  }else{
    /* Page not in cache.  Acquire it. */
    if( pgno>pagerPagecount(pBt) ){
      return SQLITE_CORRUPT_BKPT; 
    }
    rc = sqlite3BtreeGetPage(pBt, pgno, ppPage, 0);

  sqlite3BtreeEnter(p);
  pBt->db = p->db;
  assert( pBt->inTransaction==TRANS_WRITE && p->inTrans==TRANS_WRITE );
  if( !pBt->autoVacuum ){
    rc = SQLITE_DONE;
  }else{
    invalidateAllOverflowCache(pBt);
    rc = incrVacuumStep(pBt, 0, pagerPagecount(pBt));
  }
  sqlite3BtreeLeave(p);
  return rc;
}

/*
** This routine is called prior to sqlite3PagerCommit when a transaction

      pBt->inTransaction = TRANS_NONE;
    }
  }

  /* Set the handles current transaction state to TRANS_NONE and unlock
  ** the pager if this call closed the only read or write transaction.
  */
  btreeClearHasContent(pBt);
  p->inTrans = TRANS_NONE;
  unlockBtreeIfUnused(pBt);

  btreeIntegrity(p);
  sqlite3BtreeLeave(p);
  return SQLITE_OK;
}

    assert( pBt->nTransaction>0 );
    pBt->nTransaction--;
    if( 0==pBt->nTransaction ){
      pBt->inTransaction = TRANS_NONE;
    }
  }

  btreeClearHasContent(pBt);
  p->inTrans = TRANS_NONE;
  pBt->inStmt = 0;
  unlockBtreeIfUnused(pBt);

  btreeIntegrity(p);
  sqlite3BtreeLeave(p);
  return rc;

** Given the page number of an overflow page in the database (parameter
** ovfl), this function finds the page number of the next page in the 
** linked list of overflow pages. If possible, it uses the auto-vacuum
** pointer-map data instead of reading the content of page ovfl to do so. 
**
** If an error occurs an SQLite error code is returned. Otherwise:
**
** The page number of the next overflow page in the linked list is 
** written to *pPgnoNext. If page ovfl is the last page in its linked 
** list, *pPgnoNext is set to zero. 
**
** If ppPage is not NULL, and a reference to the MemPage object corresponding
** to page number pOvfl was obtained, then *ppPage is set to point to that
** reference. It is the responsibility of the caller to call releasePage()
** on *ppPage to free the reference. In no reference was obtained (because
** the pointer-map was used to obtain the value for *pPgnoNext), then
** *ppPage is set to zero.
*/
static int getOverflowPage(
  BtShared *pBt, 
  Pgno ovfl,                   /* Overflow page */
  MemPage **ppPage,            /* OUT: MemPage handle (may be NULL) */
  Pgno *pPgnoNext              /* OUT: Next overflow page number */
){
  Pgno next = 0;
  MemPage *pPage = 0;
  int rc = SQLITE_OK;

  assert( sqlite3_mutex_held(pBt->mutex) );






  assert(pPgnoNext);



#ifndef SQLITE_OMIT_AUTOVACUUM
  /* Try to find the next page in the overflow list using the
  ** autovacuum pointer-map pages. Guess that the next page in 
  ** the overflow list is page number (ovfl+1). If that guess turns 
  ** out to be wrong, fall back to loading the data of page 
  ** number ovfl to determine the next page number.
  */
  if( pBt->autoVacuum ){
    Pgno pgno;
    Pgno iGuess = ovfl+1;
    u8 eType;

    while( PTRMAP_ISPAGE(pBt, iGuess) || iGuess==PENDING_BYTE_PAGE(pBt) ){
      iGuess++;
    }

    if( iGuess<=pagerPagecount(pBt) ){
      rc = ptrmapGet(pBt, iGuess, &eType, &pgno);



      if( rc==SQLITE_OK && eType==PTRMAP_OVERFLOW2 && pgno==ovfl ){
        next = iGuess;
        rc = SQLITE_DONE;
      }
    }
  }
#endif



  if( rc==SQLITE_OK ){
    rc = sqlite3BtreeGetPage(pBt, ovfl, &pPage, 0);
    assert(rc==SQLITE_OK || pPage==0);
    if( next==0 && rc==SQLITE_OK ){
      next = get4byte(pPage->aData);
    }
  }

  *pPgnoNext = next;
  if( ppPage ){
    *ppPage = pPage;
  }else{
    releasePage(pPage);
  }



  return (rc==SQLITE_DONE ? SQLITE_OK : rc);
}

/*
** Copy data from a buffer to a page, or from a page to a buffer.
**
** pPayload is a pointer to data stored on database page pDbPage.
** If argument eOp is false, then nByte bytes of data are copied

          }
        }else{
          closest = 0;
        }

        iPage = get4byte(&aData[8+closest*4]);
        if( !searchList || iPage==nearby ){
          int noContent;
          Pgno nPage;
          *pPgno = iPage;
          nPage = pagerPagecount(pBt);
          if( *pPgno>nPage ){
            /* Free page off the end of the file */
            rc = SQLITE_CORRUPT_BKPT;
            goto end_allocate_page;
          }
          TRACE(("ALLOCATE: %d was leaf %d of %d on trunk %d"
                 ": %d more free pages\n",
                 *pPgno, closest+1, k, pTrunk->pgno, n-1));
          if( closest<k-1 ){
            memcpy(&aData[8+closest*4], &aData[4+k*4], 4);
          }
          put4byte(&aData[4], k-1);
          assert( sqlite3PagerIswriteable(pTrunk->pDbPage) );
          noContent = !btreeGetHasContent(pBt, *pPgno);
          rc = sqlite3BtreeGetPage(pBt, *pPgno, ppPage, noContent);
          if( rc==SQLITE_OK ){

            rc = sqlite3PagerWrite((*ppPage)->pDbPage);
            if( rc!=SQLITE_OK ){
              releasePage(*ppPage);
            }
          }
          searchList = 0;
        }
      }
      releasePage(pPrevTrunk);
      pPrevTrunk = 0;
    }while( searchList );
  }else{
    /* There are no pages on the freelist, so create a new page at the
    ** end of the file */
    int nPage = pagerPagecount(pBt);
    *pPgno = nPage + 1;

    if( *pPgno==PENDING_BYTE_PAGE(pBt) ){
      (*pPgno)++;
    }

#ifndef SQLITE_OMIT_AUTOVACUUM
    if( pBt->autoVacuum && PTRMAP_ISPAGE(pBt, *pPgno) ){
      /* If *pPgno refers to a pointer-map page, allocate two new pages
      ** at the end of the file instead of one. The first allocated page
      ** becomes a new pointer-map page, the second is used by the caller.
      */
      TRACE(("ALLOCATE: %d from end of file (pointer-map page)\n", *pPgno));

    }
    (*ppPage)->isInit = 0;
  }
  return rc;
}

/*
** This function is used to add page iPage to the database file free-list. 
** It is assumed that the page is not already a part of the free-list.
**
** The value passed as the second argument to this function is optional.
** If the caller happens to have a pointer to the MemPage object 
** corresponding to page iPage handy, it may pass it as the second value. 
** Otherwise, it may pass NULL.
**
** If a pointer to a MemPage object is passed as the second argument,
** its reference count is not altered by this function.

*/
static int freePage2(BtShared *pBt, MemPage *pMemPage, Pgno iPage){
  MemPage *pTrunk = 0;                /* Free-list trunk page */
  Pgno iTrunk = 0;                    /* Page number of free-list trunk page */ 

  MemPage *pPage1 = pBt->pPage1;      /* Local reference to page 1 */
  MemPage *pPage;                     /* Page being freed. May be NULL. */
  int rc;                             /* Return Code */
  int nFree;                          /* Initial number of pages on free-list */


  assert( sqlite3_mutex_held(pBt->mutex) );
  assert( iPage>1 );
  assert( !pMemPage || pMemPage->pgno==iPage );

  if( pMemPage ){
    pPage = pMemPage;
    sqlite3PagerRef(pPage->pDbPage);
  }else{
    pPage = btreePageLookup(pBt, iPage);
  }

  /* Increment the free page count on pPage1 */
  rc = sqlite3PagerWrite(pPage1->pDbPage);
  if( rc ) goto freepage_out;
  nFree = get4byte(&pPage1->aData[36]);
  put4byte(&pPage1->aData[36], nFree+1);

#ifdef SQLITE_SECURE_DELETE
  /* If the SQLITE_SECURE_DELETE compile-time option is enabled, then
  ** always fully overwrite deleted information with zeros.
  */
  if( (!pPage && (rc = sqlite3BtreeGetPage(pBt, iPage, &pPage, 0)))
   ||            (rc = sqlite3PagerWrite(pPage->pDbPage))
  ){
    goto freepage_out;
  }
  memset(pPage->aData, 0, pPage->pBt->pageSize);
#endif

  /* If the database supports auto-vacuum, write an entry in the pointer-map
  ** to indicate that the page is free.
  */
  if( ISAUTOVACUUM ){
    rc = ptrmapPut(pBt, iPage, PTRMAP_FREEPAGE, 0);
    if( rc ) goto freepage_out;
  }

  /* Now manipulate the actual database free-list structure. There are two
  ** possibilities. If the free-list is currently empty, or if the first
  ** trunk page in the free-list is full, then this page will become a
  ** new free-list trunk page. Otherwise, it will become a leaf of the
  ** first trunk page in the current free-list. This block tests if it
  ** is possible to add the page as a new free-list leaf.
  */
  if( nFree!=0 ){
    int nLeaf;                /* Initial number of leaf cells on trunk page */



    iTrunk = get4byte(&pPage1->aData[32]);





    rc = sqlite3BtreeGetPage(pBt, iTrunk, &pTrunk, 0);
    if( rc!=SQLITE_OK ){
      goto freepage_out;
    }

    nLeaf = get4byte(&pTrunk->aData[4]);
    if( nLeaf<0 ){
      rc = SQLITE_CORRUPT_BKPT;
      goto freepage_out;
    }
    if( nLeaf<pBt->usableSize/4 - 8 ){
      /* In this case there is room on the trunk page to insert the page
      ** being freed as a new leaf.

      **
      ** Note that the trunk page is not really full until it contains
      ** usableSize/4 - 2 entries, not usableSize/4 - 8 entries as we have
      ** coded.  But due to a coding error in versions of SQLite prior to
      ** 3.6.0, databases with freelist trunk pages holding more than
      ** usableSize/4 - 8 entries will be reported as corrupt.  In order
      ** to maintain backwards compatibility with older versions of SQLite,
      ** we will contain to restrict the number of entries to usableSize/4 - 8
      ** for now.  At some point in the future (once everyone has upgraded
      ** to 3.6.0 or later) we should consider fixing the conditional above
      ** to read "usableSize/4-2" instead of "usableSize/4-8".
      */












      rc = sqlite3PagerWrite(pTrunk->pDbPage);
      if( rc==SQLITE_OK ){
        put4byte(&pTrunk->aData[4], nLeaf+1);
        put4byte(&pTrunk->aData[8+nLeaf*4], iPage);
#ifndef SQLITE_SECURE_DELETE
        if( pPage ){
          sqlite3PagerDontWrite(pPage->pDbPage);
        }
#endif
        rc = btreeSetHasContent(pBt, iPage);
      }
      TRACE(("FREE-PAGE: %d leaf on trunk page %d\n",pPage->pgno,pTrunk->pgno));
      goto freepage_out;
    }
  }

  /* If control flows to this point, then it was not possible to add the
  ** the page being freed as a leaf page of the first trunk in the free-list.
  ** Possibly because the free-list is empty, or possibly because the 
  ** first trunk in the free-list is full. Either way, the page being freed
  ** will become the new first trunk page in the free-list.
  */
  if( (!pPage && (rc = sqlite3BtreeGetPage(pBt, iPage, &pPage, 0)))
   ||            (rc = sqlite3PagerWrite(pPage->pDbPage))
  ){
    goto freepage_out;
  }
  put4byte(pPage->aData, iTrunk);
  put4byte(&pPage->aData[4], 0);
  put4byte(&pPage1->aData[32], iPage);
  TRACE(("FREE-PAGE: %d new trunk page replacing %d\n", pPage->pgno, iTrunk));

freepage_out:
  if( pPage ){
    pPage->isInit = 0;
  }
  releasePage(pPage);
  releasePage(pTrunk);
  return rc;
}
static int freePage(MemPage *pPage){
  return freePage2(pPage->pBt, pPage, pPage->pgno);
}

/*
** Free any overflow pages associated with the given Cell.
*/
static int clearCell(MemPage *pPage, unsigned char *pCell){
  BtShared *pBt = pPage->pBt;

    return SQLITE_OK;  /* No overflow pages. Return without doing anything */
  }
  ovflPgno = get4byte(&pCell[info.iOverflow]);
  ovflPageSize = pBt->usableSize - 4;
  nOvfl = (info.nPayload - info.nLocal + ovflPageSize - 1)/ovflPageSize;
  assert( ovflPgno==0 || nOvfl>0 );
  while( nOvfl-- ){
    Pgno iNext;
    MemPage *pOvfl = 0;
    if( ovflPgno==0 || ovflPgno>pagerPagecount(pBt) ){
      return SQLITE_CORRUPT_BKPT;
    }
    if( nOvfl ){
      rc = getOverflowPage(pBt, ovflPgno, &pOvfl, &iNext);
      if( rc ) return rc;
    }
    rc = freePage2(pBt, pOvfl, ovflPgno);
    if( pOvfl ){
      sqlite3PagerUnref(pOvfl->pDbPage);
    }
    if( rc ) return rc;
    ovflPgno = iNext;
  }
  return SQLITE_OK;
}

/*
** Create the byte sequence used to represent a cell on page pPage
** and write that byte sequence into pCell[].  Overflow pages are

      rc = ptrmapGet(pBt, pgnoRoot, &eType, &iPtrPage);
      if( rc!=SQLITE_OK || eType==PTRMAP_ROOTPAGE || eType==PTRMAP_FREEPAGE ){
        releasePage(pRoot);
        return rc;
      }
      assert( eType!=PTRMAP_ROOTPAGE );
      assert( eType!=PTRMAP_FREEPAGE );





      rc = relocatePage(pBt, pRoot, eType, iPtrPage, pgnoMove, 0);
      releasePage(pRoot);

      /* Obtain the page at pgnoRoot */
      if( rc!=SQLITE_OK ){
        return rc;
      }

** The pager filename is invariant as long as the pager is
** open so it is safe to access without the BtShared mutex.
*/
const char *sqlite3BtreeGetFilename(Btree *p){
  assert( p->pBt->pPager!=0 );
  return sqlite3PagerFilename(p->pBt->pPager);
}












/*
** Return the pathname of the journal file for this database. The return
** value of this routine is the same regardless of whether the journal file
** has been created or not.
**
** The pager journal filename is invariant as long as the pager is

          ** represent what they do.  Write() really means "put this page in the
          ** rollback journal and mark it as dirty so that it will be written
          ** to the database file later."  DontWrite() undoes the second part of
          ** that and prevents the page from being written to the database. The
          ** page is still on the rollback journal, though.  And that is the 
          ** whole point of this block: to put pages on the rollback journal. 
          */
          sqlite3PagerDontWrite(pDbPage);
        }
        sqlite3PagerUnref(pDbPage);
      }
    }

    /* Overwrite the data in page i of the target database */
    if( rc==SQLITE_OK && i!=iSkip && i<=nNewPage ){

/*
** 2004 April 6
**
** The author disclaims copyright to this source code.  In place of
** a legal notice, here is a blessing:
**
**    May you do good and not evil.
**    May you find forgiveness for yourself and forgive others.
**    May you share freely, never taking more than you give.
**
*************************************************************************
** $Id: btreeInt.h,v 1.39 2009/01/16 15:21:06 danielk1977 Exp $
**
** This file implements a external (disk-based) database using BTrees.
** For a detailed discussion of BTrees, refer to
**
**     Donald E. Knuth, THE ART OF COMPUTER PROGRAMMING, Volume 3:
**     "Sorting And Searching", pages 473-480. Addison-Wesley
**     Publishing Company, Reading, Massachusetts.

**
**    SIZE    DESCRIPTION
**      4     Page number of next trunk page
**      4     Number of leaf pointers on this page
**      *     zero or more pages numbers of leaves
*/
#include "sqliteInt.h"





/* Round up a number to the next larger multiple of 8.  This is used
** to force 8-byte alignment on 64-bit architectures.
*/
#define ROUND8(x)   ((x+7)&~7)

  u16 maxLeaf;          /* Maximum local payload in a LEAFDATA table */
  u16 minLeaf;          /* Minimum local payload in a LEAFDATA table */
  u8 inTransaction;     /* Transaction state */
  int nTransaction;     /* Number of open transactions (read + write) */
  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 struct */
  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 *pExclusive;    /* Btree with an EXCLUSIVE lock on the whole db */
#endif
  u8 *pTmpSpace;        /* BtShared.pageSize bytes of space for tmp use */

**   should return the error code stored in BtCursor.skip
*/
#define CURSOR_INVALID           0
#define CURSOR_VALID             1
#define CURSOR_REQUIRESEEK       2
#define CURSOR_FAULT             3

/* 
** The database page the PENDING_BYTE occupies. This page is never used.





*/

# define PENDING_BYTE_PAGE(pBt) PAGER_MJ_PGNO(pBt)




/*
** A linked list of the following structures is stored at BtShared.pLock.
** Locks are added (or upgraded from READ_LOCK to WRITE_LOCK) when a cursor 
** is opened on the table with root page BtShared.iTable. Locks are removed
** from this list when a transaction is committed or rolled back, or when
** a btree handle is closed.

** 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 1.552 2009/01/16 15:21:06 danielk1977 Exp $
*/
#ifndef SQLITE_OMIT_DISKIO
#include "sqliteInt.h"

/*
** Macros for troubleshooting.  Normally turned off
*/

  Pgno nOrig;                  /* Original number of pages in file */
  Pgno iSubRec;                /* Index of first record in sub-journal */
};

/*
** A open page cache is an instance of the following structure.
**
** errCode
**
**   Pager.errCode may be set to SQLITE_IOERR, SQLITE_CORRUPT, or
**   or SQLITE_FULL. Once one of the first three errors occurs, it persists
**   and is returned as the result of every major pager API call.  The
**   SQLITE_FULL return code is slightly different. It persists only until the
**   next successful rollback is performed on the pager cache. Also,
**   SQLITE_FULL does not affect the sqlite3PagerGet() and sqlite3PagerLookup()
**   APIs, they may still be used successfully.
**
** dbSizeValid, dbSize, dbOrigSize, dbFileSize
**
**   Managing the size of the database file in pages is a little complicated.
**   The variable Pager.dbSize contains the number of pages that the database
**   image currently contains. As the database image grows or shrinks this
**   variable is updated. The variable Pager.dbFileSize contains the number
**   of pages in the database file. This may be different from Pager.dbSize
**   if some pages have been appended to the database image but not yet written
**   out from the cache to the actual file on disk. Or if the image has been
**   truncated by an incremental-vacuum operation. The Pager.dbOrigSize variable
**   contains the number of pages in the database image when the current
**   transaction was opened. The contents of all three of these variables is
**   only guaranteed to be correct if the boolean Pager.dbSizeValid is true.
**
**   TODO: Under what conditions is dbSizeValid set? Cleared?
**
** changeCountDone
**
**   This boolean variable is used to make sure that the change-counter 
**   (the 4-byte header field at byte offset 24 of the database file) is 
**   not updated more often than necessary. 
**
**   It is set to true when the change-counter field is updated, which 
**   can only happen if an exclusive lock is held on the database file.
**   It is cleared (set to false) whenever an exclusive lock is 
**   relinquished on the database file. Each time a transaction is committed,
**   The changeCountDone flag is inspected. If it is true, the work of
**   updating the change-counter is omitted for the current transaction.
**
**   This mechanism means that when running in exclusive mode, a connection 
**   need only update the change-counter once, for the first transaction
**   committed.
**
** dbModified
**
**   The dbModified flag is set whenever a database page is dirtied.
**   It is cleared at the end of each transaction.
**
**   It is used when committing or otherwise ending a transaction. If
**   the dbModified flag is clear then less work has to be done.
**
**   TODO: Check some of the logic surrounding this optimization.
**
** journalStarted
**
**   This flag is set whenever the the main journal is synced. 
**
**   The point of this flag is that it must be set after the 
**   first journal header in a journal file has been synced to disk.
**   After this has happened, new pages appended to the database 
**   do not need the PGHDR_NEED_SYNC flag set, as they do not need
**   to wait for a journal sync before they can be written out to

**   the database file (see function pager_write()).
**   
** setMaster
**
**   This variable is used to ensure that the master journal file name
**   (if any) is only written into the journal file once.
**
**   When committing a transaction, the master journal file name (if any)
**   may be written into the journal file while the pager is still in
**   PAGER_RESERVED state (see CommitPhaseOne() for the action). It
**   then attempts to upgrade to an exclusive lock. If this attempt
**   fails, then SQLITE_BUSY may be returned to the user and the user


**   may attempt to commit the transaction again later (calling
**   CommitPhaseOne() again). This flag is used to ensure that the 
**   master journal name is only written to the journal file the first
**   time CommitPhaseOne() is called.
**
** doNotSync
**
**   This variable is set and cleared by sqlite3PagerWrite().
**
** needSync
**
**   TODO: It might be easier to set this variable in writeJournalHdr()
**   and writeMasterJournal() only. Change its meaning to "unsynced data
**   has been written to the journal".

*/
struct Pager {
  sqlite3_vfs *pVfs;          /* OS functions to use for IO */
  u8 exclusiveMode;           /* Boolean. True if locking_mode==EXCLUSIVE */
  u8 journalMode;             /* On of the PAGER_JOURNALMODE_* values */
  u8 useJournal;              /* Use a rollback journal on this file */
  u8 noReadlock;              /* Do not bother to obtain readlocks */
  u8 noSync;                  /* Do not sync the journal if true */
  u8 fullSync;                /* Do extra syncs of the journal for robustness */
  u8 sync_flags;              /* One of SYNC_NORMAL or SYNC_FULL */

  u8 tempFile;                /* zFilename is a temporary file */
  u8 readOnly;                /* True for a read-only database */


  u8 memDb;                   /* True to inhibit all file I/O */

  /* The following block contains those class members that are dynamically
  ** modified during normal operations. The other variables in this structure
  ** are either constant throughout the lifetime of the pager, or else
  ** used to store configuration parameters that affect the way the pager 
  ** operates.
  **
  ** The 'state' variable is described in more detail along with the
  ** descriptions of the values it may take - PAGER_UNLOCK etc. Many of the
  ** other variables in this block are described in the comment directly 
  ** above this class definition.
  */
  u8 state;                   /* PAGER_UNLOCK, _SHARED, _RESERVED, etc. */
  u8 dbModified;              /* True if there are any changes to the Db */
  u8 needSync;                /* True if an fsync() is needed on the journal */
  u8 journalStarted;          /* True if header of journal is synced */
  u8 changeCountDone;         /* Set after incrementing the change-counter */
  u8 setMaster;               /* True if a m-j name has been written to jrnl */
  u8 doNotSync;               /* Boolean. While true, do not spill the cache */
  u8 dbSizeValid;             /* Set when dbSize is correct */
  Pgno dbSize;                /* Number of pages in the database */
  Pgno dbOrigSize;            /* dbSize before the current transaction */
  Pgno dbFileSize;            /* Number of pages in the database file */

  int errCode;                /* One of several kinds of errors */
  int nRec;                   /* Pages journalled since last j-header written */
  u32 cksumInit;              /* Quasi-random value added to every checksum */
  u32 nSubRec;                /* Number of records written to sub-journal */
  Bitvec *pInJournal;         /* One bit for each page in the database file */
  sqlite3_file *fd;           /* File descriptor for database */
  sqlite3_file *jfd;          /* File descriptor for main journal */
  sqlite3_file *sjfd;         /* File descriptor for sub-journal */
  i64 journalOff;             /* Current write offset in the journal file */
  i64 journalHdr;             /* Byte offset to previous journal header */
  PagerSavepoint *aSavepoint; /* Array of active savepoints */
  int nSavepoint;             /* Number of elements in aSavepoint[] */
  char dbFileVers[16];        /* Changes whenever database file changes */
  u32 sectorSize;             /* Assumed sector size during rollback */

  int nExtra;                 /* Add this many bytes to each in-memory page */
  u32 vfsFlags;               /* Flags for sqlite3_vfs.xOpen() */
  int pageSize;               /* Number of bytes in a page */


  Pgno mxPgno;                /* Maximum allowed size of the database */


  char *zFilename;            /* Name of the database file */
  char *zJournal;             /* Name of the journal file */



  int (*xBusyHandler)(void*); /* Function to call when busy */
  void *pBusyHandlerArg;      /* Context argument for xBusyHandler */



#ifdef SQLITE_TEST
  int nHit, nMiss;            /* Cache hits and missing */
  int nRead, nWrite;          /* Database pages read/written */
#endif
  void (*xReiniter)(DbPage*); /* Call this routine when reloading pages */
#ifdef SQLITE_HAS_CODEC
  void *(*xCodec)(void*,void*,Pgno,int); /* Routine for en/decoding data */
  void *pCodecArg;            /* First argument to xCodec() */
#endif
  char *pTmpSpace;            /* Pager.pageSize bytes of space for tmp use */

  i64 journalSizeLimit;       /* Size limit for persistent journal files */
  PCache *pPCache;            /* Pointer to page cache object */


};

/*
** The following global variables hold counters used for
** testing purposes only.  These variables do not exist in
** a non-testing build.  These variables are not thread-safe.
*/

** is different for every journal, we minimize that risk.
*/
static const unsigned char aJournalMagic[] = {
  0xd9, 0xd5, 0x05, 0xf9, 0x20, 0xa1, 0x63, 0xd7,
};

/*
** The size of the of each page record in the journal is given by
** the following macro.
*/
#define JOURNAL_PG_SZ(pPager)  ((pPager->pageSize) + 8)

/*
** The journal header size for this pager. This is usually the same 

** size as a single disk sector. See also setSectorSize().
*/
#define JOURNAL_HDR_SZ(pPager) (pPager->sectorSize)

/*
** The macro MEMDB is true if we are dealing with an in-memory database.
** We do this as a macro so that if the SQLITE_OMIT_MEMORYDB macro is set,
** the value of MEMDB will be a constant and the compiler will optimize
** out code that would never execute.
*/
#ifdef SQLITE_OMIT_MEMORYDB
# define MEMDB 0
#else
# define MEMDB pPager->memDb
#endif











/*
** The maximum legal page number is (2^31 - 1).
*/
#define PAGER_MAX_PGNO 2147483647

/*
** Return true if it is necessary to write page *pPg into the sub-journal.

*/
static int write32bits(sqlite3_file *fd, i64 offset, u32 val){
  char ac[4];
  put32bits(ac, val);
  return sqlite3OsWrite(fd, ac, 4, offset);
}

/*
** The argument to this macro is a file descriptor (type sqlite3_file*).
** Return 0 if it is not open, or non-zero (but not 1) if it is.
**
** This is so that expressions can be written as:
**
**   if( isOpen(pPager->jfd) ){ ...
**
** instead of
**
**   if( pPager->jfd->pMethods ){ ...
*/
#define isOpen(pFd) ((pFd)->pMethods)

/*
** If file pFd is open, call sqlite3OsUnlock() on it.
*/
static int osUnlock(sqlite3_file *pFd, int eLock){
  if( !isOpen(pFd) ){
    return SQLITE_OK;
  }
  return sqlite3OsUnlock(pFd, eLock);
}

/*
** This function determines whether or not the atomic-write optimization
** can be used with this pager. The optimization can be used if:
**
**  (a) the value returned by OsDeviceCharacteristics() indicates that
**      a database page may be written atomically, and
**  (b) the value returned by OsSectorSize() is less than or equal
**      to the page size.
**
** The optimization is also always enabled for temporary files. It is
** an error to call this function if pPager is opened on an in-memory
** database.
**
** If the optimization cannot be used, 0 is returned. If it can be used,
** then the value returned is the size of the journal file when it
** contains rollback data for exactly one page.
*/
#ifdef SQLITE_ENABLE_ATOMIC_WRITE
static int jrnlBufferSize(Pager *pPager){
  assert( !MEMDB );
  if( !pPager->tempFile ){
    int dc;                           /* Device characteristics */
    int nSector;                      /* Sector size */
    int szPage;                       /* Page size */


    assert( isOpen(pPager->fd) );

    dc = sqlite3OsDeviceCharacteristics(pPager->fd);
    nSector = pPager->sectorSize;
    szPage = pPager->pageSize;


    assert(SQLITE_IOCAP_ATOMIC512==(512>>8));
    assert(SQLITE_IOCAP_ATOMIC64K==(65536>>8));
    if( 0==(dc&(SQLITE_IOCAP_ATOMIC|(szPage>>8)) || nSector>szPage) ){
      return 0;
    }
  }


  return JOURNAL_HDR_SZ(pPager) + JOURNAL_PG_SZ(pPager);


}
#endif

/*
** This function should be called when an IOERR, CORRUPT or FULL error
** may have occured. 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. 
**
** If the second argument is SQLITE_IOERR, SQLITE_CORRUPT, or SQLITE_FULL
** the error becomes persistent. Until the persisten error is cleared,
** subsequent API calls on this Pager will immediately return the same 
** error code.
**
** A persistent error indicates that the contents of the pager-cache 
** cannot be trusted. This state can be cleared by completely discarding 
** the contents of the pager-cache. If a transaction was active when
** the persistent error occured, then the rollback journal may need
** to be replayed to restore the contents of the database file (as if
** it were a hot-journal).
*/
static void pager_unlock(Pager *pPager);
static int pager_error(Pager *pPager, int rc){
  int rc2 = rc & 0xff;
  assert(
       pPager->errCode==SQLITE_FULL ||
       pPager->errCode==SQLITE_OK ||

#define pager_datahash(X,Y)  0
#define pager_pagehash(X)  0
#define CHECK_PAGE(x)
#endif  /* SQLITE_CHECK_PAGES */

/*
** When this is called the journal file for pager pPager must be open.
** This function attempts to read a master journal file name from the 
** end of the file and, if successful, copies it into memory supplied 
** by the caller. See comments above writeMasterJournal() for the format
** used to store a master journal file name at the end of a journal file.
**
** zMaster must point to a buffer of at least nMaster bytes allocated by
** the caller. This should be sqlite3_vfs.mxPathname+1 (to ensure there is
** enough space to write the master journal name). If the master journal
** name in the journal is longer than nMaster bytes (including a
** nul-terminator), then this is handled as if no master journal name
** were present in the journal.
**
** If a master journal file name is present at the end of the journal
** file, then it is copied into the buffer pointed to by zMaster. A
** nul-terminator byte is appended to the buffer following the master
** journal file name.
**
** If it is determined that no master journal file name is present 
** zMaster[0] is set to 0 and SQLITE_OK returned.
**
** If an error occurs while reading from the journal file, an SQLite
** error code is returned.
*/
static int readMasterJournal(sqlite3_file *pJrnl, char *zMaster, u32 nMaster){
  int rc;                    /* Return code */
  u32 len;                   /* Length in bytes of master journal name */
  i64 szJ;                   /* Total size in bytes of journal file pJrnl */
  u32 cksum;                 /* MJ checksum value read from journal */
  u32 u;                     /* Unsigned loop counter */
  unsigned char aMagic[8];   /* A buffer to hold the magic header */

  zMaster[0] = '\0';

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

   || SQLITE_OK!=(rc = read32bits(pJrnl, szJ-16, &len))


   || len>=nMaster 



   || SQLITE_OK!=(rc = read32bits(pJrnl, szJ-12, &cksum))


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

   || SQLITE_OK!=(rc = sqlite3OsRead(pJrnl, zMaster, len, szJ-16-len))
  ){
    return rc;
  }


  /* See if the checksum matches the master journal name */
  for(u=0; u<len; u++){
    cksum -= zMaster[u];
  }
  if( cksum ){
    /* If the checksum doesn't add up, then one or more of the disk sectors
    ** containing the master journal filename is corrupted. This means
    ** definitely roll back, so just return SQLITE_OK and report a (nul)
    ** master-journal filename.
    */
    len = 0;
  }
  zMaster[len] = '\0';
   
  return SQLITE_OK;
}

/*
** Return the offset of the sector boundary at or immediately 
** following the value in pPager->journalOff, assuming a sector 
** size of pPager->sectorSize bytes.
**
** i.e for a sector size of 512:
**
**   Pager.journalOff          Return value
**   ---------------------------------------
**   0                         0
**   512                       512
**   100                       512
**   2000                      2048
** 
*/
static i64 journalHdrOffset(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) );
  return offset;
}




/*
** The journal file must be open when this function is called.
**
** This function is a no-op if the journal file has not been written to
** within the current transaction (i.e. if Pager.journalOff==0).
**
** If doTruncate is non-zero or the Pager.journalSizeLimit variable is
** set to 0, then truncate the journal file to zero bytes in size. Otherwise,
** zero the 28-byte header at the start of the journal file. In either case, 
** if the pager is not in no-sync mode, sync the journal file immediately 
** after writing or truncating it.
**
** If Pager.journalSizeLimit is set to a positive, non-zero value, and
** following the truncation or zeroing described above the size of the 
** journal file in bytes is larger than this value, then truncate the
** journal file to Pager.journalSizeLimit bytes. The journal file does
** not need to be synced following this operation.
**
** If an IO error occurs, abandon processing and return the IO error code.
** Otherwise, return SQLITE_OK.
*/
static int zeroJournalHdr(Pager *pPager, int doTruncate){
  int rc = SQLITE_OK;                               /* Return code */

  assert( isOpen(pPager->jfd) );
  if( pPager->journalOff ){
    const i64 iLimit = pPager->journalSizeLimit;    /* Local cache of jsl */

    IOTRACE(("JZEROHDR %p\n", pPager))
    if( doTruncate || iLimit==0 ){
      rc = sqlite3OsTruncate(pPager->jfd, 0);
    }else{
      static const char zeroHdr[28] = {0};
      rc = sqlite3OsWrite(pPager->jfd, zeroHdr, sizeof(zeroHdr), 0);
    }
    if( rc==SQLITE_OK && !pPager->noSync ){
      rc = sqlite3OsSync(pPager->jfd, SQLITE_SYNC_DATAONLY|pPager->sync_flags);
    }

    /* At this point the transaction is committed but the write lock 

** - 4 bytes: Initial database page count.
** - 4 bytes: Sector size used by the process that wrote this journal.
** - 4 bytes: Database page size.
** 
** Followed by (JOURNAL_HDR_SZ - 28) bytes of unused space.
*/
static int writeJournalHdr(Pager *pPager){
  int rc = SQLITE_OK;                 /* Return code */
  char *zHeader = pPager->pTmpSpace;  /* Temporary space used to build header */
  u32 nHeader = pPager->pageSize;     /* Size of buffer pointed to by zHeader */
  u32 nWrite;                         /* Bytes of header sector written */
  int ii;                             /* Loop counter */

  assert( isOpen(pPager->jfd) );      /* Journal file must be open. */

  if( nHeader>JOURNAL_HDR_SZ(pPager) ){
    nHeader = JOURNAL_HDR_SZ(pPager);
  }

  /* If there are active savepoints and any of them were created 
  ** since the most recent journal header was written, update the 
  ** PagerSavepoint.iHdrOffset fields now.
  */
  for(ii=0; ii<pPager->nSavepoint; ii++){
    if( pPager->aSavepoint[ii].iHdrOffset==0 ){
      pPager->aSavepoint[ii].iHdrOffset = pPager->journalOff;
    }
  }


  pPager->journalHdr = pPager->journalOff = journalHdrOffset(pPager);

  memcpy(zHeader, aJournalMagic, sizeof(aJournalMagic));

  /* 
  ** Write the nRec Field - the number of page records that follow this
  ** journal header. Normally, zero is written to this value at this time.
  ** After the records are added to the journal (and the journal synced, 
  ** if in full-sync mode), the zero is overwritten with the true number

  **
  **   * When the pager is in no-sync mode. Corruption can follow a
  **     power failure in this case anyway.
  **
  **   * When the SQLITE_IOCAP_SAFE_APPEND flag is set. This guarantees
  **     that garbage data is never appended to the journal file.
  */
  assert( isOpen(pPager->fd) || pPager->noSync );
  if( (pPager->noSync) || (pPager->journalMode==PAGER_JOURNALMODE_MEMORY)
   || (sqlite3OsDeviceCharacteristics(pPager->fd)&SQLITE_IOCAP_SAFE_APPEND) 
  ){
    put32bits(&zHeader[sizeof(aJournalMagic)], 0xffffffff);
  }else{
    put32bits(&zHeader[sizeof(aJournalMagic)], 0);
  }

  /* The random check-hash initialiser */ 
  sqlite3_randomness(sizeof(pPager->cksumInit), &pPager->cksumInit);
  put32bits(&zHeader[sizeof(aJournalMagic)+4], pPager->cksumInit);
  /* The initial database size */
  put32bits(&zHeader[sizeof(aJournalMagic)+8], pPager->dbOrigSize);
  /* The assumed sector size for this process */
  put32bits(&zHeader[sizeof(aJournalMagic)+12], pPager->sectorSize);

  /* The page size */
  put32bits(&zHeader[sizeof(aJournalMagic)+16], pPager->pageSize);

  /* Initializing the tail of the buffer is not necessary.  Everything
  ** works find if the following memset() is omitted.  But initializing
  ** the memory prevents valgrind from complaining, so we are willing to
  ** take the performance hit.
  */
  memset(&zHeader[sizeof(aJournalMagic)+20], 0,
         nHeader-(sizeof(aJournalMagic)+20));

  /* In theory, it is only necessary to write the 28 bytes that the 
  ** journal header consumes to the journal file here. Then increment the 
  ** Pager.journalOff variable by JOURNAL_HDR_SZ so that the next 
  ** record is written to the following sector (leaving a gap in the file
  ** that will be implicitly filled in by the OS).
  **
  ** However it has been discovered that on some systems this pattern can 
  ** be significantly slower than contiguously writing data to the file,
  ** even if that means explicitly writing data to the block of 
  ** (JOURNAL_HDR_SZ - 28) bytes that will not be used. So that is what
  ** is done. 
  **
  ** The loop is required here in case the sector-size is larger than the 
  ** database page size. Since the zHeader buffer is only Pager.pageSize


  ** bytes in size, more than one call to sqlite3OsWrite() may be required
  ** to populate the entire journal header sector.
  */ 
  for(nWrite=0; rc==SQLITE_OK&&nWrite<JOURNAL_HDR_SZ(pPager); nWrite+=nHeader){
    IOTRACE(("JHDR %p %lld %d\n", pPager, pPager->journalHdr, nHeader))
    rc = sqlite3OsWrite(pPager->jfd, zHeader, nHeader, pPager->journalOff);
    pPager->journalOff += nHeader;
  }

  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
** file. The current location in the journal file is given by
** pPager->journalOff. See comments above function writeJournalHdr() for
** a description of the journal header format.
**
** If the header is read successfully, *pNRec is set to the number of
** page records following this header and *pDbSize is set to the size of the
** database before the transaction began, in pages. Also, pPager->cksumInit
** is set to the value read from the journal header. SQLITE_OK is returned
** in this case.
**
** If the journal header file appears to be corrupted, SQLITE_DONE is
** returned and *pNRec and *PDbSize are undefined.  If JOURNAL_HDR_SZ bytes
** cannot be read from the journal file an error code is returned.
*/
static int readJournalHdr(
  Pager *pPager,               /* Pager object */
  i64 journalSize,             /* Size of the open journal file in bytes */
  u32 *pNRec,                  /* OUT: Value read from the nRec field */
  u32 *pDbSize                 /* OUT: Value of original database size field */
){
  int rc;                      /* Return code */
  unsigned char aMagic[8];     /* A buffer to hold the magic header */
  i64 iHdrOff;                 /* Offset of journal header being read */

  assert( isOpen(pPager->jfd) );      /* Journal file must be open. */

  /* Advance Pager.journalOff to the start of the next sector. If the
  ** journal file is too small for there to be a header stored at this
  ** point, return SQLITE_DONE.
  */
  pPager->journalOff = journalHdrOffset(pPager);
  if( pPager->journalOff+JOURNAL_HDR_SZ(pPager) > journalSize ){
    return SQLITE_DONE;
  }
  iHdrOff = pPager->journalOff;

  /* Read in the first 8 bytes of the journal header. If they do not match
  ** the  magic string found at the start of each journal header, return
  ** SQLITE_DONE. If an IO error occurs, return an error code. Otherwise,
  ** proceed.
  */
  rc = sqlite3OsRead(pPager->jfd, aMagic, sizeof(aMagic), iHdrOff);
  if( rc ){
    return rc;
  }
  if( memcmp(aMagic, aJournalMagic, sizeof(aMagic))!=0 ){
    return SQLITE_DONE;
  }

  /* Read the first three 32-bit fields of the journal header: The nRec
  ** field, the checksum-initializer and the database size at the start
  ** of the transaction. Return an error code if anything goes wrong.
  */
  if( SQLITE_OK!=(rc = read32bits(pPager->jfd, iHdrOff+8, pNRec))

   || SQLITE_OK!=(rc = read32bits(pPager->jfd, iHdrOff+12, &pPager->cksumInit))
   || SQLITE_OK!=(rc = read32bits(pPager->jfd, iHdrOff+16, pDbSize))
  ){
    return rc;
  }



  if( pPager->journalOff==0 ){
    u32 iPageSize;               /* Page-size field of journal header */
    u32 iSectorSize;             /* Sector-size field of journal header */
    u16 iPageSize16;             /* Copy of iPageSize in 16-bit variable */

    /* Read the page-size and sector-size journal header fields. */
    if( SQLITE_OK!=(rc = read32bits(pPager->jfd, iHdrOff+20, &iSectorSize))
     || SQLITE_OK!=(rc = read32bits(pPager->jfd, iHdrOff+24, &iPageSize))
    ){
      return rc;
    }

    /* Check that the values read from the page-size and sector-size fields
    ** are within range. To be 'in range', both values need to be a power
    ** of two greater than or equal to 512, and not greater than their 
    ** respective compile time maximum limits.
    */
    if( iPageSize<512                  || iSectorSize<512
     || iPageSize>SQLITE_MAX_PAGE_SIZE || iSectorSize>MAX_SECTOR_SIZE
     || ((iPageSize-1)&iPageSize)!=0   || ((iSectorSize-1)&iSectorSize)!=0 
    ){
      /* If the either the page-size or sector-size in the journal-header is 
      ** invalid, then the process that wrote the journal-header must have 
      ** crashed before the header was synced. In this case stop reading 
      ** the journal file here.
      */
      return SQLITE_DONE;
    }

    /* Update the page-size to match the value read from the journal. 
    ** Use a testcase() macro to make sure that malloc failure within 
    ** PagerSetPagesize() is tested.
    */
    iPageSize16 = (u16)iPageSize;
    rc = sqlite3PagerSetPagesize(pPager, &iPageSize16);
    testcase( rc!=SQLITE_OK );

    assert( rc!=SQLITE_OK || iPageSize16==(u16)iPageSize );

    /* 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.
    */








    pPager->sectorSize = iSectorSize;
  }

  pPager->journalOff += JOURNAL_HDR_SZ(pPager);
  return rc;
}


/*
** 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
** journal file descriptor is advanced to the next sector boundary before
** anything is written. The format is:
**
**   + 4 bytes: PAGER_MJ_PGNO.
**   + N bytes: Master journal filename in utf-8.
**   + 4 bytes: N (length of master journal name in bytes, no nul-terminator).
**   + 4 bytes: Master journal name checksum.
**   + 8 bytes: aJournalMagic[].
**
** The master journal page checksum is the sum of the bytes in the master
** journal name, where each byte is interpreted as a signed 8-bit integer.
**
** 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;                          /* Return code */

  int nMaster;                     /* Length of string zMaster */
  i64 iHdrOff;                     /* Offset of header in journal file */
  i64 jrnlSize;                    /* Size of journal file on disk */
  u32 cksum = 0;                   /* Checksum of string zMaster */

  assert( isOpen(pPager->jfd) );
  assert( !pPager->setMaster );

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

  /* Calculate the length in bytes and the checksum of zMaster */
  for(nMaster=0; zMaster[nMaster]; nMaster++){
    cksum += zMaster[nMaster];
  }

  /* 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 ){
    pPager->journalOff = journalHdrOffset(pPager);
  }
  iHdrOff = pPager->journalOff;


  /* Write the master journal data to the end of the journal file. If
  ** an error occurs, return the error code to the caller.
  */
  if( (rc = write32bits(pPager->jfd, iHdrOff, PAGER_MJ_PGNO(pPager)))
   || (rc = sqlite3OsWrite(pPager->jfd, zMaster, nMaster, iHdrOff+4))
   || (rc = write32bits(pPager->jfd, iHdrOff+4+nMaster, nMaster))


   || (rc = write32bits(pPager->jfd, iHdrOff+4+nMaster+4, cksum))
   || (rc = sqlite3OsWrite(pPager->jfd, aJournalMagic, 8, iHdrOff+4+nMaster+8))
  ){
    return rc;

  }




  pPager->journalOff += (nMaster+20);
  pPager->needSync = !pPager->noSync;

  /* If the pager is in peristent-journal mode, then the physical 
  ** journal-file may extend past the end of the master-journal name
  ** and 8 bytes of magic data just written to the file. This is 
  ** dangerous because the code to rollback a hot-journal file
  ** will not be able to find the master-journal name to determine 
  ** whether or not the journal is hot. 
  **
  ** Easiest thing to do in this scenario is to truncate the journal 
  ** file to the required size.
  */ 

  if( SQLITE_OK==(rc = sqlite3OsFileSize(pPager->jfd, &jrnlSize))
   && jrnlSize>pPager->journalOff
  ){
    rc = sqlite3OsTruncate(pPager->jfd, pPager->journalOff);
  }
  return rc;
}

/*
** Find a page in the hash table given its page number. Return
** a pointer to the page or NULL if the requested page is not 
** already in memory.
*/
static PgHdr *pager_lookup(Pager *pPager, Pgno pgno){
  PgHdr *p;                         /* Return value */

  /* It is not possible for a call to PcacheFetch() with createFlag==0 to
  ** fail, since no attempt to allocate dynamic memory will be made.
  */
  (void)sqlite3PcacheFetch(pPager->pPCache, pgno, 0, &p);
  return p;
}

/*
** Unless the pager is in error-state, discard all in-memory pages. If
** the pager is in error-state, then this call is a no-op.


*/
static void pager_reset(Pager *pPager){
  if( SQLITE_OK==pPager->errCode ){
    sqlite3PcacheClear(pPager->pPCache);
  }
}

/*
** Free all structures in the Pager.aSavepoint[] array and set both
** Pager.aSavepoint and Pager.nSavepoint to zero. Close the sub-journal
** if it is open and the pager is not in exclusive mode.
*/
static void releaseAllSavepoints(Pager *pPager){
  int ii;               /* Iterator for looping through Pager.aSavepoint */
  for(ii=0; ii<pPager->nSavepoint; ii++){
    sqlite3BitvecDestroy(pPager->aSavepoint[ii].pInSavepoint);
  }
  if( !pPager->exclusiveMode ){
    sqlite3OsClose(pPager->sjfd);
  }
  sqlite3_free(pPager->aSavepoint);
  pPager->aSavepoint = 0;
  pPager->nSavepoint = 0;
  pPager->nSubRec = 0;
}

/*
** Set the bit number pgno in the PagerSavepoint.pInSavepoint 
** bitvecs of all open savepoints. Return SQLITE_OK if successful
** or SQLITE_NOMEM if a malloc failure occurs.
*/
static int addToSavepointBitvecs(Pager *pPager, Pgno pgno){
  int ii;                   /* Loop counter */
  int rc = SQLITE_OK;       /* Result code */

  for(ii=0; ii<pPager->nSavepoint; ii++){
    PagerSavepoint *p = &pPager->aSavepoint[ii];
    if( pgno<=p->nOrig ){
      rc |= sqlite3BitvecSet(p->pInSavepoint, pgno);
      testcase( rc==SQLITE_NOMEM );
      assert( rc==SQLITE_OK || rc==SQLITE_NOMEM );
    }
  }
  return rc;
}

/*
** Unlock the database file. This function is a no-op if the pager
** is in exclusive mode.
**
** If the pager is currently in error state, discard the contents of 
** the cache and reset the Pager structure internal state. If there is
** an open journal-file, then the next time a shared-lock is obtained
** on the pager file (by this or any other process), it will be
** treated as a hot-journal and rolled back.
*/
static void pager_unlock(Pager *pPager){
  if( !pPager->exclusiveMode ){
    int rc;                      /* Return code */

    /* Always close the journal file when dropping the database lock.
    ** Otherwise, another connection with journal_mode=delete might
    ** delete the file out from under us.
    */

    sqlite3OsClose(pPager->jfd);

    sqlite3BitvecDestroy(pPager->pInJournal);
    pPager->pInJournal = 0;

    releaseAllSavepoints(pPager);

    /* If the file is unlocked, somebody else might change it. The
    ** values stored in Pager.dbSize etc. might become invalid if
    ** this happens. TODO: Really, this doesn't need to be cleared
    ** until the change-counter check fails in pagerSharedLock().
    */
    pPager->dbSizeValid = 0;

    rc = osUnlock(pPager->fd, NO_LOCK);
    if( rc ){
      pPager->errCode = rc;
    }
    IOTRACE(("UNLOCK %p\n", pPager))

    /* If Pager.errCode is set, the contents of the pager cache cannot be
    ** trusted. Now that the pager file is unlocked, the contents of the
    ** cache can be discarded and the error code safely cleared.
    */
    if( pPager->errCode ){
      if( rc==SQLITE_OK ){
        pPager->errCode = SQLITE_OK;
      }
      pager_reset(pPager);




    }

    pPager->changeCountDone = 0;
    pPager->state = PAGER_UNLOCK;

  }
}

/*
** Execute a rollback if a transaction is active and unlock the 
** database file. 
**
** If the pager has already entered the error state, do not attempt 
** the rollback at this time. Instead, pager_unlock() is called. The
** call to pager_unlock() will discard all in-memory pages, unlock
** the database file and clear the error state. If this means that
** there is a hot-journal left in the file-system, the next connection
** to obtain a shared lock on the pager (which may be this one) will
** roll it back.
**
** If the pager has not already entered the error state, but an IO or
** malloc error occurs during a rollback, then this will itself cause 
** the pager to enter the error state. Which will be cleared by the
** call to pager_unlock(), as described above.
*/
static void pagerUnlockAndRollback(Pager *pPager){
  if( pPager->errCode==SQLITE_OK && pPager->state>=PAGER_RESERVED ){
    sqlite3BeginBenignMalloc();
    sqlite3PagerRollback(pPager);
    sqlite3EndBenignMalloc();
  }
  pager_unlock(pPager);
}

/*
** This routine ends a transaction. A transaction is usually ended by 
** either a COMMIT or a ROLLBACK operation. This routine may be called 
** after rollback of a hot-journal, or if an error occurs while opening
** the journal file or writing the very first journal-header of a
** database transaction.
** 
** If the pager is in PAGER_SHARED or PAGER_UNLOCK state when this
** routine is called, it is a no-op (returns SQLITE_OK).
**
** Otherwise, any active savepoints are released.
**
** If the journal file is open, then it is "finalized". Once a journal 
** file has been finalized it is not possible to use it to roll back a 
** transaction. Nor will it be considered to be a hot-journal by this

** or any other database connection. Exactly how a journal is finalized

** depends on whether or not the pager is running in exclusive mode and
** the current journal-mode (Pager.journalMode value), as follows:

**
**   journalMode==MEMORY
**     Journal file descriptor is simply closed. This destroys an 
**     in-memory journal.
**
**   journalMode==TRUNCATE
**     Journal file is truncated to zero bytes in size.
**
**   journalMode==PERSIST
**     The first 28 bytes of the journal file are zeroed. This invalidates
**     the first journal header in the file, and hence the entire journal
**     file. An invalid journal file cannot be rolled back.
**
**   journalMode==DELETE
**     The journal file is closed and deleted using sqlite3OsDelete().
**
**     If the pager is running in exclusive mode, this method of finalizing
**     the journal file is never used. Instead, if the journalMode is
**     DELETE and the pager is in exclusive mode, the method described under
**     journalMode==PERSIST is used instead.
**
** After the journal is finalized, if running in non-exclusive mode, the
** pager moves to PAGER_SHARED state (and downgrades the lock on the
** database file accordingly).
**
** If the pager is running in exclusive mode and is in PAGER_SYNCED state,
** it moves to PAGER_EXCLUSIVE. No locks are downgraded when running in
** exclusive mode.
**
** SQLITE_OK is returned if no error occurs. If an error occurs during
** any of the IO operations to finalize the journal file or unlock the
** database then the IO error code is returned to the user. If the 

** operation to finalize the journal file fails, then the code still
** tries to unlock the database file if not in exclusive mode. If the
** unlock operation fails as well, then the first error code related
** to the first error encountered (the journal finalization one) is
** returned.
*/
static int pager_end_transaction(Pager *pPager, int hasMaster){
  int rc = SQLITE_OK;      /* Error code from journal finalization operation */
  int rc2 = SQLITE_OK;     /* Error code from db file unlock operation */

  if( pPager->state<PAGER_RESERVED ){
    return SQLITE_OK;
  }
  releaseAllSavepoints(pPager);

  assert( isOpen(pPager->jfd) || pPager->pInJournal==0 );
  if( isOpen(pPager->jfd) ){

    /* TODO: There's a problem here if a journal-file was opened in MEMORY
    ** mode and then the journal-mode is changed to TRUNCATE or PERSIST
    ** during the transaction. This code should be changed to assume
    ** that the journal mode has not changed since the transaction was
    ** started. And the sqlite3PagerJournalMode() function should be
    ** changed to make sure that this is the case too.
    */

    /* Finalize the journal file. */
    if( pPager->journalMode==PAGER_JOURNALMODE_MEMORY ){
      int isMemoryJournal = sqlite3IsMemJournal(pPager->jfd);
      sqlite3OsClose(pPager->jfd);

      if( !isMemoryJournal ){
        rc = sqlite3OsDelete(pPager->pVfs, pPager->zJournal, 0);
      }
    }else if( pPager->journalMode==PAGER_JOURNALMODE_TRUNCATE
         && (rc = sqlite3OsTruncate(pPager->jfd, 0))==SQLITE_OK ){
      pPager->journalOff = 0;
      pPager->journalStarted = 0;
    }else if( pPager->exclusiveMode 
     || pPager->journalMode==PAGER_JOURNALMODE_PERSIST
    ){
      rc = zeroJournalHdr(pPager, hasMaster);
      pager_error(pPager, rc);
      pPager->journalOff = 0;
      pPager->journalStarted = 0;
    }else{
      assert( pPager->journalMode==PAGER_JOURNALMODE_DELETE || rc );
      sqlite3OsClose(pPager->jfd);

      if( rc==SQLITE_OK && !pPager->tempFile ){
        rc = sqlite3OsDelete(pPager->pVfs, pPager->zJournal, 0);
      }
    }




#ifdef SQLITE_CHECK_PAGES
    sqlite3PcacheIterateDirty(pPager->pPCache, pager_set_pagehash);
#endif

    sqlite3PcacheCleanAll(pPager->pPCache);
    sqlite3BitvecDestroy(pPager->pInJournal);
    pPager->pInJournal = 0;
    pPager->nRec = 0;


  }

  if( !pPager->exclusiveMode ){
    rc2 = osUnlock(pPager->fd, SHARED_LOCK);
    pPager->state = PAGER_SHARED;
    pPager->changeCountDone = 0;
  }else if( pPager->state==PAGER_SYNCED ){
    pPager->state = PAGER_EXCLUSIVE;
  }

  pPager->setMaster = 0;
  pPager->needSync = 0;
  pPager->dbModified = 0;

  /* TODO: Is this optimal? Why is the db size invalidated here 
  ** when the database file is not unlocked? */
  pPager->dbOrigSize = 0;
  sqlite3PcacheTruncate(pPager->pPCache, pPager->dbSize);
  if( !MEMDB ){
    pPager->dbSizeValid = 0;
  }


  return (rc==SQLITE_OK?rc2:rc);
}

/*
** Parameter aData must point to a buffer of pPager->pageSize bytes
** of data. Compute and return a checksum based ont the contents of the 
** page of data and the current value of pPager->cksumInit.
**
** This is not a real checksum. It is really just the sum of the 
** random initial value (pPager->cksumInit) and every 200th byte
** of the page data, starting with byte offset (pPager->pageSize%200).
** Each byte is interpreted as an 8-bit unsigned integer.
**

** Changing the formula used to compute this checksum results in an
** incompatible journal file format.
**
** If journal corruption occurs due to a power failure, the most likely 
** scenario is that one end or the other of the record will be changed. 
** It is much less likely that the two ends of the journal record will be
** correct and the middle be corrupt.  Thus, this "checksum" scheme,
** though fast and simple, catches the mostly likely kind of corruption.





*/
static u32 pager_cksum(Pager *pPager, const u8 *aData){
  u32 cksum = pPager->cksumInit;         /* Checksum value to return */
  int i = pPager->pageSize-200;          /* Loop counter */
  while( i>0 ){
    cksum += aData[i];
    i -= 200;
  }
  return cksum;
}

/*
** Read a single page from either the journal file (if isMainJrnl==1) or
** from the sub-journal (if isMainJrnl==0) and playback that page.
** The page begins at offset *pOffset into the file. The *pOffset
** value is increased to the start of the next page in the journal.
**
** The isMainJrnl flag is true if this is the main rollback journal and
** false for the statement journal.  The main rollback journal uses
** checksums - the statement journal does not.
**
** If the page number of the page record read from the (sub-)journal file
** is greater than the current value of Pager.dbSize, then playback is
** skipped and SQLITE_OK is returned.
**
** If pDone is not NULL, then it is a record of pages that have already
** been played back.  If the page at *pOffset has already been played back
** (if the corresponding pDone bit is set) then skip the playback.
** Make sure the pDone bit corresponding to the *pOffset page is set
** prior to returning.
**
** If the page record is successfully read from the (sub-)journal file
** and played back, then SQLITE_OK is returned. If an IO error occurs
** while reading the record from the (sub-)journal file or while writing
** to the database file, then the IO error code is returned. If data
** is successfully read from the (sub-)journal file but appears to be
** corrupted, SQLITE_DONE is returned. Data is considered corrupted in
** two circumstances:
** 
**   * If the record page-number is illegal (0 or PAGER_MJ_PGNO), or
**   * If the record is being rolled back from the main journal file
**     and the checksum field does not match the record content.
**
** Neither of these two scenarios are possible during a savepoint rollback.
**
** If this is a savepoint rollback, then memory may have to be dynamically
** allocated by this function. If this is the case and an allocation fails,
** SQLITE_NOMEM is returned.
*/
static int pager_playback_one_page(
  Pager *pPager,                /* The pager being played back */
  int isMainJrnl,               /* 1 -> main journal. 0 -> sub-journal. */
  i64 *pOffset,                 /* Offset of record to playback */
  int isSavepnt,                /* True for a savepoint rollback */
  Bitvec *pDone                 /* Bitvec of pages already played back */

  assert( (isSavepnt&~1)==0 );       /* isSavepnt is 0 or 1 */
  assert( isMainJrnl || pDone );     /* pDone always used on sub-journals */
  assert( isSavepnt || pDone==0 );   /* pDone never used on non-savepoint */

  aData = (u8*)pPager->pTmpSpace;
  assert( aData );         /* Temp storage must have already been allocated */

  /* Read the page number and page data from the journal or sub-journal
  ** file. Return an error code to the caller if an IO error occurs.
  */
  jfd = isMainJrnl ? pPager->jfd : pPager->sjfd;

  rc = read32bits(jfd, *pOffset, &pgno);
  if( rc!=SQLITE_OK ) return rc;
  rc = sqlite3OsRead(jfd, aData, pPager->pageSize, (*pOffset)+4);
  if( rc!=SQLITE_OK ) return rc;
  *pOffset += pPager->pageSize + 4 + isMainJrnl*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) ){
    assert( !isSavepnt );
    return SQLITE_DONE;
  }
  if( pgno>(Pgno)pPager->dbSize || sqlite3BitvecTest(pDone, pgno) ){
    return SQLITE_OK;
  }
  if( isMainJrnl ){
    rc = read32bits(jfd, (*pOffset)-4, &cksum);
    if( rc ) return rc;
    if( !isSavepnt && pager_cksum(pPager, aData)!=cksum ){
      return SQLITE_DONE;
    }
  }

  if( pDone && (rc = sqlite3BitvecSet(pDone, pgno)) ){
    return rc;
  }

  assert( pPager->state==PAGER_RESERVED || pPager->state>=PAGER_EXCLUSIVE );

  /* If the pager is in RESERVED state, then there must be a copy of this

  pPg = pager_lookup(pPager, pgno);
  PAGERTRACE(("PLAYBACK %d page %d hash(%08x) %s\n",
               PAGERID(pPager), pgno, pager_datahash(pPager->pageSize, aData),
               (isMainJrnl?"main-journal":"sub-journal")
  ));
  if( (pPager->state>=PAGER_EXCLUSIVE)
   && (pPg==0 || 0==(pPg->flags&PGHDR_NEED_SYNC))
   && isOpen(pPager->fd)
  ){
    i64 ofst = (pgno-1)*(i64)pPager->pageSize;
    rc = sqlite3OsWrite(pPager->fd, aData, pPager->pageSize, ofst);
    if( pgno>pPager->dbFileSize ){
      pPager->dbFileSize = pgno;
    }
  }else if( !isMainJrnl && pPg==0 ){

    */
    void *pData;
    pData = pPg->pData;
    memcpy(pData, aData, pPager->pageSize);
    if( pPager->xReiniter ){
      pPager->xReiniter(pPg);
    }
    if( isMainJrnl && (!isSavepnt || *pOffset<=pPager->journalHdr) ){
      /* If the contents of this page were just restored from the main 
      ** journal file, then its content must be as they were when the 
      ** transaction was first opened. In this case we can mark the page
      ** as clean, since there will be no need to write it out to the.
      **
      ** There is one exception to this rule. If the page is being rolled
      ** back as part of a savepoint (or statement) rollback from an 

** file that referred to the master journal file has just been rolled back.
** This routine checks if it is possible to delete the master journal file,
** and does so if it is.
**
** Argument zMaster may point to Pager.pTmpSpace. So that buffer is not 
** available for use within this function.
**
** When a master journal file is created, it is populated with the names 
** of all of its child journals, one after another, formatted as utf-8 
** encoded text. The end of each child journal file is marked with a 
** nul-terminator byte (0x00). i.e. the entire contents of a master journal
** file for a transaction involving two databases might be:
**
**   "/home/bill/a.db-journal\x00/home/bill/b.db-journal\x00"
**
** A master journal file may only be deleted once all of its child 
** journals have been rolled back.
**
** This function reads the contents of the master-journal file into 
** memory and loops through each of the child journal names. For
** each child journal, it checks if:
**
**   * if the child journal exists, and if so
**   * if the child journal contains a reference to master journal 
**     file zMaster
**
** If a child journal can be found that matches both of the criteria
** above, this function returns without doing anything. Otherwise, if
** no such child journal can be found, file zMaster is deleted from
** the file-system using sqlite3OsDelete().
**
** If an IO error within this function, an error code is returned. This
** function allocates memory by calling sqlite3Malloc(). If an allocation
** fails, SQLITE_NOMEM is returned. Otherwise, if no IO or malloc errors 
** occur, SQLITE_OK is returned.
**
** TODO: This function allocates a single block of memory to load
** the entire contents of the master journal file. This could be
** a couple of kilobytes or so - potentially larger than the page 
** size.
*/
static int pager_delmaster(Pager *pPager, const char *zMaster){
  sqlite3_vfs *pVfs = pPager->pVfs;
  int rc;                   /* Return code */

  sqlite3_file *pMaster;    /* Malloc'd master-journal file descriptor */
  sqlite3_file *pJournal;   /* Malloc'd child-journal file descriptor */
  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.
  */
  pMaster = (sqlite3_file *)sqlite3Malloc(pVfs->szOsFile * 2);
  pJournal = (sqlite3_file *)(((u8 *)pMaster) + pVfs->szOsFile);
  if( !pMaster ){
    rc = SQLITE_NOMEM;
  }else{
    int flags = (SQLITE_OPEN_READONLY|SQLITE_OPEN_MASTER_JOURNAL);
    rc = sqlite3OsOpen(pVfs, zMaster, pMaster, flags, 0);
  }
  if( rc!=SQLITE_OK ) goto delmaster_out;


  rc = sqlite3OsFileSize(pMaster, &nMasterJournal);
  if( rc!=SQLITE_OK ) goto delmaster_out;

  if( nMasterJournal>0 ){
    char *zJournal;
    char *zMasterPtr = 0;
    int nMasterPtr = pVfs->mxPathname+1;

    /* Load the entire master journal file into space obtained from
    ** sqlite3_malloc() and pointed to by zMasterJournal. 
    */
    zMasterJournal = (char *)sqlite3Malloc((int)nMasterJournal + nMasterPtr);
    if( !zMasterJournal ){
      rc = SQLITE_NOMEM;

  
  rc = sqlite3OsDelete(pVfs, zMaster, 0);

delmaster_out:
  if( zMasterJournal ){
    sqlite3_free(zMasterJournal);
  }  
  if( pMaster ){
    sqlite3OsClose(pMaster);
    assert( !isOpen(pJournal) );
  }
  sqlite3_free(pMaster);
  return rc;
}


/*
** This function is used to change the actual size of the database 
** file in the file-system. This only happens when committing a transaction,
** or rolling back a transaction (including rolling back a hot-journal).
**
** If the main database file is not open, or an exclusive lock is not
** held, this function is a no-op. Otherwise, the size of the file is
** changed to nPage pages (nPage*pPager->pageSize bytes). If the file
** on disk is currently larger than nPage pages, then use the VFS
** xTruncate() method to truncate it.
**
** Or, it might might be the case that the file on disk is smaller than 



** nPage pages. Some operating system implementations can get confused if 
** you try to truncate a file to some size that is larger than it 
** currently is, so detect this case and write a single zero byte to 
** the end of the new file instead.
**
** If successful, return SQLITE_OK. If an IO error occurs while modifying
** the database file, return the error code to the caller.
*/
static int pager_truncate(Pager *pPager, Pgno nPage){
  int rc = SQLITE_OK;
  if( pPager->state>=PAGER_EXCLUSIVE && isOpen(pPager->fd) ){
    i64 currentSize, newSize;
    /* TODO: Is it safe to use Pager.dbFileSize here? */
    rc = sqlite3OsFileSize(pPager->fd, &currentSize);
    newSize = pPager->pageSize*(i64)nPage;
    if( rc==SQLITE_OK && currentSize!=newSize ){
      if( currentSize>newSize ){
        rc = sqlite3OsTruncate(pPager->fd, newSize);
      }else{
        rc = sqlite3OsWrite(pPager->fd, "", 1, newSize-1);
      }
      if( rc==SQLITE_OK ){
        pPager->dbFileSize = nPage;
      }
    }
  }
  return rc;
}

/*
** Set the value of the Pager.sectorSize variable for the given
** pager based on the value returned by the xSectorSize method
** of the open database file. The sector size will be used used 
** to determine the size and alignment of journal header and 
** master journal pointers within created journal files.
**
** For temporary files the effective sector size is always 512 bytes.
**
** Otherwise, for non-temporary files, the effective sector size is
** the value returned by the xSectorSize() method rounded up to 512 if
** it is less than 512, or rounded down to MAX_SECTOR_SIZE if it
** is greater than MAX_SECTOR_SIZE.
*/
static void setSectorSize(Pager *pPager){
  assert( isOpen(pPager->fd) || pPager->tempFile );

  if( !pPager->tempFile ){
    /* Sector size doesn't matter for temporary files. Also, the file
    ** may not have been opened yet, in which case the OsSectorSize()
    ** call will segfault.
    */
    pPager->sectorSize = sqlite3OsSectorSize(pPager->fd);
  }
  if( pPager->sectorSize<512 ){
    pPager->sectorSize = 512;
  }
  if( pPager->sectorSize>MAX_SECTOR_SIZE ){
    assert( MAX_SECTOR_SIZE>=512 );
    pPager->sectorSize = MAX_SECTOR_SIZE;
  }
}

/*
** Playback the journal and thus restore the database file to
** the state it was in before we started making changes.  

  int rc;                  /* Result code of a subroutine */
  int res = 1;             /* Value returned by sqlite3OsAccess() */
  char *zMaster = 0;       /* Name of master journal file if any */

  /* Figure out how many records are in the journal.  Abort early if
  ** the journal is empty.
  */
  assert( isOpen(pPager->jfd) );
  rc = sqlite3OsFileSize(pPager->jfd, &szJ);
  if( rc!=SQLITE_OK || szJ==0 ){
    goto end_playback;
  }

  /* Read the master journal name from the journal, if it is present.
  ** If a master journal file name is specified, but the file is not
  ** present on disk, then the journal is not hot and does not need to be
  ** played back.
  **
  ** TODO: Technically the following is an error because it assumes that
  ** buffer Pager.pTmpSpace is (mxPathname+1) bytes or larger. i.e. that
  ** (pPager->pageSize >= pPager->pVfs->mxPathname+1). Using os_unix.c,
  **  mxPathname is 512, which is the same as the minimum allowable value
  ** for pageSize.
  */
  zMaster = pPager->pTmpSpace;
  rc = readMasterJournal(pPager->jfd, zMaster, pPager->pVfs->mxPathname+1);
  if( rc==SQLITE_OK && zMaster[0] ){
    rc = sqlite3OsAccess(pVfs, zMaster, SQLITE_ACCESS_EXISTS, &res);
  }
  zMaster = 0;
  if( rc!=SQLITE_OK || !res ){
    goto end_playback;
  }
  pPager->journalOff = 0;

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

    /* Read the next journal header from the journal file.  If there are
    ** not enough bytes left in the journal file for a complete header, or
    ** it is corrupted, then a process must of failed while writing it.
    ** This indicates nothing more needs to be rolled back.
    */

      rc = pager_truncate(pPager, mxPg);
      if( rc!=SQLITE_OK ){
        goto end_playback;
      }
      pPager->dbSize = mxPg;
    }

    /* Copy original pages out of the journal and back into the 
    ** database file and/or page cache.
    */
    for(u=0; u<nRec; u++){
      rc = pager_playback_one_page(pPager, 1, &pPager->journalOff, 0, 0);
      if( rc!=SQLITE_OK ){
        if( rc==SQLITE_DONE ){
          rc = SQLITE_OK;
          pPager->journalOff = szJ;

    pPager->fd->pMethods==0 ||
    sqlite3OsFileControl(pPager->fd,SQLITE_FCNTL_DB_UNCHANGED,0)>=SQLITE_OK
  );

  if( rc==SQLITE_OK ){
    zMaster = pPager->pTmpSpace;
    rc = readMasterJournal(pPager->jfd, zMaster, pPager->pVfs->mxPathname+1);
    testcase( rc!=SQLITE_OK );
  }
  if( rc==SQLITE_OK ){
    rc = pager_end_transaction(pPager, zMaster[0]!='\0');
    testcase( rc!=SQLITE_OK );
  }
  if( rc==SQLITE_OK && zMaster[0] && res ){
    /* If there was a master journal and this routine will return success,
    ** see if it is possible to delete the master journal.
    */
    rc = pager_delmaster(pPager, zMaster);
    testcase( rc!=SQLITE_OK );
  }

  /* The Pager.sectorSize variable may have been updated while rolling
  ** back a journal created by a process with a different sector size
  ** value. Reset it to the correct value for this process.
  */
  setSectorSize(pPager);
  return rc;
}

/*
** Playback savepoint pSavepoint. Or, if pSavepoint==NULL, then playback
** the entire master journal file. The case pSavepoint==NULL occurs when 
** a ROLLBACK TO command is invoked on a SAVEPOINT that is a transaction 
** savepoint.
**

** When pSavepoint is not NULL (meaning a non-transaction savepoint is 
** being rolled back), then the rollback consists of up to three stages,
** performed in the order specified:
**
**   * Pages are played back from the main journal starting at byte
**     offset PagerSavepoint.iOffset and continuing to 
**     PagerSavepoint.iHdrOffset, or to the end of the main journal
**     file if PagerSavepoint.iHdrOffset is zero.
**
**   * If PagerSavepoint.iHdrOffset is not zero, then pages are played
**     back starting from the journal header immediately following 
**     PagerSavepoint.iHdrOffset to the end of the main journal file.
**
**   * Pages are then played back from the sub-journal file, starting
**     with the PagerSavepoint.iSubRec and continuing to the end of
**     the journal file.
**
** Throughout the rollback process, each time a page is rolled back, the
** corresponding bit is set in a bitvec structure (variable pDone in the
** implementation below). This is used to ensure that a page is only
** rolled back the first time it is encountered in either journal.
**
** If pSavepoint is NULL, then pages are only played back from the main
** journal file. There is no need for a bitvec in this case.
**
** In either case, before playback commences the Pager.dbSize variable
** is reset to the value that it held at the start of the savepoint 
** (or transaction). No page with a page-number greater than this value
** is played back. If one is encountered it is simply skipped.
*/
static int pagerPlaybackSavepoint(Pager *pPager, PagerSavepoint *pSavepoint){
  i64 szJ;                 /* Effective size of the main journal */
  i64 iHdrOff;             /* End of first segment of main-journal records */

  int rc = SQLITE_OK;      /* Return code */
  Bitvec *pDone = 0;       /* Bitvec to ensure pages played back only once */

  assert( pPager->state>=PAGER_SHARED );

  /* Allocate a bitvec to use to store the set of pages rolled back */
  if( pSavepoint ){
    pDone = sqlite3BitvecCreate(pSavepoint->nOrig);
    if( !pDone ){
      return SQLITE_NOMEM;
    }
  }

  /* Set the database size back to the value it was before the savepoint 
  ** being reverted was opened.
  */
  pPager->dbSize = pSavepoint ? pSavepoint->nOrig : pPager->dbOrigSize;


  /* Use pPager->journalOff as the effective size of the main rollback
  ** journal.  The actual file might be larger than this in
  ** PAGER_JOURNALMODE_TRUNCATE or PAGER_JOURNALMODE_PERSIST.  But anything
  ** past pPager->journalOff is off-limits to us.
  */
  szJ = pPager->journalOff;

  /* Begin by rolling back records from the main journal starting at
  ** PagerSavepoint.iOffset and continuing to the next journal header.
  ** There might be records in the main journal that have a page number
  ** greater than the current database size (pPager->dbSize) but those
  ** will be skipped automatically.  Pages are added to pDone as they
  ** are played back.
  */
  if( pSavepoint ){
    iHdrOff = pSavepoint->iHdrOffset ? pSavepoint->iHdrOffset : szJ;
    pPager->journalOff = pSavepoint->iOffset;
    while( rc==SQLITE_OK && pPager->journalOff<iHdrOff ){
      rc = pager_playback_one_page(pPager, 1, &pPager->journalOff, 1, pDone);
    }
    assert( rc!=SQLITE_DONE );

  }else{
    pPager->journalOff = 0;
  }

  /* Continue rolling back records out of the main journal starting at
  ** the first journal header seen and continuing until the effective end
  ** of the main journal file.  Continue to skip out-of-range pages and
  ** continue adding pages rolled back to pDone.
  */
  while( rc==SQLITE_OK && pPager->journalOff<szJ ){
    u32 ii;            /* Loop counter */
    u32 nJRec = 0;     /* Number of Journal Records */
    u32 dummy;
    rc = readJournalHdr(pPager, szJ, &nJRec, &dummy);
    assert( rc!=SQLITE_DONE );

    /*
    ** The "pPager->journalHdr+JOURNAL_HDR_SZ(pPager)==pPager->journalOff"

    if( nJRec==0 
     && pPager->journalHdr+JOURNAL_HDR_SZ(pPager)==pPager->journalOff
    ){
      nJRec = (szJ - pPager->journalOff)/JOURNAL_PG_SZ(pPager);
    }
    for(ii=0; rc==SQLITE_OK && ii<nJRec && pPager->journalOff<szJ; ii++){
      rc = pager_playback_one_page(pPager, 1, &pPager->journalOff, 1, pDone);
    }
    assert( rc!=SQLITE_DONE );

  }
  assert( rc!=SQLITE_OK || pPager->journalOff==szJ );

  /* Finally,  rollback pages from the sub-journal.  Page that were
  ** previously rolled back out of the main journal (and are hence in pDone)
  ** will be skipped.  Out-of-range pages are also skipped.
  */
  if( pSavepoint ){
    u32 ii;            /* Loop counter */
    i64 offset = pSavepoint->iSubRec*(4+pPager->pageSize);
    for(ii=pSavepoint->iSubRec; rc==SQLITE_OK && ii<pPager->nSubRec; ii++){
      assert( offset==ii*(4+pPager->pageSize) );
      rc = pager_playback_one_page(pPager, 0, &offset, 1, pDone);
    }
    assert( rc!=SQLITE_DONE );

  }

  sqlite3BitvecDestroy(pDone);
  if( rc==SQLITE_OK ){
    pPager->journalOff = szJ;
  }
  return rc;

** testing and analysis only.  
*/
#ifdef SQLITE_TEST
int sqlite3_opentemp_count = 0;
#endif

/*
** Open a temporary file.
**
** Write the file descriptor into *pFile. 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.
**
** The flags passed to the VFS layer xOpen() call are those specified
** by parameter vfsFlags ORed with the following:
**
**     SQLITE_OPEN_READWRITE
**     SQLITE_OPEN_CREATE
**     SQLITE_OPEN_EXCLUSIVE
**     SQLITE_OPEN_DELETEONCLOSE
*/
static int pagerOpentemp(
  Pager *pPager,        /* The pager object */
  sqlite3_file *pFile,  /* Write the file descriptor here */
  int vfsFlags          /* Flags passed through to the VFS */
){
  int rc;               /* Return code */

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

  vfsFlags |=  SQLITE_OPEN_READWRITE | SQLITE_OPEN_CREATE |
            SQLITE_OPEN_EXCLUSIVE | SQLITE_OPEN_DELETEONCLOSE;
  rc = sqlite3OsOpen(pPager->pVfs, 0, pFile, vfsFlags, 0);
  assert( rc!=SQLITE_OK || isOpen(pFile) );
  return rc;
}

static int pagerStress(void *,PgHdr *);

/*
** Allocate and initialize a new Pager object and put a pointer to it
** in *ppPager. The pager should eventually be freed by passing it
** to sqlite3PagerClose().

**
** The zFilename argument is the path to the database file to open.
** If zFilename is NULL then a randomly-named temporary file is created
** and used as the file to be cached. Temporary files are be deleted
** automatically when they are closed. If zFilename is ":memory:" then 
** all information is held in cache. It is never written to disk. 
** This can be used to implement an in-memory database.
**
** The nExtra parameter specifies the number of bytes of space allocated
** along with each page reference. This space is available to the user
** via the sqlite3PagerGetExtra() API.
**

** The flags argument is used to specify properties that affect the
** operation of the pager. It should be passed some bitwise combination
** of the PAGER_OMIT_JOURNAL and PAGER_NO_READLOCK flags.
**
** The vfsFlags parameter is a bitmask to pass to the flags parameter
** of the xOpen() method of the supplied VFS when opening files. 
**
** If the pager object is allocated and the specified file opened 
** successfully, SQLITE_OK is returned and *ppPager set to point to
** the new pager object. If an error occurs, *ppPager is set to NULL
** and error code returned. This function may return SQLITE_NOMEM
** (sqlite3Malloc() is used to allocate memory), SQLITE_CANTOPEN or 
** various SQLITE_IO_XXX errors.
*/
int sqlite3PagerOpen(
  sqlite3_vfs *pVfs,       /* The virtual file system to use */
  Pager **ppPager,         /* OUT: Return the Pager structure here */
  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 */
  int vfsFlags             /* flags passed through to sqlite3_vfs.xOpen() */
){
  u8 *pPtr;
  Pager *pPager = 0;       /* Pager object to allocate and return */
  int rc = SQLITE_OK;      /* Return code */

  int tempFile = 0;        /* True for temp files (incl. in-memory files) */
  int memDb = 0;           /* True if this is an in-memory file */
  int readOnly = 0;        /* True if this is a read-only file */
  int journalFileSize;     /* Bytes to allocate for each journal fd */
  char *zPathname = 0;     /* Full path to database file */
  int nPathname = 0;       /* Number of bytes in zPathname */
  int useJournal = (flags & PAGER_OMIT_JOURNAL)==0; /* False to omit journal */
  int noReadlock = (flags & PAGER_NO_READLOCK)!=0;  /* True to omit read-lock */

  int pcacheSize = sqlite3PcacheSize();       /* Bytes to allocate for PCache */
  u16 szPageDflt = SQLITE_DEFAULT_PAGE_SIZE;  /* Default page size */



  /* Figure out how much space is required for each journal file-handle
  ** (there are two of them, the main journal and the sub-journal). This
  ** is the maximum space required for an in-memory journal file handle 
  ** and a regular journal file-handle. Note that a "regular journal-handle"
  ** may be a wrapper capable of caching the first portion of the journal
  ** file in memory to implement the atomic-write optimization (see 
  ** source file journal.c).
  */
  if( sqlite3JournalSize(pVfs)>sqlite3MemJournalSize() ){
    journalFileSize = sqlite3JournalSize(pVfs);
  }else{
    journalFileSize = sqlite3MemJournalSize();
  }

  /* Set the output variable to NULL in case an error occurs. */
  *ppPager = 0;

  /* Compute and store the full pathname in an allocated buffer pointed
  ** to by zPathname, length nPathname. Or, if this is a temporary file,
  ** leave both nPathname and zPathname set to 0.
  */
  if( zFilename && zFilename[0] ){

      memDb = 1;
      zPathname[0] = 0;
    }else
#endif
    {
      rc = sqlite3OsFullPathname(pVfs, zFilename, nPathname, zPathname);
    }

    nPathname = sqlite3Strlen30(zPathname);
    if( rc==SQLITE_OK && nPathname+8>pVfs->mxPathname ){
      /* This branch is taken when the journal path required by
      ** the database being opened will be more than pVfs->mxPathname
      ** bytes in length. This means the database cannot be opened,
      ** as it will not be possible to open the journal file or even
      ** check for a hot-journal before reading.
      */
      rc = SQLITE_CANTOPEN;
    }
    if( rc!=SQLITE_OK ){
      sqlite3_free(zPathname);
      return rc;
    }

  }

  /* Allocate memory for the Pager structure, PCache object, the
  ** three file descriptors, the database file name and the journal 
  ** file name. The layout in memory is as follows:
  **
  **     Pager object                    (sizeof(Pager) bytes)
  **     PCache object                   (sqlite3PcacheSize() bytes)
  **     Database file handle            (pVfs->szOsFile bytes)
  **     Sub-journal file handle         (journalFileSize bytes)
  **     Main journal file handle        (journalFileSize bytes)
  **     Database file name              (nPathname+1 bytes)
  **     Journal file name               (nPathname+8+1 bytes)
  */
  pPtr = (u8 *)sqlite3MallocZero(
    sizeof(*pPager) +           /* Pager structure */
    pcacheSize      +           /* PCache object */

    pVfs->szOsFile  +           /* The main db file */
    journalFileSize * 2 +       /* The two journal files */ 
    nPathname + 1 +             /* zFilename */
    nPathname + 8 + 1           /* zJournal */
  );
  if( !pPtr ){
    sqlite3_free(zPathname);
    return SQLITE_NOMEM;
  }
  pPager =              (Pager*)(pPtr);
  pPager->pPCache =    (PCache*)(pPtr += sizeof(*pPager));


  pPager->fd =   (sqlite3_file*)(pPtr += pcacheSize);
  pPager->sjfd = (sqlite3_file*)(pPtr += pVfs->szOsFile);
  pPager->jfd =  (sqlite3_file*)(pPtr += journalFileSize);
  pPager->zFilename =    (char*)(pPtr += journalFileSize);

  /* Fill in the Pager.zFilename and Pager.zJournal buffers, if required. */

  if( zPathname ){
    pPager->zJournal =   (char*)(pPtr += nPathname + 1);
    memcpy(pPager->zFilename, zPathname, nPathname);
    memcpy(pPager->zJournal, zPathname, nPathname);
    memcpy(&pPager->zJournal[nPathname], "-journal", 8);
    sqlite3_free(zPathname);
  }
  pPager->pVfs = pVfs;
  pPager->vfsFlags = vfsFlags;

  /* Open the pager file.
  */
  if( zFilename && zFilename[0] && !memDb ){



    int fout = 0;                    /* VFS flags returned by xOpen() */
    rc = sqlite3OsOpen(pVfs, pPager->zFilename, pPager->fd, vfsFlags, &fout);

    readOnly = (fout&SQLITE_OPEN_READONLY);

    /* If the file was successfully opened for read/write access,
    ** choose a default page size in case we have to create the
    ** database file. The default page size is the maximum of:
    **
    **    + SQLITE_DEFAULT_PAGE_SIZE,
    **    + The value returned by sqlite3OsSectorSize()
    **    + The largest page size that can be written atomically.
    */
    if( rc==SQLITE_OK && !readOnly ){
      setSectorSize(pPager);
      if( szPageDflt<pPager->sectorSize ){
        szPageDflt = pPager->sectorSize;
      }
#ifdef SQLITE_ENABLE_ATOMIC_WRITE
      {
        int iDc = sqlite3OsDeviceCharacteristics(pPager->fd);
        int ii;
        assert(SQLITE_IOCAP_ATOMIC512==(512>>8));
        assert(SQLITE_IOCAP_ATOMIC64K==(65536>>8));
        assert(SQLITE_MAX_DEFAULT_PAGE_SIZE<=65536);
        for(ii=szPageDflt; ii<=SQLITE_MAX_DEFAULT_PAGE_SIZE; ii=ii*2){
          if( iDc&(SQLITE_IOCAP_ATOMIC|(ii>>8)) ){
            szPageDflt = ii;
          }
        }
      }
#endif
      if( szPageDflt>SQLITE_MAX_DEFAULT_PAGE_SIZE ){
        szPageDflt = SQLITE_MAX_DEFAULT_PAGE_SIZE;

      }
    }
  }else{
    /* If a temporary file is requested, it is not opened immediately.
    ** In this case we accept the default page size and delay actually
    ** opening the file until the first call to OsWrite().
    **
    ** This branch is also run for an in-memory database. An in-memory
    ** database is the same as a temp-file that is never written out to
    ** disk and uses an in-memory rollback journal.
    */ 
    tempFile = 1;
    pPager->state = PAGER_EXCLUSIVE;
  }

  /* The following call to PagerSetPagesize() serves to set the value of 
  ** Pager.pageSize and to allocate the Pager.pTmpSpace buffer.
  */
  if( rc==SQLITE_OK ){
    assert( pPager->memDb==0 );
    rc = sqlite3PagerSetPagesize(pPager, &szPageDflt);
    testcase( rc!=SQLITE_OK );
  }

  /* If an error occured in either of the blocks above, free the 
  ** Pager structure and close the file.


  */
  if( rc!=SQLITE_OK ){
    assert( !pPager->pTmpSpace );
    sqlite3OsClose(pPager->fd);
    sqlite3_free(pPager);
    return rc;
  }

  /* Initialize the PCache object. */
  nExtra = FORCE_ALIGNMENT(nExtra);
  sqlite3PcacheOpen(szPageDflt, nExtra, !memDb,
                    !memDb?pagerStress:0, (void *)pPager, pPager->pPCache);

  PAGERTRACE(("OPEN %d %s\n", FILEHANDLEID(pPager->fd), pPager->zFilename));
  IOTRACE(("OPEN %p %s\n", pPager, pPager->zFilename))
















  pPager->useJournal = (u8)useJournal;
  pPager->noReadlock = (noReadlock && readOnly) ?1:0;
  /* pPager->stmtOpen = 0; */
  /* pPager->stmtInUse = 0; */
  /* pPager->nRef = 0; */
  pPager->dbSizeValid = (u8)memDb;

  /* pPager->stmtSize = 0; */
  /* pPager->stmtJSize = 0; */
  /* pPager->nPage = 0; */

  pPager->mxPgno = SQLITE_MAX_PAGE_COUNT;
  /* pPager->state = PAGER_UNLOCK; */
  assert( pPager->state == (tempFile ? PAGER_EXCLUSIVE : PAGER_UNLOCK) );
  /* pPager->errMask = 0; */
  pPager->tempFile = (u8)tempFile;
  assert( tempFile==PAGER_LOCKINGMODE_NORMAL 
          || tempFile==PAGER_LOCKINGMODE_EXCLUSIVE );
  assert( PAGER_LOCKINGMODE_EXCLUSIVE==1 );
  pPager->exclusiveMode = (u8)tempFile; 
  pPager->memDb = (u8)memDb;
  pPager->readOnly = (u8)readOnly;
  /* pPager->needSync = 0; */
  pPager->noSync = (pPager->tempFile || !useJournal) ?1:0;
  pPager->fullSync = pPager->noSync ?0:1;
  pPager->sync_flags = SQLITE_SYNC_NORMAL;
  /* pPager->pFirst = 0; */
  /* pPager->pFirstSynced = 0; */
  /* pPager->pLast = 0; */
  pPager->nExtra = nExtra;
  pPager->journalSizeLimit = SQLITE_DEFAULT_JOURNAL_SIZE_LIMIT;
  assert( isOpen(pPager->fd) || tempFile );
  setSectorSize(pPager);
  if( memDb ){
    pPager->journalMode = PAGER_JOURNALMODE_MEMORY;
  }
  /* pPager->xBusyHandler = 0; */
  /* pPager->pBusyHandlerArg = 0; */
  /* memset(pPager->aHash, 0, sizeof(pPager->aHash)); */
  *ppPager = pPager;
  return SQLITE_OK;
}

/*
** Set the busy handler function.
**
** The pager invokes the busy-handler if sqlite3OsLock() returns 
** SQLITE_BUSY when trying to upgrade from no-lock to a SHARED lock,
** or when trying to upgrade from a RESERVED lock to an EXCLUSIVE 
** lock. It does *not* invoke the busy handler when upgrading from
** SHARED to RESERVED, or when upgrading from SHARED to EXCLUSIVE
** (which occurs during hot-journal rollback). Summary:
**
**   Transition                        | Invokes xBusyHandler
**   --------------------------------------------------------
**   NO_LOCK       -> SHARED_LOCK      | Yes
**   SHARED_LOCK   -> RESERVED_LOCK    | No
**   SHARED_LOCK   -> EXCLUSIVE_LOCK   | No
**   RESERVED_LOCK -> EXCLUSIVE_LOCK   | Yes
**
** If the busy-handler callback returns non-zero, the lock is 
** retried. If it returns zero, then the SQLITE_BUSY error is
** returned to the caller of the pager API function.
*/
void sqlite3PagerSetBusyhandler(
  Pager *pPager,                       /* Pager object */
  int (*xBusyHandler)(void *),         /* Pointer to busy-handler function */
  void *pBusyHandlerArg                /* Argument to pass to xBusyHandler */
){  
  pPager->xBusyHandler = xBusyHandler;
  pPager->pBusyHandlerArg = pBusyHandlerArg;
}

/*
** Set the reinitializer for this pager. If not NULL, the reinitializer
** is called when the content of a page in cache is modified (restored)
** as part of a transaction or savepoint rollback. The callback gives 
** higher-level code an opportunity to restore the EXTRA section to 
** agree with the restored page data.
*/
void sqlite3PagerSetReiniter(Pager *pPager, void (*xReinit)(DbPage*)){
  pPager->xReiniter = xReinit;
}

/*
** Change the page size used by the Pager object. The new page size 
** is passed in *pPageSize.
**
** If the pager is in the error state when this function is called, it
** is a no-op. The value returned is the error state error code (i.e. 
** one of SQLITE_IOERR, SQLITE_CORRUPT or SQLITE_FULL).
**
** Otherwise, if all of the following are true:
**
**   * the new page size (value of *pPageSize) is valid (a power 
**     of two between 512 and SQLITE_MAX_PAGE_SIZE, inclusive), and
**
**   * there are no outstanding page references, and
**
**   * the database is either not an in-memory database or it is
**     an in-memory database that currently consists of zero pages.
**
** then the pager object page size is set to *pPageSize.
**
** If the page size is changed, then this function uses sqlite3PagerMalloc() 
** to obtain a new Pager.pTmpSpace buffer. If this allocation attempt 
** fails, SQLITE_NOMEM is returned and the page size remains unchanged. 
** In all other cases, SQLITE_OK is returned.
**
** If the page size is not changed, either because one of the enumerated
** conditions above is not true, the pager was in error state when this
** function was called, or because the memory allocation attempt failed, 
** then *pPageSize is set to the old, retained page size before returning.
*/
int sqlite3PagerSetPagesize(Pager *pPager, u16 *pPageSize){
  int rc = pPager->errCode;
  if( rc==SQLITE_OK ){
    u16 pageSize = *pPageSize;
    assert( pageSize==0 || (pageSize>=512 && pageSize<=SQLITE_MAX_PAGE_SIZE) );
    if( pageSize && pageSize!=pPager->pageSize 
     && (pPager->memDb==0 || pPager->dbSize==0)
     && sqlite3PcacheRefCount(pPager->pPCache)==0 
    ){
      char *pNew = (char *)sqlite3PageMalloc(pageSize);
      if( !pNew ){
        rc = SQLITE_NOMEM;
      }else{
        pager_reset(pPager);
        pPager->pageSize = pageSize;

        sqlite3PageFree(pPager->pTmpSpace);
        pPager->pTmpSpace = pNew;
        sqlite3PcacheSetPageSize(pPager->pPCache, pageSize);
      }
    }
    *pPageSize = (u16)pPager->pageSize;
  }

# define enable_simulated_io_errors()
#endif

/*
** Read the first N bytes from the beginning of the file into memory
** that pDest points to. 
**
** If the pager was opened on a transient file (zFilename==""), or
** opened on a file less than N bytes in size, the output buffer is
** zeroed and SQLITE_OK returned. The rationale for this is that this 
** function is used to read database headers, and a new transient or
** zero sized database has a header than consists entirely of zeroes.
**


** If any IO error apart from SQLITE_IOERR_SHORT_READ is encountered,
** the error code is returned to the caller and the contents of the
** output buffer undefined.

*/
int sqlite3PagerReadFileheader(Pager *pPager, int N, unsigned char *pDest){
  int rc = SQLITE_OK;
  memset(pDest, 0, N);
  assert( isOpen(pPager->fd) || pPager->tempFile );
  if( isOpen(pPager->fd) ){
    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;
}

/*
** Return the total number of pages in the database file associated 
** with pPager. Normally, this is calculated as (<db file size>/<page-size>).
** However, if the file is between 1 and <page-size> bytes in size, then 
** this is considered a 1 page file.
**
** If the pager is in error state when this function is called, then the
** error state error code is returned and *pnPage left unchanged. Or,

** if the file system has to be queried for the size of the file and
** the query attempt returns an IO error, the IO error code is returned
** and *pnPage is left unchanged.
**
** Otherwise, if everything is successful, then SQLITE_OK is returned
** and *pnPage is set to the number of pages in the database.
*/
int sqlite3PagerPagecount(Pager *pPager, int *pnPage){
  Pgno nPage;               /* Value to return via *pnPage */

  /* If the pager is already in the error state, return the error code. */


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

  }

  /* Determine the number of pages in the file. Store this in nPage. */
  if( pPager->dbSizeValid ){
    nPage = pPager->dbSize;
  }else{
    int rc;                 /* Error returned by OsFileSize() */
    i64 n = 0;              /* File size in bytes returned by OsFileSize() */

    assert( isOpen(pPager->fd) || pPager->tempFile );
    if( isOpen(pPager->fd) && (rc = sqlite3OsFileSize(pPager->fd, &n)) ){
      pager_error(pPager, rc);
      return rc;
    }
    if( n>0 && n<pPager->pageSize ){
      nPage = 1;
    }else{
      nPage = n / pPager->pageSize;
    }
    if( pPager->state!=PAGER_UNLOCK ){
      pPager->dbSize = (Pgno)nPage;
      pPager->dbFileSize = (Pgno)nPage;
      pPager->dbSizeValid = 1;
    }
  }



  /* If the current number of pages in the file is greater than the 
  ** configured maximum pager number, increase the allowed limit so
  ** that the file can be read.
  */
  if( nPage>pPager->mxPgno ){
    pPager->mxPgno = (Pgno)nPage;
  }

  /* Set the output variable and return SQLITE_OK */
  if( pnPage ){
    *pnPage = nPage;
  }
  return SQLITE_OK;
}

/*
** Forward declaration.
*/
static int syncJournal(Pager*);

/*
** Try to obtain a lock of type locktype on the database file. If
** a similar or greater lock is already held, this function is a no-op
** (returning SQLITE_OK immediately).
**
** Otherwise, attempt to obtain the lock using sqlite3OsLock(). Invoke 
** the busy callback if the lock is currently not available. Repeat 
** until the busy callback returns false or until the attempt to 
** obtain the lock succeeds.
**
** Return SQLITE_OK on success and an error code if we cannot obtain
** the lock. If the lock is obtained successfully, set the Pager.state 
** variable to locktype before returning.
*/
static int pager_wait_on_lock(Pager *pPager, int locktype){
  int rc;                              /* Return code */

  /* The OS lock values must be the same as the Pager lock values */
  assert( PAGER_SHARED==SHARED_LOCK );
  assert( PAGER_RESERVED==RESERVED_LOCK );
  assert( PAGER_EXCLUSIVE==EXCLUSIVE_LOCK );

  /* If the file is currently unlocked then the size must be unknown */
  assert( pPager->state>=PAGER_SHARED || pPager->dbSizeValid==0 );

  /* Check that this is either a no-op (because the requested lock is 
  ** already held, or one of the transistions that the busy-handler
  ** may be invoked during, according to the comment above
  ** sqlite3PagerSetBusyhandler().
  */
  assert( (pPager->state>=locktype)
       || (pPager->state==PAGER_UNLOCK && locktype==PAGER_SHARED)
       || (pPager->state==PAGER_RESERVED && locktype==PAGER_EXCLUSIVE)
  );

  if( pPager->state>=locktype ){
    rc = SQLITE_OK;
  }else{
    do {
      rc = sqlite3OsLock(pPager->fd, locktype);
    }while( rc==SQLITE_BUSY && pPager->xBusyHandler(pPager->pBusyHandlerArg) );
    if( rc==SQLITE_OK ){

** function does not actually modify the database file on disk. It 
** just sets the internal state of the pager object so that the 
** truncation will be done when the current transaction is committed.
*/
void sqlite3PagerTruncateImage(Pager *pPager, Pgno nPage){
  assert( pPager->dbSizeValid );
  assert( pPager->dbSize>=nPage );














  assert( pPager->state>=PAGER_RESERVED );
  pPager->dbSize = nPage;
}
#endif  /* ifndef SQLITE_OMIT_AUTOVACUUM */

/*
** Shutdown the page cache.  Free all memory and close all files.
**
** If a transaction was in progress when this routine is called, that
** transaction is rolled back.  All outstanding pages are invalidated
** and their memory is freed.  Any attempt to use a page associated
** with this page cache after this function returns will likely
** result in a coredump.
**
** This function always succeeds. If a transaction is active an attempt
** is made to roll it back. If an error occurs during the rollback 
** a hot journal may be left in the filesystem but no error is returned
** to the caller.
*/
int sqlite3PagerClose(Pager *pPager){

  disable_simulated_io_errors();
  sqlite3BeginBenignMalloc();
  pPager->errCode = 0;
  pPager->exclusiveMode = 0;
  pager_reset(pPager);
  if( MEMDB ){
    pager_unlock(pPager);
  }else{
    /* Set Pager.journalHdr to -1 for the benefit of the pager_playback() 
    ** call which may be made from within pagerUnlockAndRollback(). If it
    ** is not -1, then the unsynced portion of an open journal file may
    ** be played back into the database. If a power failure occurs while
    ** this is happening, the database may become corrupt.
    */
    pPager->journalHdr = -1;
    pagerUnlockAndRollback(pPager);
  }
  sqlite3EndBenignMalloc();
  enable_simulated_io_errors();

  PAGERTRACE(("CLOSE %d\n", PAGERID(pPager)));
  IOTRACE(("CLOSE %p\n", pPager))

  sqlite3OsClose(pPager->fd);
  sqlite3PageFree(pPager->pTmpSpace);
  sqlite3PcacheClose(pPager->pPCache);

  assert( !pPager->aSavepoint && !pPager->pInJournal );


  assert( !isOpen(pPager->jfd) && !isOpen(pPager->sjfd) );








  sqlite3_free(pPager);
  return SQLITE_OK;
}

#if !defined(NDEBUG) || defined(SQLITE_TEST)
/*
** Return the page number for page pPg.
*/
Pgno sqlite3PagerPagenumber(DbPage *pPg){
  return pPg->pgno;
}
#endif

/*
** Increment the reference count for page pPg.

*/
void sqlite3PagerRef(DbPage *pPg){
  sqlite3PcacheRef(pPg);

}

/*
** Sync the journal. In other words, make sure all the pages that have
** been written to the journal have actually reached the surface of the
** disk and can be restored in the event of a hot-journal rollback.
**
** If the Pager.needSync flag is not set, then this function is a
** no-op. Otherwise, the actions required depend on the journal-mode

** and the device characteristics of the the file-system, as follows:
**
**   * If the journal file is an in-memory journal file, no action need
**     be taken.
**
**   * Otherwise, if the device does not support the SAFE_APPEND property,
**     then the nRec field of the most recently written journal header
**     is updated to contain the number of journal records that have
**     been written following it. If the pager is operating in full-sync
**     mode, then the journal file is synced before this field is updated.
**
**   * If the device does not support the SEQUENTIAL property, then 
**     journal file is synced.
**
** Or, in pseudo-code:


**

**   if( NOT <in-memory journal> ){
**     if( NOT SAFE_APPEND ){
**       if( <full-sync mode> ) xSync(<journal file>);
**       <update nRec field>
**     } 
**     if( NOT SEQUENTIAL ) xSync(<journal file>);
**   }
**
** The Pager.needSync flag is never be set for temporary files, or any
** file operating in no-sync mode (Pager.noSync set to non-zero).
**
** If successful, this routine clears the PGHDR_NEED_SYNC flag of every 
** page currently held in memory before returning SQLITE_OK. If an IO
** error is encountered, then the IO error code is returned to the caller.
*/
static int syncJournal(Pager *pPager){





  if( pPager->needSync ){
    assert( !pPager->tempFile );
    if( pPager->journalMode!=PAGER_JOURNALMODE_MEMORY ){
      int rc;                              /* Return code */
      const int iDc = sqlite3OsDeviceCharacteristics(pPager->fd);
      assert( isOpen(pPager->jfd) );

      if( 0==(iDc&SQLITE_IOCAP_SAFE_APPEND) ){
        /* Variable iNRecOffset is set to the offset in the journal file
        ** of the nRec field of the most recently written journal header.
        ** This field will be updated following the xSync() operation
        ** on the journal file. */
        i64 iNRecOffset = pPager->journalHdr + sizeof(aJournalMagic);

        /* This block deals with an obscure problem. If the last connection
        ** that wrote to this database was operating in persistent-journal
        ** mode, then the journal file may at this point actually be larger
        ** than Pager.journalOff bytes. If the next thing in the journal
        ** file happens to be a journal-header (written as part of the
        ** previous connections transaction), and a crash or power-failure 
        ** occurs after nRec is updated but before this connection writes 
        ** anything else to the journal file (or commits/rolls back its 
        ** transaction), then SQLite may become confused when doing the 
        ** hot-journal rollback following recovery. It may roll back all
        ** of this connections data, then proceed to rolling back the old,
        ** out-of-date data that follows it. Database corruption.
        **
        ** To work around this, if the journal file does appear to contain
        ** a valid header following Pager.journalOff, then write a 0x00
        ** byte to the start of it to prevent it from being recognized.
        **
        ** Variable iNextHdrOffset is set to the offset at which this
        ** problematic header will occur, if it exists. aMagic is used 
        ** as a temporary buffer to inspect the first couple of bytes of
        ** the potential journal header.
        */
        i64 iNextHdrOffset = journalHdrOffset(pPager);
        u8 aMagic[8];
        rc = sqlite3OsRead(pPager->jfd, aMagic, 8, iNextHdrOffset);
        if( rc==SQLITE_OK && 0==memcmp(aMagic, aJournalMagic, 8) ){
          static const u8 zerobyte = 0;
          rc = sqlite3OsWrite(pPager->jfd, &zerobyte, 1, iNextHdrOffset);
        }
        if( rc!=SQLITE_OK && rc!=SQLITE_IOERR_SHORT_READ ){
          return rc;
        }

        /* 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.
        **
        ** This is not required if the persistent media supports the
        ** SAFE_APPEND property. Because in this case it is not possible 
        ** for garbage data to be appended to the file, the nRec field
        ** is populated with 0xFFFFFFFF when the journal header is written
        ** and never needs to be updated.
        */
        if( pPager->fullSync && 0==(iDc&SQLITE_IOCAP_SEQUENTIAL) ){
          PAGERTRACE(("SYNC journal of %d\n", PAGERID(pPager)));
          IOTRACE(("JSYNC %p\n", pPager))
          rc = sqlite3OsSync(pPager->jfd, pPager->sync_flags);
          if( rc!=SQLITE_OK ) return rc;
        }


        IOTRACE(("JHDR %p %lld %d\n", pPager, iNRecOffset, 4));
        rc = write32bits(pPager->jfd, iNRecOffset, pPager->nRec);
        if( rc!=SQLITE_OK ) return rc;
      }
      if( 0==(iDc&SQLITE_IOCAP_SEQUENTIAL) ){
        PAGERTRACE(("SYNC journal of %d\n", PAGERID(pPager)));
        IOTRACE(("JSYNC %p\n", pPager))
        rc = sqlite3OsSync(pPager->jfd, pPager->sync_flags| 
          (pPager->sync_flags==SQLITE_SYNC_FULL?SQLITE_SYNC_DATAONLY:0)
        );
        if( rc!=SQLITE_OK ) return rc;
      }

    }


    /* The journal file was just successfully synced. Set Pager.needSync 
    ** to zero and clear the PGHDR_NEED_SYNC flag on all pagess.
    */
    pPager->needSync = 0;
    pPager->journalStarted = 1;
    sqlite3PcacheClearSyncFlags(pPager->pPCache);
  }

  return SQLITE_OK;
}

/*
** The argument is the first in a linked list of dirty pages connected
** by the PgHdr.pDirty pointer. This function writes each one of the
** in-memory pages in the list to the database file. The argument may
** be NULL, representing an empty list. In this case this function is
** a no-op.
**
** The pager must hold at least a RESERVED lock when this function
** is called. Before writing anything to the database file, this lock
** is upgraded to an EXCLUSIVE lock. If the lock cannot be obtained,
** SQLITE_BUSY is returned and no data is written to the database file.
** 
** If the pager is a temp-file pager and the actual file-system file
** is not yet open, it is created and opened before any data is 
** written out.
**
** Once the lock has been upgraded and, if necessary, the file opened,
** the pages are written out to the database file in list order. Writing
** a page is skipped if it meets either of the following criteria:
**
**   * The page number is greater than Pager.dbSize, or
**   * The PGHDR_DONT_WRITE flag is set on the page.
**
** If writing out a page causes the database file to grow, Pager.dbFileSize
** is updated accordingly. If page 1 is written out, then the value cached
** in Pager.dbFileVers[] is updated to match the new value stored in
** the database file.
**
** If everything is successful, SQLITE_OK is returned. If an IO error 
** occurs, an IO error code is returned. Or, if the EXCLUSIVE lock cannot
** be obtained, SQLITE_BUSY is returned.
*/
static int pager_write_pagelist(PgHdr *pList){
  Pager *pPager;                       /* Pager object */
  int rc;                              /* Return code */

  if( pList==0 ) return SQLITE_OK;
  pPager = pList->pPager;

  /* At this point there may be either a RESERVED or EXCLUSIVE lock on the
  ** database file. If there is already an EXCLUSIVE lock, the following
  ** call is a no-op.
  **
  ** Moving the lock from RESERVED to EXCLUSIVE actually involves going
  ** through an intermediate state PENDING.   A PENDING lock prevents new
  ** readers from attaching to the database but is unsufficient for us to
  ** write.  The idea of a PENDING lock is to prevent new readers from
  ** coming in while we wait for existing readers to clear.
  **
  ** While the pager is in the RESERVED state, the original database file
  ** is unchanged and we can rollback without having to playback the
  ** journal into the original database file.  Once we transition to
  ** EXCLUSIVE, it means the database file has been changed and any rollback
  ** will require a journal playback.
  */
  assert( pPager->state>=PAGER_RESERVED );
  rc = pager_wait_on_lock(pPager, EXCLUSIVE_LOCK);



  /* If the file is a temp-file has not yet been opened, open it now. It
  ** is not possible for rc to be other than SQLITE_OK if this branch
  ** is taken, as pager_wait_on_lock() is a no-op for temp-files.
  */

  if( !isOpen(pPager->fd) ){
    assert( pPager->tempFile && rc==SQLITE_OK );
    rc = pagerOpentemp(pPager, pPager->fd, pPager->vfsFlags);

  }

  while( rc==SQLITE_OK && pList ){
    Pgno pgno = pList->pgno;

    /* If there are dirty pages in the page cache with page numbers greater
    ** than Pager.dbSize, this means sqlite3PagerTruncateImage() was called to
    ** make the file smaller (presumably by auto-vacuum code). Do not write
    ** any such pages to the file.
    **
    ** Also, do not write out any page that has the PGHDR_DONT_WRITE flag
    ** set (set by sqlite3PagerDontWrite()).
    */
    if( pgno<=pPager->dbSize && 0==(pList->flags&PGHDR_DONT_WRITE) ){
      i64 offset = (pgno-1)*(i64)pPager->pageSize;         /* Offset to write */
      char *pData = CODEC2(pPager, pList->pData, pgno, 6); /* Data to write */

      /* Write out the page data. */


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

      /* If page 1 was just written, update Pager.dbFileVers to match
      ** the value now stored in the database file. If writing this 
      ** page caused the database file to grow, update dbFileSize. 
      */

      if( pgno==1 ){
        memcpy(&pPager->dbFileVers, &pData[24], sizeof(pPager->dbFileVers));
      }
      if( pgno>pPager->dbFileSize ){
        pPager->dbFileSize = pgno;
      }

      PAGERTRACE(("STORE %d page %d hash(%08x)\n",
                   PAGERID(pPager), pgno, pager_pagehash(pList)));
      IOTRACE(("PGOUT %p %d\n", pPager, pgno));
      PAGER_INCR(sqlite3_pager_writedb_count);
      PAGER_INCR(pPager->nWrite);

    }else{
      PAGERTRACE(("NOSTORE %d page %d\n", PAGERID(pPager), pgno));
    }


#ifdef SQLITE_CHECK_PAGES
    pList->pageHash = pager_pagehash(pList);
#endif
    pList = pList->pDirty;
  }

  return rc;
}

/*
** Append a record of the current state of page pPg to the sub-journal. 
** It is the callers responsibility to use subjRequiresPage() to check 
** that it is really required before calling this function.
**
** If successful, set the bit corresponding to pPg->pgno in the bitvecs
** for all open savepoints before returning.
**
** This function returns SQLITE_OK if everything is successful, an IO
** error code if the attempt to write to the sub-journal fails, or 
** SQLITE_NOMEM if a malloc fails while setting a bit in a savepoint
** bitvec.
*/
static int subjournalPage(PgHdr *pPg){
  int rc;
  void *pData = pPg->pData;
  Pager *pPager = pPg->pPager;
  i64 offset = pPager->nSubRec*(4+pPager->pageSize);
  char *pData2 = CODEC2(pPager, pData, pPg->pgno, 7);

  PAGERTRACE(("STMT-JOURNAL %d page %d\n", PAGERID(pPager), pPg->pgno));

  assert( pageInJournal(pPg) || pPg->pgno>pPager->dbOrigSize );
  rc = write32bits(pPager->sjfd, offset, pPg->pgno);
  if( rc==SQLITE_OK ){
    rc = sqlite3OsWrite(pPager->sjfd, pData2, pPager->pageSize, offset+4);
  }
  if( rc==SQLITE_OK ){
    pPager->nSubRec++;
    assert( pPager->nSavepoint>0 );
    rc = addToSavepointBitvecs(pPager, pPg->pgno);
    testcase( rc!=SQLITE_OK );
  }
  return rc;
}


/*
** This function is called by the pcache layer when it has reached some
** soft memory limit. The first argument is a pointer to a Pager object
** (cast as a void*). The pager is always 'purgeable' (not an in-memory
** database). The second argument is a reference to a page that is 
** currently dirty but has no outstanding references. The page
** is always associated with the Pager object passed as the first 
** argument.
**
** The job of this function is to make pPg clean by writing its contents
** out to the database file, if possible. This may involve syncing the
** journal file. 
**
** If successful, sqlite3PcacheMakeClean() is called on the page and
** SQLITE_OK returned. If an IO error occurs while trying to make the
** page clean, the IO error code is returned. If the page cannot be
** made clean for some other reason, but no error occurs, then SQLITE_OK
** is returned by sqlite3PcacheMakeClean() is not called.
*/
static int pagerStress(void *p, PgHdr *pPg){
  Pager *pPager = (Pager *)p;
  int rc = SQLITE_OK;

  assert( pPg->pPager==pPager );
  assert( pPg->flags&PGHDR_DIRTY );

  /* The doNotSync flag is set by the sqlite3PagerWrite() function while it
  ** is journalling a set of two or more database pages that are stored
  ** on the same disk sector. Syncing the journal is not allowed while
  ** this is happening as it is important that all members of such a
  ** set of pages are synced to disk together. So, if the page this function
  ** is trying to make clean will require a journal sync and the doNotSync
  ** flag is set, return without doing anything. The pcache layer will
  ** just have to go ahead and allocate a new page buffer instead of
  ** reusing pPg.
  **
  ** Similarly, if the pager has already entered the error state, do not
  ** try to write the contents of pPg to disk.
  */
  if( pPager->errCode || (pPager->doNotSync && pPg->flags&PGHDR_NEED_SYNC) ){
    return SQLITE_OK;
  }

  /* Sync the journal file if required. */

  if( pPg->flags&PGHDR_NEED_SYNC ){
    rc = syncJournal(pPager);
    if( rc==SQLITE_OK && pPager->fullSync && 
      !(pPager->journalMode==PAGER_JOURNALMODE_MEMORY) &&
      !(sqlite3OsDeviceCharacteristics(pPager->fd)&SQLITE_IOCAP_SAFE_APPEND)
    ){
      pPager->nRec = 0;
      rc = writeJournalHdr(pPager);
    }
  }

  /* If the page number of this page is larger than the current size of
  ** the database image, it may need to be written to the sub-journal.
  ** This is because the call to pager_write_pagelist() below will not
  ** actually write data to the file in this case.
  **
  ** Consider the following sequence of events:
  **
  **   BEGIN;
  **     <journal page X>
  **     <modify page X>
  **     SAVEPOINT sp;
  **       <shrink database file to Y pages>
  **       pagerStress(page X)
  **     ROLLBACK TO sp;
  **
  ** If (X>Y), then when pagerStress is called page X will not be written
  ** out to the database file, but will be dropped from the cache. Then,
  ** following the "ROLLBACK TO sp" statement, reading page X will read
  ** data from the database file. This will be the copy of page X as it
  ** was when the transaction started, not as it was when "SAVEPOINT sp"
  ** was executed.
  **
  ** The solution is to write the current data for page X into the 
  ** sub-journal file now (if it is not already there), so that it will
  ** be restored to its current value when the "ROLLBACK TO sp" is 
  ** executed.
  */
  if( rc==SQLITE_OK && pPg->pgno>pPager->dbSize && subjRequiresPage(pPg) ){
    rc = subjournalPage(pPg);
  }

  /* Write the contents of the page out to the database file. */
  if( rc==SQLITE_OK ){
    pPg->pDirty = 0;
    rc = pager_write_pagelist(pPg);
  }





  /* Mark the page as clean. */
  if( rc==SQLITE_OK ){
    PAGERTRACE(("STRESS %d page %d\n", PAGERID(pPager), pPg->pgno));
    sqlite3PcacheMakeClean(pPg);
  }

  return pager_error(pPager, rc);
}


/*
** This function is called after transitioning from PAGER_UNLOCK to
** PAGER_SHARED state. It tests if there is a hot journal present in
** the file-system for the given pager. A hot journal is one that 
** needs to be played back. According to this function, a hot-journal
** file exists if the following three criteria are met:
**
**   * The journal file exists in the file system, and
**   * No process holds a RESERVED or greater lock on the database file, and
**   * The database file itself is greater than 0 bytes in size.
**
** If the current size of the database file is 0 but a journal file
** exists, that is probably an old journal left over from a prior
** database with the same name. In this case the journal file is
** just deleted using OsDelete, *pExists is set to 0 and SQLITE_OK
** is returned.

**
** This routine does not open the journal file to examine its
** content.  Hence, the journal might contain the name of a master
** journal file that has been deleted, and hence not be hot.  Or
** the header of the journal might be zeroed out.  This routine
** does not discover these cases of a non-hot journal - if the
** journal file exists and is not empty this routine assumes it
** is hot.  The pager_playback() routine will discover that the
** journal file is not really hot and will no-op.
**
** If a hot-journal file is found to exist, *pExists is set to 1 and 
** SQLITE_OK returned. If no hot-journal file is present, *pExists is
** set to 0 and SQLITE_OK returned. If an IO error occurs while trying
** to determine whether or not a hot-journal file exists, the IO error
** code is returned and the value of *pExists is undefined.
*/
static int hasHotJournal(Pager *pPager, int *pExists){
  sqlite3_vfs * const pVfs = pPager->pVfs;
  int rc;                       /* Return code */
  int exists = 0;               /* True if a journal file is present */
  int locked = 0;               /* True if some process holds a RESERVED lock */

  assert( pPager!=0 );
  assert( pPager->useJournal );
  assert( isOpen(pPager->fd) );

  *pExists = 0;
  rc = sqlite3OsAccess(pVfs, pPager->zJournal, SQLITE_ACCESS_EXISTS, &exists);
  if( rc==SQLITE_OK && exists ){
    rc = sqlite3OsCheckReservedLock(pPager->fd, &locked);

    if( rc==SQLITE_OK && !locked ){
      int nPage;
      rc = sqlite3PagerPagecount(pPager, &nPage);
      if( rc==SQLITE_OK ){
       if( nPage==0 ){
          sqlite3OsDelete(pVfs, pPager->zJournal, 0);
        }else{
          *pExists = 1;
        }
      }
    }
  }
  return rc;
}

/*
** Read the content for page pPg out of the database file and into 
** pPg->pData. A shared lock or greater must be held on the database
** file before this function is called.
**
** If page 1 is read, then the value of Pager.dbFileVers[] is set to
** the value read from the database file.
**
** If an IO error occurs, then the IO error is returned to the caller.
** Otherwise, SQLITE_OK is returned.
*/
static int readDbPage(PgHdr *pPg){
  Pager *pPager = pPg->pPager; /* Pager object associated with page pPg */
  Pgno pgno = pPg->pgno;       /* Page number to read */
  int rc;                      /* Return code */
  i64 iOffset;                 /* Byte offset of file to read from */

  assert( pPager->state>=PAGER_SHARED && !MEMDB );

  if( !isOpen(pPager->fd) ){
    assert( pPager->tempFile );
    return SQLITE_IOERR_SHORT_READ;
  }
  iOffset = (pgno-1)*(i64)pPager->pageSize;
  rc = sqlite3OsRead(pPager->fd, pPg->pData, pPager->pageSize, iOffset);
  if( pgno==1 ){
    u8 *dbFileVers = &((u8*)pPg->pData)[24];
    memcpy(&pPager->dbFileVers, dbFileVers, sizeof(pPager->dbFileVers));
  }
  CODEC1(pPager, pPg->pData, pgno, 3);

  PAGER_INCR(sqlite3_pager_readdb_count);
  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;
}


/*
** This function is called to obtain the shared lock required before
** data may be read from the pager cache. If the shared lock has already
** been obtained, this function is a no-op.
**
** Immediately after obtaining the shared lock (if required), this function
** checks for a hot-journal file. If one is found, an emergency rollback
** is performed immediately.
*/
static int pagerSharedLock(Pager *pPager){
  int rc = SQLITE_OK;                /* Return code */
  int isErrorReset = 0;              /* True if recovering from error state */

  /* If this database is opened for exclusive access, has no outstanding 
  ** page references and is in an error-state, this is a chance to clear
  ** the error. Discard the contents of the pager-cache and treat any
  ** open journal file as a hot-journal.
  */
  if( !MEMDB && pPager->exclusiveMode 
   && sqlite3PcacheRefCount(pPager->pPCache)==0 && pPager->errCode 
  ){
    if( isOpen(pPager->jfd) ){
      isErrorReset = 1;
    }
    pPager->errCode = SQLITE_OK;
    pager_reset(pPager);
  }

  /* If the pager is still in an error state, do not proceed. The error 
  ** state will be cleared at some point in the future when all page 
  ** references are dropped and the cache can be discarded.
  */
  if( pPager->errCode && pPager->errCode!=SQLITE_FULL ){
    return pPager->errCode;
  }

  if( pPager->state==PAGER_UNLOCK || isErrorReset ){
    sqlite3_vfs * const pVfs = pPager->pVfs;
    int isHotJournal = 0;
    assert( !MEMDB );
    assert( sqlite3PcacheRefCount(pPager->pPCache)==0 );
    if( !pPager->noReadlock ){
      rc = pager_wait_on_lock(pPager, SHARED_LOCK);
      if( rc!=SQLITE_OK ){
        assert( pPager->state==PAGER_UNLOCK );

      }
    }
    if( isErrorReset || isHotJournal ){
      /* Get an EXCLUSIVE lock on the database file. At this point it is
      ** important that a RESERVED lock is not obtained on the way to the
      ** EXCLUSIVE lock. If it were, another process might open the
      ** database file, detect the RESERVED lock, and conclude that the
      ** database is safe to read while this process is still rolling the 
      ** hot-journal back.
      ** 
      ** Because the intermediate RESERVED lock is not requested, any
      ** other process attempting to access the database file will get to 
      ** this point in the code and fail to obtain its own EXCLUSIVE lock 
      ** on the database file.
      */
      if( pPager->state<EXCLUSIVE_LOCK ){
        rc = sqlite3OsLock(pPager->fd, EXCLUSIVE_LOCK);
        if( rc!=SQLITE_OK ){
          rc = pager_error(pPager, rc);
          goto failed;
        }
        pPager->state = PAGER_EXCLUSIVE;
      }
 
      /* Open the journal for read/write access. This is because in 
      ** exclusive-access mode the file descriptor will be kept open and
      ** possibly used for a transaction later on. On some systems, the
      ** OsTruncate() call used in exclusive-access mode also requires
      ** a read/write file handle.
      */
      if( !isOpen(pPager->jfd) ){
        int res;
        rc = sqlite3OsAccess(pVfs,pPager->zJournal,SQLITE_ACCESS_EXISTS,&res);
        if( rc==SQLITE_OK ){
          if( res ){
            int fout = 0;
            int f = SQLITE_OPEN_READWRITE|SQLITE_OPEN_MAIN_JOURNAL;
            assert( !pPager->tempFile );
            rc = sqlite3OsOpen(pVfs, pPager->zJournal, pPager->jfd, f, &fout);
            assert( rc!=SQLITE_OK || isOpen(pPager->jfd) );
            if( rc==SQLITE_OK && fout&SQLITE_OPEN_READONLY ){
              rc = SQLITE_CANTOPEN;
              sqlite3OsClose(pPager->jfd);
            }
          }else{
            /* If the journal does not exist, that means some other process
            ** has already rolled it back */
            rc = SQLITE_BUSY;
          }
        }
      }
      if( rc!=SQLITE_OK ){
        goto failed;
      }

      /* TODO: Why are these cleared here? Is it necessary? */
      pPager->journalStarted = 0;
      pPager->journalOff = 0;
      pPager->setMaster = 0;
      pPager->journalHdr = 0;
 
      /* Playback and delete the journal.  Drop the database write
      ** lock and reacquire the read lock. Purge the cache before
      ** playing back the hot-journal so that we don't end up with
      ** an inconsistent cache.
      */
      sqlite3PcacheClear(pPager->pPCache);
      rc = pager_playback(pPager, 1);
      if( rc!=SQLITE_OK ){
        rc = pager_error(pPager, rc);
        goto failed;
      }
      assert( (pPager->state==PAGER_SHARED)
           || (pPager->exclusiveMode && pPager->state>PAGER_SHARED)
      );
    }

    if( sqlite3PcachePagecount(pPager->pPCache)>0 ){
      /* The shared-lock has just been acquired on the database file
      ** and there are already pages in the cache (from a previous
      ** read or write transaction).  Check to see if the database

 failed:
  if( rc!=SQLITE_OK ){
    /* pager_unlock() is a no-op for exclusive mode and in-memory databases. */
    pager_unlock(pPager);
  }
  return rc;
}




















/*
** If the reference count has reached zero, and the pager is not in the
** middle of a write transaction or opened in exclusive mode, unlock it.
*/ 
static void pagerUnlockIfUnused(Pager *pPager){
  if( (sqlite3PcacheRefCount(pPager->pPCache)==0)

** just returns 0.  This routine acquires a read-lock the first time it
** has to go to disk, and could also playback an old journal if necessary.
** Since Lookup() never goes to disk, it never has to deal with locks
** or journal files.
**
** If noContent is false, the page contents are actually read from disk.
** If noContent is true, it means that we do not care about the contents
** of the page. This occurs in two seperate scenarios:
**
**   a) When reading a free-list leaf page from the database, and
**
**   b) When a savepoint is being rolled back and we need to load
**      a new page into the cache to populate with the data read
**      from the savepoint journal.
**
** If noContent is true, then the data returned is zeroed instead of
** being read from the database. Additionally, the bits corresponding
** to pgno in Pager.pInJournal (bitvec of pages already written to the
** journal file) and the PagerSavepoint.pInSavepoint bitvecs of any open
** savepoints are set. This means if the page is made writable at any
** point in the future, using a call to sqlite3PagerWrite(), its contents
** will not be journaled. This saves IO.


*/
int sqlite3PagerAcquire(
  Pager *pPager,      /* The pager open on the database file */
  Pgno pgno,          /* Page number to fetch */
  DbPage **ppPage,    /* Write a pointer to the page here */
  int noContent       /* Do not bother reading content from disk if true */
){

    if( nMax<(int)pgno || MEMDB || noContent ){
      if( pgno>pPager->mxPgno ){
        sqlite3PagerUnref(pPg);
        return SQLITE_FULL;
      }
      memset(pPg->pData, 0, pPager->pageSize);
      if( noContent ){
        /* Failure to set the bits in the InJournal bit-vectors is benign.
        ** It merely means that we might do some extra work to journal a 
        ** page that does not need to be journaled.  Nevertheless, be sure 
        ** to test the case where a malloc error occurs while trying to set 
        ** a bit in a bit vector.
        */
        sqlite3BeginBenignMalloc();
        TESTONLY( rc = ) sqlite3BitvecSet(pPager->pInJournal, pPg->pgno);
        testcase( rc==SQLITE_NOMEM );
        TESTONLY( rc = ) addToSavepointBitvecs(pPager, pPg->pgno);
        testcase( rc==SQLITE_NOMEM );
        sqlite3EndBenignMalloc();
      }
      IOTRACE(("ZERO %p %d\n", pPager, pgno));
    }else{
      assert( pPg->pPager==pPager && pPg->pgno==pgno );
      rc = readDbPage(pPg);
      if( rc!=SQLITE_OK && rc!=SQLITE_IOERR_SHORT_READ ){

        pagerDropPage(pPg);
        return rc;
      }
    }
#ifdef SQLITE_CHECK_PAGES
    pPg->pageHash = pager_pagehash(pPg);
#endif
  }else{
    /* The requested page is in the page cache. */

    PAGER_INCR(pPager->nHit);







  }

  *ppPage = pPg;
  return SQLITE_OK;
}

/*

** Release a page.
**
** If the number of references to the page drop to zero, then the
** page is added to the LRU list.  When all references to all pages
** are released, a rollback occurs and the lock on the database is
** removed.
*/
void sqlite3PagerUnref(DbPage *pPg){
  if( pPg ){
    Pager *pPager = pPg->pPager;
    sqlite3PcacheRelease(pPg);
    pagerUnlockIfUnused(pPager);
  }

}

/*
** If the main journal file has already been opened, ensure that the
** sub-journal file is open too. If the main journal is not open,
** this function is a no-op.
**
** SQLITE_OK is returned if everything goes according to plan. An 
** SQLITE_IOERR_XXX error code is returned if the call to 
** sqlite3OsOpen() fails.
*/
static int openSubJournal(Pager *pPager){
  int rc = SQLITE_OK;
  if( isOpen(pPager->jfd) && !isOpen(pPager->sjfd) ){
    if( pPager->journalMode==PAGER_JOURNALMODE_MEMORY ){
      sqlite3MemJournalOpen(pPager->sjfd);
    }else{
      rc = pagerOpentemp(pPager, pPager->sjfd, SQLITE_OPEN_SUBJOURNAL);
    }
  }
  return rc;
}

/*
** Create a journal file for pPager.  There should already be a RESERVED

  sqlite3PagerPagecount(pPager, 0);
  pPager->pInJournal = sqlite3BitvecCreate(pPager->dbSize);
  if( pPager->pInJournal==0 ){
    rc = SQLITE_NOMEM;
    goto failed_to_open_journal;
  }

  if( !isOpen(pPager->jfd) ){
    if( pPager->tempFile ){
      flags |= (SQLITE_OPEN_DELETEONCLOSE|SQLITE_OPEN_TEMP_JOURNAL);
    }else{
      flags |= (SQLITE_OPEN_MAIN_JOURNAL);
    }
    if( pPager->journalMode==PAGER_JOURNALMODE_MEMORY ){
      sqlite3MemJournalOpen(pPager->jfd);
      rc = SQLITE_OK;
    }else{
#ifdef SQLITE_ENABLE_ATOMIC_WRITE
      rc = sqlite3JournalOpen(
          pVfs, pPager->zJournal, pPager->jfd, flags, jrnlBufferSize(pPager)
      );
#else
      rc = sqlite3OsOpen(pVfs, pPager->zJournal, pPager->jfd, flags, 0);
#endif
    }
    assert( rc!=SQLITE_OK || isOpen(pPager->jfd) );
    pPager->journalOff = 0;
    pPager->setMaster = 0;
    pPager->journalHdr = 0;
    if( rc!=SQLITE_OK ){
      if( rc==SQLITE_NOMEM ){
        sqlite3OsDelete(pVfs, pPager->zJournal, 0);
      }
      goto failed_to_open_journal;
    }
  }

  pPager->journalStarted = 0;
  pPager->needSync = 0;
  pPager->nRec = 0;
  if( pPager->errCode ){
    rc = pPager->errCode;
    goto failed_to_open_journal;
  }

      if( exFlag ){
        rc = pager_wait_on_lock(pPager, EXCLUSIVE_LOCK);
      }
    }
    if( rc!=SQLITE_OK ){
      return rc;
    }

    PAGERTRACE(("TRANSACTION %d\n", PAGERID(pPager)));
    if( pPager->useJournal && !pPager->tempFile
           && pPager->journalMode!=PAGER_JOURNALMODE_OFF ){
      rc = pager_open_journal(pPager);
    }
  }else if( isOpen(pPager->jfd) && pPager->journalOff==0 ){
    /* This happens when the pager was in exclusive-access mode the last
    ** time a (read or write) transaction was successfully concluded
    ** by this connection. Instead of deleting the journal file it was 
    ** kept open and either was truncated to 0 bytes or its header was
    ** overwritten with zeros.
    */
    assert( pPager->nRec==0 );
    assert( pPager->dbOrigSize==0 );
    assert( pPager->pInJournal==0 );
    sqlite3PagerPagecount(pPager, 0);
    pPager->pInJournal = sqlite3BitvecCreate( pPager->dbSize );
    if( !pPager->pInJournal ){
      rc = SQLITE_NOMEM;
    }else{
      pPager->dbOrigSize = pPager->dbSize;
      rc = writeJournalHdr(pPager);
    }
  }
  assert( !isOpen(pPager->jfd) || pPager->journalOff>0 || rc!=SQLITE_OK );
  return rc;
}

/*
** Mark a data page as writeable.  The page is written into the journal 
** if it is not there already.  This routine must be called before making
** changes to a page.

  if( pPager->readOnly ){
    return SQLITE_PERM;
  }

  assert( !pPager->setMaster );

  CHECK_PAGE(pPg);














  /* Mark the page as dirty.  If the page has already been written
  ** to the journal then we can return right away.
  */
  sqlite3PcacheMakeDirty(pPg);
  if( pageInJournal(pPg) && !subjRequiresPage(pPg) ){

    pPager->dbModified = 1;
  }else{

    /* If we get this far, it means that the page needs to be
    ** written to the transaction journal or the ckeckpoint journal
    ** or both.
    **
    ** First check to see that the transaction journal exists and
    ** create it if it does not.
    */
    assert( pPager->state!=PAGER_UNLOCK );
    rc = sqlite3PagerBegin(pPg, 0);
    if( rc!=SQLITE_OK ){
      return rc;
    }
    assert( pPager->state>=PAGER_RESERVED );
    if( !isOpen(pPager->jfd) && pPager->useJournal
          && pPager->journalMode!=PAGER_JOURNALMODE_OFF ){
      rc = pager_open_journal(pPager);
      if( rc!=SQLITE_OK ) return rc;
    }

    pPager->dbModified = 1;
  
    /* The transaction journal now exists and we have a RESERVED or an
    ** EXCLUSIVE lock on the main database file.  Write the current page to
    ** the transaction journal if it is not there already.
    */
    if( !pageInJournal(pPg) && isOpen(pPager->jfd) ){
      if( pPg->pgno<=pPager->dbOrigSize ){
        u32 cksum;
        char *pData2;

        /* We should never write to the journal file the page that
        ** contains the database locks.  The following assert verifies
        ** that we do not. */

** The overlying software layer calls this routine when all of the data
** on the given page is unused.  The pager marks the page as clean so
** that it does not get written to disk.
**
** Tests show that this optimization, together with the
** sqlite3PagerDontRollback() below, more than double the speed
** of large INSERT operations and quadruple the speed of large DELETEs.











*/
void sqlite3PagerDontWrite(PgHdr *pPg){

  Pager *pPager = pPg->pPager;














  if( (pPg->flags&PGHDR_DIRTY) && pPager->nSavepoint==0 ){











    PAGERTRACE(("DONT_WRITE page %d of %d\n", pPg->pgno, PAGERID(pPager)));
    IOTRACE(("CLEAN %p %d\n", pPager, pPg->pgno))
    pPg->flags |= PGHDR_DONT_WRITE;
#ifdef SQLITE_CHECK_PAGES
    pPg->pageHash = pager_pagehash(pPg);







































#endif
  }











}






















/*
** This routine is called to increment the database file change-counter,
** stored at byte 24 of the pager file.
*/
static int pager_incr_changecounter(Pager *pPager, int isDirect){
  PgHdr *pPgHdr;

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

#ifdef SQLITE_ENABLE_ATOMIC_WRITE
    if( isDirect && isOpen(pPager->fd) ){
      const void *zBuf = pPgHdr->pData;
      assert( pPager->dbFileSize>0 );
      rc = sqlite3OsWrite(pPager->fd, zBuf, pPager->pageSize, 0);
    }
#endif

    /* Release the page reference. */

    return pPager->errCode;
  }

  /* If no changes have been made, we can leave the transaction early.
  */
  if( pPager->dbModified==0 &&
        (pPager->journalMode!=PAGER_JOURNALMODE_DELETE ||
          pPager->exclusiveMode) ){
    assert( pPager->dbModified==0 || !isOpen(pPager->jfd) );
    return SQLITE_OK;
  }

  PAGERTRACE(("DATABASE SYNC: File=%s zMaster=%s nSize=%d\n", 
      pPager->zFilename, zMaster, pPager->dbSize));

  /* If this is an in-memory db, or no pages have been written to, or this
  ** function has already been called, it is a no-op.
  */
  if( pPager->state!=PAGER_SYNCED && !MEMDB && pPager->dbModified ){
    PgHdr *pPg;

#ifdef SQLITE_ENABLE_ATOMIC_WRITE
    /* The atomic-write optimization can be used if all of the
    ** following are true:
    **
    **    + The file-system supports the atomic-write property for
    **      blocks of size page-size, and
    **    + This commit is not part of a multi-file transaction, and
    **    + Exactly one page has been modified and store in the journal file.
    **
    ** If the optimization can be used, then the journal file will never
    ** be created for this transaction.
    */
    int useAtomicWrite;
    pPg = sqlite3PcacheDirtyList(pPager->pPCache);
    useAtomicWrite = (
        !zMaster && 
        isOpen(pPager->jfd) &&
        pPager->journalOff==jrnlBufferSize(pPager) && 
        pPager->dbSize>=pPager->dbFileSize && 
        (pPg==0 || pPg->pDirty==0)
    );
    assert( isOpen(pPager->jfd) || pPager->journalMode==PAGER_JOURNALMODE_OFF );
    if( useAtomicWrite ){
      /* Update the nRec field in the journal file. */
      int offset = pPager->journalHdr + sizeof(aJournalMagic);
      assert(pPager->nRec==1);
      rc = write32bits(pPager->jfd, offset, pPager->nRec);

      /* Update the db file change counter. The following call will modify

      ** pager_get_all_dirty_pages() that verifies that no attempt
      ** is made to use an invalid dirty list.
      */
      goto sync_exit;
    }
    sqlite3PcacheCleanAll(pPager->pPCache);

    if( pPager->dbSize!=pPager->dbFileSize ){
      Pgno nNew = pPager->dbSize - (pPager->dbSize==PAGER_MJ_PGNO(pPager));
      assert( pPager->state>=PAGER_EXCLUSIVE );
      rc = pager_truncate(pPager, nNew);
      if( rc!=SQLITE_OK ) goto sync_exit;
    }

    /* Sync the database file. */
    if( !pPager->noSync && !noSync ){
      rc = sqlite3OsSync(pPager->fd, pPager->sync_flags);
    }

  }
  if( pPager->state<PAGER_RESERVED ){
    return SQLITE_ERROR;
  }
  if( pPager->dbModified==0 &&
        (pPager->journalMode!=PAGER_JOURNALMODE_DELETE ||
          pPager->exclusiveMode!=0) ){
    assert( pPager->dbModified==0 || isOpen(pPager->jfd)==0 );
    return SQLITE_OK;
  }
  PAGERTRACE(("COMMIT %d\n", PAGERID(pPager)));
  assert( pPager->state==PAGER_SYNCED || MEMDB || !pPager->dbModified );
  rc = pager_end_transaction(pPager, pPager->setMaster);
  rc = pager_error(pPager, rc);
  return rc;
}

/*
** Rollback all changes.  The database falls back to PAGER_SHARED mode.
** All in-memory cache pages revert to their original data contents.
** The journal is deleted.
**
** This routine cannot fail unless some other process is not following
** the correct locking protocol or unless some other
** process is writing trash into the journal file (SQLITE_CORRUPT) or
** unless a prior malloc() failed (SQLITE_NOMEM).  Appropriate error
** codes are returned for all these occasions.  Otherwise,
** SQLITE_OK is returned.
*/
int sqlite3PagerRollback(Pager *pPager){
  int rc = SQLITE_OK;
  PAGERTRACE(("ROLLBACK %d\n", PAGERID(pPager)));
  if( !pPager->dbModified || !isOpen(pPager->jfd) ){
    rc = pager_end_transaction(pPager, pPager->setMaster);
  }else if( pPager->errCode && pPager->errCode!=SQLITE_FULL ){
    if( pPager->state>=PAGER_EXCLUSIVE ){
      pager_playback(pPager, 0);
    }
    rc = pPager->errCode;
  }else{

  if( nSavepoint>pPager->nSavepoint && pPager->useJournal ){
    int ii;
    PagerSavepoint *aNew;

    /* Either there is no active journal or the sub-journal is open or 
    ** the journal is always stored in memory */
    assert( pPager->nSavepoint==0 || isOpen(pPager->sjfd) ||
            pPager->journalMode==PAGER_JOURNALMODE_MEMORY );

    /* Grow the Pager.aSavepoint array using realloc(). Return SQLITE_NOMEM
    ** if the allocation fails. Otherwise, zero the new portion in case a 
    ** malloc failure occurs while populating it in the for(...) loop below.
    */
    aNew = (PagerSavepoint *)sqlite3Realloc(

    ii = pPager->nSavepoint;
    pPager->nSavepoint = nSavepoint;

    /* Populate the PagerSavepoint structures just allocated. */
    for(/* no-op */; ii<nSavepoint; ii++){
      assert( pPager->dbSizeValid );
      aNew[ii].nOrig = pPager->dbSize;
      if( isOpen(pPager->jfd) && pPager->journalOff>0 ){
        aNew[ii].iOffset = pPager->journalOff;
      }else{
        aNew[ii].iOffset = JOURNAL_HDR_SZ(pPager);
      }
      aNew[ii].iSubRec = pPager->nSubRec;
      aNew[ii].pInSavepoint = sqlite3BitvecCreate(pPager->dbSize);
      if( !aNew[ii].pInSavepoint ){
        return SQLITE_NOMEM;
      }
    }

    /* Open the sub-journal, if it is not already opened. */

    int ii;
    int nNew = iSavepoint + (op==SAVEPOINT_ROLLBACK);
    for(ii=nNew; ii<pPager->nSavepoint; ii++){
      sqlite3BitvecDestroy(pPager->aSavepoint[ii].pInSavepoint);
    }
    pPager->nSavepoint = nNew;

    if( op==SAVEPOINT_ROLLBACK && isOpen(pPager->jfd) ){
      PagerSavepoint *pSavepoint = (nNew==0)?0:&pPager->aSavepoint[nNew-1];
      rc = pagerPlaybackSavepoint(pPager, pSavepoint);
      assert(rc!=SQLITE_DONE);
    }
  
    /* If this is a release of the outermost savepoint, truncate 
    ** the sub-journal. */
    if( nNew==0 && op==SAVEPOINT_RELEASE && isOpen(pPager->sjfd) ){
      assert( rc==SQLITE_OK );
      rc = sqlite3OsTruncate(pPager->sjfd, 0);
      pPager->nSubRec = 0;
    }
  }
  return rc;
}

/*
** Return the full pathname of the database file.

** Return the file handle for the database file associated
** with the pager.  This might return NULL if the file has
** not yet been opened.
*/
sqlite3_file *sqlite3PagerFile(Pager *pPager){
  return pPager->fd;
}








/*
** Return the full pathname of the journal file.
*/
const char *sqlite3PagerJournalname(Pager *pPager){
  return pPager->zJournal;
}

  ){
    return rc;
  }

  PAGERTRACE(("MOVE %d page %d (needSync=%d) moves to %d\n", 
      PAGERID(pPager), pPg->pgno, (pPg->flags&PGHDR_NEED_SYNC)?1:0, pgno));
  IOTRACE(("MOVE %p %d %d\n", pPager, pPg->pgno, pgno))



  /* If the journal needs to be sync()ed before page pPg->pgno can
  ** be written to, store pPg->pgno in local variable needSyncPgno.
  **
  ** If the isCommit flag is set, there is no need to remember that
  ** the journal needs to be sync()ed before database page pPg->pgno 
  ** can be written to. The caller has already promised not to write to it.

  sqlite3PcacheMove(pPg, pgno);
  if( pPgOld ){
    sqlite3PcacheDrop(pPgOld);
  }

  sqlite3PcacheMakeDirty(pPg);

  pPager->dbModified = 1;

  if( needSyncPgno ){
    /* If needSyncPgno is non-zero, then the journal file needs to be 
    ** sync()ed before any data is written to database file page needSyncPgno.
    ** Currently, no such page exists in the page-cache and the 
    ** "is journaled" bitvec flag has been set. This needs to be remedied by

**    PAGER_JOURNALMODE_TRUNCATE
**    PAGER_JOURNALMODE_PERSIST
**    PAGER_JOURNALMODE_OFF
**
** If the parameter is not _QUERY, then the journal-mode is set to the
** value specified.
**
** The returned indicate the current (possibly updated) journal-mode.

*/
int sqlite3PagerJournalMode(Pager *pPager, int eMode){
  if( !MEMDB ){
    assert( eMode==PAGER_JOURNALMODE_QUERY
              || eMode==PAGER_JOURNALMODE_DELETE
              || eMode==PAGER_JOURNALMODE_TRUNCATE
              || eMode==PAGER_JOURNALMODE_PERSIST

**    May you share freely, never taking more than you give.
**
*************************************************************************
** This header file defines the interface that the sqlite page cache
** subsystem.  The page cache subsystem reads and writes a file a page
** at a time and provides a journal for rollback.
**
** @(#) $Id: pager.h,v 1.94 2009/01/16 15:21:06 danielk1977 Exp $
*/

#ifndef _PAGER_H_
#define _PAGER_H_

/*
** Default maximum size for persistent journal files. A negative 
** value means no limit. This value may be overridden using the 
** sqlite3PagerJournalSizeLimit() API. See also "PRAGMA journal_size_limit".
*/
#ifndef SQLITE_DEFAULT_JOURNAL_SIZE_LIMIT
  #define SQLITE_DEFAULT_JOURNAL_SIZE_LIMIT -1
#endif

/*
** The type used to represent a page number.  The first page in a file

typedef struct Pager Pager;

/*
** Handle type for pages.
*/
typedef struct PgHdr DbPage;

/*
** Page number PAGER_MJ_PGNO is never used in an SQLite database (it is
** reserved for working around a windows/posix incompatibility). It is
** used in the journal to signify that the remainder of the journal file 
** is devoted to storing a master journal name - there are no more pages to
** roll back. See comments for function writeMasterJournal() in pager.c 
** for details.
*/
#define PAGER_MJ_PGNO(x) ((Pgno)((PENDING_BYTE/((x)->pageSize))+1))

/*
** Allowed values for the flags parameter to sqlite3PagerOpen().
**
** NOTE: These values must match the corresponding BTREE_ values in btree.h.
*/
#define PAGER_OMIT_JOURNAL  0x0001    /* Do not use a rollback journal */
#define PAGER_NO_READLOCK   0x0002    /* Omit readlocks on readonly files */

/*
** Valid values for the second argument to sqlite3PagerLockingMode().
*/

#define PAGER_JOURNALMODE_DELETE      0   /* Commit by deleting journal file */
#define PAGER_JOURNALMODE_PERSIST     1   /* Commit by zeroing journal header */
#define PAGER_JOURNALMODE_OFF         2   /* Journal omitted.  */
#define PAGER_JOURNALMODE_TRUNCATE    3   /* Commit by truncating journal */
#define PAGER_JOURNALMODE_MEMORY      4   /* In-memory journal file */

/*
** The remainder of this file contains the declarations of the functions
** that make up the Pager sub-system API. See source code comments for 
** a detailed description of each routine.
*/

/* Open and close a Pager connection. */ 
int sqlite3PagerOpen(sqlite3_vfs *, Pager **ppPager, const char*, int,int,int);
int sqlite3PagerClose(Pager *pPager);

/* Functions used to configure a Pager object. */
void sqlite3PagerSetBusyhandler(Pager*, int(*)(void *), void *);
void sqlite3PagerSetReiniter(Pager*, void(*)(DbPage*));
int sqlite3PagerSetPagesize(Pager*, u16*);
int sqlite3PagerMaxPageCount(Pager*, int);
int sqlite3PagerReadFileheader(Pager*, int, unsigned char*);
void sqlite3PagerSetCachesize(Pager*, int);
void sqlite3PagerSetSafetyLevel(Pager*,int,int);
int sqlite3PagerLockingMode(Pager *, int);
int sqlite3PagerJournalMode(Pager *, int);
i64 sqlite3PagerJournalSizeLimit(Pager *, i64);

/* Functions used to obtain and release page references. */ 
int sqlite3PagerAcquire(Pager *pPager, Pgno pgno, DbPage **ppPage, int clrFlag);
#define sqlite3PagerGet(A,B,C) sqlite3PagerAcquire(A,B,C,0)
DbPage *sqlite3PagerLookup(Pager *pPager, Pgno pgno);

void sqlite3PagerRef(DbPage*);
void sqlite3PagerUnref(DbPage*);

/* Operations on page references. */
int sqlite3PagerWrite(DbPage*);
void sqlite3PagerDontWrite(DbPage*);
int sqlite3PagerMovepage(Pager*,DbPage*,Pgno,int);
int sqlite3PagerPageRefcount(DbPage*);
void *sqlite3PagerGetData(DbPage *); 
void *sqlite3PagerGetExtra(DbPage *); 

/* Functions used to manage pager transactions and savepoints. */
int sqlite3PagerPagecount(Pager*, int*);
int sqlite3PagerBegin(DbPage*, int exFlag);
int sqlite3PagerCommitPhaseOne(Pager*,const char *zMaster, int);
int sqlite3PagerSync(Pager *pPager);
int sqlite3PagerCommitPhaseTwo(Pager*);
int sqlite3PagerRollback(Pager*);
int sqlite3PagerOpenSavepoint(Pager *pPager, int n);
int sqlite3PagerSavepoint(Pager *pPager, int op, int iSavepoint);

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


int sqlite3PagerRefcount(Pager*);

const char *sqlite3PagerFilename(Pager*);
const sqlite3_vfs *sqlite3PagerVfs(Pager*);
sqlite3_file *sqlite3PagerFile(Pager*);
const char *sqlite3PagerDirname(Pager*);
const char *sqlite3PagerJournalname(Pager*);
int sqlite3PagerNosync(Pager*);






void *sqlite3PagerTempSpace(Pager*);




/* Functions used in auto-vacuum mode to truncate the database file. */
#ifndef SQLITE_OMIT_AUTOVACUUM
  void sqlite3PagerTruncateImage(Pager*,Pgno);

#endif

/* Used by encryption extensions. */
#ifdef SQLITE_HAS_CODEC
  void sqlite3PagerSetCodec(Pager*,void*(*)(void*,void*,Pgno,int),void*);
#endif

/* Functions to support testing and debugging. */
#if !defined(NDEBUG) || defined(SQLITE_TEST)
  Pgno sqlite3PagerPagenumber(DbPage*);
  int sqlite3PagerIswriteable(DbPage*);
#endif

#ifdef SQLITE_TEST
  int *sqlite3PagerStats(Pager*);
  void sqlite3PagerRefdump(Pager*);
  int sqlite3PagerIsMemdb(Pager*);



  void disable_simulated_io_errors(void);
  void enable_simulated_io_errors(void);
#else
# define disable_simulated_io_errors()
# define enable_simulated_io_errors()
#endif

#endif /* _PAGER_H_ */

/*
** 2008 August 05
**
** The author disclaims copyright to this source code.  In place of
** a legal notice, here is a blessing:
**
**    May you do good and not evil.
**    May you find forgiveness for yourself and forgive others.
**    May you share freely, never taking more than you give.
**
*************************************************************************
** This file implements that page cache.
**
** @(#) $Id: pcache.c,v 1.40 2009/01/16 15:21:06 danielk1977 Exp $
*/
#include "sqliteInt.h"

/*
** A complete page cache is an instance of this structure.
*/
struct PCache {

    sqlite3GlobalConfig.pcache.xDestroy(pCache->pCache);
  }
}

/* 
** Discard the contents of the cache.
*/
void sqlite3PcacheClear(PCache *pCache){
  sqlite3PcacheTruncate(pCache, 0);

}

/*
** Merge two lists of pages connected by pDirty and in pgno order.
** Do not both fixing the pDirtyPrev pointers.
*/
static PgHdr *pcacheMergeDirtyList(PgHdr *pA, PgHdr *pB){

**    May you find forgiveness for yourself and forgive others.
**    May you share freely, never taking more than you give.
**
*************************************************************************
** This header file defines the interface that the sqlite page cache
** subsystem. 
**
** @(#) $Id: pcache.h,v 1.17 2009/01/16 15:21:06 danielk1977 Exp $
*/

#ifndef _PCACHE_H_

typedef struct PgHdr PgHdr;
typedef struct PCache PCache;

/* Reset and close the cache object */
void sqlite3PcacheClose(PCache*);

/* Clear flags from pages of the page cache */
void sqlite3PcacheClearSyncFlags(PCache *);

/* Discard the contents of the cache */
void sqlite3PcacheClear(PCache*);

/* Return the total number of outstanding page references */
int sqlite3PcacheRefCount(PCache*);

/* Increment the reference count of an existing page */
void sqlite3PcacheRef(PgHdr*);

/*
** 2001 September 15
**
** The author disclaims copyright to this source code.  In place of
** a legal notice, here is a blessing:
**
**    May you do good and not evil.
**    May you find forgiveness for yourself and forgive others.
**    May you share freely, never taking more than you give.
**
*************************************************************************
** Internal interface definitions for SQLite.
**
** @(#) $Id: sqliteInt.h,v 1.826 2009/01/16 15:21:06 danielk1977 Exp $
*/
#ifndef _SQLITEINT_H_
#define _SQLITEINT_H_

/*
** Include the configuration header output by 'configure' if we're using the
** autoconf-based build

void sqlite3EndTable(Parse*,Token*,Token*,Select*);

Bitvec *sqlite3BitvecCreate(u32);
int sqlite3BitvecTest(Bitvec*, u32);
int sqlite3BitvecSet(Bitvec*, u32);
void sqlite3BitvecClear(Bitvec*, u32);
void sqlite3BitvecDestroy(Bitvec*);
u32 sqlite3BitvecSize(Bitvec*);
int sqlite3BitvecBuiltinTest(int,int*);

RowSet *sqlite3RowSetInit(sqlite3*, void*, unsigned int);
void sqlite3RowSetClear(RowSet*);
void sqlite3RowSetInsert(RowSet*, i64);
int sqlite3RowSetNext(RowSet*, i64*);

**    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 1.66 2009/01/16 15:21:06 danielk1977 Exp $
*/
#include "sqliteInt.h"
#include "tcl.h"
#include <stdlib.h>
#include <string.h>
#include <ctype.h>

static int page_unref(
  void *NotUsed,
  Tcl_Interp *interp,    /* The TCL interpreter that invoked this command */
  int argc,              /* Number of arguments */
  const char **argv      /* Text of each argument */
){
  DbPage *pPage;

  if( argc!=2 ){
    Tcl_AppendResult(interp, "wrong # args: should be \"", argv[0],
       " PAGE\"", 0);
    return TCL_ERROR;
  }
  pPage = (DbPage *)sqlite3TestTextToPtr(argv[1]);
  sqlite3PagerUnref(pPage);




  return TCL_OK;
}

/*
** Usage:   page_read PAGE
**
** Return the content of a page

**
*************************************************************************
** This file contains code used for creating, destroying, and populating
** a VDBE (or an "sqlite3_stmt" as it is known to the outside world.)  Prior
** to version 2.8.7, all this code was combined into the vdbe.c source file.
** But that file was getting too big so this subroutines were split out.
**
** $Id: vdbeaux.c,v 1.431 2009/01/16 15:21:06 danielk1977 Exp $
*/
#include "sqliteInt.h"
#include <ctype.h>
#include "vdbeInt.h"

        }
      }
    }

    /* Sync the master journal file. If the IOCAP_SEQUENTIAL device
    ** flag is set this is not required.
    */

    if( needSync 
     && 0==(sqlite3OsDeviceCharacteristics(pMaster)&SQLITE_IOCAP_SEQUENTIAL)
     && SQLITE_OK!=(rc = sqlite3OsSync(pMaster, SQLITE_SYNC_NORMAL))
    ){
      sqlite3OsCloseFree(pMaster);
      sqlite3OsDelete(pVfs, zMaster, 0);
      sqlite3DbFree(db, zMaster);
      return rc;
    }

    /* Sync all the db files involved in the transaction. The same call