SQLite

** 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 1.163 2004/09/08 20:13:05 drh Exp $
*/
#include "os.h"         /* Must be first to enable large file support */
#include "sqliteInt.h"
#include "pager.h"
#include <assert.h>
#include <string.h>

  u8 *aInStmt;                /* One bit for each page in the database */
  u8 setMaster;               /* True if a m-j name has been written to jrnl */
  BusyHandler *pBusyHandler;  /* Pointer to sqlite.busyHandler */
  PgHdr *pFirst, *pLast;      /* List of free pages */
  PgHdr *pFirstSynced;        /* First free page with PgHdr.needSync==0 */
  PgHdr *pAll;                /* List of all pages */
  PgHdr *pStmt;               /* List of pages in the statement subjournal */
  PgHdr *aHash[N_PG_HASH];    /* Hash table to map page number to PgHdr */
  off_t journalOff;           /* Current byte offset in the journal file */
  off_t journalHdr;           /* Byte offset to previous journal header */
  off_t stmtHdrOff;           /* First journal header written this statement */
  off_t stmtCksum;            /* cksumInit when statement was started */
  int sectorSize;             /* Assumed sector size during rollback */

};

/*
** These are bits that can be set in Pager.errMask.
*/
#define PAGER_ERR_FULL     0x01  /* a write() failed */
#define PAGER_ERR_MEM      0x02  /* malloc() failed */

** by an instance of the following structure
*/
typedef struct et_info {   /* Information about each format field */
  char fmttype;            /* The format field code letter */
  etByte base;             /* The base for radix conversion */
  etByte flags;            /* One or more of FLAG_ constants below */
  etByte type;             /* Conversion paradigm */
  char *charset;           /* The character set for conversion */
  char *prefix;            /* Prefix on non-zero values in alt format */
} et_info;

/*
** Allowed values for et_info.flags
*/
#define FLAG_SIGNED  1     /* True if the value to convert is signed */
#define FLAG_INTERN  2     /* True if for internal use only */


/*
** The following table is searched linearly, so it is good to put the
** most frequently used conversion types first.
*/


static et_info fmtinfo[] = {
  {  'd', 10, 1, etRADIX,      "0123456789",       0    },
  {  's',  0, 0, etSTRING,     0,                  0    },
  {  'z',  0, 2, etDYNSTRING,  0,                  0    },
  {  'q',  0, 0, etSQLESCAPE,  0,                  0    },
  {  'Q',  0, 0, etSQLESCAPE2, 0,                  0    },
  {  'c',  0, 0, etCHARX,      0,                  0    },
  {  'o',  8, 0, etRADIX,      "01234567",         "0"  },
  {  'u', 10, 0, etRADIX,      "0123456789",       0    },
  {  'x', 16, 0, etRADIX,      "0123456789abcdef", "x0" },
  {  'X', 16, 0, etRADIX,      "0123456789ABCDEF", "X0" },
  {  'f',  0, 1, etFLOAT,      0,                  0    },
  {  'e',  0, 1, etEXP,        "e",                0    },
  {  'E',  0, 1, etEXP,        "E",                0    },
  {  'g',  0, 1, etGENERIC,    "e",                0    },
  {  'G',  0, 1, etGENERIC,    "E",                0    },
  {  'i', 10, 1, etRADIX,      "0123456789",       0    },
  {  'n',  0, 0, etSIZE,       0,                  0    },
  {  '%',  0, 0, etPERCENT,    0,                  0    },
  {  'p', 16, 0, etPOINTER,    "0123456789abcdef", "x0" },
  {  'T',  0, 2, etTOKEN,      0,                  0    },
  {  'S',  0, 2, etSRCLIST,    0,                  0    },
};
#define etNINFO  (sizeof(fmtinfo)/sizeof(fmtinfo[0]))

/*
** If NOFLOATINGPOINT is defined, then none of the floating point
** conversions will work.
*/

        }
        if( longvalue==0 ) flag_alternateform = 0;
        if( flag_zeropad && precision<width-(prefix!=0) ){
          precision = width-(prefix!=0);
        }
        bufpt = &buf[etBUFSIZE-1];
        {
          register char *cset;      /* Use registers for speed */
          register int base;
          cset = infop->charset;
          base = infop->base;
          do{                                           /* Convert to ascii */
            *(--bufpt) = cset[longvalue%base];
            longvalue = longvalue/base;
          }while( longvalue>0 );
        }
        length = &buf[etBUFSIZE-1]-bufpt;
        for(idx=precision-length; idx>0; idx--){
          *(--bufpt) = '0';                             /* Zero pad */
        }
        if( prefix ) *(--bufpt) = prefix;               /* Add sign */
        if( flag_alternateform && infop->prefix ){      /* Add "0" or "0x" */
          char *pre, x;

          pre = infop->prefix;
          if( *bufpt!=pre[0] ){
            for(pre=infop->prefix; (x=(*pre))!=0; pre++) *(--bufpt) = x;
          }
        }
        length = &buf[etBUFSIZE-1]-bufpt;
        break;
      case etFLOAT:
      case etEXP:
      case etGENERIC:

          bufpt--;                            /* point to last digit */
          if( flag_rtz && flag_dp ){          /* Remove tail zeros */
            while( bufpt>=buf && *bufpt=='0' ) *(bufpt--) = 0;
            if( bufpt>=buf && *bufpt=='.' ) *(bufpt--) = 0;
          }
          bufpt++;                            /* point to next free slot */
          if( exp || flag_exp ){
            *(bufpt++) = infop->charset[0];
            if( exp<0 ){ *(bufpt++) = '-'; exp = -exp; } /* sign of exp */
            else       { *(bufpt++) = '+'; }
            if( exp>=100 ){
              *(bufpt++) = (exp/100)+'0';                /* 100's digit */
              exp %= 100;
            }
            *(bufpt++) = exp/10+'0';                     /* 10's digit */

#if defined(SQLITE_TEST) || defined(SQLITE_DEBUG)
/*
** A version of printf() that understands %lld.  Used for debugging.
** The printf() built into some versions of windows does not understand %lld
** and segfaults if you give it a long long int.
*/
void sqlite3DebugPrintf(const char *zFormat, ...){

  va_list ap;
  char zBuf[500];
  va_start(ap, zFormat);
  base_vprintf(0, 0, zBuf, sizeof(zBuf), zFormat, ap);
  va_end(ap);
  fprintf(stdout,"%d: %s", getpid(), zBuf);
  fflush(stdout);
}
#endif

/*
** Turn tracing on or off
*/
void sqlite3VdbeTrace(Vdbe *p, FILE *trace){
  p->trace = trace;
}
















/*
** Add a new instruction to the list of instructions current in the
** VDBE.  Return the address of the new instruction.
**
** Parameters:
**

int sqlite3VdbeAddOp(Vdbe *p, int op, int p1, int p2){
  int i;
  VdbeOp *pOp;

  i = p->nOp;
  p->nOp++;
  assert( p->magic==VDBE_MAGIC_INIT );
  if( i>=p->nOpAlloc ){
    int oldSize = p->nOpAlloc;
    Op *aNew;
    p->nOpAlloc = p->nOpAlloc*2 + 100;
    aNew = sqliteRealloc(p->aOp, p->nOpAlloc*sizeof(Op));
    if( aNew==0 ){
      p->nOpAlloc = oldSize;
      return 0;
    }
    p->aOp = aNew;
    memset(&p->aOp[oldSize], 0, (p->nOpAlloc-oldSize)*sizeof(Op));
  }
  pOp = &p->aOp[i];
  pOp->opcode = op;
  pOp->p1 = p1;
  if( p2<0 && (-1-p2)<p->nLabel && p->aLabel[-1-p2]>=0 ){
    p2 = p->aLabel[-1-p2];
  }
  pOp->p2 = p2;
  pOp->p3 = 0;
  pOp->p3type = P3_NOTUSED;
#ifndef NDEBUG
  if( sqlite3_vdbe_addop_trace ) sqlite3VdbePrintOp(0, i, &p->aOp[i]);
#endif
  return i;

** Zero is returned if a malloc() fails.
*/
int sqlite3VdbeMakeLabel(Vdbe *p){
  int i;
  i = p->nLabel++;
  assert( p->magic==VDBE_MAGIC_INIT );
  if( i>=p->nLabelAlloc ){
    int *aNew;
    p->nLabelAlloc = p->nLabelAlloc*2 + 10;
    aNew = sqliteRealloc( p->aLabel, p->nLabelAlloc*sizeof(p->aLabel[0]));
    if( aNew==0 ){
      sqliteFree(p->aLabel);
    }
    p->aLabel = aNew;
  }
  if( p->aLabel==0 ){
    p->nLabel = 0;
    p->nLabelAlloc = 0;
    return 0;
  }
  p->aLabel[i] = -1;
  return -1-i;
}

/*
** Resolve label "x" to be the address of the next instruction to
** be inserted.  The parameter "x" must have been obtained from
** a prior call to sqlite3VdbeMakeLabel().
*/
void sqlite3VdbeResolveLabel(Vdbe *p, int x){
  int j;
  assert( p->magic==VDBE_MAGIC_INIT );
  if( x<0 && (-x)<=p->nLabel && p->aOp ){
    if( p->aLabel[-1-x]==p->nOp ) return;
    assert( p->aLabel[-1-x]<0 );
    p->aLabel[-1-x] = p->nOp;
    for(j=0; j<p->nOp; j++){
      if( p->aOp[j].p2==x ) p->aOp[j].p2 = p->nOp;
    }
  }




















}

/*
** Return the address of the next instruction to be inserted.
*/
int sqlite3VdbeCurrentAddr(Vdbe *p){
  assert( p->magic==VDBE_MAGIC_INIT );
  return p->nOp;
}

/*
** Add a whole list of operations to the operation stack.  Return the
** address of the first operation added.
*/
int sqlite3VdbeAddOpList(Vdbe *p, int nOp, VdbeOpList const *aOp){
  int addr;
  assert( p->magic==VDBE_MAGIC_INIT );
  if( p->nOp + nOp >= p->nOpAlloc ){
    int oldSize = p->nOpAlloc;
    Op *aNew;
    p->nOpAlloc = p->nOpAlloc*2 + nOp + 10;
    aNew = sqliteRealloc(p->aOp, p->nOpAlloc*sizeof(Op));
    if( aNew==0 ){
      p->nOpAlloc = oldSize;
      return 0;
    }
    p->aOp = aNew;
    memset(&p->aOp[oldSize], 0, (p->nOpAlloc-oldSize)*sizeof(Op));
  }
  addr = p->nOp;
  if( nOp>0 ){
    int i;
    VdbeOpList const *pIn = aOp;
    for(i=0; i<nOp; i++, pIn++){
      int p2 = pIn->p2;

/*
** Replace the P3 field of the most recently coded instruction with
** comment text.
*/
void sqlite3VdbeComment(Vdbe *p, const char *zFormat, ...){
  va_list ap;
  assert( p->nOp>0 );
  assert( p->aOp[p->nOp-1].p3==0 );
  va_start(ap, zFormat);
  sqlite3VdbeChangeP3(p, -1, sqlite3VMPrintf(zFormat, ap), P3_DYNAMIC);
  va_end(ap);
}
#endif

/*

  if( pOut==0 ) pOut = stdout;
  zP3 = displayP3(pOp, zPtr, sizeof(zPtr));
  fprintf(pOut, zFormat1,
      pc, sqlite3OpcodeNames[pOp->opcode], pOp->p1, pOp->p2, zP3);
  fflush(pOut);
}
#endif












/*
** Give a listing of the program in the virtual machine.
**
** The interface is the same as sqlite3VdbeExec().  But instead of
** running the code, it invokes the callback once for each instruction.
** This feature is used to implement "EXPLAIN".

  assert( p->explain );

  /* Even though this opcode does not put dynamic strings onto the
  ** the stack, they may become dynamic if the user calls
  ** sqlite3_column_text16(), causing a translation to UTF-16 encoding.
  */
  if( p->pTos==&p->aStack[4] ){
    for(i=0; i<5; i++){
      sqlite3VdbeMemRelease(&p->aStack[i]);
      p->aStack[i].flags = 0;
    }
  }
  p->resOnStack = 0;

  i = p->pc++;
  if( i>=p->nOp ){
    p->rc = SQLITE_OK;
    rc = SQLITE_DONE;

  ** stack.  And the stack never grows on successive executions of the
  ** same loop.  So the total number of instructions is an upper bound
  ** on the maximum stack depth required.
  **
  ** Allocation all the stack space we will ever need.
  */
  if( p->aStack==0 ){

    assert( nVar>=0 );
    n = isExplain ? 10 : p->nOp;
    p->aStack = sqliteMalloc(
      n*(sizeof(p->aStack[0])+sizeof(Mem*))          /* aStack, apArg */
      + nVar*sizeof(Mem)                             /* aVar */
      + nVar*sizeof(char*)                           /* azVar */
      + nMem*sizeof(Mem)                             /* aMem */

** This routine will automatically close any cursors, lists, and/or
** sorters that were left open.  It also deletes the values of
** variables in the aVar[] array.
*/
static void Cleanup(Vdbe *p){
  int i;
  if( p->aStack ){
    Mem *pTos = p->pTos;
    while( pTos>=p->aStack ){
      sqlite3VdbeMemRelease(pTos);
      pTos--;
    }
    p->pTos = pTos;
  }
  closeAllCursors(p);
  for(i=0; i<p->nMem; i++){
    sqlite3VdbeMemRelease(&p->aMem[i]);
  }
  if( p->pList ){
    sqlite3VdbeKeylistFree(p->pList);
    p->pList = 0;
  }

  for(i=0; i<p->contextStackTop; i++){
    sqlite3VdbeKeylistFree(p->contextStack[i].pList);


  }
  sqlite3VdbeSorterReset(p);
  sqlite3VdbeAggReset(0, &p->agg, 0);
  sqliteFree(p->contextStack);
  p->contextStack = 0;
  p->contextStackDepth = 0;
  p->contextStackTop = 0;
  sqliteFree(p->zErrMsg);
  p->zErrMsg = 0;
}

/*
** Set the number of result columns that will be returned by this SQL
** statement. This is now set at compile time, rather than during
** execution of the vdbe program so that sqlite3_column_count() can
** be called on an SQL statement before sqlite3_step().
*/
void sqlite3VdbeSetNumCols(Vdbe *p, int nResColumn){


  assert( 0==p->nResColumn );
  p->nResColumn = nResColumn;






}

/*
** Set the name of the idx'th column to be returned by the SQL statement.
** zName must be a pointer to a nul terminated string.
**
** This call must be made after a call to sqlite3VdbeSetNumCols().
**
** If N==P3_STATIC  it means that zName is a pointer to a constant static
** string and we can just copy the pointer. If it is P3_DYNAMIC, then 
** the string is freed using sqliteFree() when the vdbe is finished with
** it. Otherwise, N bytes of zName are copied.
*/
int sqlite3VdbeSetColName(Vdbe *p, int idx, const char *zName, int N){
  int rc;
  Mem *pColName;
  assert( idx<(2*p->nResColumn) );

  /* If the Vdbe.aColName array has not yet been allocated, allocate
  ** it now.
  */
  if( !p->aColName ){
    int i;
    p->aColName = (Mem *)sqliteMalloc(sizeof(Mem)*p->nResColumn*2);
    if( !p->aColName ){
      return SQLITE_NOMEM;
    }
    for(i=0; i<(2*p->nResColumn); i++){
      p->aColName[i].flags = MEM_Null;
    }
  }

  pColName = &(p->aColName[idx]);
  if( N==P3_DYNAMIC || N==P3_STATIC ){
    rc = sqlite3VdbeMemSetStr(pColName, zName, -1, SQLITE_UTF8, SQLITE_STATIC);
  }else{
    rc = sqlite3VdbeMemSetStr(pColName, zName, N, SQLITE_UTF8,SQLITE_TRANSIENT);
  }
  if( rc==SQLITE_OK && N==P3_DYNAMIC ){

  }else{
    assert( p->db->pVdbe==p );
    p->db->pVdbe = p->pNext;
  }
  if( p->pNext ){
    p->pNext->pPrev = p->pPrev;
  }
  p->pPrev = p->pNext = 0;
  if( p->nOpAlloc==0 ){
    p->aOp = 0;
    p->nOp = 0;
  }
  for(i=0; i<p->nOp; i++){
    Op *pOp = &p->aOp[i];
    if( pOp->p3type==P3_DYNAMIC || pOp->p3type==P3_KEYINFO ){
      sqliteFree(pOp->p3);
    }
    if( pOp->p3type==P3_VDBEFUNC ){
      VdbeFunc *pVdbeFunc = (VdbeFunc *)pOp->p3;
      sqlite3VdbeDeleteAuxData(pVdbeFunc, 0);
      sqliteFree(pVdbeFunc);
    }
  }
  for(i=0; i<p->nVar; i++){
    sqlite3VdbeMemRelease(&p->aVar[i]);
  }
  sqliteFree(p->aOp);
  sqliteFree(p->aLabel);
  sqliteFree(p->aStack);
  if( p->aColName ){
    for(i=0; i<(p->nResColumn)*2; i++){
      sqlite3VdbeMemRelease(&(p->aColName[i]));
    }
    sqliteFree(p->aColName);
  }
  p->magic = VDBE_MAGIC_DEAD;
  sqliteFree(p);
}

/*
** If a MoveTo operation is pending on the given cursor, then do that
** MoveTo now.  Return an error code.  If no MoveTo is pending, this