SQLite

      CREATE INDEX i1 ON t1(b, c);
      CREATE INDEX i2 ON t1(c, b);
      CREATE INDEX i3 ON t1(a, b, c, a, b, c);
    }
  } {
    reset_db
    execsql $schema
  

    do_test 1.$tn2.$tn.1 {
      create_ota1 ota.db
      $cmd test.db ota.db
    } {SQLITE_DONE}
    

    do_execsql_test 1.$tn2.$tn.2 {
      SELECT * FROM t1 ORDER BY a ASC;
    } {
      1 2 3 
      2 two three 
      3 {} 8.2
    }
  
 
    do_execsql_test 1.$tn2.$tn.3 { PRAGMA integrity_check } ok
  }
}

#-------------------------------------------------------------------------
# Check that an OTA cannot be applied to a table that has no PK.
#

#include "sqlite3.h"
#include "sqlite3ota.h"


/*
** The ota_state table is used to save the state of a partially applied
** update so that it can be resumed later. The table contains at most a
** single row:
** update so that it can be resumed later. The table consists of integer
** keys mapped to values as follows:
**
** OTA_STATE_STAGE:
**   May be set to integer values 1, 2 or 3. As follows:
**       0: Nothing at all has been done.
**       1: the *-ota file is currently under construction.
**       2: the *-ota file has been constructed, but not yet moved 
**          to the *-wal path.
**       3: the checkpoint is underway.
**
** OTA_STATE_TBL:
**   Only valid if STAGE==1. The target database name of the table 
**   "tbl"       -> Table currently being written (target database names).
**   currently being written.
**
** OTA_STATE_IDX:
**   "idx"       -> Index currently being written (target database names).
**                  Or, if the main table is being written, a NULL value.
**   Only valid if STAGE==1. The target database name of the index 
**   currently being written, or NULL if the main table is currently being
**   updated.
**
** OTA_STATE_ROW:
**   "row"       -> Number of rows for this object already processed
**   Only valid if STAGE==1. Number of rows already processed for the current
**   table/index.
**
** OTA_STATE_PROGRESS:
**   "progress"  -> total number of key/value b-tree operations performed
**                  so far as part of this ota update.
**   Total number of sqlite3ota_step() calls made so far as part of this
**   ota update.
**
** OTA_STATE_CKPT:
**   Valid if STAGE==3. The blob to pass to sqlite3ckpt_start() to resume
**   the incremental checkpoint.
**
*/
#define OTA_STATE_STAGE       1
#define OTA_STATE_TBL         2
#define OTA_STATE_IDX         3
#define OTA_STATE_ROW         4
#define OTA_STATE_PROGRESS    5
#define OTA_STATE_CKPT        6

#define OTA_STAGE_OAL         1
#define OTA_STAGE_COPY        2
#define OTA_STAGE_CKPT        3
#define OTA_STAGE_DONE        4


#define OTA_CREATE_STATE "CREATE TABLE IF NOT EXISTS ota.ota_state"        \
                             "(tbl, idx, row, progress)"
                             "(k INTEGER PRIMARY KEY, v)"

typedef struct OtaState OtaState;
typedef struct OtaObjIter OtaObjIter;

/*
** A structure to store values read from the ota_state table in memory.
*/
struct OtaState {
  int eStage;
  char *zTbl;
  char *zIdx;
  unsigned char *pCkptState;
  int nCkptState;
  int nRow;
  sqlite3_int64 nProgress;
};

/*
** An iterator of this type is used to iterate through all objects in
** the target database that require updating. For each such table, the

  sqlite3_stmt *pUpdate;          /* Last update statement (or NULL) */
};

/*
** OTA handle.
*/
struct sqlite3ota {
  int eStage;                     /* Value of OTA_STATE_STAGE field */
  sqlite3 *db;                    /* "main" -> target db, "ota" -> ota db */
  char *zTarget;                  /* Path to target db */
  char *zOta;                     /* Path to ota db */
  int rc;                         /* Value returned by last ota_step() call */
  char *zErrmsg;                  /* Error message if rc!=SQLITE_OK */
  int nStep;                      /* Rows processed for current object */
  int nProgress;                  /* Rows processed for all objects */
  OtaObjIter objiter;
  OtaObjIter objiter;             /* Iterator for skipping through tbl/idx */
  sqlite3_ckpt *pCkpt;            /* Incr-checkpoint handle */
};

/*
** Prepare the SQL statement in buffer zSql against database handle db.
** If successful, set *ppStmt to point to the new statement and return
** SQLITE_OK. 
**

      memcpy(pIter->zMask, zMask, pIter->nTblCol);
    }
    sqlite3_free(zWhere);
    sqlite3_free(zSet);
  }
  return p->rc;
}

static void otaOpenDatabase(sqlite3ota *p){
  assert( p->rc==SQLITE_OK );
  sqlite3_close(p->db);
  p->db = 0;

  p->rc = sqlite3_open(p->zTarget, &p->db);
  if( p->rc ){
    p->zErrmsg = sqlite3_mprintf("%s", sqlite3_errmsg(p->db));
  }
  otaMPrintfExec(p, "ATTACH %Q AS ota", p->zOta);
}

/*
** This routine is a copy of the sqlite3FileSuffix3() routine from the core.
** It is a no-op unless SQLITE_ENABLE_8_3_NAMES is defined.
**
** If SQLITE_ENABLE_8_3_NAMES is set at compile-time and if the database
** filename in zBaseFilename is a URI with the "8_3_names=1" parameter and
** if filename in z[] has a suffix (a.k.a. "extension") that is longer than
** three characters, then shorten the suffix on z[] to be the last three
** characters of the original suffix.
**
** If SQLITE_ENABLE_8_3_NAMES is set to 2 at compile-time, then always
** do the suffix shortening regardless of URI parameter.
**
** Examples:
**
**     test.db-journal    =>   test.nal
**     test.db-wal        =>   test.wal
**     test.db-shm        =>   test.shm
**     test.db-mj7f3319fa =>   test.9fa
*/
static void otaFileSuffix3(const char *zBase, char *z){
#ifdef SQLITE_ENABLE_8_3_NAMES
#if SQLITE_ENABLE_8_3_NAMES<2
  if( sqlite3_uri_boolean(zBase, "8_3_names", 0) )
#endif
  {
    int i, sz;
    sz = sqlite3Strlen30(z);
    for(i=sz-1; i>0 && z[i]!='/' && z[i]!='.'; i--){}
    if( z[i]=='.' && ALWAYS(sz>i+4) ) memmove(&z[i+1], &z[sz-3], 4);
  }
#endif
}

/*
** Move the "*-oal" file corresponding to the target database to the
** "*-wal" location. If an error occurs, leave an error code and error 
** message in the ota handle.
*/
static void otaMoveOalFile(sqlite3ota *p){
  const char *zBase = sqlite3_db_filename(p->db, "main");

  char *zWal = sqlite3_mprintf("%s-wal", zBase);
  char *zOal = sqlite3_mprintf("%s-oal", zBase);

  assert( p->rc==SQLITE_OK && p->zErrmsg==0 );
  if( zWal==0 || zOal==0 ){
    p->rc = SQLITE_NOMEM;
  }else{
    /* Move the *-oal file to *-wal. At this point connection p->db is
    ** holding a SHARED lock on the target database file (because it is
    ** in WAL mode). So no other connection may be writing the db.  */
    otaFileSuffix3(zBase, zWal);
    otaFileSuffix3(zBase, zOal);
    rename(zOal, zWal);

    /* Re-open the databases. */
    otaObjIterFinalize(&p->objiter);
    otaOpenDatabase(p);
    p->eStage = OTA_STAGE_CKPT;
  }

  sqlite3_free(zWal);
  sqlite3_free(zOal);
}

/*
** The SELECT statement iterating through the keys for the current object
** (p->objiter.pSelect) currently points to a valid row. This function
** determines the type of operation requested by this row and returns
** one of the following values to indicate the result:
**

      /* no-op */
      assert( eType==OTA_DELETE && pIter->zIdx );
    }
  }

  return p->rc;
}

/*
** Increment the schema cookie of the main database opened by p->db.
*/
static void otaIncrSchemaCookie(sqlite3ota *p){
  int iCookie = 1000000;
  sqlite3_stmt *pStmt;

  assert( p->rc==SQLITE_OK && p->zErrmsg==0 );
  p->rc = prepareAndCollectError(p->db, &pStmt, &p->zErrmsg, 
      "PRAGMA schema_version"
  );
  if( p->rc==SQLITE_OK ){
    if( SQLITE_ROW==sqlite3_step(pStmt) ){
      iCookie = sqlite3_column_int(pStmt, 0);
    }
    p->rc = sqlite3_finalize(pStmt);
  }
  if( p->rc==SQLITE_OK ){
    otaMPrintfExec(p, "PRAGMA schema_version = %d", iCookie+1);
  }
}

/*
** Step the OTA object.
*/
int sqlite3ota_step(sqlite3ota *p){
  if( p ){
    switch( p->eStage ){
      case OTA_STAGE_OAL: {
    OtaObjIter *pIter = &p->objiter;
    while( p && p->rc==SQLITE_OK && pIter->zTbl ){
        OtaObjIter *pIter = &p->objiter;
        while( p && p->rc==SQLITE_OK && pIter->zTbl ){

      if( pIter->bCleanup ){
        /* Clean up the ota_tmp_xxx table for the previous table. It 
        ** cannot be dropped as there are currently active SQL statements.
        ** But the contents can be deleted.  */
        otaMPrintfExec(p, "DELETE FROM ota.'ota_tmp_%q'", pIter->zTbl);
      }else{
        otaObjIterPrepareAll(p, pIter, 0);
        
        /* Advance to the next row to process. */
        if( p->rc==SQLITE_OK ){
          int rc = sqlite3_step(pIter->pSelect);
          if( rc==SQLITE_ROW ){
          if( pIter->bCleanup ){
            /* Clean up the ota_tmp_xxx table for the previous table. It 
            ** cannot be dropped as there are currently active SQL statements.
            ** But the contents can be deleted.  */
            otaMPrintfExec(p, "DELETE FROM ota.'ota_tmp_%q'", pIter->zTbl);
          }else{
            otaObjIterPrepareAll(p, pIter, 0);

            /* Advance to the next row to process. */
            if( p->rc==SQLITE_OK ){
              int rc = sqlite3_step(pIter->pSelect);
              if( rc==SQLITE_ROW ){
            p->nStep++;
            p->nProgress++;
            return otaStep(p);
          }
          p->rc = sqlite3_reset(pIter->pSelect);
          p->nStep = 0;
        }
      }
                p->nProgress++;
                p->nStep++;
                return otaStep(p);
              }
              p->rc = sqlite3_reset(pIter->pSelect);
              p->nStep = 0;
            }
          }

      otaObjIterNext(p, pIter);
    }
          otaObjIterNext(p, pIter);
        }

    if( p->rc==SQLITE_OK && pIter->zTbl==0 ){
      p->rc = SQLITE_DONE;
        if( p->rc==SQLITE_OK && pIter->zTbl==0 ){
          p->nProgress++;
          otaIncrSchemaCookie(p);
          if( p->rc==SQLITE_OK ){
            p->rc = sqlite3_exec(p->db, "COMMIT", 0, 0, &p->zErrmsg);
          }
          if( p->rc==SQLITE_OK ){
            otaMoveOalFile(p);
          }
        }
        break;
      }

      case OTA_STAGE_CKPT: {

        if( p->rc==SQLITE_OK && p->pCkpt==0 ){
          p->rc = sqlite3_ckpt_open(p->db, 0, 0, &p->pCkpt);
        }
        if( p->rc==SQLITE_OK ){
          if( SQLITE_OK!=sqlite3_ckpt_step(p->pCkpt) ){
            p->rc = sqlite3_ckpt_close(p->pCkpt, 0, 0);
            p->pCkpt = 0;
            if( p->rc==SQLITE_OK ){
              p->eStage = OTA_STAGE_DONE;
              p->rc = SQLITE_DONE;
            }
          }
          p->nProgress++;
        }

        break;
      }

      default:
        break;
    }
  }
  return p->rc;
}

static void otaSaveTransactionState(sqlite3ota *p){
  sqlite3_stmt *pInsert;
  int rc;
  otaMPrintfExec(p, 
    "INSERT OR REPLACE INTO ota.ota_state(rowid, tbl, idx, row, progress)"
    "VALUES(1, %Q, %Q, %d, %lld)",
    p->objiter.zTbl, p->objiter.zIdx, p->nStep, p->nProgress

  assert( (p->rc==SQLITE_OK || p->rc==SQLITE_DONE) && p->zErrmsg==0 );
  rc = prepareFreeAndCollectError(p->db, &pInsert, &p->zErrmsg, 
      sqlite3_mprintf(
        "INSERT OR REPLACE INTO ota.ota_state(k, v) VALUES "
        "(%d, %d), "
        "(%d, %Q), "
        "(%d, %Q), "
        "(%d, %d), "
        "(%d, %lld), "
        "(%d, ?) ",
        OTA_STATE_STAGE, p->eStage,
        OTA_STATE_TBL, p->objiter.zTbl, 
        OTA_STATE_IDX, p->objiter.zIdx, 
        OTA_STATE_ROW, p->nStep, 
        OTA_STATE_PROGRESS, p->nProgress,
        OTA_STATE_CKPT
      )
  );
  assert( pInsert==0 || rc==SQLITE_OK );
  if( rc==SQLITE_OK ){
    if( p->pCkpt ){
      unsigned char *pCkptState = 0;
      int nCkptState = 0;
      rc = sqlite3_ckpt_close(p->pCkpt, &pCkptState, &nCkptState);
      p->pCkpt = 0;
      sqlite3_bind_blob(pInsert, 1, pCkptState, nCkptState, SQLITE_TRANSIENT);
      sqlite3_free(pCkptState);
    }
  }
  if( rc==SQLITE_OK ){
    sqlite3_step(pInsert);
    rc = sqlite3_finalize(pInsert);
  }else{
    sqlite3_finalize(pInsert);
  }

  if( rc!=SQLITE_OK ){
    p->rc = rc;
  }
}

static char *otaStrndup(char *zStr, int nStr, int *pRc){
  char *zRet = 0;
  assert( *pRc==SQLITE_OK );

  if( zStr ){
    int nCopy = nStr;
    if( nCopy<0 ) nCopy = strlen(zStr) + 1;
    zRet = (char*)sqlite3_malloc(nCopy);
    if( zRet ){
      memcpy(zRet, zStr, nCopy);
    }else{
      *pRc = SQLITE_NOMEM;
    }
  }

  return zRet;
}

static void otaFreeState(OtaState *p){
  sqlite3_free(p->zTbl);
  sqlite3_free(p->zIdx);
  sqlite3_free(p->pCkptState);
  sqlite3_free(p);
}

/*
** Allocate an OtaState object and load the contents of the ota_state 
** table into it. Return a pointer to the new object. It is the 
** responsibility of the caller to eventually free the object using
** sqlite3_free().
**
** If an error occurs, leave an error code and message in the ota handle
** and return NULL.
*/
static OtaState *otaLoadState(sqlite3ota *p){
  const char *zSelect = "SELECT tbl, idx, row, progress FROM ota.ota_state";
  const char *zSelect = "SELECT k, v FROM ota.ota_state";
  OtaState *pRet = 0;
  sqlite3_stmt *pStmt;
  int rc;
  int rc2;

  assert( p->rc==SQLITE_OK );
  pRet = (OtaState*)sqlite3_malloc(sizeof(OtaState));
  if( pRet==0 ){
    rc = SQLITE_NOMEM;
  }else{
    memset(pRet, 0, sizeof(OtaState));
  rc = prepareAndCollectError(p->db, &pStmt, &p->zErrmsg, zSelect);
  if( rc==SQLITE_OK ){
    if( sqlite3_step(pStmt)==SQLITE_ROW ){
      const char *zIdx = (const char*)sqlite3_column_text(pStmt, 1);
      const char *zTbl = (const char*)sqlite3_column_text(pStmt, 0);
    rc = prepareAndCollectError(p->db, &pStmt, &p->zErrmsg, zSelect);
  }

  while( rc==SQLITE_OK && SQLITE_ROW==sqlite3_step(pStmt) ){
    switch( sqlite3_column_int(pStmt, 0) ){
      case OTA_STATE_STAGE:
        pRet->eStage = sqlite3_column_int(pStmt, 1);
      int nIdx = zIdx ? (strlen(zIdx) + 1) : 0;
      int nTbl = strlen(zTbl) + 1;
        if( pRet->eStage!=OTA_STAGE_OAL
         && pRet->eStage!=OTA_STAGE_COPY
         && pRet->eStage!=OTA_STAGE_CKPT
        ){
          p->rc = SQLITE_CORRUPT;
      int nByte = sizeof(OtaState) + nTbl + nIdx;

        }
      pRet = (OtaState*)sqlite3_malloc(nByte);
      if( pRet ){
        pRet->zTbl = (char*)&pRet[1];
        memcpy(pRet->zTbl, sqlite3_column_text(pStmt, 0), nTbl);
        if( zIdx ){
        break;

      case OTA_STATE_TBL:
        pRet->zTbl = otaStrndup((char*)sqlite3_column_text(pStmt, 1), -1, &rc);
        break;

          pRet->zIdx = &pRet->zTbl[nTbl];
          memcpy(pRet->zIdx, zIdx, nIdx);
        }else{
          pRet->zIdx = 0;
        }
        pRet->nRow = sqlite3_column_int(pStmt, 2);
        pRet->nProgress = sqlite3_column_int64(pStmt, 3);
      }
    }else{
      pRet = (OtaState*)sqlite3_malloc(sizeof(OtaState));
      if( pRet ){
        memset(pRet, 0, sizeof(*pRet));
      }
    }
    rc = sqlite3_finalize(pStmt);
    if( rc==SQLITE_OK && pRet==0 ) rc = SQLITE_NOMEM;
      case OTA_STATE_IDX:
        pRet->zIdx = otaStrndup((char*)sqlite3_column_text(pStmt, 1), -1, &rc);
        break;

      case OTA_STATE_ROW:
        pRet->nRow = sqlite3_column_int(pStmt, 1);
        break;

      case OTA_STATE_PROGRESS:
        pRet->nProgress = sqlite3_column_int64(pStmt, 1);
        break;

      case OTA_STATE_CKPT:
        pRet->nCkptState = sqlite3_column_bytes(pStmt, 1);
        pRet->pCkptState = (unsigned char*)otaStrndup(
            (char*)sqlite3_column_blob(pStmt, 1), pRet->nCkptState, &rc
        );
        break;

      default:
        rc = SQLITE_CORRUPT;
        break;
    }
  }
  rc2 = sqlite3_finalize(pStmt);
  if( rc==SQLITE_OK ) rc = rc2;
    if( rc!=SQLITE_OK ){
      sqlite3_free(pRet);
      pRet = 0;
    }
  }

  p->rc = rc;
  return pRet;
}

static int otaStrCompare(const char *z1, const char *z2){
  if( z1==0 && z2==0 ) return 0;

      rc = otaObjIterPrepareAll(p, &p->objiter, p->nStep);
    }

    p->rc = rc;
  }
}

/*
** This routine is a copy of the sqlite3FileSuffix3() routine from the core.
** It is a no-op unless SQLITE_ENABLE_8_3_NAMES is defined.
**
** If SQLITE_ENABLE_8_3_NAMES is set at compile-time and if the database
** filename in zBaseFilename is a URI with the "8_3_names=1" parameter and
** if filename in z[] has a suffix (a.k.a. "extension") that is longer than
** three characters, then shorten the suffix on z[] to be the last three
** characters of the original suffix.
**
** If SQLITE_ENABLE_8_3_NAMES is set to 2 at compile-time, then always
** do the suffix shortening regardless of URI parameter.
**
** Examples:
**
**     test.db-journal    =>   test.nal
**     test.db-wal        =>   test.wal
**     test.db-shm        =>   test.shm
**     test.db-mj7f3319fa =>   test.9fa
*/
static void otaFileSuffix3(const char *zBase, char *z){
#ifdef SQLITE_ENABLE_8_3_NAMES
#if SQLITE_ENABLE_8_3_NAMES<2
  if( sqlite3_uri_boolean(zBase, "8_3_names", 0) )
#endif
  {
    int i, sz;
    sz = sqlite3Strlen30(z);
    for(i=sz-1; i>0 && z[i]!='/' && z[i]!='.'; i--){}
    if( z[i]=='.' && ALWAYS(sz>i+4) ) memmove(&z[i+1], &z[sz-3], 4);
  }
#endif
}

/*
** Move the "*-oal" file corresponding to the target database to the
** "*-wal" location. If an error occurs, leave an error code and error 
** message in the ota handle.
*/
static void otaMoveOalFile(const char *zBase, sqlite3ota *p){
  char *zWal = sqlite3_mprintf("%s-wal", p->zTarget);
  char *zOal = sqlite3_mprintf("%s-oal", p->zTarget);

  assert( p->rc==SQLITE_DONE && p->zErrmsg==0 );
  if( zWal==0 || zOal==0 ){
    p->rc = SQLITE_NOMEM;
  }else{
    otaFileSuffix3(zBase, zWal);
    otaFileSuffix3(zBase, zOal);
    rename(zOal, zWal);
  }

  sqlite3_free(zWal);
  sqlite3_free(zOal);
}

/*
** If there is a "*-oal" file in the file-system corresponding to the
** target database in the file-system, delete it. If an error occurs,
** leave an error code and error message in the ota handle.
*/
static void otaDeleteOalFile(sqlite3ota *p){
  char *zOal = sqlite3_mprintf("%s-oal", p->zTarget);
  assert( p->rc==SQLITE_OK && p->zErrmsg==0 );
  unlink(zOal);
  sqlite3_free(zOal);
}

/*
** Open and return a new OTA handle. 
*/
sqlite3ota *sqlite3ota_open(const char *zTarget, const char *zOta){
  sqlite3ota *p;
  int nTarget = strlen(zTarget);
  int nOta = strlen(zOta);

  p = (sqlite3ota*)sqlite3_malloc(sizeof(sqlite3ota)+nTarget+1);
  p = (sqlite3ota*)sqlite3_malloc(sizeof(sqlite3ota)+nTarget+1+nOta+1);
  if( p ){
    OtaState *pState = 0;

    /* Open the target database */
    memset(p, 0, sizeof(sqlite3ota));
    p->zTarget = (char*)&p[1];
    memcpy(p->zTarget, zTarget, nTarget+1);
    p->rc = sqlite3_open(zTarget, &p->db);
    if( p->rc ){
    p->zOta = &p->zTarget[nTarget+1];
    memcpy(p->zOta, zOta, nOta+1);
      p->zErrmsg = sqlite3_mprintf("%s", sqlite3_errmsg(p->db));
    }
    otaMPrintfExec(p, "ATTACH %Q AS ota", zOta);
    otaOpenDatabase(p);

    /* If it has not already been created, create the ota_state table */
    if( p->rc==SQLITE_OK ){
      p->rc = sqlite3_exec(p->db, OTA_CREATE_STATE, 0, 0, &p->zErrmsg);
    }

    if( p->rc==SQLITE_OK ){
      pState = otaLoadState(p);
      assert( pState || p->rc!=SQLITE_OK );
      if( pState && pState->zTbl==0 ){
        otaDeleteOalFile(p);
      }
    }

    if( p->rc==SQLITE_OK ){
      const char *zScript =
        "PRAGMA journal_mode=off;"
        "PRAGMA pager_ota_mode=1;"
        "PRAGMA ota_mode=1;"
        "BEGIN IMMEDIATE;"
      ;
      p->rc = sqlite3_exec(p->db, zScript, 0, 0, &p->zErrmsg);
    }
      if( pState ){
        if( pState->eStage==0 ){ 
          otaDeleteOalFile(p);
          p->eStage = 1;
        }else{
          p->eStage = pState->eStage;
        }
        p->nProgress = pState->nProgress;
      }
    }
    assert( p->rc!=SQLITE_OK || p->eStage!=0 );

    if( p->eStage==OTA_STAGE_OAL ){
      if( p->rc==SQLITE_OK ){
        const char *zScript =
          "PRAGMA journal_mode=off;"
          "PRAGMA pager_ota_mode=1;"
          "PRAGMA ota_mode=1;"
          "BEGIN IMMEDIATE;"
          ;
        p->rc = sqlite3_exec(p->db, zScript, 0, 0, &p->zErrmsg);
      }

    /* Point the object iterator at the first object */
    if( p->rc==SQLITE_OK ){
      p->rc = otaObjIterFirst(p, &p->objiter);
    }
      /* Point the object iterator at the first object */
      if( p->rc==SQLITE_OK ){
        p->rc = otaObjIterFirst(p, &p->objiter);
      }

    if( p->rc==SQLITE_OK ){
      if( p->rc==SQLITE_OK ){
      p->nProgress = pState->nProgress;
      otaLoadTransactionState(p, pState);
    }

    sqlite3_free(pState);
        otaLoadTransactionState(p, pState);
      }
    }else if( p->rc==SQLITE_OK && p->eStage==OTA_STAGE_CKPT ){
      p->rc = sqlite3_ckpt_open(
          p->db, pState->pCkptState, pState->nCkptState, &p->pCkpt
      );
    }

    otaFreeState(pState);
  }

  return p;
}

/*
** Close the OTA handle.
*/
int sqlite3ota_close(sqlite3ota *p, char **pzErrmsg){
  int rc;
  if( p ){
    const char *zBase = sqlite3_db_filename(p->db, "main");

    /* If the update has not been fully applied, save the state in 
    ** the ota db. If successful, this call also commits the open 
    ** transaction on the ota db. */
    assert( p->rc!=SQLITE_ROW );
    if( p->rc==SQLITE_OK ){
    if( p->rc==SQLITE_OK || p->rc==SQLITE_DONE ){
      assert( p->zErrmsg==0 );
      otaSaveTransactionState(p);
    }

    /* Close all open statement handles. */
    /* Close any open statement handles. */
    otaObjIterFinalize(&p->objiter);

    /* Commit the transaction to the *-oal file. */
    if( p->rc==SQLITE_OK || p->rc==SQLITE_DONE ){
      rc = sqlite3_exec(p->db, "COMMIT", 0, 0, &p->zErrmsg);
      if( rc!=SQLITE_OK ) p->rc = rc;
    if( p->rc==SQLITE_OK && p->eStage==OTA_STAGE_OAL ){
      p->rc = sqlite3_exec(p->db, "COMMIT", 0, 0, &p->zErrmsg);
    }

    if( p->rc==SQLITE_OK && p->eStage==OTA_STAGE_CKPT ){
      p->rc = sqlite3_exec(p->db, "PRAGMA pager_ota_mode=2", 0, 0, &p->zErrmsg);
    }

    /* Close the open database handles */
    /* Close the open database handle */
    if( p->pCkpt ) sqlite3_ckpt_close(p->pCkpt, 0, 0);
    sqlite3_close(p->db);

    /* If the OTA has been completely applied and no error occurred, move
    ** the *-oal file to *-wal. */
    if( p->rc==SQLITE_DONE ){
      otaMoveOalFile(zBase, p);
    }

    rc = p->rc;
    *pzErrmsg = p->zErrmsg;
    sqlite3_free(p);
  }else{
    rc = SQLITE_NOMEM;
    *pzErrmsg = 0;
  }

    sqlite3Error(db, rc);
  }
  rc = sqlite3ApiExit(db, rc);
  sqlite3_mutex_leave(db->mutex);
  return rc;
#endif
}

int sqlite3_ckpt_open(
  sqlite3 *db, 
  unsigned char *a, int n, 
  sqlite3_ckpt **ppCkpt
){
  Pager *pPager;
  Btree *pBt;
  int rc;

  *ppCkpt = 0;
  sqlite3_mutex_enter(db->mutex);
  pBt = db->aDb[0].pBt;
  pPager = sqlite3BtreePager(pBt);
  rc = sqlite3PagerWalCheckpointStart(db, pPager, a, n, ppCkpt);
  sqlite3_mutex_leave(db->mutex);
  return rc;
}


/*
** Checkpoint database zDb. If zDb is NULL, or if the buffer zDb points
** to contains a zero-length string, all attached databases are 
** checkpointed.
*/

** otaMode
**   This variable is normally 0. It is set to 1 by the PagerSetOtaMode()
**   function - as a result of a "PRAGMA pager_ota_mode=1" command. Once 
**   the *-oal file has been opened and it has been determined that the 
**   database file has not been modified since it was created, this variable 
**   is set to 2.
**
** noCkptOnClose
**
**
*/
struct Pager {
  sqlite3_vfs *pVfs;          /* OS functions to use for IO */
  u8 exclusiveMode;           /* Boolean. True if locking_mode==EXCLUSIVE */
  u8 journalMode;             /* One of the PAGER_JOURNALMODE_* values */
  u8 useJournal;              /* Use a rollback journal on this file */

  return sqlite3WalFramesize(pPager->pWal);
}
#endif

/*
** Set or clear the "OTA mode" flag.
*/
int sqlite3PagerSetOtaMode(Pager *pPager, int bOta){
  if( pPager->pWal || pPager->eState!=PAGER_OPEN ){
int sqlite3PagerSetOtaMode(Pager *pPager, int iOta){
  if( iOta==1 && (pPager->pWal || pPager->eState!=PAGER_OPEN) ){
    return SQLITE_ERROR;
  }
  pPager->otaMode = 1;
  pPager->otaMode = iOta;
  return SQLITE_OK;
}

int sqlite3PagerWalCheckpointStart(
  sqlite3 *db, 
  Pager *pPager,
  u8 *a, int n, 
  sqlite3_ckpt **ppCkpt
){
  return sqlite3WalCheckpointStart(db, pPager->pWal, a, n,
      pPager->xBusyHandler, pPager->pBusyHandlerArg,
      pPager->ckptSyncFlags, ppCkpt
  );
}

#endif /* SQLITE_OMIT_DISKIO */

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

int sqlite3PagerSetOtaMode(Pager *pPager, int bOta);
void sqlite3PagerWalSalt(Pager *pPager, u32 *aSalt);
int sqlite3PagerWalCheckpointStart(sqlite3*, Pager*, u8*, int, sqlite3_ckpt**);

#endif /* _PAGER_H_ */

  ** Other clients see a rollback-mode database on which the pager_ota_mode
  ** client is holding a SHARED lock.
  */
  case PragTyp_PAGER_OTA_MODE: {
    Btree *pBt = pDb->pBt;
    assert( pBt!=0 );
    if( zRight ){
      int iArg = !!sqlite3Atoi(zRight);
      int iArg = sqlite3Atoi(zRight);
      if( sqlite3BtreeIsInReadTrans(pBt) ){
        sqlite3ErrorMsg(pParse, 
            "cannot set pager_ota_mode with open transaction"
        );
      }else if( sqlite3PagerSetOtaMode(sqlite3BtreePager(pBt), iArg) ){
        sqlite3ErrorMsg(pParse, "cannot set pager_ota_mode in wal mode");
      }

  int bDelete,                    /* Zero for insert, non-zero for delete */
  const char *zIndex,             /* Index to write to */
  sqlite3_stmt**,                 /* OUT: New statement handle */
  const char ***pazColl,          /* OUT: Collation sequences for each column */
  int **paiCol, int *pnCol        /* OUT: See above */
);

/*
** Incremental checkpoint API.
**
** An incremental checkpoint handle is opened using the sqlite3_ckpt_open()
** API. To begin a new checkpoint, the second and third arguments should both
** be passed zero. To resume an earlier checkpoint, the second and third
** arguments should specify a buffer returned by an earlier call to
** sqlite3_ckpt_close(). When resuming a checkpoint, if the database or WAL 
** file has been modified since the checkpoint was suspended, the 
** sqlite3_ckpt_open() call fails with SQLITE_MISMATCH.
**
** Each time sqlite3_ckpt_step() is called on an open checkpoint handle, a
** single page is copied from the WAL file to the database. If no error 
** occurs, but the checkpoint is not finished, SQLITE_OK is returned. If the
** checkpoint has been finished (and so sqlite3_ckpt_step() should not be
** called again), SQLITE_DONE is returned. Otherwise, if an error occurs,
** some other SQLite error code is returned.
**
** Calling sqlite3_ckpt_close() closes an open checkpoint handle. If the
** checkpoint has finished and no error has occurred, SQLITE_OK is returned
** and the two output parameters zeroed. Or, if an error has occurred, an
** error code is returned and the two output parameters are zeroed. Finally,
** if the checkpoint is not finished but no error has occurred, SQLITE_OK is
** returned and the first output variable set to point to a buffer allocated 
** using sqlite3_malloc() containing the serialized state of the checkpoint. 
** The contents of this buffer may be passed to a later call to
** sqlite3_ckpt_open() to restart the checkpoint. The second output variable 
** is set to the size of the buffer in bytes.
*/
typedef struct sqlite3_ckpt sqlite3_ckpt;
int sqlite3_ckpt_open(sqlite3*, unsigned char *a, int n, sqlite3_ckpt **ppCkpt);
int sqlite3_ckpt_step(sqlite3_ckpt*);
int sqlite3_ckpt_close(sqlite3_ckpt*, unsigned char **pa, int *pn);

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

    ht_slot *aIndex;              /* i0, i1, i2... such that aPgno[iN] ascend */
    u32 *aPgno;                   /* Array of page numbers. */
    int nEntry;                   /* Nr. of entries in aPgno[] and aIndex[] */
    int iZero;                    /* Frame number associated with aPgno[0] */
  } aSegment[1];                  /* One for every 32KB page in the wal-index */
};

/*
** walCheckpoint
*/
typedef struct WalCkpt WalCkpt;
struct WalCkpt {
  sqlite3 *db;                    /* Database pointer (incremental only) */
  int szPage;                     /* Database page-size */
  int sync_flags;                 /* Flags for OsSync() (or 0) */
  u32 mxSafeFrame;                /* Max frame that can be backfilled */
  u32 mxPage;                     /* Max database page to write */
  volatile WalCkptInfo *pInfo;    /* The checkpoint status information */
  WalIterator *pIter;             /* Wal iterator context */
  Wal *pWal;                      /* Pointer to owner object */                 
  u8 *aBuf;                       /* Temporary page-sized buffer to use */
  int rc;                         /* Error code. SQLITE_DONE -> finished */
  int nStep;                      /* Number of times pIter has been stepped */
};

/*
** Define the parameters of the hash tables in the wal-index file. There
** is a hash-table following every HASHTABLE_NPAGE page numbers in the
** wal-index.
**
** Changing any of these constants will alter the wal-index format and
** create incompatibilities.

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

static int walCheckpointStart(
  Wal *pWal, 
  u8 *aBuf,                       /* Page-sized temporary buffer */
  int nBuf,                       /* Size of aBuf[] in bytes */
  int (*xBusy)(void*),            /* Function to call when busy (or NULL) */
  void *pBusyArg,                 /* Context argument for xBusyHandler */
  int sync_flags,                 /* Flags for OsSync() (or 0) */
  WalCkpt *p                      /* Allocated object to populate */
){
  int rc;                         /* Return code */
  int i;                          /* Iterator variable */

  memset(p, 0, sizeof(WalCkpt));

  if( pWal->hdr.mxFrame && walPagesize(pWal)!=nBuf ){
    return SQLITE_CORRUPT_BKPT;
  }

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

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

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

  if( p->pInfo->nBackfill>=p->mxSafeFrame
   || (rc = walBusyLock(pWal, xBusy, pBusyArg, WAL_READ_LOCK(0), 1))!=SQLITE_OK
  ){
    walIteratorFree(p->pIter);
    p->pIter = 0;
  }
  if( rc==SQLITE_BUSY ) rc = SQLITE_OK;

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

    /* If the database may grow as a result of this checkpoint, hint
    ** about the eventual size of the db file to the VFS layer.  */
    if( rc==SQLITE_OK ){
      i64 nSize;                  /* Current size of database file */
      i64 nReq = ((i64)p->mxPage * p->szPage);
      rc = sqlite3OsFileSize(pWal->pDbFd, &nSize);
      if( rc==SQLITE_OK && nSize<nReq ){
        sqlite3OsFileControlHint(pWal->pDbFd, SQLITE_FCNTL_SIZE_HINT, &nReq);
      }
    }
  }

  return rc;
}

static int walCheckpointStep(WalCkpt *p){
  u32 iDbpage = 0;                /* Next database page to write */
  u32 iFrame = 0;                 /* Wal frame containing data for iDbpage */
  int rc = SQLITE_DONE;

  assert( p->rc==SQLITE_OK );
  while( p->pIter && 0==walIteratorNext(p->pIter, &iDbpage, &iFrame) ){
    i64 iOffset;
    assert( walFramePgno(p->pWal, iFrame)==iDbpage );
    p->nStep++;
    if( iFrame<=p->pInfo->nBackfill 
     || iFrame>p->mxSafeFrame 
     || iDbpage>p->mxPage 
    ){
      continue;
    }

    iOffset = walFrameOffset(iFrame, p->szPage) + WAL_FRAME_HDRSIZE;
    /* testcase( IS_BIG_INT(iOffset) ); // requires a 4GiB WAL file */
    rc = sqlite3OsRead(p->pWal->pWalFd, p->aBuf, p->szPage, iOffset);
    if( rc!=SQLITE_OK ) break;
    iOffset = (iDbpage-1)*(i64)p->szPage;
    testcase( IS_BIG_INT(iOffset) );
    rc = sqlite3OsWrite(p->pWal->pDbFd, p->aBuf, p->szPage, iOffset);
    break;
  }

  p->rc = rc;
  return rc;
}

static int walCheckpointFinalize(WalCkpt *p){
  if( p->pIter ){
    int rc = p->rc;
    Wal *pWal = p->pWal;

    /* If work was completed */
    if( p->pIter && rc==SQLITE_DONE ){
      rc = SQLITE_OK;
      if( p->mxSafeFrame==walIndexHdr(pWal)->mxFrame ){
        i64 szDb = pWal->hdr.nPage*(i64)p->szPage;
        testcase( IS_BIG_INT(szDb) );
        rc = sqlite3OsTruncate(pWal->pDbFd, szDb);
        if( rc==SQLITE_OK && p->sync_flags ){
          rc = sqlite3OsSync(pWal->pDbFd, p->sync_flags);
        }
      }
      if( rc==SQLITE_OK ){
        p->pInfo->nBackfill = p->mxSafeFrame;
      }
      p->rc = rc;
    }

    /* Release the reader lock held while backfilling */
    walUnlockExclusive(pWal, WAL_READ_LOCK(0), 1);
    walIteratorFree(p->pIter);
    p->pIter = 0;
  }

  return p->rc;
}

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

*/
static int walCheckpoint(
  Wal *pWal,                      /* Wal connection */
  int eMode,                      /* One of PASSIVE, FULL or RESTART */
  int (*xBusyCall)(void*),        /* Function to call when busy */
  void *pBusyArg,                 /* Context argument for xBusyHandler */
  int sync_flags,                 /* Flags for OsSync() (or 0) */
  u8 *zBuf,                       /* Temporary buffer to use */
  u8 *zBuf                        /* Temporary buffer to use */
  int nBuf                        /* Size of zBuf in bytes */
){
  int rc;                         /* Return code */
  int szPage;                     /* Database page-size */
  WalIterator *pIter = 0;         /* Wal iterator context */
  u32 iDbpage = 0;                /* Next database page to write */
  u32 iFrame = 0;                 /* Wal frame containing data for iDbpage */
  u32 mxSafeFrame;                /* Max frame that can be backfilled */
  u32 mxPage;                     /* Max database page to write */
  int i;                          /* Loop counter */
  volatile WalCkptInfo *pInfo;    /* The checkpoint status information */
  int (*xBusy)(void*) = 0;        /* Function to call when waiting for locks */

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

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

  if( eMode!=SQLITE_CHECKPOINT_PASSIVE ) xBusy = xBusyCall;

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

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

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

    /* If the database may grow as a result of this checkpoint, hint
    ** about the eventual size of the db file to the VFS layer.
    */
    if( rc==SQLITE_OK ){
  assert( rc==SQLITE_OK );
      i64 nReq = ((i64)mxPage * szPage);
      rc = sqlite3OsFileSize(pWal->pDbFd, &nSize);
      if( rc==SQLITE_OK && nSize<nReq ){
        sqlite3OsFileControlHint(pWal->pDbFd, SQLITE_FCNTL_SIZE_HINT, &nReq);
      }

    }

  /* Step the checkpoint object until it reports something other than 

    /* Iterate through the contents of the WAL, copying data to the db file. */
    while( rc==SQLITE_OK && 0==walIteratorNext(pIter, &iDbpage, &iFrame) ){
      i64 iOffset;
      assert( walFramePgno(pWal, iFrame)==iDbpage );
      if( iFrame<=nBackfill || iFrame>mxSafeFrame || iDbpage>mxPage ) continue;
      iOffset = walFrameOffset(iFrame, szPage) + WAL_FRAME_HDRSIZE;
      /* testcase( IS_BIG_INT(iOffset) ); // requires a 4GiB WAL file */
      rc = sqlite3OsRead(pWal->pWalFd, zBuf, szPage, iOffset);
      if( rc!=SQLITE_OK ) break;
  ** SQLITE_OK.  */
      iOffset = (iDbpage-1)*(i64)szPage;
      testcase( IS_BIG_INT(iOffset) );
      rc = sqlite3OsWrite(pWal->pDbFd, zBuf, szPage, iOffset);
      if( rc!=SQLITE_OK ) break;
    }

  while( SQLITE_OK==(rc = walCheckpointStep(&sC)) );
    /* If work was actually accomplished... */
    if( rc==SQLITE_OK ){
  if( rc==SQLITE_DONE ) rc = SQLITE_OK;
      if( mxSafeFrame==walIndexHdr(pWal)->mxFrame ){
        i64 szDb = pWal->hdr.nPage*(i64)szPage;
        testcase( IS_BIG_INT(szDb) );
        rc = sqlite3OsTruncate(pWal->pDbFd, szDb);
        if( rc==SQLITE_OK && sync_flags ){
          rc = sqlite3OsSync(pWal->pDbFd, sync_flags);
        }
      }
      if( rc==SQLITE_OK ){
        pInfo->nBackfill = mxSafeFrame;
      }
    }

  rc = walCheckpointFinalize(&sC);
    /* Release the reader lock held while backfilling */
    walUnlockExclusive(pWal, WAL_READ_LOCK(0), 1);
  }

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

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

 walcheckpoint_out:
  walIteratorFree(pIter);
  walIteratorFree(sC.pIter);
  return rc;
}

/*
** If the WAL file is currently larger than nMax bytes in size, truncate
** it to exactly nMax bytes. If an error occurs while doing so, ignore it.
*/

    if( isChanged && pWal->pDbFd->pMethods->iVersion>=3 ){
      sqlite3OsUnfetch(pWal->pDbFd, 0, 0);
    }
  }

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

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

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

int sqlite3_ckpt_step(sqlite3_ckpt *pCkpt){
  int rc;
  WalCkpt *p = (WalCkpt*)pCkpt;
  sqlite3_mutex_enter(p->db->mutex);
  rc = walCheckpointStep(p);
  sqlite3_mutex_leave(p->db->mutex);
  return rc;
}

int sqlite3_ckpt_close(sqlite3_ckpt *pCkpt, u8 **paState, int *pnState){
  int rc;
  WalCkpt *p = (WalCkpt*)pCkpt;
  sqlite3 *db = p->db;
  Wal *pWal = p->pWal;
  sqlite3_mutex_enter(db->mutex);
  if( paState ){
    *paState = 0;
    *pnState = 0;
    if( p->rc==SQLITE_OK ){
      u8 *aState = sqlite3_malloc(sizeof(u32) * 3);
      if( aState==0 ){
        p->rc = SQLITE_NOMEM;
      }else{
        *pnState = sizeof(u32)*3;
        sqlite3Put4byte(&aState[0], p->nStep);
        sqlite3Put4byte(&aState[4], p->pWal->hdr.aCksum[0]);
        sqlite3Put4byte(&aState[8], p->pWal->hdr.aCksum[1]);
        *paState = aState;
      }
    }
  }
  rc = walCheckpointFinalize(p);
  walUnlockExclusive(pWal, WAL_CKPT_LOCK, 1);
  pWal->ckptLock = 0;
  sqlite3_free(p);
  memset(&pWal->hdr, 0, sizeof(WalIndexHdr));
  sqlite3_mutex_leave(db->mutex);
  return rc;
}

int sqlite3WalCheckpointStart(
  sqlite3 *db,                    /* Database connection */
  Wal *pWal,                      /* Wal connection */
  u8 *aState, int nState,         /* Checkpoint state to restore */
  int (*xBusy)(void*),            /* Function to call when busy */
  void *pBusyArg,                 /* Context argument for xBusyHandler */
  int sync_flags,                 /* Flags to sync db file with (or 0) */
  sqlite3_ckpt **ppCkpt           /* OUT: Incremental checkpoint object */
){
  WalCkpt *p = 0;
  int isChanged = 0;
  int rc;
  int pgsz;

  *ppCkpt = 0;
  if( pWal->readOnly ) return SQLITE_READONLY;
  WALTRACE(("WAL%p: checkpoint begins\n", pWal));
  rc = walLockExclusive(pWal, WAL_CKPT_LOCK, 1);
  if( rc ){
    /* Usually this is SQLITE_BUSY meaning that another thread or process
    ** is already running a checkpoint, or maybe a recovery.  But it might
    ** also be SQLITE_IOERR. */
    return rc;
  }
  pWal->ckptLock = 1;

  /* Read the wal-index header. */
  rc = walIndexReadHdr(pWal, &isChanged);
  if( rc!=SQLITE_OK ) goto ckptstart_out;
  if( isChanged && pWal->pDbFd->pMethods->iVersion>=3 ){
    sqlite3OsUnfetch(pWal->pDbFd, 0, 0);
  }

  pgsz = walPagesize(pWal);
  p = sqlite3_malloc(sizeof(WalCkpt) + pgsz);
  if( p==0 ){
    rc = SQLITE_NOMEM;
    goto ckptstart_out;
  }

  rc = walCheckpointStart(
      pWal, (u8*)&p[1], pgsz, xBusy, pBusyArg, sync_flags, p
  );
  p->db = db;

  if( rc==SQLITE_OK && aState ){
    if( nState!=sizeof(u32)*3 ){
      rc = SQLITE_CORRUPT_BKPT;
    }else{
      int i;
      if( pWal->hdr.aCksum[0]!=sqlite3Get4byte(&aState[4])
       || pWal->hdr.aCksum[1]!=sqlite3Get4byte(&aState[8])
      ){
        rc = SQLITE_MISMATCH;
      }else{
        p->nStep = (int)sqlite3Get4byte(aState);
        sqlite3Put4byte(&aState[4], pWal->hdr.aCksum[0]);
        sqlite3Put4byte(&aState[8], pWal->hdr.aCksum[1]);
        for(i=0; rc==SQLITE_OK && i<p->nStep; i++){
          u32 dummy1, dummy2; 
          rc = walIteratorNext(p->pIter, &dummy1, &dummy2);
        }
      }
    }
  }

 ckptstart_out:
  if( rc!=SQLITE_OK ){
    if( p ) walIteratorFree(p->pIter);
    walUnlockExclusive(pWal, WAL_CKPT_LOCK, 1);
    pWal->ckptLock = 0;
    sqlite3_free(p);
    p = 0;
  }
  *ppCkpt = (sqlite3_ckpt*)p;
  return rc;
}

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

/* Return true if the argument is non-NULL and the WAL module is using
** heap-memory for the wal-index. Otherwise, if the argument is NULL or the
** WAL module is using shared-memory, return false. 
*/
int sqlite3WalHeapMemory(Wal *pWal);

int sqlite3WalCheckSalt(Wal *pWal, sqlite3_file*);

int sqlite3WalCheckpointStart(sqlite3 *,
  Wal *pWal,                      /* Wal connection */
  u8 *aState, int nState,         /* Checkpoint state to restore */
  int (*xBusy)(void*),            /* Function to call when busy */
  void *pBusyArg,                 /* Context argument for xBusyHandler */
  int sync_flags,                 /* Flags to sync db file with (or 0) */
  sqlite3_ckpt **ppCkpt           /* OUT: Incremental checkpoint object */
);

#ifdef SQLITE_ENABLE_ZIPVFS
/* If the WAL file is not empty, return the number of bytes of content
** stored in each frame (i.e. the db page-size when the WAL was created).
*/
int sqlite3WalFramesize(Wal *pWal);
#endif

  list [file size test.db] [file size test.db-wal]
} [list 2048 [wal_file_size 3 1024]]

# Execute some transactions in auto-vacuum mode to test database file
# truncation.
#
do_test wal-8.1 {
breakpoint
  reopen_db
  catch { db close }
  forcedelete test.db test.db-wal

  sqlite3 db test.db
  db function blob blob
  execsql {