SQLite

# 2014 October 22
#
# 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.
#
#***********************************************************************
#

if {![info exists testdir]} {
  set testdir [file join [file dirname [info script]] .. .. test]
}
source $testdir/tester.tcl
set ::testprefix otacrash

# Set up a target database and an ota update database. The target
# db is the usual "test.db", the ota db is "test.db2".
#
forcedelete test.db2
do_execsql_test 1.0 {
  CREATE TABLE t1(a, b, c, PRIMARY KEY(a), UNIQUE(b));
  INSERT INTO t1 VALUES(1, 2, 3);
  INSERT INTO t1 VALUES(4, 5, 6);
  INSERT INTO t1 VALUES(7, 8, 9);

  ATTACH 'test.db2' AS ota;
  CREATE TABLE ota.data_t1(a, b, c, ota_control);
  INSERT INTO data_t1 VALUES(10, 11, 12, 0);
  INSERT INTO data_t1 VALUES(13, 14, 15, 0);
  INSERT INTO data_t1 VALUES(4, NULL, NULL, 1);
  INSERT INTO data_t1 VALUES(1, NULL, 100, '..x');
}
db_save_and_close


for {set i 0} {$i < 10} {incr i} {
  forcedelete test.db2 test.db2-journal test.db test.db-oal test.db-wal
  db_restore
  do_test 2.$i.1 {
    crashsql -file test.db2 -delay 3 -tclbody {

      sqlite3ota ota test.db file:test.db2?vfs=crash
      ota step
      ota close

      sqlite3ota ota test.db file:test.db2?vfs=crash
      ota step
      ota step
      ota close
    } {}
  } {1 {child process exited abnormally}}

  do_test 2.$i.2 {
    sqlite3ota ota test.db test.db2
    while {[ota step]=="SQLITE_OK"} {}
    ota close
  } {SQLITE_DONE}

  sqlite3 db test.db
  do_execsql_test 2.$i.3 {
    SELECT * FROM t1;
  } {1 2 100  7 8 9  10 11 12  13 14 15}
  do_execsql_test 2.$i.4 { PRAGMA integrity_check } {ok}
  db close
}

finish_test

**   cases when another client appends frames to the *-wal file in the
**   middle of an incremental checkpoint (an incremental checkpoint cannot
**   be continued if this happens).
**
** OTA_STATE_COOKIE:
**   Valid if STAGE==1. The current change-counter cookie value in the 
**   target db file.
**
** OTA_STATE_OALSZ:
**   Valid if STAGE==1. The size in bytes of the *-oal file.
*/
#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_STATE_COOKIE      7
#define OTA_STATE_OALSZ       8

#define OTA_STAGE_OAL         1
#define OTA_STAGE_MOVE        2
#define OTA_STAGE_CAPTURE     3
#define OTA_STAGE_CKPT        4
#define OTA_STAGE_DONE        5


#define OTA_CREATE_STATE "CREATE TABLE IF NOT EXISTS ota_state"        \
                             "(k INTEGER PRIMARY KEY, v)"

typedef struct OtaFrame OtaFrame;
typedef struct OtaObjIter OtaObjIter;
typedef struct OtaState OtaState;
typedef struct ota_vfs ota_vfs;
typedef struct ota_file ota_file;

  int eStage;
  char *zTbl;
  char *zIdx;
  i64 iWalCksum;
  int nRow;
  i64 nProgress;
  u32 iCookie;
  i64 iOalSz;
};

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

  int bUnique;                    /* Current index is unique */

  /* Statements created by otaObjIterPrepareAll() */
  int nCol;                       /* Number of columns in current object */
  sqlite3_stmt *pSelect;          /* Source data */
  sqlite3_stmt *pInsert;          /* Statement for INSERT operations */
  sqlite3_stmt *pDelete;          /* Statement for DELETE ops */
  sqlite3_stmt *pTmpInsert;       /* Insert into ota_tmp_$zTbl */

  /* Last UPDATE used (for PK b-tree updates only), or NULL. */
  char *zMask;                    /* Copy of update mask used with pUpdate */
  sqlite3_stmt *pUpdate;          /* Last update statement (or NULL) */
};

/*

};

/*
** OTA handle.
*/
struct sqlite3ota {
  int eStage;                     /* Value of OTA_STATE_STAGE field */
  sqlite3 *dbMain;                /* target database handle */
  sqlite3 *dbOta;                 /* ota database handle */
  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;             /* Iterator for skipping through tbl/idx */
  const char *zVfsName;           /* Name of automatically created ota vfs */
  ota_file *pTargetFd;            /* File handle open on target db */
  i64 iOalSz;

  /* The following state variables are used as part of the incremental
  ** checkpoint stage (eStage==OTA_STAGE_CKPT). See comments surrounding
  ** function otaSetupCheckpoint() for details.  */
  u32 iMaxFrame;                  /* Largest iWalFrame value in aFrame[] */
  u32 mLock;
  int nFrame;                     /* Entries in aFrame[] array */

** the current object (table/index pair).
*/
static void otaObjIterClearStatements(OtaObjIter *pIter){
  sqlite3_finalize(pIter->pSelect);
  sqlite3_finalize(pIter->pInsert);
  sqlite3_finalize(pIter->pDelete);
  sqlite3_finalize(pIter->pUpdate);
  sqlite3_finalize(pIter->pTmpInsert);
  pIter->pSelect = 0;
  pIter->pInsert = 0;
  pIter->pDelete = 0;
  pIter->pUpdate = 0;
  pIter->pTmpInsert = 0;
  pIter->nCol = 0;
}

/*
** Clean up any resources allocated as part of the iterator object passed
** as the only argument.
*/

static int otaObjIterNext(sqlite3ota *p, OtaObjIter *pIter){
  int rc = p->rc;
  if( rc==SQLITE_OK ){

    /* Free any SQLite statements used while processing the previous object */ 
    otaObjIterClearStatements(pIter);
    if( pIter->zIdx==0 ){
      rc = sqlite3_exec(p->dbMain,
          "DROP TRIGGER IF EXISTS temp.ota_insert_tr;"
          "DROP TRIGGER IF EXISTS temp.ota_update1_tr;"
          "DROP TRIGGER IF EXISTS temp.ota_update2_tr;"
          "DROP TRIGGER IF EXISTS temp.ota_delete_tr;"
          , 0, 0, &p->zErrmsg
      );
    }

    if( rc==SQLITE_OK ){
      if( pIter->bCleanup ){
        otaObjIterFreeCols(pIter);
        pIter->bCleanup = 0;
        rc = sqlite3_step(pIter->pTblIter);
        if( rc!=SQLITE_ROW ){
          rc = resetAndCollectError(pIter->pTblIter, &p->zErrmsg);
          pIter->zTbl = 0;
        }else{
          pIter->zTbl = (const char*)sqlite3_column_text(pIter->pTblIter, 0);

          rc = pIter->zTbl ? SQLITE_OK : SQLITE_NOMEM;
        }
      }else{
        if( pIter->zIdx==0 ){
          sqlite3_stmt *pIdx = pIter->pIdxIter;
          rc = sqlite3_bind_text(pIdx, 1, pIter->zTbl, -1, SQLITE_STATIC);
        }

** left in the OTA handle passed as the first argument. A copy of the 
** error code is returned.
*/
static int otaObjIterFirst(sqlite3ota *p, OtaObjIter *pIter){
  int rc;
  memset(pIter, 0, sizeof(OtaObjIter));

  rc = prepareAndCollectError(p->dbOta, &pIter->pTblIter, &p->zErrmsg, 

      "SELECT substr(name, 6) FROM sqlite_master "

      "WHERE type='table' AND name LIKE 'data_%'"
  );

  if( rc==SQLITE_OK ){
    rc = prepareAndCollectError(p->dbMain, &pIter->pIdxIter, &p->zErrmsg,
        "SELECT name, rootpage, sql IS NULL OR substr(8, 6)=='UNIQUE' "
        "  FROM main.sqlite_master "
        "  WHERE type='index' AND tbl_name = ?"
    );
  }

  pIter->bCleanup = 1;

** the printf() style substitutions and executes the result as an SQL
** statement on the OTA handles database.
**
** If an error occurs, an error code and error message is stored in the
** OTA handle. If an error has already occurred when this function is
** called, it is a no-op.
*/
static int otaMPrintfExec(sqlite3ota *p, sqlite3 *db, const char *zFmt, ...){
  va_list ap;
  va_start(ap, zFmt);
  char *zSql = sqlite3_vmprintf(zFmt, ap);
  if( p->rc==SQLITE_OK ){
    if( zSql==0 ){
      p->rc = SQLITE_NOMEM;
    }else{
      p->rc = sqlite3_exec(db, zSql, 0, 0, &p->zErrmsg);
    }
  }
  sqlite3_free(zSql);
  va_end(ap);
  return p->rc;
}

**     return OTA_PK_NONE
**   }
*/
static void otaTableType(
  sqlite3ota *p,
  const char *zTab,
  int *peType,
  int *piTnum,
  int *piPk
){
  /*
  ** 0) SELECT count(*) FROM sqlite_master where name=%Q AND IsVirtual(%Q)
  ** 1) PRAGMA index_list = ?
  ** 2) SELECT count(*) FROM sqlite_master where name=%Q 
  ** 3) PRAGMA table_info = ?
  */
  sqlite3_stmt *aStmt[4] = {0, 0, 0, 0};

  *peType = OTA_PK_NOTABLE;
  *piPk = 0;

  assert( p->rc==SQLITE_OK );
  p->rc = prepareFreeAndCollectError(p->dbMain, &aStmt[0], &p->zErrmsg, 
    sqlite3_mprintf(
          "SELECT (sql LIKE 'create virtual%%'), rootpage"
          "  FROM sqlite_master"
          " WHERE name=%Q", zTab
  ));
  if( p->rc!=SQLITE_OK || sqlite3_step(aStmt[0])!=SQLITE_ROW ){
    /* Either an error, or no such table. */
    goto otaTableType_end;
  }
  if( sqlite3_column_int(aStmt[0], 0) ){
    *peType = OTA_PK_VTAB;                     /* virtual table */
    goto otaTableType_end;
  }
  *piTnum = sqlite3_column_int(aStmt[0], 1);

  p->rc = prepareFreeAndCollectError(p->dbMain, &aStmt[1], &p->zErrmsg, 
    sqlite3_mprintf("PRAGMA index_list=%Q",zTab)
  );
  if( p->rc ) goto otaTableType_end;
  while( sqlite3_step(aStmt[1])==SQLITE_ROW ){
    const u8 *zOrig = sqlite3_column_text(aStmt[1], 3);
    const u8 *zIdx = sqlite3_column_text(aStmt[1], 1);
    if( zOrig && zIdx && zOrig[0]=='p' ){
      p->rc = prepareFreeAndCollectError(p->dbMain, &aStmt[2], &p->zErrmsg, 
          sqlite3_mprintf(
            "SELECT rootpage FROM sqlite_master WHERE name = %Q", zIdx
      ));
      if( p->rc==SQLITE_OK ){
        if( sqlite3_step(aStmt[2])==SQLITE_ROW ){
          *piPk = sqlite3_column_int(aStmt[2], 0);
          *peType = OTA_PK_EXTERNAL;
        }else{
          *peType = OTA_PK_WITHOUT_ROWID;
        }
      }
      goto otaTableType_end;
    }
  }

  p->rc = prepareFreeAndCollectError(p->dbMain, &aStmt[3], &p->zErrmsg, 
    sqlite3_mprintf("PRAGMA table_info=%Q",zTab)
  );
  if( p->rc==SQLITE_OK ){
    while( sqlite3_step(aStmt[3])==SQLITE_ROW ){
      if( sqlite3_column_int(aStmt[3],5)>0 ){
        *peType = OTA_PK_IPK;                /* explicit IPK column */
        goto otaTableType_end;

static int otaObjIterCacheTableInfo(sqlite3ota *p, OtaObjIter *pIter){
  if( pIter->azTblCol==0 ){
    sqlite3_stmt *pStmt = 0;
    int nCol = 0;
    int i;                        /* for() loop iterator variable */
    int bOtaRowid = 0;            /* If input table has column "ota_rowid" */
    int iOrder = 0;
    int iTnum = 0;

    /* Figure out the type of table this step will deal with. */
    assert( pIter->eType==0 );
    otaTableType(p, pIter->zTbl, &pIter->eType, &iTnum, &pIter->iPkTnum);
    if( p->rc ) return p->rc;
    if( pIter->zIdx==0 ) pIter->iTnum = iTnum;

    assert( pIter->eType==OTA_PK_NONE || pIter->eType==OTA_PK_IPK 
         || pIter->eType==OTA_PK_EXTERNAL || pIter->eType==OTA_PK_WITHOUT_ROWID
         || pIter->eType==OTA_PK_VTAB
    );

    /* Populate the azTblCol[] and nTblCol variables based on the columns
    ** of the input table. Ignore any input table columns that begin with
    ** "ota_".  */
    p->rc = prepareFreeAndCollectError(p->dbOta, &pStmt, &p->zErrmsg, 
        sqlite3_mprintf("SELECT * FROM 'data_%q'", pIter->zTbl)
    );
    if( p->rc==SQLITE_OK ){
      nCol = sqlite3_column_count(pStmt);
      otaAllocateIterArrays(p, pIter, nCol);
    }
    for(i=0; p->rc==SQLITE_OK && i<nCol; i++){

      );
    }

    /* Check that all non-HIDDEN columns in the destination table are also
    ** present in the input table. Populate the abTblPk[], azTblType[] and
    ** aiTblOrder[] arrays at the same time.  */
    if( p->rc==SQLITE_OK ){
      p->rc = prepareFreeAndCollectError(p->dbMain, &pStmt, &p->zErrmsg, 
          sqlite3_mprintf("PRAGMA table_info(%Q)", pIter->zTbl)
      );
    }
    while( p->rc==SQLITE_OK && SQLITE_ROW==sqlite3_step(pStmt) ){
      const char *zName = (const char*)sqlite3_column_text(pStmt, 1);
      if( zName==0 ) break;  /* An OOM - finalize() below returns S_NOMEM */
      for(i=iOrder; i<pIter->nTblCol; i++){
        if( 0==strcmp(zName, pIter->azTblCol[i]) ) break;

  int nBind = 0;                  /* Value to return via *pnBind */
  const char *zCom = "";          /* Set to ", " later on */
  const char *zAnd = "";          /* Set to " AND " later on */
  sqlite3_stmt *pXInfo = 0;       /* PRAGMA index_xinfo = ? */

  if( rc==SQLITE_OK ){
    assert( p->zErrmsg==0 );
    rc = prepareFreeAndCollectError(p->dbMain, &pXInfo, &p->zErrmsg,
        sqlite3_mprintf("PRAGMA main.index_xinfo = %Q", pIter->zIdx)
    );
  }

  while( rc==SQLITE_OK && SQLITE_ROW==sqlite3_step(pXInfo) ){
    int iCid = sqlite3_column_int(pXInfo, 1);
    int bDesc = sqlite3_column_int(pXInfo, 3);

  char *z = 0;
  assert( pIter->zIdx==0 );
  if( p->rc==SQLITE_OK ){
    const char *zSep = "PRIMARY KEY(";
    sqlite3_stmt *pXList = 0;     /* PRAGMA index_list = (pIter->zTbl) */
    sqlite3_stmt *pXInfo = 0;     /* PRAGMA index_xinfo = <pk-index> */
   
    p->rc = prepareFreeAndCollectError(p->dbMain, &pXList, &p->zErrmsg,
        sqlite3_mprintf("PRAGMA main.index_list = %Q", pIter->zTbl)
    );
    while( p->rc==SQLITE_OK && SQLITE_ROW==sqlite3_step(pXList) ){
      const char *zOrig = (const char*)sqlite3_column_text(pXList,3);
      if( zOrig && strcmp(zOrig, "pk")==0 ){
        const char *zIdx = (const char*)sqlite3_column_text(pXList,1);
        if( zIdx ){
          p->rc = prepareFreeAndCollectError(p->dbMain, &pXInfo, &p->zErrmsg,
              sqlite3_mprintf("PRAGMA main.index_xinfo = %Q", zIdx)
          );
        }
        break;
      }
    }
    otaFinalize(p, pXList);

    const char *zComma = "";
    char *zCols = 0;              /* Used to build up list of table cols */
    char *zPk = 0;                /* Used to build up table PK declaration */

    /* Figure out the name of the primary key index for the current table.
    ** This is needed for the argument to "PRAGMA index_xinfo". Set
    ** zIdx to point to a nul-terminated string containing this name. */
    p->rc = prepareAndCollectError(p->dbMain, &pQuery, &p->zErrmsg, 
        "SELECT name FROM sqlite_master WHERE rootpage = ?"
    );
    if( p->rc==SQLITE_OK ){
      sqlite3_bind_int(pQuery, 1, tnum);
      if( SQLITE_ROW==sqlite3_step(pQuery) ){
        zIdx = (const char*)sqlite3_column_text(pQuery, 0);
      }
    }
    if( zIdx ){
      p->rc = prepareFreeAndCollectError(p->dbMain, &pXInfo, &p->zErrmsg,
          sqlite3_mprintf("PRAGMA main.index_xinfo = %Q", zIdx)
      );
    }
    otaFinalize(p, pQuery);

    while( p->rc==SQLITE_OK && SQLITE_ROW==sqlite3_step(pXInfo) ){
      int bKey = sqlite3_column_int(pXInfo, 5);

        zPk = otaMPrintf(p, "%z%sc%d%s", zPk, zComma, iCid, bDesc?" DESC":"");
        zComma = ", ";
      }
    }
    zCols = otaMPrintf(p, "%z, id INTEGER", zCols);
    otaFinalize(p, pXInfo);

    sqlite3_test_control(SQLITE_TESTCTRL_IMPOSTER, p->dbMain, "main", 1, tnum);
    otaMPrintfExec(p, p->dbMain,
        "CREATE TABLE ota_imposter2(%z, PRIMARY KEY(%z)) WITHOUT ROWID", 
        zCols, zPk
    );
    sqlite3_test_control(SQLITE_TESTCTRL_IMPOSTER, p->dbMain, "main", 0, 0);
  }
}

/*
** If an error has already occurred when this function is called, it 
** immediately returns zero (without doing any work). Or, if an error
** occurs during the execution of this function, it sets the error code

*/
static void otaCreateImposterTable(sqlite3ota *p, OtaObjIter *pIter){
  if( p->rc==SQLITE_OK && pIter->eType!=OTA_PK_VTAB ){
    int tnum = pIter->iTnum;
    const char *zComma = "";
    char *zSql = 0;
    int iCol;
    sqlite3_test_control(SQLITE_TESTCTRL_IMPOSTER, p->dbMain, "main", 0, 1);

    for(iCol=0; p->rc==SQLITE_OK && iCol<pIter->nTblCol; iCol++){
      const char *zPk = "";
      const char *zCol = pIter->azTblCol[iCol];
      const char *zColl = 0;

      p->rc = sqlite3_table_column_metadata(
          p->dbMain, "main", pIter->zTbl, zCol, 0, &zColl, 0, 0, 0
      );

      if( pIter->eType==OTA_PK_IPK && pIter->abTblPk[iCol] ){
        /* If the target table column is an "INTEGER PRIMARY KEY", add
        ** "PRIMARY KEY" to the imposter table column declaration. */
        zPk = "PRIMARY KEY ";
      }

    if( pIter->eType==OTA_PK_WITHOUT_ROWID ){
      char *zPk = otaWithoutRowidPK(p, pIter);
      if( zPk ){
        zSql = otaMPrintf(p, "%z, %z", zSql, zPk);
      }
    }

    sqlite3_test_control(SQLITE_TESTCTRL_IMPOSTER, p->dbMain, "main", 1, tnum);
    otaMPrintfExec(p, p->dbMain, "CREATE TABLE \"ota_imp_%w\"(%z)%s", 
        pIter->zTbl, zSql, 
        (pIter->eType==OTA_PK_WITHOUT_ROWID ? " WITHOUT ROWID" : "")
    );
    sqlite3_test_control(SQLITE_TESTCTRL_IMPOSTER, p->dbMain, "main", 0, 0);
  }
}

/*
** Prepare a statement used to insert rows into the "ota_tmp_xxx" table.
** Specifically a statement of the form:
**
**     INSERT INTO ota_tmp_xxx VALUES(?, ?, ? ...);
**
** The number of bound variables is equal to the number of columns in
** the target table, plus one (for the ota_control column), plus one more 
** (for the ota_rowid column) if the target table is an implicit IPK or 
** virtual table.
*/
static void otaObjIterPrepareTmpInsert(
  sqlite3ota *p, 
  OtaObjIter *pIter,
  const char *zCollist,
  const char *zOtaRowid
){
  int bOtaRowid = (pIter->eType==OTA_PK_EXTERNAL || pIter->eType==OTA_PK_NONE);
  char *zBind = otaObjIterGetBindlist(p, pIter->nTblCol + 1 + bOtaRowid);
  if( zBind ){
    assert( pIter->pTmpInsert==0 );
    p->rc = prepareFreeAndCollectError(
        p->dbOta, &pIter->pTmpInsert, &p->zErrmsg, sqlite3_mprintf(
          "INSERT INTO 'ota_tmp_%q'(ota_control,%s%s) VALUES(%z)", 
          pIter->zTbl, zCollist, zOtaRowid, zBind
    ));
  }
}

static void otaTmpInsertFunc(
  sqlite3_context *pCtx, 
  int nVal,
  sqlite3_value **apVal
){
  sqlite3ota *p = sqlite3_user_data(pCtx);
  int rc = SQLITE_OK;
  int i;

  for(i=0; rc==SQLITE_OK && i<nVal; i++){
    rc = sqlite3_bind_value(p->objiter.pTmpInsert, i+1, apVal[i]);
  }
  if( rc==SQLITE_OK ){
    sqlite3_step(p->objiter.pTmpInsert);
    rc = sqlite3_reset(p->objiter.pTmpInsert);
  }

  if( rc!=SQLITE_OK ){
    sqlite3_result_error_code(pCtx, rc);
  }
}

/*
** Ensure that the SQLite statement handles required to update the 
** target database object currently indicated by the iterator passed 
** as the second argument are available.

      assert( pIter->eType!=OTA_PK_VTAB );
      zCollist = otaObjIterGetIndexCols(
          p, pIter, &zImposterCols, &zImposterPK, &zWhere, &nBind
      );
      zBind = otaObjIterGetBindlist(p, nBind);

      /* Create the imposter table used to write to this index. */
      sqlite3_test_control(SQLITE_TESTCTRL_IMPOSTER, p->dbMain, "main", 0, 1);
      sqlite3_test_control(SQLITE_TESTCTRL_IMPOSTER, p->dbMain, "main", 1,tnum);
      otaMPrintfExec(p, p->dbMain,
          "CREATE TABLE \"ota_imp_%w\"( %s, PRIMARY KEY( %s ) ) WITHOUT ROWID",
          zTbl, zImposterCols, zImposterPK
      );
      sqlite3_test_control(SQLITE_TESTCTRL_IMPOSTER, p->dbMain, "main", 0, 0);

      /* Create the statement to insert index entries */
      pIter->nCol = nBind;
      if( p->rc==SQLITE_OK ){
        p->rc = prepareFreeAndCollectError(
            p->dbMain, &pIter->pInsert, &p->zErrmsg,
          sqlite3_mprintf("INSERT INTO \"ota_imp_%w\" VALUES(%s)", zTbl, zBind)
        );
      }

      /* And to delete index entries */
      if( p->rc==SQLITE_OK ){
        p->rc = prepareFreeAndCollectError(
            p->dbMain, &pIter->pDelete, &p->zErrmsg,
          sqlite3_mprintf("DELETE FROM \"ota_imp_%w\" WHERE %s", zTbl, zWhere)
        );
      }

      /* Create the SELECT statement to read keys in sorted order */
      if( p->rc==SQLITE_OK ){
        char *zSql;
        if( pIter->eType==OTA_PK_EXTERNAL || pIter->eType==OTA_PK_NONE ){
          zSql = sqlite3_mprintf(
              "SELECT %s, ota_control FROM 'ota_tmp_%q' ORDER BY %s%s",
              zCollist, pIter->zTbl,
              zCollist, zLimit
          );
        }else{
          zSql = sqlite3_mprintf(
              "SELECT %s, ota_control FROM 'data_%q' "
              "WHERE typeof(ota_control)='integer' AND ota_control!=1 "
              "UNION ALL "
              "SELECT %s, ota_control FROM 'ota_tmp_%q' "
              "ORDER BY %s%s",
              zCollist, pIter->zTbl, 
              zCollist, pIter->zTbl, 
              zCollist, zLimit
          );
        }
        p->rc = prepareFreeAndCollectError(p->dbOta, &pIter->pSelect, pz, zSql);
      }

      sqlite3_free(zImposterCols);
      sqlite3_free(zImposterPK);
      sqlite3_free(zWhere);
      sqlite3_free(zBind);
    }else{

      char *zNewlist = otaObjIterGetOldlist(p, pIter, "new");

      zCollist = otaObjIterGetCollist(p, pIter);
      pIter->nCol = pIter->nTblCol;

      /* Create the SELECT statement to read keys from data_xxx */
      if( p->rc==SQLITE_OK ){
        p->rc = prepareFreeAndCollectError(p->dbOta, &pIter->pSelect, pz,
            sqlite3_mprintf(
              "SELECT %s, ota_control%s FROM 'data_%q'%s", 
              zCollist, (bOtaRowid ? ", ota_rowid" : ""), zTbl, zLimit
            )
        );
      }

      /* Create the imposter table or tables (if required). */
      otaCreateImposterTable(p, pIter);
      otaCreateImposterTable2(p, pIter);
      zWrite = (pIter->eType==OTA_PK_VTAB ? "" : "ota_imp_");

      /* Create the INSERT statement to write to the target PK b-tree */
      if( p->rc==SQLITE_OK ){
        p->rc = prepareFreeAndCollectError(p->dbMain, &pIter->pInsert, pz,
            sqlite3_mprintf(
              "INSERT INTO \"%s%w\"(%s%s) VALUES(%s)", 
              zWrite, zTbl, zCollist, (bOtaRowid ? ", _rowid_" : ""), zBindings
            )
        );
      }

      /* Create the DELETE statement to write to the target PK b-tree */
      if( p->rc==SQLITE_OK ){
        p->rc = prepareFreeAndCollectError(p->dbMain, &pIter->pDelete, pz,
            sqlite3_mprintf(
              "DELETE FROM \"%s%w\" WHERE %s", zWrite, zTbl, zWhere
            )
        );
      }

      if( pIter->eType!=OTA_PK_VTAB ){
        const char *zOtaRowid = "";
        if( pIter->eType==OTA_PK_EXTERNAL || pIter->eType==OTA_PK_NONE ){
          zOtaRowid = ", ota_rowid";
        }

        /* Create the ota_tmp_xxx table and the triggers to populate it. */
        otaMPrintfExec(p, p->dbOta,
            "CREATE TABLE IF NOT EXISTS 'ota_tmp_%q' AS "
            "SELECT *%s FROM 'data_%q' WHERE 0;"
            , zTbl, (pIter->eType==OTA_PK_EXTERNAL ? ", 0 AS ota_rowid" : "")
            , zTbl
        );

        otaMPrintfExec(p, p->dbMain,
            "CREATE TEMP TRIGGER ota_delete_tr BEFORE DELETE ON \"%s%w\" "
            "BEGIN "
            "  SELECT ota_tmp_insert(2, %s);"
            "END;"

            "CREATE TEMP TRIGGER ota_update1_tr BEFORE UPDATE ON \"%s%w\" "
            "BEGIN "
            "  SELECT ota_tmp_insert(2, %s);"
            "END;"

            "CREATE TEMP TRIGGER ota_update2_tr AFTER UPDATE ON \"%s%w\" "
            "BEGIN "
            "  SELECT ota_tmp_insert(3, %s);"
            "END;",


            zWrite, zTbl, zOldlist,
            zWrite, zTbl, zOldlist,
            zWrite, zTbl, zNewlist
        );

        if( pIter->eType==OTA_PK_EXTERNAL || pIter->eType==OTA_PK_NONE ){
          otaMPrintfExec(p, p->dbMain,
              "CREATE TEMP TRIGGER ota_insert_tr AFTER INSERT ON \"%s%w\" "
              "BEGIN "

              "  SELECT ota_tmp_insert(0, %s);"
              "END;",
              zWrite, zTbl, zNewlist
          );
        }

        otaObjIterPrepareTmpInsert(p, pIter, zCollist, zOtaRowid);
      }

      /* Allocate space required for the zMask field. */
      pIter->zMask = (char*)otaMalloc(p, pIter->nTblCol+1);

      sqlite3_free(zWhere);
      sqlite3_free(zOldlist);

      const char *zPrefix = "";

      if( pIter->eType!=OTA_PK_VTAB ) zPrefix = "ota_imp_";
      zUpdate = sqlite3_mprintf("UPDATE \"%s%w\" SET %s WHERE %s", 
          zPrefix, pIter->zTbl, zSet, zWhere
      );
      p->rc = prepareFreeAndCollectError(
          p->dbMain, &pIter->pUpdate, &p->zErrmsg, zUpdate
      );
      *ppStmt = pIter->pUpdate;
    }
    if( p->rc==SQLITE_OK ){
      memcpy(pIter->zMask, zMask, pIter->nTblCol);
    }
    sqlite3_free(zWhere);
    sqlite3_free(zSet);
  }
  return p->rc;
}

static sqlite3 *otaOpenDbhandle(sqlite3ota *p, const char *zName){
  sqlite3 *db = 0;
  if( p->rc==SQLITE_OK ){
    const int flags = SQLITE_OPEN_READWRITE | SQLITE_OPEN_CREATE;
    p->rc = sqlite3_open_v2(zName, &db, flags, p->zVfsName);
    if( p->rc ){
      p->zErrmsg = sqlite3_mprintf("%s", sqlite3_errmsg(db));
      sqlite3_close(db);
      db = 0;
    }
  }
  return db;
}

/*
** Open the database handle and attach the OTA database as "ota". If an
** error occurs, leave an error code and message in the OTA handle.
*/
static void otaOpenDatabase(sqlite3ota *p){

  assert( p->rc==SQLITE_OK );
  assert( p->dbMain==0 && p->dbOta==0 );

  p->eStage = 0;
  p->dbMain = otaOpenDbhandle(p, p->zTarget);
  p->dbOta = otaOpenDbhandle(p, p->zOta);

  if( p->rc==SQLITE_OK ){
    p->rc = sqlite3_create_function(p->dbMain, 
        "ota_tmp_insert", -1, SQLITE_UTF8, (void*)p, otaTmpInsertFunc, 0, 0
    );
  }


  if( p->rc==SQLITE_OK ){

    p->rc = sqlite3_file_control(p->dbMain, "main", SQLITE_FCNTL_OTA, (void*)p);
  }
  otaMPrintfExec(p, p->dbMain, "SELECT * FROM sqlite_master");

  /* Mark the database file just opened as an OTA target database. If 
  ** this call returns SQLITE_NOTFOUND, then the OTA vfs is not in use.
  ** This is an error.  */
  if( p->rc==SQLITE_OK ){
    p->rc = sqlite3_file_control(p->dbMain, "main", SQLITE_FCNTL_OTA, (void*)p);
  }

  if( p->rc==SQLITE_NOTFOUND ){
    p->rc = SQLITE_ERROR;
    p->zErrmsg = sqlite3_mprintf("ota vfs not found");

  }
}

/*
** 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 pState is NULL, then the wal file may not have been opened and
  ** recovered. Running a read-statement here to ensure that doing so
  ** does not interfere with the "capture" process below.  */
  if( pState==0 ){
    p->eStage = 0;
    if( p->rc==SQLITE_OK ){
      p->rc = sqlite3_exec(p->dbMain, "SELECT * FROM sqlite_master", 0, 0, 0);
    }
  }

  /* Assuming no error has occurred, run a "restart" checkpoint with the
  ** sqlite3ota.eStage variable set to CAPTURE. This turns on the following
  ** special behaviour in the ota VFS:
  **

  ** WRITER, CHECKPOINT and READ0 locks are still held, it is safe to copy 
  ** data from the wal file into the database file according to the 
  ** contents of aFrame[].
  */
  if( p->rc==SQLITE_OK ){
    int rc2;
    p->eStage = OTA_STAGE_CAPTURE;
    rc2 = sqlite3_exec(p->dbMain, "PRAGMA main.wal_checkpoint=restart", 0, 0,0);
    if( rc2!=SQLITE_INTERNAL ) p->rc = rc2;
  }

  if( p->rc==SQLITE_OK ){
    p->eStage = OTA_STAGE_CKPT;
    p->nStep = (pState ? pState->nRow : 0);
    p->aBuf = otaMalloc(p, p->pgsz);

** The OTA handle is currently in OTA_STAGE_OAL state, with a SHARED lock
** on the database file. This proc moves the *-oal file to the *-wal path,
** then reopens the database file (this time in vanilla, non-oal, WAL mode).
** 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->dbMain, "main");

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

  assert( p->eStage==OTA_STAGE_MOVE );
  assert( p->rc==SQLITE_OK && p->zErrmsg==0 );
  if( zWal==0 || zOal==0 ){

    if( p->rc==SQLITE_OK ){
      otaFileSuffix3(zBase, zWal);
      otaFileSuffix3(zBase, zOal);
      rename(zOal, zWal);

      /* Re-open the databases. */
      otaObjIterFinalize(&p->objiter);
      sqlite3_close(p->dbMain);
      sqlite3_close(p->dbOta);
      p->dbMain = 0;
      p->dbOta = 0;
      otaOpenDatabase(p);
      otaSetupCheckpoint(p, 0);
    }
  }

  sqlite3_free(zWal);
  sqlite3_free(zOal);

  }

 step_out:
  return p->rc;
}

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

    p->rc = prepareAndCollectError(p->dbMain, &pStmt, &p->zErrmsg, 
        "PRAGMA schema_version"
    );
    if( p->rc==SQLITE_OK ){
      /* Coverage: it may be that this sqlite3_step() cannot fail. There
      ** is already a transaction open, so the prepared statement cannot
      ** throw an SQLITE_SCHEMA exception. The only database page the
      ** statement reads is page 1, which is guaranteed to be in the cache.
      ** And no memory allocations are required.  */
      if( SQLITE_ROW==sqlite3_step(pStmt) ){
        iCookie = sqlite3_column_int(pStmt, 0);
      }
      otaFinalize(p, pStmt);
    }
    if( p->rc==SQLITE_OK ){
      otaMPrintfExec(p, p->dbMain, "PRAGMA schema_version = %d", iCookie+1);
    }
  }
}

/*
** Update the contents of the ota_state table within the ota database. The
** value stored in the OTA_STATE_STAGE column is eStage. All other values
** are determined by inspecting the ota handle passed as the first argument.
*/
static void otaSaveState(sqlite3ota *p, int eStage){
  if( p->rc==SQLITE_OK || p->rc==SQLITE_DONE ){
    sqlite3_stmt *pInsert = 0;
    int rc;

    assert( p->zErrmsg==0 );
    rc = prepareFreeAndCollectError(p->dbOta, &pInsert, &p->zErrmsg, 
        sqlite3_mprintf(
          "INSERT OR REPLACE INTO ota_state(k, v) VALUES "
          "(%d, %d), "
          "(%d, %Q), "
          "(%d, %Q), "
          "(%d, %d), "
          "(%d, %lld), "
          "(%d, %lld), "
          "(%d, %lld), "
          "(%d, %lld) ",
          OTA_STATE_STAGE, 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, p->iWalCksum,
          OTA_STATE_COOKIE, (i64)p->pTargetFd->iCookie,
          OTA_STATE_OALSZ, p->iOalSz
      )
    );
    assert( pInsert==0 || rc==SQLITE_OK );

    if( rc==SQLITE_OK ){
      sqlite3_step(pInsert);
      rc = sqlite3_finalize(pInsert);
    }

        while( 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.  */
            if( pIter->eType!=OTA_PK_VTAB ){
              const char *zTbl = pIter->zTbl;
              otaMPrintfExec(p, p->dbOta, "DELETE FROM 'ota_tmp_%q'", 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( p->rc==SQLITE_OK ){
          assert( pIter->zTbl==0 );
          otaSaveState(p, OTA_STAGE_MOVE);
          otaIncrSchemaCookie(p);
          if( p->rc==SQLITE_OK ){
            p->rc = sqlite3_exec(p->dbMain, "COMMIT", 0, 0, &p->zErrmsg);
          }
          if( p->rc==SQLITE_OK ){
            p->rc = sqlite3_exec(p->dbOta, "COMMIT", 0, 0, &p->zErrmsg);
          }
          p->eStage = OTA_STAGE_MOVE;
        }
        break;
      }

      case OTA_STAGE_MOVE: {

** 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 k, v FROM ota_state";
  OtaState *pRet = 0;
  sqlite3_stmt *pStmt = 0;
  int rc;
  int rc2;

  pRet = (OtaState*)otaMalloc(p, sizeof(OtaState));
  if( pRet==0 ) return 0;

  rc = prepareAndCollectError(p->dbOta, &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);
        if( pRet->eStage!=OTA_STAGE_OAL
         && pRet->eStage!=OTA_STAGE_MOVE
         && pRet->eStage!=OTA_STAGE_CKPT

      case OTA_STATE_CKPT:
        pRet->iWalCksum = sqlite3_column_int64(pStmt, 1);
        break;

      case OTA_STATE_COOKIE:
        pRet->iCookie = (u32)sqlite3_column_int64(pStmt, 1);
        break;

      case OTA_STATE_OALSZ:
        pRet->iOalSz = (u32)sqlite3_column_int64(pStmt, 1);
        break;

      default:
        rc = SQLITE_CORRUPT;
        break;
    }
  }
  rc2 = sqlite3_finalize(pStmt);

      p->zOta = &p->zTarget[nTarget+1];
      memcpy(p->zOta, zOta, nOta+1);
      otaOpenDatabase(p);
    }

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

    if( p->rc==SQLITE_OK ){
      pState = otaLoadState(p);
      assert( pState || p->rc!=SQLITE_OK );
      if( p->rc==SQLITE_OK ){

        if( pState->eStage==0 ){ 
          otaDeleteOalFile(p);
          p->eStage = OTA_STAGE_OAL;
        }else{
          p->eStage = pState->eStage;
        }
        p->nProgress = pState->nProgress;
        p->iOalSz = pState->iOalSz;
      }
    }
    assert( p->rc!=SQLITE_OK || p->eStage!=0 );

    if( p->rc==SQLITE_OK 
     && (p->eStage==OTA_STAGE_OAL || p->eStage==OTA_STAGE_MOVE)
    ){   

        p->zErrmsg = sqlite3_mprintf("database modified during ota update");
      }
    }

    if( p->rc==SQLITE_OK ){
      if( p->eStage==OTA_STAGE_OAL ){

        /* Open transactions both databases. The *-oal file is opened or
        ** created at this point. */
        p->rc = sqlite3_exec(p->dbMain, "BEGIN IMMEDIATE", 0, 0, &p->zErrmsg);
        if( p->rc==SQLITE_OK ){
          p->rc = sqlite3_exec(p->dbOta, "BEGIN IMMEDIATE", 0, 0, &p->zErrmsg);
        }
        assert( p->rc!=SQLITE_OK || p->pTargetFd->pWalFd );
  
        /* Point the object iterator at the first object */
        if( p->rc==SQLITE_OK ){
          p->rc = otaObjIterFirst(p, &p->objiter);
        }
  
        if( p->rc==SQLITE_OK ){

  return p;
}

/*
** Return the database handle used by pOta.
*/
sqlite3 *sqlite3ota_db(sqlite3ota *pOta){
  return (pOta ? pOta->dbMain : 0);
}


/*
** If the error code currently stored in the OTA handle is SQLITE_CONSTRAINT,
** then edit any error message string so as to remove all occurrences of
** the pattern "ota_imp_[0-9]*".

/*
** Close the OTA handle.
*/
int sqlite3ota_close(sqlite3ota *p, char **pzErrmsg){
  int rc;
  if( p ){



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

    otaSaveState(p, p->eStage);

    if( p->rc==SQLITE_OK && p->eStage==OTA_STAGE_OAL ){
      p->rc = sqlite3_exec(p->dbOta, "COMMIT", 0, 0, &p->zErrmsg);
    }

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






    /* Close the open database handle and VFS object. */
    sqlite3_close(p->dbMain);
    sqlite3_close(p->dbOta);
    otaDeleteVfs(p);
    sqlite3_free(p->aBuf);
    sqlite3_free(p->aFrame);

    otaEditErrmsg(p);
    rc = p->rc;
    *pzErrmsg = p->zErrmsg;

static int otaVfsRead(
  sqlite3_file *pFile, 
  void *zBuf, 
  int iAmt, 
  sqlite_int64 iOfst
){
  ota_file *p = (ota_file*)pFile;
  sqlite3ota *pOta = p->pOta;
  int rc;

  if( pOta && pOta->eStage==OTA_STAGE_CAPTURE ){
    assert( p->openFlags & SQLITE_OPEN_WAL );
    rc = otaCaptureWalRead(p->pOta, iOfst, iAmt);
  }else{
    if( pOta && pOta->eStage==OTA_STAGE_OAL 
     && (p->openFlags & SQLITE_OPEN_WAL) 
     && iOfst>=pOta->iOalSz 
    ){
      rc = SQLITE_OK;
      memset(zBuf, 0, iAmt);
    }else{
      rc = p->pReal->pMethods->xRead(p->pReal, zBuf, iAmt, iOfst);
    }
    if( rc==SQLITE_OK && iOfst==0 && (p->openFlags & SQLITE_OPEN_MAIN_DB) ){
      /* These look like magic numbers. But they are stable, as they are part
       ** of the definition of the SQLite file format, which may not change. */
      u8 *pBuf = (u8*)zBuf;
      p->iCookie = otaGetU32(&pBuf[24]);
      p->iWriteVer = pBuf[19];
    }

static int otaVfsWrite(
  sqlite3_file *pFile, 
  const void *zBuf, 
  int iAmt, 
  sqlite_int64 iOfst
){
  ota_file *p = (ota_file*)pFile;
  sqlite3ota *pOta = p->pOta;
  int rc;

  if( pOta && pOta->eStage==OTA_STAGE_CAPTURE ){
    assert( p->openFlags & SQLITE_OPEN_MAIN_DB );
    rc = otaCaptureDbWrite(p->pOta, iOfst);
  }else{
    if( pOta && pOta->eStage==OTA_STAGE_OAL 
     && (p->openFlags & SQLITE_OPEN_WAL) 
     && iOfst>=pOta->iOalSz
    ){
      pOta->iOalSz = iAmt + iOfst;
    }
    rc = p->pReal->pMethods->xWrite(p->pReal, zBuf, iAmt, iOfst);
    if( rc==SQLITE_OK && iOfst==0 && (p->openFlags & SQLITE_OPEN_MAIN_DB) ){
      /* These look like magic numbers. But they are stable, as they are part
      ** of the definition of the SQLite file format, which may not change. */
      u8 *pBuf = (u8*)zBuf;
      p->iCookie = otaGetU32(&pBuf[24]);
      p->iWriteVer = pBuf[19];