Fix a problem that could cause a spurious SQLITE_NOMEM error when attempting

to resume an RBU operation if the previous client failed right after
completing the incremental checkpoint. Also a "cannot vacuum wal db" error
that could occur when resuming an RBU vacuum if an error (OOM or IO error)
occurs during the incremental checkpoint.

FossilOrigin-Name: 681d96eb822e606da53700867191d4738bda20c8

ext/rbu/rbufault4.test

@@ -0,0 +1,66 @@
+# 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
+source $testdir/malloc_common.tcl
+set ::testprefix rbufault4
+
+for {set tn 1} {1} {incr tn} {
+  reset_db
+  do_execsql_test 1.0 {
+    CREATE TABLE t1(a INTEGER PRIMARY KEY, b, c);
+    CREATE INDEX i1b ON t1(b);
+    CREATE INDEX i1c ON t1(c);
+    INSERT INTO t1 VALUES(1, 2, 3);
+    INSERT INTO t1 VALUES(4, 5, 6);
+  }
+
+  forcedelete test.db2
+  sqlite3rbu_vacuum rbu test.db test.db2
+  for {set i 0} {$i < $tn} {incr i} { rbu step }
+  set rc [rbu close]
+  if {$rc!="SQLITE_OK"} { 
+    if {$rc!="SQLITE_DONE"} {error $rc}
+    break
+  }
+  faultsim_save
+
+  do_faultsim_test $tn -faults oom-t* -prep {
+    faultsim_restore
+  } -body {
+    sqlite3rbu_vacuum rbu test.db test.db2
+    while 1 {
+      set rc [rbu step]
+      if {$rc=="SQLITE_DONE"} break
+      if {$rc!="SQLITE_OK"} { error $rc }
+    }
+  } -test {
+    catch {rbu close}
+    faultsim_test_result {0 {}} {1 SQLITE_NOMEM} {1 SQLITE_IOERR_NOMEM}
+
+    sqlite3rbu_vacuum rbu test.db test.db2
+    while {[rbu step]=="SQLITE_OK"} {}
+    set trc [rbu close]
+    if {$trc!="SQLITE_DONE"} { error "Got $trc instead of SQLITE_DONE!" }
+
+    set rc [db one {PRAGMA integrity_check}]
+    if {$rc!="ok"} { error "Got $rc instead of ok!" }
+  }
+}
+
+
+
+finish_test
+

ext/rbu/rbuvacuum2.test

@@ -199,6 +199,37 @@ if {$::tcl_platform(platform)=="unix"} {
   }
 }
 
+#-------------------------------------------------------------------------
+# Test the outcome of some other connection running a checkpoint while
+# the incremental checkpoint is suspended.
+#
+reset_db
+do_execsql_test 6.0 {
+  CREATE TABLE t1(a INTEGER PRIMARY KEY, b, c);
+  CREATE INDEX i1b ON t1(b);
+  CREATE INDEX i1c ON t1(c);
+  INSERT INTO t1 VALUES(1, 2, 3);
+  INSERT INTO t1 VALUES(4, 5, 6);
+}
+forcedelete test.db2
+
+do_test 6.1 {
+  sqlite3rbu_vacuum rbu test.db test.db2
+  while {[rbu state]!="checkpoint"} { rbu step }
+  rbu close
+} {SQLITE_OK}
+
+do_execsql_test 6.2 {
+  SELECT 1 FROM sqlite_master LIMIT 1;
+  PRAGMA wal_checkpoint;
+} {1 0 4 4}
+
+do_test 6.3 {
+  sqlite3rbu_vacuum rbu test.db test.db2
+  while {[rbu step]!="SQLITE_DONE"} { rbu step }
+  rbu close
+  execsql { PRAGMA integrity_check }
+} {ok}
 
 finish_test
 

ext/rbu/sqlite3rbu.c

@@ -2333,7 +2333,7 @@ static RbuState *rbuLoadState(sqlite3rbu *p){
 ** Open the database handle and attach the RBU database as "rbu". If an
 ** error occurs, leave an error code and message in the RBU handle.
 */
-static void rbuOpenDatabase(sqlite3rbu *p){
+static void rbuOpenDatabase(sqlite3rbu *p, int *pbRetry){
   assert( p->rc || (p->dbMain==0 && p->dbRbu==0) );
   assert( p->rc || rbuIsVacuum(p) || p->zTarget!=0 );
 
@@ -2420,6 +2420,15 @@ static void rbuOpenDatabase(sqlite3rbu *p){
     if( !rbuIsVacuum(p) ){
       p->dbMain = rbuOpenDbhandle(p, p->zTarget, 1);
     }else if( p->pRbuFd->pWalFd ){
+      if( pbRetry ){
+        p->pRbuFd->bNolock = 0;
+        sqlite3_close(p->dbRbu);
+        sqlite3_close(p->dbMain);
+        p->dbMain = 0;
+        p->dbRbu = 0;
+        *pbRetry = 1;
+        return;
+      }
       p->rc = SQLITE_ERROR;
       p->zErrmsg = sqlite3_mprintf("cannot vacuum wal mode database");
     }else{
@@ -2600,16 +2609,18 @@ static void rbuSetupCheckpoint(sqlite3rbu *p, RbuState *pState){
     if( rc2!=SQLITE_INTERNAL ) p->rc = rc2;
   }
 
-  if( p->rc==SQLITE_OK ){
+  if( p->rc==SQLITE_OK && p->nFrame>0 ){
     p->eStage = RBU_STAGE_CKPT;
     p->nStep = (pState ? pState->nRow : 0);
     p->aBuf = rbuMalloc(p, p->pgsz);
     p->iWalCksum = rbuShmChecksum(p);
   }
 
-  if( p->rc==SQLITE_OK && pState && pState->iWalCksum!=p->iWalCksum ){
-    p->rc = SQLITE_DONE;
-    p->eStage = RBU_STAGE_DONE;
+  if( p->rc==SQLITE_OK ){
+    if( p->nFrame==0 || (pState && pState->iWalCksum!=p->iWalCksum) ){
+      p->rc = SQLITE_DONE;
+      p->eStage = RBU_STAGE_DONE;
+    }
   }
 }
 
@@ -2782,7 +2793,7 @@ static void rbuMoveOalFile(sqlite3rbu *p){
 #endif
 
       if( p->rc==SQLITE_OK ){
-        rbuOpenDatabase(p);
+        rbuOpenDatabase(p, 0);
         rbuSetupCheckpoint(p, 0);
       }
     }
@@ -3493,6 +3504,7 @@ static sqlite3rbu *openRbuHandle(
     /* Open the target, RBU and state databases */
     if( p->rc==SQLITE_OK ){
       char *pCsr = (char*)&p[1];
+      int bRetry = 0;
       if( zTarget ){
         p->zTarget = pCsr;
         memcpy(p->zTarget, zTarget, nTarget+1);
@@ -3504,7 +3516,18 @@ static sqlite3rbu *openRbuHandle(
       if( zState ){
         p->zState = rbuMPrintf(p, "%s", zState);
       }
-      rbuOpenDatabase(p);
+
+      /* If the first attempt to open the database file fails and the bRetry
+      ** flag it set, this means that the db was not opened because it seemed
+      ** to be a wal-mode db. But, this may have happened due to an earlier
+      ** RBU vacuum operation leaving an old wal file in the directory.
+      ** If this is the case, it will have been checkpointed and deleted
+      ** when the handle was closed and a second attempt to open the 
+      ** database may succeed.  */
+      rbuOpenDatabase(p, &bRetry);
+      if( bRetry ){
+        rbuOpenDatabase(p, 0);
+      }
     }
 
     if( p->rc==SQLITE_OK ){