Re: bk commit into 5.1 tree (Justin.He:1.2589)

[Jonas Oreland]

Hi,

General comment 1: looks like a good start
General comment 2: using quilt to separate logical changes in patch would be good
1) backwards scan
2) move-tuple
3) optimize table

General (not) comment 3: I did not look at all at the
NdbOptimizeTableHandleOrSomethingWithLongName implementation

General comment 4: i think the DbtupScan changes looks quite big...can it be made smaller?

Comment 5: Can't this be removed
NdbOperation::setOptimize(Uint64 options)

Comment 6: Isnt below dead code, i.e lets remove it for now and add it as part of a
"real-move" patch

@@ -1144,6 +1145,21 @@ void Dbacc::execACCKEYREQ(Signal* signal
 	operationRecPtr.p->m_op_bits = opbits;
 	return;
       } else {
+        if (op == ZMOVE) {
+          //TODO, for fixpart moving
+          /**
+           * We need update element data since we need to update
+           * acc key after moving fixpart of optimizing table,
+           * and this should not influnce lock issue.
+           * On NON-primary fragment node, need to update forcely,
+           * because normally cannot get lock in commit phase.
+           */
+          Uint32 eh = gePageptr.p->word32[tgeElementptr];
+          operationRecPtr.p->elementPage = gePageptr.i;
+          operationRecPtr.p->elementContainer = tgeContainerptr;
+          operationRecPtr.p->elementPointer = tgeElementptr;
+          operationRecPtr.p->elementIsforward = tgeForward;
+        }