Commit580bde7

committed

Initialize shared memory copy of ckptXidEpoch correctly when not in recovery.

This bug was introduced by commit20d98ab,so backpatch this to 9.0-9.2 like that one.This fixes bug #6710, reported by Tarvi Pillessaar

1 parent38a9052 commit580bde7Copy full SHA for 580bde7

File tree

-4

lines changed

-4

lines changed

Lines changed: 3 additions & 4 deletions

Original file line number	Diff line number	Diff line change
`@@ -5941,11 +5941,14 @@ StartupXLOG(void)`
`5941`	`5941`	`ereport(PANIC,`
`5942`	`5942`	`(errmsg("invalid next transaction ID")));`
`5943`	`5943`
	`5944`	`+/* initialize shared memory variables from the checkpoint record */`
`5944`	`5945`	`ShmemVariableCache->nextXid=checkPoint.nextXid;`
`5945`	`5946`	`ShmemVariableCache->nextOid=checkPoint.nextOid;`
`5946`	`5947`	`ShmemVariableCache->oidCount=0;`
`5947`	`5948`	`MultiXactSetNextMXact(checkPoint.nextMulti,checkPoint.nextMultiOffset);`
`5948`	`5949`	`SetTransactionIdLimit(checkPoint.oldestXid,checkPoint.oldestXidDB);`
	`5950`	`+XLogCtl->ckptXidEpoch=checkPoint.nextXidEpoch;`
	`5951`	`+XLogCtl->ckptXid=checkPoint.nextXid;`
`5949`	`5952`
`5950`	`5953`	`/*`
`5951`	`5954`	`* We must replay WAL entries using the same TimeLineID they were created`
`@@ -6022,10 +6025,6 @@ StartupXLOG(void)`
`6022`	`6025`	`/* No need to hold ControlFileLock yet, we aren't up far enough */`
`6023`	`6026`	`UpdateControlFile();`
`6024`	`6027`
`6025`		`-/* initialize shared-memory copy of latest checkpoint XID/epoch */`
`6026`		`-XLogCtl->ckptXidEpoch=ControlFile->checkPointCopy.nextXidEpoch;`
`6027`		`-XLogCtl->ckptXid=ControlFile->checkPointCopy.nextXid;`
`6028`		`-`
`6029`	`6028`	`/* initialize our local copy of minRecoveryPoint */`
`6030`	`6029`	`minRecoveryPoint=ControlFile->minRecoveryPoint;`
`6031`	`6030`

Comments

(0)