"Oliver Elphick" <olly@lfix.co.uk> writes:
> This is where things are going wrong:
> the pg_vlock file gets deleted in vc_abort with this backtrace:
> (gdb) bt
> #0 vc_abort () at vacuum.c:252
> #1 0x8078f28 in TransactionIdAbort (transactionId=156164) at transam.c:578
> #2 0x80e0269 in XactLockTableWait (xid=156164) at lmgr.c:332
> #3 0x806c9a9 in heap_delete (relation=0x8216ac8, tid=0x8225c48, ctid=0x0) at heapam.c:1149
Ah-hah, I *knew* that code was bogus: TransactionIdAbort() has no
business calling vc_abort(). I fixed that about two days ago
in both current and REL6_5 branches...
regards, tom lane