Alfred Perlstein <bright@wintelcom.net> writes:
> I posted earlier this week about a table of ours getting corrupted
> after some time. The table looks like this:
> ...
> After a while we get this crash apparently, followed by crashes of any
> backend that scans this table, not updating this specific table makes
> the crashes go away, the problem seems pretty isolated to this data.
Hmm. It looks like one of the variable-length fields is getting
clobbered, but I have no idea why. It will probably take some digging
in the corefile and clobbered table to learn much. The good thing is
that you seem to have a fairly easily reproducible bug, so tracking it
down should be possible.
Is the data+program needed to trigger the bug self-contained enough that
you could wrap it up and send it to me? The most convenient thing from
my end would be to reproduce and study it here, if possible.
> I know there's been a couple of updates to the source since this date
> (compiled on Aug 3), any idea if:
> a) an upgrade is a good idea
> b) an upgrade is a safe idea
I see no reason that you shouldn't update to the tip of the
REL7_0_PATCHES branch, but I don't have a lot of hope that it will cure
your problem. I doubt this is related to anything we've fixed.
regards, tom lane