Re: heap_mark4update: (am)invalid tid

"Robert Creager" <pgsql-bugs(at)logicalchaos(dot)org> writes:
>> If (a) obs_v_trigger is a BEFORE trigger, *and* (b) the problem UPDATE
>> could try to update a row of obs_v more than once (ie, there are
>> multiple rows in obs_i/files that match a row in obs_v), then I think
>> this is a known problem. Otherwise, we need to dig more ...

> Yes, and yes. Any work arounds (ie current sources, smarter SQL),
> other than to break this up into multiple updates?

The fix is in CVS. I've been pestering Marc to wrap up a snapshot
tarball of REL7_3_STABLE branch tip, but he hasn't got round to it.
So probably the easiest answer for you is to check out the branch
from our anonymous-CVS server --- see basic instructions at
http://www.ca.postgresql.org/users-lounge/docs/7.3/postgres/cvs.html
but note the initial checkout command should be
cvs ... co -P -r REL7_3_STABLE pgsql
to get the stable branch rather than development tip. Note you will
need a fairly current version of bison to build from source.

regards, tom lane