On Jan 14, 2010, at 10:44 AM, Tom Lane wrote:
> Justin Pitts <jpitts@bplglobal.net> writes:
>> On Jan 14, 2010, at 10:16 AM, Tom Lane wrote:
>>> The 100 temp table creations probably will do that just fine.
>=20
>> Is there a way to verify this?
>=20
> You could add an elog(LOG, "message") into ResetPlanCache so you could
> tell when it had been called.
>=20
Done. Sometimes I see it, sometimes not.
>> I don't follow. Are you suggesting I begin another transaction on connec=
tion 1 with a read, and that
>> would provoke the crash?
>=20
> Yes. The rollback only sets the stage for the next transaction to try
> to use a snapshot that isn't there anymore.
>=20
Oh, duh. A read from the same session that rolled-back. That didn't get it =
working (failing?) however.
Running concurrent instances of this test reliably provokes the crash on un=
-patched 8.4.2.
They do not provoke a crash with the patch. That's what i was looking for.
Thanks!
> regards, tom lane