On Fri, 31 Oct 2008, Gregory Stark wrote:
> Tom Lane <tgl@sss.pgh.pa.us> writes:
>
>> Jeff Frost <jeff@frostconsultingllc.com> writes:
>>> Tom Lane wrote:
>>>> Huh. That does sound like it's a version-to-version difference.
>>>> There's nothing in the CVS log that seems related though. Are you
>>>> willing to post your test case?
>>>>
>>> It's a customer DB, so I'll contact them and see if we can boil it down
>>> to a test case with no sensitive data.
>>
>> Well, if there was a change it seems to have been in the right direction
>> ;-) so this is mostly just idle curiosity. Don't jump through hoops to
>> get a test case.
>
> Assuming it's not a bug...
Well, after boiling down my test case to the bare essentials, I was unable to
reproduce the different behavior between 8.3.3 and 8.3.4. Now, I've gone back
to the original script and can't reproduce the behavior I previously saw on
8.3.4 and my screen session doesn't have enough scrollback to look at what
happened previously. I was thinking perhaps I had inadvertently committed the
transaction, but then the act would have been dropped as it's a temp table
created with ON COMMIT DROP. But, I've tested 3 times in a row and every time
8.3.4 uses the seq scan just like 8.3.3 now, so I must've done something
differently to get that result as Tom had originally suggested. I just can't
think what it might have been. Perhaps it's time to buy some glasses. :-/
--
Jeff Frost, Owner <jeff@frostconsultingllc.com>
Frost Consulting, LLC http://www.frostconsultingllc.com/
Phone: 916-647-6411 FAX: 916-405-4032
