Peter Eisentraut <peter_e@gmx.net> writes:
> On Tue, 2012-09-04 at 20:46 -0400, Andrew Dunstan wrote:
>> OK, nobody else has reacted. I've spoken to Bruce and he seems happy
>> with it, although, TBH, whe I talked to him I thought I understood it
>> and now I'm not so sure. So we have 3 possibilities: leave it as is with
>> an error-hiding hack in the test script, apply this patch which removes
>> the hack and applies a fix that apparently works but which confuses us a
>> bit, or go back to generating errors. The last choice would mean I would
>> need to turn off pg_ugrade testing on Windows pending a fix. And we have
>> to decide pretty much now so we can get 9.2 out the door.
> I think now is not the time to cram in poorly understood changes into a
> release candidate. There is no requirement to have the tests running
> now or in time for the release, seeing also that no one has been
> particularly bothered about it for the past 11 months.
Also, the tests *are* passing right now. I agree, let's not risk
destabilizing it. pg_upgrade is way overdue for some quiet time so we
can verify a full day's buildfarm cycle on it before the release wrap.
regards, tom lane