Re: --single-transaction hack to pg_upgrade does not work

On Sat, Dec 1, 2012 at 03:41:15PM -0500, Andrew Dunstan wrote:
>
> On 12/01/2012 02:34 PM, Bruce Momjian wrote:
> >On Sat, Dec 1, 2012 at 02:31:03PM -0500, Tom Lane wrote:
> >>Andres Freund <andres(at)2ndquadrant(dot)com> writes:
> >>>On 2012-12-01 12:14:37 -0500, Tom Lane wrote:
> >>>>It could do with some comments ;-)
> >>>Hehe, yes. Hopefully this version has enough of that.
> >>Hm, maybe too many --- I don't really think it's necessary for utility.c
> >>to provide a redundant explanation of what's happening.
> >>
> >>Committed with adjustments --- mainly, the
> >>TransactionIdIsCurrentTransactionId test was flat out wrong, because it
> >>would accept a parent transaction ID as well as a subcommitted
> >>subtransaction ID. We could safely allow the latter, but I don't think
> >>it's worth the trouble to add another xact.c test function.
> >Thanks everyone. I can confirm that pg_upgrades make "check now"
> >passes, so this should green the buildfarm. Again, I aplogize for the
> >fire drill.
> >
>
>
> I've added better logging of pg_upgrade testing to the buildfarm
> module: <https://github.com/PGBuildFarm/client-code/commit/83834cceaea95ba42c03a1079a8c768782e32a6b>
> example is at <http://www.pgbuildfarm.org/cgi-bin/show_log.pl?nm=crake&dt=2012-12-01%2017%3A44%3A03>
> This will be in the next buildfarm client release.

Wow, that looks great. You even show the last few lines from the log
files!

--
Bruce Momjian <bruce(at)momjian(dot)us> http://momjian.us
EnterpriseDB http://enterprisedb.com

+ It's impossible for everything to be true. +