Re: installcheck-world concurrency issues

Hi,

On 2022-10-04 17:05:40 +0900, Michael Paquier wrote:
> On Mon, Oct 03, 2022 at 04:41:11PM -0700, Andres Freund wrote:
> > There's a few further roles that seem to pose some danger goign forward:
> 
> I have never seen that myself, but 0001 is a nice cleanup.
> generated.sql includes a user named "regress_user11".  Perhaps that's
> worth renaming while on it?

I think regress_* without a "namespace" is what's src/test/regress uses, so
there's not really a need?


> > A second issue I noticed is that advisory_lock.sql often fails, because the
> > pg_locks queries don't restrict to the current database. Patch attached.
> 
> As in prepared_xacts.sql or just advisory locks taken in an installed
> cluster?  Or both?

There's various isolation tests, including several in src/test/isolation, that
use advisory locks.

prepared_xacts.sql shouldn't be an issue, because it's scheduled in a separate
group.


> > I attached the meson patch as well, but just because I used it to to get to
> > these patches.
> 
> I am still studying a lot of this area, but it seems like all the
> spots requiring a custom configuration (aka NO_INSTALLCHECK) are
> covered.  --setup running is working here with 0003.

Thanks for checking.

Greetings,

Andres Freund