On Wed, Feb 11, 2026 at 3:34 AM Tom Lane <tgl@sss.pgh.pa.us> wrote:
> Nathan Bossart <nathandbossart@gmail.com> writes:
> > On Tue, Feb 10, 2026 at 12:06:02PM -0500, Tom Lane wrote:
> >> It's not clear to me that it's worth running this to ground in any
> >> more detail than that. The behavior is not wrong; it's the test's
> >> fault to assume that these rows will be returned in a deterministic
> >> order. So I think the right fix is to adjust the test query,
> >> along the lines of
> >>
> >> -UPDATE ft2 SET c3 = 'bar' WHERE postgres_fdw_abs(c1) > 2000 RETURNING *;
> >> +WITH cte AS (
> >> + UPDATE ft2 SET c3 = 'bar' WHERE postgres_fdw_abs(c1) > 2000 RETURNING *
> >> +) SELECT * FROM cte ORDER BY c1;
>
> > +1. I faintly recall looking into this a while ago and, for some reason, I
> > was worried that this would become a game of Whac-A-Mole, so apparently I
> > didn't follow through. But fixing this query is still an improvement over
> > the status quo.
>
> Yeah, it's certainly fair to wonder where else we have
> even-lower-probability test interactions. But I don't think
> getting rid of the interaction is realistic, especially given
> Alexander's results (which I confess to having forgotten about)
> that show that autovacuum is involved in this somehow despite
> being disabled on this particular table. So the answer has to
> be to make the test case more robust against such things.
+1 for that. I noticed this problem because of Alexander's report,
but I completely forgot about it...
Thanks!
Best regards,
Etsuro Fujita
