Re: BUG #18950: pgsql function that worked in Postgresql 16 does not return in Postgresql 17

Lowell Hought <lowell.hought@gmail.com>

From: Lowell Hought <lowell.hought@gmail.com>
To: Tom Lane <tgl@sss.pgh.pa.us>
Cc: Laurenz Albe <laurenz.albe@cybertec.at>, pgsql-bugs@lists.postgresql.org
Date: 2025-06-08T03:08:06Z
Lists: pgsql-bugs
My initial response that changing the parameter worked was mistaken.  I
forgot to shut down version 16 so the function call happened on that
instance.  I now shut down 16, fired up 17 and set the parameter, and the
behavior did not change.  The query itself returns results in about 1
second.  The function never returns at all, it just sits there.

Lowell


On Sat, Jun 7, 2025 at 10:02 PM Tom Lane <tgl@sss.pgh.pa.us> wrote:

> Lowell Hought <lowell.hought@gmail.com> writes:
> > Yes!  I just changed the parameter as  you suggested and the query
> returned
> > as expected.
> > So I guess something changed between version 16 and version 17?  Perhaps
> > the default for that setting?
>
> No, that default has not changed.  What apparently happened is that
> cost estimates or something changed enough to persuade the planner
> to use a generic plan that's considerably inferior to what it was
> choosing before.  There's nowhere near enough info in your report
> to pin it down more closely than that.
>
> Just to eliminate the obvious --- you did run ANALYZE after the
> upgrade, right?  If you had any nondefault settings of
> default_statistics_target or similar planner parameters,
> did you make sure they got transferred to the new installation?
>
>                         regards, tom lane
>