MIME-Version: 1.0
References: <E1u0lJN-002fn9-0n@gemulon.postgresql.org>
 <CAHgHdKtLFWZcjr87hMH0hYDHgcifu4Tj7iHz-xh8qsJREt5cqA@mail.gmail.com>
 <CAH2-Wz=1eJZMSMoACDW2oLcxd8QUtKQS7HuYtqx98LLji+WKuQ@mail.gmail.com>
 <CAHgHdKuWD2h3n6MHv2SOp8VLcKsKM1R+tLBHZrnE_WDOmDaqPg@mail.gmail.com>
 <CAH2-WzmDF0Rcnvi2vhhYwLdF5bcs=8+rvScXWieBVkR2kezn6Q@mail.gmail.com>
 <CAHgHdKsn2W=gPBmj7p6MjQFvxB+zZDBkwTSg0o3f5Hh8rkRrsA@mail.gmail.com>
In-Reply-To: 
 <CAHgHdKsn2W=gPBmj7p6MjQFvxB+zZDBkwTSg0o3f5Hh8rkRrsA@mail.gmail.com>
From: Peter Geoghegan <pg@bowt.ie>
Date: Wed, 30 Apr 2025 16:25:17 -0400
Message-ID: 
 <CAH2-WzkLSMvgGVTbj0g-1yHoQd8X2pjMAQgqjssJd3VBxmNhXw@mail.gmail.com>
Subject: Re: pgsql: Improve nbtree skip scan primitive scan scheduling.
To: Mark Dilger <mark.dilger@enterprisedb.com>
Cc: Matthias van de Meent <boekewurm+postgres@gmail.com>,
 pgsql-committers@lists.postgresql.org
Content-Type: text/plain; charset="UTF-8"
Content-Transfer-Encoding: quoted-printable
Archived-At: 
 <https://www.postgresql.org/message-id/CAH2-WzkLSMvgGVTbj0g-1yHoQd8X2pjMAQgqjssJd3VBxmNhXw%40mail.gmail.com>
Precedence: bulk

On Wed, Apr 30, 2025 at 4:11=E2=80=AFPM Mark Dilger
<mark.dilger@enterprisedb.com> wrote:
> A similar assertion can still be reached on HEAD using the attached (rath=
er contrived) regression test as of f60420cff66a9089a9b431f9c07f4a29aae4990=
a:

I independently found the same bug just a few hours ago, using an
enhanced version of my python fuzzing script (posted to the main skip
scan thread) that now performs backwards scans. I'm not at all
surprised to see that your repro also uses ORDER BY ... DESC.

Would you post to that other thread going forward, rather than posting
to this -committers thread?

I'm going to talk about this on the main skip scan -hackers thread,
when I have a proper fix. I already have a fix for the issue that you
reported, but there seems to be an independent remaining issue with
backwards scans that I haven't quite got to the bottom of just yet.

Thanks
--=20
Peter Geoghegan