Re: Prevent object capture in CREATE/ALTER EXTENSION scripts

From: Jan Nidzwetzki <jan(at)planetscale(dot)com>
To: solai v <solai(dot)cdac(at)gmail(dot)com>, Osama Abdul Qader <osamaabdulqader(dot)cs(at)gmail(dot)com>
Cc: PostgreSQL Hackers <pgsql-hackers(at)lists(dot)postgresql(dot)org>
Subject: Re: Prevent object capture in CREATE/ALTER EXTENSION scripts
Date: 2026-11-13 10:31:05
Message-ID: 6c5a1cb1-b799-4e40-bda1-525afb8c82aa@planetscale.com
Views: Whole Thread | Raw Message | Download mbox | Resend email
Thread:
Lists: pgsql-hackers

Hi Hackers,

To confirm the patch works with existing extensions, I ran additional
tests. I installed various extensions on both a patched and an unpatched
server. I tested contrib and src/test/modules on master (95 extensions),
and all postgresql-18-* packages from apt.postgresql.org plus contrib
(171 extensions) on a backport of this patch to 18.6 (since there are no
extension packages for master).

I installed each extension as a superuser and, when possible (trusted or
superuser = false), also as a non-superuser. I also did this in
databases that already had objects owned by a non-superuser. I updated
extensions from older versions to the current one and repeated the
installs with debug_parallel_query set to regress. Only a few scripts
actually ran queries in parallel workers, and they behaved the same on
both servers. For each run, I compared the results of every statement
and checked which objects the extension's members pointed to using
pg_depend.

I found no differences between the patched and unpatched servers. The
regression tests for contrib and 35 third-party extensions also gave the
same results. For the third-party regression tests, I only included
extensions that run their tests with a plain PGXS "make installcheck".

I'm happy to share the scripts I used if anyone wants to rerun the tests.

Best regards
Jan

--
Jan Nidzwetzki
PlanetScale

In response to

Browse pgsql-hackers by date

  From Date Subject
Next Message Peter Eisentraut 2026-11-13 10:37:40 -Wformat-nonliteral
Previous Message Hayato Kuroda (Fujitsu) 2026-11-13 10:21:27 RE: Hash index: stop registering pin-only bucket pages in WAL (remove REGBUF_NO_CHANGE)