From: Tom Lane <tgl@sss.pgh.pa.us>
To: David Rowley <dgrowleyml@gmail.com>
cc: pgsql-hackers@lists.postgresql.org
Subject: Surely this code in setrefs.c is wrong?
MIME-Version: 1.0
Content-Type: text/plain; charset="us-ascii"
Content-ID: <2277536.1694301772.1@sss.pgh.pa.us>
Date: Sat, 09 Sep 2023 19:22:52 -0400
Message-ID: <2277537.1694301772@sss.pgh.pa.us>
Archived-At: <https://www.postgresql.org/message-id/2277537.1694301772%40sss.pgh.pa.us>
Precedence: bulk

I happened to notice this bit in fix_expr_common's processing
of ScalarArrayOpExprs:

        set_sa_opfuncid(saop);
        record_plan_function_dependency(root, saop->opfuncid);

        if (!OidIsValid(saop->hashfuncid))
            record_plan_function_dependency(root, saop->hashfuncid);

        if (!OidIsValid(saop->negfuncid))
            record_plan_function_dependency(root, saop->negfuncid);

Surely those if-conditions are exactly backward, and we should be
recording nonzero hashfuncid and negfuncid entries, not zero ones.
As-is, the code's a no-op because record_plan_function_dependency
will ignore OIDs less than FirstUnpinnedObjectId, including zero.

"git blame" blames 50e17ad28 and 29f45e299 for these, so v14
has only half the problem of later branches.

			regards, tom lane