From: Tom Lane <tgl@sss.pgh.pa.us>
To: Dean Rasheed <dean.a.rasheed@gmail.com>
cc: Timo Stolz <timo.stolz@nullachtvierzehn.de>,
        PostgreSQL mailing lists <pgsql-bugs@lists.postgresql.org>,
        Jonas Reinsch <jonas.reinsch@pitchview.de>,
        Laura Schlimmer <laura.schlimmer@pitchview.de>
Subject: Re: If a row-level security policy contains a set returning function,
 pg_dump returns an incorrect serialization of that policy if the return type
 of the function was altered
In-reply-to: 
 <CAEZATCUJk88LWb5VNaW4qz6u2NgL1wvJQEHg=jJ4LK8kV-299A@mail.gmail.com>
References: <5c91267e-3b6d-5795-189c-d15a55d61dbb@nullachtvierzehn.de>
 <1983818.1658350464@sss.pgh.pa.us>
 <CAEZATCUy6Uqtv=_PYbc-CmpkfF2vBxE0ZhyXRr_u1nt=wjHCkw@mail.gmail.com>
 <130689.1658426324@sss.pgh.pa.us> <182492.1658431155@sss.pgh.pa.us>
 <CAEZATCUJk88LWb5VNaW4qz6u2NgL1wvJQEHg=jJ4LK8kV-299A@mail.gmail.com>
Comments: In-reply-to Dean Rasheed <dean.a.rasheed@gmail.com>
	message dated "Thu, 21 Jul 2022 22:27:38 +0100"
MIME-Version: 1.0
Content-Type: text/plain; charset="us-ascii"
Content-ID: <667072.1658508629.1@sss.pgh.pa.us>
Date: Fri, 22 Jul 2022 12:50:29 -0400
Message-ID: <667073.1658508629@sss.pgh.pa.us>
Archived-At: 
 <https://www.postgresql.org/message-id/667073.1658508629%40sss.pgh.pa.us>
Precedence: bulk

Dean Rasheed <dean.a.rasheed@gmail.com> writes:
> On Thu, 21 Jul 2022 at 20:19, Tom Lane <tgl@sss.pgh.pa.us> wrote:
>> Here's a proposed patch for that.  I wouldn't consider pushing this
>> into released branches, but maybe it's not too late for v15?

> That looks reasonable to me. It covers all the cases I could think of
> to try, and I can't see any loopholes. +1 for applying it to v15.

Thanks for checking it!

I had second thoughts about removing the old test cases: that would
leave us with no test coverage for the executor's defenses against
bad plans.  I'm not so foolish as to imagine we'll never introduce
another bug that would reach those defenses.  So what I did was to
adjust those cases to manually delete the new pg_depend entries,
allowing us to still test what happens without 'em.

Pushed that way.

			regards, tom lane