From: Tom Lane <tgl@sss.pgh.pa.us>
To: Stephen Frost <sfrost@snowman.net>
cc: cam.daniel@gmail.com, pgsql-bugs@lists.postgresql.org
Subject: Re: BUG #16655: pg_dump segfault when excluding postgis table
In-reply-to: <1995213.1602011218@sss.pgh.pa.us>
References: <16655-5c92d6b3a9438137@postgresql.org>
 <1819571.1601937068@sss.pgh.pa.us>
 <20201006163318.GX3063@tamriel.snowman.net>
 <1909344.1602003707@sss.pgh.pa.us> <1995213.1602011218@sss.pgh.pa.us>
Comments: In-reply-to Tom Lane <tgl@sss.pgh.pa.us>
	message dated "Tue, 06 Oct 2020 15:06:58 -0400"
MIME-Version: 1.0
Content-Type: text/plain; charset="us-ascii"
Content-ID: <2007399.1602019564.1@sss.pgh.pa.us>
Content-Transfer-Encoding: quoted-printable
Date: Tue, 06 Oct 2020 17:26:04 -0400
Message-ID: <2007400.1602019564@sss.pgh.pa.us>
Precedence: bulk

I wrote:
> 0001 is sufficient to get past the proposed test case in 0002.  However,
> I think processExtensionTables is also buggy here.  It does need to set
> the "interesting" flag to true if it decides we need to dump the table's
> data, because we have to have the per-attribute data to support that,
> and "interesting" might not ever get set true on the table otherwise.
> But *it is not OK to set "interesting" to false just because we don't
> decide to dump the data*.  If it's been set true for some other reason,
> that other reason didn't just vanish.  So I think we also need 0003.

Upon excavating in the git history, I discovered that that error was
introduced in 3eb3d3e78 [1], which explains why the OP thought it was
new in v13 --- no other branch has shipped the buggy code yet.

It's not entirely clear to me why we don't get a crash on the case in
pre-v12 branches, but I'm inclined to apply these patches all the way
back anyway, as they definitely seem like robustness improvements
whether or not a given branch accidentally fails to fail.

			regards, tom lane

[1] https://www.postgresql.org/message-id/flat/18048b44-3414-b983-8c7c-916=
5b177900d%402ndQuadrant.com