From: Tom Lane <tgl@sss.pgh.pa.us>
To: Tomas Vondra <tomas.vondra@2ndquadrant.com>
cc: pgsql-hackers@lists.postgresql.org
Subject: Re: Strange coding in mvdistinct.c
In-reply-to: <29785.1555365602@sss.pgh.pa.us>
References: <29785.1555365602@sss.pgh.pa.us>
Comments: In-reply-to Tom Lane <tgl@sss.pgh.pa.us>
	message dated "Mon, 15 Apr 2019 18:00:02 -0400"
MIME-Version: 1.0
Content-Type: text/plain; charset="us-ascii"
Content-ID: <30399.1555366344.1@sss.pgh.pa.us>
Content-Transfer-Encoding: quoted-printable
Date: Mon, 15 Apr 2019 18:12:24 -0400
Message-ID: <30400.1555366344@sss.pgh.pa.us>
Precedence: bulk

Oh, and as I continue to grep, I found this in dependencies.c:

            dependencies =3D (MVDependencies *) repalloc(dependencies,
                                                       offsetof(MVDependen=
cies, deps)
                                                       + dependencies->nde=
ps * sizeof(MVDependency));

I'm pretty sure this is an actual bug: the calculation should be

                       offsetof(MVDependencies, deps)
                       + dependencies->ndeps * sizeof(MVDependency *));

because deps is an array of MVDependency* not MVDependency.

This would lead to an overallocation not underallocation, and it's
probably pretty harmless because ndeps can't get too large (I hope;
if it could, this would have O(N^2) performance problems).  Still,
you oughta fix it.

(There's a similar calculation later in the file that gets it right.)

			regards, tom lane