Feedback-ID: i0fe9450f:Fastmail
Date: Thu, 26 Mar 2026 11:22:25 +0900
From: Michael Paquier <michael@paquier.xyz>
To: Nathan Bossart <nathandbossart@gmail.com>
Cc: Andres Freund <andres@anarazel.de>,
	shihao zhong <zhong950419@gmail.com>,
	PostgreSQL-development <pgsql-hackers@postgresql.org>
Subject: Re: Fixes inconsistent behavior in vacuum when it processes multiple
 relations
Message-ID: <acSYYcbfDk0bSgGq@paquier.xyz>
References: <aFrSlSz9fi47j5Sr@nathan>
 <aFtRd1Q4QuCWcg5k@paquier.xyz>
 <aFwT31Rw-Oj-bgO9@nathan>
 <aGIzjYsRrMCEADc5@paquier.xyz>
 <rzxpxod4c4la62yvutyrvgoyilrl2fx55djaf2suidy7np5m6c@3l2ln476eadh>
 <ab2iXwhRe3fFNPfp@nathan>
 <aglpy2jbnb75vufdx6hqkgokquzmoshcbnuxlcxhum4k7tyi7m@27m4hbxo2u2g>
 <acMGcoUwA9CiPPU1@nathan>
 <loig4i4hc4vuybn6uv6e2kgqm37sebbzujkrvdf6uhbxq5dsfm@7wbkmucts3eu>
 <acQAXPXov4cbyUx0@nathan>
MIME-Version: 1.0
Content-Type: multipart/signed; micalg=pgp-sha512;
	protocol="application/pgp-signature"; boundary="JkLjxvxfux85pUqK"
Content-Disposition: inline
In-Reply-To: <acQAXPXov4cbyUx0@nathan>
Archived-At: 
 <https://www.postgresql.org/message-id/acSYYcbfDk0bSgGq%40paquier.xyz>
Precedence: bulk


--JkLjxvxfux85pUqK
Content-Type: text/plain; charset=us-ascii
Content-Disposition: inline
Content-Transfer-Encoding: quoted-printable

On Wed, Mar 25, 2026 at 10:33:48AM -0500, Nathan Bossart wrote:
> On Tue, Mar 24, 2026 at 11:29:52PM -0400, Andres Freund wrote:
> > I wonder if more of the code in the function should be updated to use t=
he
> > copy, otherwise it seems like it could more easily happen that a part o=
f the
> > code not using the modified version is moved until after a modification=
, and
> > the code author assumes the modifications now have taken effect.
>=20
> Yeah, I was waffling on this.  I updated the rest of the function to use
> params_local instead.

Fine by me with these extra forward declarations and the pointer
manipulations across these VACUUM/ANALYZE calls.  That would be
protective enough in terms of the original intention of these const
markers added in 2252fcd4276c.

Once vacuum_rel() has its VacuumParams changed to a pointer, renaming
it to params_local and use params_local in all its code paths looks
sensible.  On top of what you are doing, I would add a big warning
comment at the top of the routine to mention that "params" should not
be used directly in the routine.  The local copies should be used.
--
Michael

--JkLjxvxfux85pUqK
Content-Type: application/pgp-signature; name=signature.asc

-----BEGIN PGP SIGNATURE-----
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=poS9
-----END PGP SIGNATURE-----

--JkLjxvxfux85pUqK--