Feedback-ID: i0fe9450f:Fastmail
Date: Thu, 4 Jun 2026 13:47:32 +0900
From: Michael Paquier <michael@paquier.xyz>
To: Chao Li <li.evan.chao@gmail.com>
Cc: Postgres hackers <pgsql-hackers@lists.postgresql.org>,
	tfoertsch123@gmail.com, kanbayashi.dev@gmail.com
Subject: Re: SERVICEFILE shows wrong file after servicefile fallback
Message-ID: <aiEDZBg0Vwh0TClB@paquier.xyz>
References: <59DD1652-4EB5-449B-8DE3-959F06561A96@gmail.com>
MIME-Version: 1.0
Content-Type: multipart/signed; micalg=pgp-sha512;
	protocol="application/pgp-signature"; boundary="IuQ61MXkfUFTutKx"
Content-Disposition: inline
In-Reply-To: <59DD1652-4EB5-449B-8DE3-959F06561A96@gmail.com>
Archived-At: 
 <https://www.postgresql.org/message-id/aiEDZBg0Vwh0TClB%40paquier.xyz>
Precedence: bulk


--IuQ61MXkfUFTutKx
Content-Type: text/plain; charset=utf-8
Content-Disposition: inline
Content-Transfer-Encoding: quoted-printable

On Tue, Jun 02, 2026 at 09:42:23PM +0800, Chao Li wrote:
> While testing =E2=80=9Cpsql: Add variable SERVICEFILE=E2=80=9D, I found a=
 small
> issue where SERVICEFILE may show the wrong value.
>=20
> While tracing the code, I noticed there is a fallback path. If the
> service file specified in the connection string does not contain the
> requested service, libpq falls back to pg_service.conf. So I tested
> the following case:

Right, as of the default file in PGSYSCONFDIR.  This qualifies as an
open item for v19.

> I think the bug was actually introduced by the previous commit
> 092f3c63efc6. In parseServiceFile(), if the service file has already
> been set in the connection options, it refuses to update the value,
> so the fallback service file is not synced to the connection
> options. Then SERVICEFILE is read from the connection options, which
> still contain the original file specified on the command line. So,
> SERVICEFILE just makes the bug visible.=20

Yep, it looks like you are right here.  It does not make sense to show
in SERVICEFILE the file that has been skipped in favor of the second
default in PGSYSCONFDIR.  We need to show the latter.  An \echo of=20
SERVICEFILE is an interesting way to show your point.  Why not.

Thanks for the report, will fix.
--
Michael

--IuQ61MXkfUFTutKx
Content-Type: application/pgp-signature; name=signature.asc

-----BEGIN PGP SIGNATURE-----

iQIzBAEBCgAdFiEEG72nH6vTowiyblFKnvQgOdbyQH0FAmohA2QACgkQnvQgOdby
QH1h1g//dfU0fxfs2ltB1F0FXXOiVo7Zao794cRT+PBPp3aUohPGW1f+ckZ9pH9F
GOC6qyz6t0ZFxOyIHcCNR8Gedw4DxfmdW3Y2UQ8DryzYuxSOA7JNzScdLIpD8BKY
QCNvOBlqCanwNMY2NQqY2UkfwDACgETrYRoRHi9sqhuRgCIiCbbkF016RDeq9UnD
Sk0mBA5//CT/wlbooLKeEshVSfrnZ6lY8T7knO1s+0B9SGQ0FZ8a4jbHw4vSmjvG
p4vWG1W3/3BynjrZOT4n7U8kSEtY+9WeWgU/4GQ+NrCt+69L4oisCk9bLSjZLW1Q
m+CeK+ZAB8F5r8QS1JfFdogyPPZzYjQgP2kDCaW421JQ2V7TcVi9o3D3W7fz6/Rl
3GaIPt0/QDKYL6Fa7todTnzlU9OyeMDRCjychom2MqPOJxN1P1uoufCi+dSnlKm2
8YR+hw8wVO+6LtpZXHLU0tn294TiMVlVRDSfVhmaHzfM+JA4O8kg+FAa9OityNPt
Sl+jnMIDdiFfjxdW54zgPk0FjZfUqhEJCKIcI+mO3IdE8zzZ6TD+EWJ4o9WIOlD3
hK3x7KDwJVaafSIS1aFfRbBGXZkPHWz9I015ENmN6dsxaUUAwOVKEPR0iDapkZNY
JfVkiCTE+Hs4lMVe1MTJtca4OMgmSgmDCUNPFFGHAZwQvapsADM=
=7v36
-----END PGP SIGNATURE-----

--IuQ61MXkfUFTutKx--