MIME-Version: 1.0
References: 
 <CAOMoQbT0VmMZFRA68Geff9MScuGRTwokCf81O-0dqja8hmjNpQ@mail.gmail.com>
 <CAGKCzNBZO-AQMfmbfbn20wvE+63j50YuPBKJupYLiDCmc84OeA@mail.gmail.com>
 <CAOMoQbROhu7WG9LnyjHb1RXJrax_HKKMBbwaou+jps0uZgqHFw@mail.gmail.com>
 <CAOMoQbQzutP0UMU3KUZAE4A9Jf8bS=uSBRdUX7g=9xxoBbosmQ@mail.gmail.com>
 <CAGKCzNAQKMKJ_BQcF=uzcfcM=vx8QYnT+jtzuv_yp=76zbhn7g@mail.gmail.com>
 <CAGKCzNAW4BFq_cSKpnPvG6hRF65G3HptjUvWVW=DefPaPMF2pg@mail.gmail.com>
 <CAGKCzNAJzhzxJ809y9e6ttYbL0k6id5KmkzmBrz5xZROd79etQ@mail.gmail.com>
 <CAGKCzNC6RkY54TrkqqQ9K83AF=vrYe+CB14-PUSLTF5YMbzrmA@mail.gmail.com>
 <CAOMoQbT3zRuL+YLeiUrOG5-vYsvrZ2MJF_DQoa3E94cBfdv6xw@mail.gmail.com>
 <Z6Is1o56psB3-TMT@msg.df7cb.de>
In-Reply-To: <Z6Is1o56psB3-TMT@msg.df7cb.de>
From: Bradford Boyle <bradford.d.boyle@gmail.com>
Date: Tue, 4 Feb 2025 23:40:16 -0800
Message-ID: 
 <CAOMoQbTA2qOo+f_4JE+oYGupZANh-0msn_Pw36qb7v3PfT1e0w@mail.gmail.com>
Subject: Re: PgBouncer 1.24.0 - New upstream version
To: Christoph Berg <cb@df7cb.de>, Bradford Boyle <bradford.d.boyle@gmail.com>,
	pgsql-pkg-debian@postgresql.org
Content-Type: text/plain; charset="UTF-8"
Content-Transfer-Encoding: quoted-printable
Archived-At: 
 <https://www.postgresql.org/message-id/CAOMoQbTA2qOo%2Bf_4JE%2BoYGupZANh-0msn_Pw36qb7v3PfT1e0w%40mail.gmail.com>
Precedence: bulk

On Tue, Feb 4, 2025 at 7:06=E2=80=AFAM Christoph Berg <cb@df7cb.de> wrote:
>
> Unfortunately that needs more work as it looks like:
>
> bookworm and others:
>
> 14:54:22 E   AttributeError: 'PGconn' object has no attribute 'close_prep=
ared'
>
> https://jengus.postgresql.org/job/pgbouncer-binaries/106/architecture=3Da=
md64,distribution=3Dbookworm/console

Yeah -- it looks like close_prepared was added in psycopg 3.2.0 and
bookworm has 3.1.7. I could patch the tests to include a skipif based on
the version of psycopg.

> bullseye doesn't have python3-psycopg (and building that doesn't look
> trivial). Do we think we could go without or with less tests there?

I think we could go without tests on bullseye since it seems like we are
nearing the trixie release and bullseye will become old old stable.
(Correct me if I am wrong on this point, I'm making this assumption
based on the trixie freeze dates). There's probably little value in
creating extra work here.

It looks like d/rules for pgbouncer will skip the tests if nocheck is in
DEB_BUILD_OPTIONS -- is this something that can be configured via the Jenki=
ns
YAML? I am not very familiar with the Jenkins job builder YAML and I didn't=
 see
any usages of that in the current set of jobs. Or would I need to update
generate-pgdg-source to include an additional tweak to disable the tests on
older platforms?

Thanks,
-- Bradford