Date: Mon, 20 Sep 2021 21:32:09 -0700
From: Andres Freund <andres@anarazel.de>
To: pgsql-committers@lists.postgresql.org,
 Michael Paquier <michael@paquier.xyz>
Subject: Re: pgsql: Support "postgres -C" with runtime-computed GUCs
User-Agent: K-9 Mail for Android
In-Reply-To: <YUlVzPkVgEPjtmIB@paquier.xyz>
References: <E1mQglq-0007xB-Jw@gemulon.postgresql.org>
 <20210921032040.lyl4lcax37aedx2x@alap3.anarazel.de>
 <YUlVzPkVgEPjtmIB@paquier.xyz>
Message-ID: <216D2155-018C-4FFB-8FE3-73CAA5A3FEE1@anarazel.de>
MIME-Version: 1.0
Content-Type: text/plain;
 charset=utf-8
Content-Transfer-Encoding: quoted-printable
Archived-At: 
 <https://www.postgresql.org/message-id/216D2155-018C-4FFB-8FE3-73CAA5A3FEE1%40anarazel.de>
Precedence: bulk

Hi,=20

On September 20, 2021 8:47:24 PM PDT, Michael Paquier <michael@paquier=2Ex=
yz> wrote:
>On Mon, Sep 20, 2021 at 08:20:40PM -0700, Andres Freund wrote:
>> So the problem is that -C data_checksums requires permissions to be dro=
pped,
>> but the way it's invoked here that's not guaranteed=2E=2E=2E And I don'=
t think we
>> should remove that requirement, given that we're actually doing stuff t=
o the
>> lock file=2E=2E=2E
>
>Hmm=2E  One way I can think of to keep the test would be something like
>that to avoid problems with pg_ctl redirecting the server's stderr to
>stdout (hence the tweak for log_min_messages):
>pg_ctl start -s -o '-C data_checksums -c log_min_messages=3Dfatal'

Yes, that should work=2E Although error should suffice=2E

FWIW, I think it'd be good to hide that message in the -C case=2E


>This would print the parameter, while getting to stderr mostly a
>"could not start server"=2E  Would that work for you?

Yes, I think so=2E

Andres
--=20
Sent from my Android device with K-9 Mail=2E Please excuse my brevity=2E