MIME-Version: 1.0
References: 
 <CAAZLFmRK+XFp=mqCeruyNVkqGq5mH45CP+e-8oNttPRtLuB5eQ@mail.gmail.com>
 <CAKFQuwbVK=oANp9XLdQknd7o9jJOUFQanUtG3QAh-3kuaM5tJw@mail.gmail.com>
 <CAAZLFmSKrcovMSauueWg=8VdZN-fHPH2ZengZRKA8AfgST6o6w@mail.gmail.com>
 <CAKFQuwaCPkb8=2_V3121Ei0MBrg02FLRNQbcAcyzCO3dAha=Gw@mail.gmail.com>
In-Reply-To: 
 <CAKFQuwaCPkb8=2_V3121Ei0MBrg02FLRNQbcAcyzCO3dAha=Gw@mail.gmail.com>
From: Robert Treat <rob@xzilla.net>
Date: Wed, 2 Jul 2025 00:31:05 -0400
Message-ID: 
 <CABV9wwOOpOijL9sY03KV3WNvgMxVFsyfvfYbfy2MYfn40z0yyQ@mail.gmail.com>
Subject: Re: [PATCH] initdb: Treat empty -U argument as unset username
To: "David G. Johnston" <david.g.johnston@gmail.com>
Cc: Jianghua Yang <yjhjstz@gmail.com>, pgsql-hackers@lists.postgresql.org
Content-Type: text/plain; charset="UTF-8"
Content-Transfer-Encoding: quoted-printable
Archived-At: 
 <https://www.postgresql.org/message-id/CABV9wwOOpOijL9sY03KV3WNvgMxVFsyfvfYbfy2MYfn40z0yyQ%40mail.gmail.com>
Precedence: bulk

On Wed, Jul 2, 2025 at 12:01=E2=80=AFAM David G. Johnston
<david.g.johnston@gmail.com> wrote:
> On Tue, Jul 1, 2025 at 8:31=E2=80=AFPM Jianghua Yang <yjhjstz@gmail.com> =
wrote:
>>
>> git show 8e673801262c66af4a54837f63ff596407835c20
>>
>>
>>         effective_user =3D get_id();
>>
>> -       if (strlen(username) =3D=3D 0)
>>
>> +       if (!username)
>>
>>                 username =3D effective_user;
>>
>>
>> The previous code already intended to treat a missing username as fallin=
g back to the system user.
>> The check was changed from strlen(username) =3D=3D 0 to !username, but t=
his inadvertently stopped handling the empty-string case. This patch restor=
es the original intent and makes the behavior consistent.
>>>
>>>
>
> At this point I'd rather take advantage of this behaveing in the "doesn't=
 work" category for the past 8 years, and thus all supported releases, and =
not change existing behavior (just improve the error message) rather than a=
ccept original intent.  Also, the amount of things it has to be consistent =
with is quite small and I'm pytr sure that some of those are also broken - =
encoding/pgdata/textsearch all exhibit the same pattern (xlog is the revers=
e so maybe ok...)
>

FWIW, I tend to agree with David; I feel like if a user passes in -U,
there was probably a reason, and a good error message would be more
useful in clarifying things rather than blindly pushing forward with
potentially the wrong thing.


Robert Treat
https://xzilla.net