MIME-Version: 1.0
References: 
 <CAHGQGwHhpizYzMo3nFP4GkNMueSNMY3QfC-gBN1VTXtuiANDvw@mail.gmail.com>
 <CAFC+b6qaCMoO7JZUx-etHJhinvPpQow8GKXZDXFfrCydeDtAdQ@mail.gmail.com>
In-Reply-To: 
 <CAFC+b6qaCMoO7JZUx-etHJhinvPpQow8GKXZDXFfrCydeDtAdQ@mail.gmail.com>
From: Fujii Masao <masao.fujii@gmail.com>
Date: Mon, 18 May 2026 19:47:15 +0900
Message-ID: 
 <CAHGQGwHz8aOjPbJd+mCb=OuYdfZQ-RkeatuQrfDCE4cqJxt69w@mail.gmail.com>
Subject: Re: pg_recvlogical: honor source cluster file permissions for output
 files
To: Srinath Reddy Sadipiralla <srinath2133@gmail.com>
Cc: PostgreSQL Hackers <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/CAHGQGwHz8aOjPbJd%2BmCb%3DOuYdfZQ-RkeatuQrfDCE4cqJxt69w%40mail.gmail.com>
Precedence: bulk

On Sun, May 17, 2026 at 12:22=E2=80=AFAM Srinath Reddy Sadipiralla
<srinath2133@gmail.com> wrote:
> i have reviewed, tested the patch and it LGTM,

Thanks for the review!


> i think we can add
> a TAP test to verify the group permission, attached a diff patch
> for the same,

Thanks for the patch!

The test you added checks that pg_recvlogical creates output files with
mode 0640 when the cluster is initialized with group access enabled.
However, it does not check the opposite case, i.e., that pg_recvlogical
creates output files with mode 0600 when group access is disabled.

It seems we should test both cases, similar to what 010_basebackup.pl does?

As far as I can tell, 010_basebackup.pl initializes the cluster without gro=
up
access and checks the backup permissions, then enables group access using
chmod_recursive() and verifies that group permissions are also applied to
the backup. I updated the TAP test following this approach and attached
a revised patch.


>> I think this should be backpatched to all supported branches.
>
>
> +1

I'm currently thinking of backpatching the fix itself to all supported bran=
ches,
but adding the test only to master. Because it does not seem worthwhile to
spend much time backporting the test to older branches, where the test code
differs much from master. Thought?

Regards,

--=20
Fujii Masao