Re: GUC parameter ACLs and physical walsender

Mark Dilger <mark.dilger@enterprisedb.com>

From: Mark Dilger <mark.dilger@enterprisedb.com>
To: Jeff Davis <pgsql@j-davis.com>
Cc: John Naylor <johncnaylorls@gmail.com>, pgsql-hackers@postgresql.org, Andrey Borodin <x4mmm@yandex-team.ru>
Date: 2026-04-23T23:41:13Z
Lists: pgsql-hackers
On Thu, Apr 23, 2026 at 2:06 PM Jeff Davis <pgsql@j-davis.com> wrote:

> On Thu, 2026-04-23 at 10:57 -0700, Mark Dilger wrote:
> > John is right that the hardcoded BKI_ROWTYPE_OID(2173) makes this
> > non-backportable as-is.
>
> Right, but that leaves the questions:
>
> (a) Is this the right fix for master?
>

Yes.  This approach has no problem in master that I can see.


> (b) Is there anything we can do in the back branches, or we just leave
> it as fix going forward only?
>

I don't see a solution.  We could try to replace the error message with
something better, but even that seems hard to phrase.  Replacing
"cannot read pg_class without having selected a database" with, say,
"permission denied" would also be confusing for a role which does have
the privilege but just can't verify it.

-- 

*Mark Dilger*

Commits

Same data as JSON: GET /api/v1/messages/:b64id/commits the thread's linked commits as JSON, with link sources. API reference →
  1. Nail pg_parameter_acl in relcache.

  2. catcache.c: use C_COLLATION_OID for texteqfast/texthashfast.