Re: missing PGDLLIMPORT in utils/acl.h after CVE-related "Invalidate plan cache after role changes"

From: Nathan Bossart <nathandbossart(at)gmail(dot)com>
To: Anton Voloshin <a(dot)voloshin(at)postgrespro(dot)ru>
Cc: Postgres hackers <pgsql-hackers(at)lists(dot)postgresql(dot)org>, Noah Misch <noah(at)leadboat(dot)com>
Subject: Re: missing PGDLLIMPORT in utils/acl.h after CVE-related "Invalidate plan cache after role changes"
Date: 2026-08-12 14:10:07
Message-ID: anx-v4VSFUe1-2cJ@nathan
Views: Whole Thread | Raw Message | Download mbox | Resend email
Thread:
Lists: pgsql-hackers

Thanks for the report.

On Wed, Aug 12, 2026 at 11:52:46AM +0300, Anton Voloshin wrote:
> So I suggest an obvious fix, which is needed in REL_14_STABLE..master,
> attached.

Yeah, that's an oversight on my part. IIUC this only affects Windows
extensions that happen to need the variable, and since it's never been
exported in any release thus far, no extension should depend on it yet.
Still, marking is the policy, and I see no reason for an exception here.

Barring objections, I'll plan on applying this change soon.

> P.S. Perhaps CI or buildfarm member running src/tools/mark_pgdllimport.pl
> would be a good idea?

FWIW I ran the script over all of src/include on master, and it flags
exactly two lines: this one, and __pg_log_level in common/logging.h,
which is deliberately unmarked (commit 8d33412665). So a check like
that seems pretty feasible, provided we have some way to note the
exceptions.

--
nathan

In response to

Browse pgsql-hackers by date

  From Date Subject
Next Message Matheus Alcantara 2026-08-12 14:23:29 Re: hashjoins vs. Bloom filters (yet again)
Previous Message Jonathan Gonzalez V. 2026-08-12 13:36:01 Re: [PATCH]Fix pg_xact corruption from subtransaction abort after subcommit