Re: [PATCH] Fix LISTEN startup race with direct advancement

From: "Joel Jacobson" <joel(at)compiler(dot)org>
To: "Arseniy Mukhin" <arseniy(dot)mukhin(dot)dev(at)gmail(dot)com>
Cc: pgsql-hackers <pgsql-hackers(at)postgresql(dot)org>
Subject: Re: [PATCH] Fix LISTEN startup race with direct advancement
Date: 2026-05-20 11:20:05
Message-ID: e46886ff-4a49-47fb-a76f-3b2f012f1948@app.fastmail.com
Views: Whole Thread | Raw Message | Download mbox | Resend email
Thread:
Lists: pgsql-hackers

On Wed, May 20, 2026, at 04:01, Arseniy Mukhin wrote:
> One point - looks like the 0003 contains the same test as 0001.

They are similar, but if you look carefully you'll see that they use
different injection points:

0001 uses async-notify-before-listen-commit
0002 uses async-notify-after-listen-commit

Another different is that if you only apply 0001 and run the tests,
that test will fail without the 0003 fix, whereas 0002 pass both
with and without 0003, since it just tests and demonstrates the
already documented quite harmless false positive race condition.

> Also while I was trying to understand the issue
...
> diagram in the head of the file). It depends on 0001. Your patch fixes
> it too.

OK, I didn't look in detail at untriggered-notification.patch.nocfbot,
but good to hear 0003 fixes it too.

Thanks for reviewing.

/Joel

In response to

Responses

Browse pgsql-hackers by date

  From Date Subject
Next Message Alexander Korotkov 2026-05-20 11:46:30 Re: Fix SPLIT PARTITION bound-overlap bug and other improvements
Previous Message Arseniy Mukhin 2026-05-20 11:01:52 Re: [PATCH] Fix LISTEN startup race with direct advancement