Thread

Re: Lost update in an ordered batch, but only with index scan

Richard Guo <guofenglinux@gmail.com> — 2025-12-29T08:12:15Z

On Mon, Dec 29, 2025 at 8:07 AM Bernice Southey
<bernice.southey@gmail.com> wrote:
> I've been following the recommendation of ordering batch updates to
> avoid deadlocks. I just added a new case, and it consistently loses
> updates in my tests.

Thanks for the report.  I think this is a bug.  git-bisect shows that:

cbc127917e04a978a788b8bc9d35a70244396d5b is the first bad commit
commit cbc127917e04a978a788b8bc9d35a70244396d5b
Author: Amit Langote <amitlan@postgresql.org>
Date:   Fri Feb 7 17:15:09 2025 +0900

    Track unpruned relids to avoid processing pruned relations

I didn't look further, but it seems like the same bug as reported in
[1].

[1] https://postgr.es/m/19355-57d7d52ea4980dc6@postgresql.org

- Richard