Re: Assertion failure in WAL summarizer

Hi Sai Pranav,

On Mon, Nov 2, 2026 at 11:56 PM Raja Sai pranav
<rajasaipranav0(at)gmail(dot)com> wrote:
> I tried the series (v2-0001 from upthread plus your v1-0001 and v1-0002) on
> current master.

Thanks for testing it on aarch64. v3 is attached, rebased on
f549ecd2660 and 316861101e8; v1's t/014 is dropped, as f549ecd2660
covers that case.

> - pg_walsummary's 004_promote_past_target times out in its setup with 0002
> applied, as you said: the standby never summarizes past the target.

0001 keeps the v2 test, which trips the assertion on master. 0003
(formerly 0002) makes it promote while the summarizer waits at replay,
since the v2 check is then only defensive.

On today's master without cassert, Shihao's test (now t/015) still
restores 4|0 instead of 4|1, 3 of 3; with the series it passes 3 of 3.
0003 now lists Shihao as co-author, since the idea and t/015 are his.

Fujii, v3 is on top of what you committed, in case it helps.

Robert, for 0003: does anything rely on summaries ahead of replay?

Regards,
Manu