Patch Committed and Back-Patched
The primary development this round is that Alexander Korotkov pushed the fix on 2026-05-14. The commit covers master, with back-patches to PG17 and PG18 requiring minor test adjustments (different log messages and different default value of log_lock_waits).
Key exchanges
-
Michael Paquier (2026-05-14 06:58) conceded he was unable to complete his promised detailed review and explicitly gave Korotkov the green light: "feel free to go ahead as you feel."
-
Korotkov (2026-05-14 13:39) confirmed the commit is done. Noted that back-patches to PG18 and PG17 required minor TAP test revisions due to differing log messages and the default value of
log_lock_waitsbetween branches. -
Davydov (2026-05-14 16:48) acknowledged the commit and flagged a separate remaining issue with temp tables — referencing Tom Lane's earlier message (4075754.1774378690@sss.pgh.pa.us) about an additional problem discussed within this thread. He indicated intent to work on that next month as a new thread.
Net state
- Thread resolved. The bufmgr-level fix is committed to master and back-patched to PG17/PG18.
- PG14–16 back-patch was not done (as previously decided — the pre-existing zero-block edge case doesn't warrant late-cycle risk).
- A separate follow-up issue regarding temp tables (referenced by Tom Lane earlier in the thread) remains open and will be handled in a future thread by Davydov.