Sep 2026

1 commit
Fix sequence synchronization failure on a concurrent refresh.
23 Sep 2026Amit KapilamasterunreleasedREL_19_STABLE19rc1
LLM analysis, 100% surelanded as postedfrom patchset #812 of 4 reviewers saw the final version

The subscription-membership check, locking, COPYSEQ_NOT_SUBSCRIBED handling, and subscriptioncmds.c comment match #81, modulo whitespace and base context.

  • shveta maliksaw the final versiontrailerShveta challenged stopping a worker shared by many sequences in #78 and verified the final approach in #82.
  • Zhijie Housaw an earlier versiontrailerZhijie reviewed #79 in #80, prompting subsequent local changes to the membership lookup and relation-lock placement.
  • Andrey Borodinsaw an earlier versiontrailerAndrey reproduced the failure and tested the worker-stop implementation in #77, before it was replaced by membership checking.
  • Amit Kapilasaw the final versioncommittercommitted the patch

Jul 2026

9 commits
Improve wording of sequence origin warning in logical replication.
31 Jul 2026Amit KapilamasterunreleasedREL_19_STABLE19beta3
LLM analysis, 100% surelanded as postedfrom patchset #711 of 1 reviewer saw the final version

All five added and five removed lines match #71: sequence-specific origin diagnostics in subscriptioncmds.c and the sequencesync.c typo correction.

  • Amit Kapilasaw the final versiontrailerAmit committed #71, establishing final review coverage despite no separate patch-review reply in the thread.
Avoid accumulating relation locks during sequence synchronization.
29 Jul 2026Amit KapilamasterunreleasedREL_19_STABLE19beta3
LLM analysis, 100% surepart of a posted patchsetlight changesfrom patchset #632 of 2 reviewers saw the final version

Lands the sequencesync.c implementation from #63 with an expanded lock-lifetime comment, omitting the separate lock-exhaustion reproducer as the sender proposed.

  • Hayato Kurodasaw the final versiontrailerHayato reproduced the failure, verified the fix, requested a clearer comment in #64, and approved the revised wording in #66.
  • Amit Kapilasaw the final versiontrailerAmit supplied the expanded explanation in #65 and confirmed committing the fix in #67.
Fix issues in logical replication sequence synchronization.
27 Jul 2026Amit KapilamasterunreleasedREL_19_STABLE19beta3
LLM analysis, 100% surelanded as postedfrom patchset #602 of 2 reviewers saw the final version

The six-file diff matches #60, including maybe_reread_subscription(), slot-drop reordering, timestamp initialization, psql completion, and documentation.

  • Hayato Kurodasaw the final versiontrailerHayato proposed rereading the subscription instead of terminating workers from the launcher in #59 and approved the revised patch in #61.
  • Amit Kapilasaw the final versiontrailerAmit stated that the revised patch looked good and had been pushed in #62.
Reject sequence synchronization against pre-PostgreSQL 19 publishers.
23 Jul 2026Amit KapilamasterunreleasedREL_19_STABLE19beta3
LLM analysis, 100% surelanded as postedfrom patchset #563 of 3 reviewers saw the final version

The four-file diff matches #56 modulo rebase; that attachment retains Vignesh's author header and explicit Author credit although Amit sent it.

  • Shveta Maliksaw the final versiontrailerShveta reviewed validation placement and documentation, corrected the function-version claim in #51, and accepted worker-side errors in #54.
  • Hayato Kurodasaw the final versiontrailerHayato exposed the ALTER SERVER bypass in #49 and verified the revised version checks in #55.
  • Amit Kapilasaw the final versiontrailerAmit reviewed validation placement and supplied the final consistent error codes and comments in #56 before committing.
Allow logical replication workers to ignore default_transaction_read_only.
21 Jul 2026Amit KapilamasterunreleasedREL_19_STABLE19beta3
LLM analysis, 87% surelanded as postedfrom patchset #421 of 1 reviewer saw the final version

The diff, also confirmed in git, matches the sequencesync-only override and test from #22, corresponding to the PostgreSQL 19 variant explicitly resent in #42. Those later attachments were not extracted; despite its broader commit message, the supplied canonical commit changes SequenceSyncWorkerMain().

  • Amit Kapilasaw the final versiontrailerAmit requested the narrow backpatch in #19, settled the GUC approach in #26, and confirmed commitment in #44.
Handle concurrent sequence refreshes.
20 Jul 2026Amit KapilamasterunreleasedREL_19_STABLE19beta3
LLM analysis, 100% suremodified before commitlight changesfrom patchset #363 of 3 reviewers saw the final version

The executable changes match #36; the explanation of the subscription object lock is reworded and rewrapped. The separately proposed injection-point test in #37 is absent.

  • vignesh Csaw the final versiontrailerVignesh verified the fix with a debugger on both branches and supplied an injection-point test in #37.
  • Shveta Maliksaw the final versiontrailerShveta tested worker termination, batch completion, and restart interlocks and reported that the patch looked good in #38.
  • Hayato Kurodasaw the final versiontrailerHayato confirmed lock-wait behavior, examined interrupt handling, and approved the main code in #41.
Revert "Reject concurrent sequence refreshes".
18 Jul 2026Amit KapilamasterunreleasedREL_19_STABLE19beta3reverts 1 commit here
LLM analysis, 100% surewritten from the discussion, no matching patch0 of 1 reviewer saw the final version

This exactly reverses the worker-presence rejection and associated test wait; Amit proposed the standalone revert in #35 and confirmed pushing it in #36. No standalone revert patch was attached.

  • Tom Lanesaw an earlier versionLLM analysis, 99% sureTom reproduced failures in #28 and argued in #30 and #33 that requiring users to avoid worker timing was unacceptable, leading to the revert; substantial redesign followed before #81.
Handle concurrent sequence drops during synchronization
16 Jul 2026Fujii MasaomasterunreleasedREL_19_STABLE19beta3Also in: Fix publisher-side sequence permission reporting (modified before commit)
Reject concurrent sequence refreshes.
15 Jul 2026Amit KapilamasterunreleasedREL_19_STABLE19beta3
LLM analysis, 100% surepart of a posted patchsetlight changesfrom patchset #223 of 3 reviewers saw the final version

Lands only the concurrency component of the three-patch submission #22, adding a translator comment to its otherwise unchanged subscriptioncmds.c logic and TAP wait.

  • Shveta Maliksaw the final versiontrailerShveta supplied the explanatory comment in #13 and approved the revised concurrency component in #16, after which only light changes followed.
  • Hayato Kurodasaw the final versiontrailerHayato requested the translator comment and analyzed a worker-start interleaving in #23.
  • Amit Kapilasaw the final versiontrailerAmit requested a shorter error message and cleaner patch separation in #21 and committed the revised component.

Credits across 10 commits

Author
vignesh C8
Amit Kapila2
Reviewer
Amit Kapilacommitter7
Hayato Kuroda (Fujitsu)6
shveta malik4
Andrey Borodin1
Fujii Masao1
vignesh C1
Zhijie Hou (Fujitsu)1
Zsolt Parragi1
Reported by
Noah Misch8
Nikolay Samokhvalov1
Peter Smith1
Tom Lane1
Committer
Amit Kapila9
Fujii Masao1