Mar - Apr 2026

9 commits
Fix injection point detach timing problem in TAP test for lock stats
7 Apr 2026Michael Paquiermaster19beta1REL_19_STABLE19beta1
LLM analysis, 100% surelanded as postedfrom patchset #570 of 1 reviewer saw the final version

The additions and deletions in src/test/modules/test_misc/t/011_lock_stats.pl match #57 after removing whitespace differences; #58 confirms the fix was committed after repeated CI runs.

  • Andres Freundsaw an earlier versionLLM analysis, 95% sureReported the detach failure in the landed test from #49 at #55, before the synchronization fix in #57.
Use single LWLock for lock statistics in pgstats
6 Apr 2026Michael Paquiermaster19beta1REL_19_STABLE19beta1
LLM analysis, 100% suremodified before commitlight changesfrom patchset #521 of 2 reviewers saw the final version

The pgstat_lock.c changes match #52 exactly; the commit simplifies the PgStatShared_Lock protection comment in pgstat_internal.h, as confirmed in #54.

  • Tomas Vondrasaw an earlier versionLLM analysis, 99% sureIdentified lost pending statistics and questioned per-entry locking in #51, prompting the callback and locking changes in #52.
  • Michael Paquiersaw the final versioncommittercommitted the patch
Add tests for lock statistics, take two
5 Apr 2026Michael Paquiermaster19beta1REL_19_STABLE19beta1
LLM analysis, 100% suremodified before commitlight changesfrom patchset #490 of 1 reviewer saw the final version

Matches the three-file TAP-test submission #49, including removal of on_error_stop; remaining differences are Perl formatting and a terminating semicolon in wait_for_pg_stat_lock().

  • Andres Freundsaw an earlier versionLLM analysis, 94% sureReported widespread Windows failures and challenged the fixed sleeps in #41, leading to replacement of the isolation tests with the TAP implementation.
Remove isolation test lock-stats
24 Mar 2026Michael Paquiermaster19beta1REL_19_STABLE19beta1
LLM analysis, 100% surewritten from the discussion, no matching patch0 of 1 reviewer saw the final version

No attached patch proposes this removal; #41 reports Windows instability and #42 announces removing the test, matching deletion of its spec, expected output and schedule entry.

  • Andres Freundsaw an earlier versionLLM analysis, 99% sureReported that ten of sixteen Windows CI runs failed and explained the sleep-based test's instability in #41; the reviewed isolation implementation predates the thread's final TAP version.
Avoid including clog.h in proc.h
24 Mar 2026Álvaro Herreramaster19beta1REL_19_STABLE19beta1
LLM analysis, 100% surepart of a posted patchsetlight changesfrom patchset #36

Takes only the fifteenth patch of #36: moving access/clog.h into proc.c and adding an XidStatus typedef in proc.h; its added and removed text is unchanged.

Don't include storage/lock.h in so many headers
24 Mar 2026Álvaro Herreramaster19beta1REL_19_STABLE19beta1
LLM analysis, 100% surepart of a posted patchsetlight changesfrom patchset #361 of 1 reviewer saw the final version

Combines the first fourteen patches of #36, leaving the clog.h cleanup separate; edits remove the redundant hsearch.h include in wait_event.c and sort forward declarations in twophase.h.

  • Bertrand Drouvotsaw the final versiontrailerTraced the affected include dependencies and flagged the redundant hsearch.h include in #37, which Alvaro removed before committing in #40.
Add support for lock statistics in pgstats
24 Mar 2026Michael Paquiermaster19beta1REL_19_STABLE19beta1
LLM analysis, 100% surepart of a posted patchsetlight changesfrom patchset #331 of 2 reviewers saw the final version

Lands the statistics and view portions of #33, with the header extraction and preparatory code movement committed separately; edits simplify the counter macro into a function, polish documentation/tests, and bump catalog and statistics formats.

  • Andres Freundsaw an earlier versiontrailerChallenged acquisition-path overhead and the extra counters/GUC in #17, #20, #24 and #26, before later changes to counter semantics and logging independence.
  • Michael Paquiersaw the final versiontrailerReviewed counter selection, timing, header dependencies and tests throughout the thread, then tested and committed the final implementation as described in #39.
Move some code blocks in lock.c and proc.c
24 Mar 2026Michael Paquiermaster19beta1REL_19_STABLE19beta1
LLM analysis, 100% surepart of a posted patchsetlight changesfrom patchset #331 of 1 reviewer saw the final version

Extracts only the LockAcquireExtended() and ProcSleep() block rearrangements from #33, omitting the statistics calls; #39 explicitly confirms this preparatory split.

  • Michael Paquiersaw the final versiontrailerReviewed the final series and extracted its lock.c/proc.c rearrangements into this commit to reduce feature-patch churn, as explained in #39.
Move declarations related to locktags from lock.h to new locktag.h
21 Mar 2026Michael Paquiermaster19beta1REL_19_STABLE19beta1
LLM analysis, 100% surepart of a posted patchsetmoderate changesfrom patchset #331 of 1 reviewer saw the final version

Lands the header-extraction portion of #33, expanded to move LOCKMETHODID, the lock-method constants and SET_LOCKTAG_* macros into locktag.h; #34 explains this expansion.

  • Michael Paquiersaw the final versiontrailerProposed the header split in #32, reviewed and expanded the attached extraction, and confirmed committing it in #34.

Credits across 9 commits

Author
Bertrand Drouvot5
Alvaro Herrera2
Michael PaquierLLM1
Reviewer
Michael Paquiercommitter4
Andres Freund1
Bertrand Drouvot1
Reported by
Andres Freund1
Tomas Vondra1
Co author
Michael Paquier1
Committer
Michael Paquier7
Alvaro Herrera2

N LLM means N of the credits come from LLM analysis, not from commit trailers.