Mar 2022

1 commit
pg_upgrade: Upgrade an Assert to a real 'if' test
23 Mar 2022Alvaro Herreramaster15beta1REL_19_STABLE15beta1REL_18_STABLE15beta1REL_17_STABLE15beta1REL_16_STABLE15beta1REL_15_STABLE15beta1REL_14_STABLE14.3REL_13_STABLE13.7+3 more
LLM analysis, 100% surepart of a posted patchsetlight changesfrom patchset #39

The src/bin/pg_rewind/parsexlog.c hunk exactly matches the corrected assertion-conversion component of #39; the other series components are absent. Despite the commit subject saying pg_upgrade, its diff changes pg_rewind.

Nov 2021

1 commit
Fix determination of broken LSN in OVERWRITTEN_CONTRECORD
26 Nov 2021Alvaro Herreramaster15beta1REL_19_STABLE15beta1REL_18_STABLE15beta1REL_17_STABLE15beta1REL_16_STABLE15beta1REL_15_STABLE15beta1REL_14_STABLE14.2REL_13_STABLE13.6+3 more
LLM analysis, 97% surewritten from the discussion, no matching patch1 of 2 reviewers saw the final version

The proposal in #116 changes abortedRecPtr and relaxes verification, but neither hunk landed. Instead, #120 explains the committed alternative: assign overwrittenRecPtr from RecPtr in xlogreader.c.

  • Tom Lanesaw the final versionLLM analysis, 95% sureReported the page-boundary failure in #108 and argued in #115 that the existing WAL value was correct, leading to the reader-side correction explained in #120.
  • Amul Sulsaw an earlier versionLLM analysis, 94% sureIdentified the inconsistent pointer assignments in #110 and questioned the compatibility adjustment in #119, although the final fix changed the opposite assignment from his initial suggestion.

Nov 2021

6 commits
Fix instability in 026_overwrite_contrecord.pl test.
9 Nov 2021Tom Lanemaster15beta1REL_19_STABLE15beta1REL_18_STABLE15beta1REL_17_STABLE15beta1REL_16_STABLE15beta1REL_15_STABLE15beta1REL_14_STABLE14.2REL_13_STABLE13.6+3 more
LLM analysis, 100% surewritten from the discussion, no matching patch2 of 2 reviewers saw the final version

Implements Tom Lane's inline proposal in #105: disable autovacuum in 026_overwrite_contrecord.pl and compare WAL filenames with Perl's string operator ne.

  • Andres Freundsaw the final versionLLM analysis, 83% sureDiagnosed autovacuum-generated WAL and explicitly raised disabling autovacuum in #88, the remedy ultimately implemented after the continuing failures in #105.
  • Alvaro Herrerasaw the final versionLLM analysis, 88% sureRecommended disabling autovacuum regardless of other test changes in #89, anticipating the configuration fix implemented from #105.
Fix typo
8 Nov 2021Alvaro HerreraREL_13_STABLE13.5
LLM analysis, 99% surewritten from the discussion, no matching patch1 of 1 reviewer saw the final version

This stable-branch correction comes from #103, with translation preservation added by Alvaro as explained in #104; no indexed patchset contains this complete change.

  • Alvaro Herrerasaw the final versioncommittercommitted the patch
Fix typo
8 Nov 2021Alvaro HerreraREL_11_STABLE11.14
LLM analysis, 99% surewritten from the discussion, no matching patch1 of 1 reviewer saw the final version

The coauthor credit and Discussion trailer identify Alexander's request in #103; the diff corrects successfully in xlog.c, the TAP test and translation catalogs.

  • Alvaro Herrerasaw the final versioncommittercommitted the patch
Fix typo
8 Nov 2021Alvaro HerreraREL_10_STABLE10.19
LLM analysis, 99% surewritten from the discussion, no matching patch1 of 1 reviewer saw the final version

Implements the stable-branch spelling correction requested in #103, including matching TAP and translation updates described in #104; the request is not an indexed patchset.

  • Alvaro Herrerasaw the final versioncommittercommitted the patch
Fix typo
8 Nov 2021Alvaro HerreraREL_12_STABLE12.9
LLM analysis, 99% surewritten from the discussion, no matching patch1 of 1 reviewer saw the final version

The commit credits Alexander Lakhin, whose correction request in #103 has no indexed patchset. It fixes the stable-branch log message, TAP expectation and translation msgids, matching Alvaro's explanation in #104.

  • Alvaro Herrerasaw the final versioncommittercommitted the patch
Fix typo
8 Nov 2021Alvaro HerreraREL9_6_STABLE9.6.24
LLM analysis, 99% surewritten from the discussion, no matching patch1 of 1 reviewer saw the final version

Corrects the same stable-branch spelling, test expectation and translation msgids requested in #103. Alvaro explicitly acknowledges an incorrect originating-commit attribution for the 9.6 correction in #104.

  • Alvaro Herrerasaw the final versioncommittercommitted the patch

Sep - Oct 2021

5 commits
Change recently added test code for stability
13 Oct 2021Alvaro Herreramaster15beta1REL_19_STABLE15beta1REL_18_STABLE15beta1REL_17_STABLE15beta1REL_16_STABLE15beta1REL_15_STABLE15beta1REL_14_STABLE14.1REL_13_STABLE13.5+4 more
LLM analysis, 100% surelanded as postedfrom patchset #931 of 1 reviewer saw the final version

The diff exactly matches #93, replacing the estimated WAL-filling INSERT with a DO loop in 026_overwrite_contrecord.pl; #96 confirms it was pushed.

  • Andres Freundsaw the final versionLLM analysis, 100% sureDiagnosed the failure and proposed the loop and wider randomized rows in #88 and #90, then verified the posted patch under valgrind in #95.
Remove unstable, unnecessary test; fix typo
1 Oct 2021Alvaro Herreramaster15beta1REL_19_STABLE15beta1REL_18_STABLE15beta1REL_17_STABLE15beta1REL_16_STABLE15beta1REL_15_STABLE15beta1REL_14_STABLE14.1
LLM analysis, 100% surelanded as postedfrom patchset #761 of 1 reviewer saw the final version

Exactly matches #76: removes the archive-based test and idiosyncratic_copy, and corrects the log message and remaining test expectation.

  • Tom Lanesaw the final versionLLM analysis, 100% sureReported the missing log message and spelling error in #75, then explicitly endorsed removing the redundant test in #77.
Repair two portability oversights of new test
30 Sep 2021Alvaro Herreramaster15beta1REL_19_STABLE15beta1REL_18_STABLE15beta1REL_17_STABLE15beta1REL_16_STABLE15beta1REL_15_STABLE15beta1REL_14_STABLE14.1Also in: pgsql: Fix WAL replay in presence of an incomplete record
LLM analysis, 100% surewritten from the discussion, no matching patch4 of 4 reviewers saw the final version

The two changes—calling set_replication_conf and uncommenting the IPC::Run finish call—come directly from #67, #68 and #69, with implementation promised in #70.

  • Tom Lanesaw the final versionLLM analysis, 100% sureReported the Windows replication-authentication failure in #62 and endorsed the exact set_replication_conf remedy in #68.
  • Andrew Dunstansaw the final versionLLM analysis, 100% sureIdentified the missing replication setup in #63 and supplied precedent for the adopted method call in #67.
  • Andres Freundsaw the final versionLLM analysis, 100% sureReported the Windows hang and verified that uncommenting the IPC::Run finish call resolved it in #69.
  • Michael Paquiersaw the final versionLLM analysis, 97% sureThe message of d03bca4d70 explicitly credits him with identifying the Windows replication-configuration omission, although he has no message in this thread.
Fix WAL replay in presence of an incomplete record
29 Sep 2021Alvaro Herreramaster15beta1REL_19_STABLE15beta1REL_18_STABLE15beta1REL_17_STABLE15beta1REL_16_STABLE15beta1REL_15_STABLE15beta1REL_14_STABLE14.1REL_13_STABLE13.5+4 more
LLM analysis, 97% suremodified before commitlight changesfrom patchset #580 of 2 reviewers saw the final version

The final core logic matches #58, whose retained 9.6 attachment also matches the listed backport apart from formatting, comments and typedef registration. The development-branch commit retains the additional archive test from #54, with test adjustments and branch-specific API differences.

  • Kyotaro Horiguchisaw an earlier versiontrailerReviewed error handling and flag checks in #29 and caught incorrect page-flag placement in #44, before subsequent verification and standby-mode logic changes.
  • Nathan Bossartsaw an earlier versiontrailerFound boundary-tracking defects in #3 and #5, then tested and reviewed the replacement in #47 and #50 before the standby-mode fix.
Document XLOG_INCLUDE_XID a little better
21 Sep 2021Alvaro Herreramaster15beta1REL_19_STABLE15beta1REL_18_STABLE15beta1REL_17_STABLE15beta1REL_16_STABLE15beta1REL_15_STABLE15beta1REL_14_STABLE14.0
LLM analysis, 100% surepart of a posted patchsetlight changesfrom patchset #54

Selects only the XLOG_INCLUDE_XID documentation component of #54; its three file changes match modulo comment whitespace.

Sep 2021

1 commit
Revert "Avoid creating archive status ".ready" files too early"
4 Sep 2021Alvaro Herreramaster15beta1REL_19_STABLE15beta1REL_18_STABLE15beta1REL_17_STABLE15beta1REL_16_STABLE15beta1REL_15_STABLE15beta1REL_14_STABLE14rc1REL_13_STABLE13.5+4 morerevert
LLM analysis, 100% surepart of a posted patchsetlight changesfrom patchset #352 of 2 reviewers saw the final version

The eight-file diff exactly matches the revert component of #35, excluding the replacement implementation and auxiliary patches; #38 confirms the revert was pushed.

  • Andres Freundsaw the final versionLLM analysis, 98% sureExplained the original approach's correctness and performance problems in #18 and explicitly requested prompt commitment of the revert in #37.
  • Kyotaro Horiguchisaw the final versionLLM analysis, 87% sureSaid the revert component did the right thing in #29, and that component remained unchanged through #35.

Credits across 14 commits

Author
Alvaro Herrera3 LLM6
Reviewer
Alvaro Herreracommitter5
Kyotaro Horiguchi1
Nathan Bossart1
Co author
Alexander Lakhin5
Committer
Alvaro Herrera13
Tom Lane1

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