Aug 2024

1 commit
Fix more holes with SLRU code in need of int64 for segment numbers
19 Aug 2024Michael Paquiermaster18beta1REL_19_STABLE18beta1REL_18_STABLE18beta1REL_17_STABLE17rc1Also in: pgsql: Fix more holes with SLRU code in need of int64 for segment numbe (landed as posted)

Jul 2024

3 commits
Fix more holes with SLRU code in need of int64 for segment numbers
26 Jul 2024Michael Paquiermaster18beta1REL_19_STABLE18beta1REL_18_STABLE18beta1REL_17_STABLE17beta3
LLM analysis, 100% surepart of a posted patchsetmoderate changesfrom patchset #1441 of 1 reviewer saw the final version

Only the three multixact.c logging hunks from #144 landed; the segment-return types, truncation variables, and slru.c pagesegno conversion are absent. Noah reports the omission in #148, and Michael confirms an incomplete git add in #149.

  • Noah Mischsaw the final versiontrailerNoah explicitly approved the complete proposed fix in #145, including the logging hunks that landed.
Use more consistently int64 for page numbers in SLRU-related code
23 Jul 2024Michael Paquiermaster18beta1REL_19_STABLE18beta1REL_18_STABLE18beta1REL_17_STABLE17beta3
LLM analysis, 100% suremodified before commitmoderate changesfrom patchset #1400 of 1 reviewer saw the final version

Includes the clog.c and async.c patch from #140, plus three additional int64 conversions in predicate.c. Michael describes finding and adding those paths in #141.

  • Noah Mischsaw an earlier versionLLM analysis, 99% sureNoah identified remaining narrowing conversions in #132 and challenged the missed asyncQueueIsFull case in #134, before further async.c and predicate.c fixes were added.
Improve comments in slru.{c,h} about segment name format
23 Jul 2024Michael Paquiermaster18beta1REL_19_STABLE18beta1REL_18_STABLE18beta1REL_17_STABLE17beta3
LLM analysis, 100% suremodified before commitlight changesfrom patchset #1402 of 2 reviewers saw the final version

Uses the comment patch from #140, simplifying SlruCtlData's description and rewording the SlruFileName examples and ranges. Michael explains this editorial adjustment in #141.

  • Noah Mischsaw the final versionLLM analysis, 100% sureNoah identified the incorrect 48-bit and 16-bit filename descriptions in #132, and the ensuing changes to this feature were confined to comments.
  • Michael Paquiersaw the final versioncommittercommitted the patch

Jan 2024

1 commit
Reorder actions in ProcArrayApplyRecoveryInfo()
19 Jan 2024Alexander Korotkovmaster17beta1REL_19_STABLE17beta1REL_18_STABLE17beta1REL_17_STABLE17beta1Also in: Assertion failure with epoch when replaying standby records for 2PC (not this thread)
LLM analysis, 100% surepart of a posted patchsetlight changesfrom patchset #1300 of 1 reviewer saw the final version

The procarray.c reordering matches #130, advancing nextXid before StandbyReleaseOldLocks. The attachment's separate nextXid-to-cutoffXid rename in twophase.c was omitted.

  • Thomas Munrosaw an earlier versionLLM analysis, 99% sureThomas reported the epoch assertion failure caused by the landed two-phase change in #129, prompting the recovery-ordering fix in #130.

Nov 2023

6 commits
Fix typo in 5a1dfde8334b
30 Nov 2023Alexander Korotkovmaster17beta1REL_19_STABLE17beta1REL_18_STABLE17beta1REL_17_STABLE17beta1
LLM analysis, 100% surewritten from the discussion, no matching patch1 of 1 reviewer saw the final version

Corrects 'occured' to 'occurred' in twophase.c exactly as reported by Alexander Lakhin in #114; no attachment proposes this spelling correction.

  • Alexander Lakhinsaw the final versionLLM analysis, 100% sureAlexander identified the exact spelling error in #114, which was corrected directly without an attached fix patch.
Fix warning due non-standard inline declaration in 4ed8f0913bfdb5f355
30 Nov 2023Alexander Korotkovmaster17beta1REL_19_STABLE17beta1REL_18_STABLE17beta1REL_17_STABLE17beta1
LLM analysis, 100% surelanded as postedfrom patchset #1233 of 3 reviewers saw the final version

Exactly matches Pavel Borisov's one-line SlruFileName declaration fix in #123; Korotkov confirms pushing it in #124.

  • Alexander Lakhinsaw the final versionLLM analysis, 100% sureAlexander compiled #112 and reported the old-style inline declaration warning in #114, with only a declaration-order correction needed.
  • Tom Lanesaw the final versionLLM analysis, 100% sureTom documented the same declaration warning across buildfarm animals in #122, directly prompting Pavel's fix.
  • Alexander Korotkovsaw the final versioncommittercommitted the patch
Add SLRU tests for 64-bit page case
28 Nov 2023Alexander Korotkovmaster17beta1REL_19_STABLE17beta1REL_18_STABLE17beta1REL_17_STABLE17beta1Also in: Add 64-bit XIDs into PostgreSQL 15 (not this thread)
LLM analysis, 100% surelanded as postedfrom patchset #1212 of 2 reviewers saw the final version

The high-page-number test patch in #121 has the same stable patch ID as the commit, adding 173 lines in test_slru SQL and expected output and identifying Maxim as author. The current verdict in thread #45306 confirms that its pre-commit attachments contain no corresponding tests and attributes this change to the final submission here.

  • Aleksander Alekseevsaw the final versiontrailerAleksander flagged the failing high-page test submission in #111, and the final patch #121 credits his review after the test adjustments.
  • Alexander Korotkovsaw the final versiontrailerKorotkov requested actual high-page tests in #101, approved the corrected series in #113, and committed the final version.
Make use FullTransactionId in 2PC filenames
28 Nov 2023Alexander Korotkovmaster17beta1REL_19_STABLE17beta1REL_18_STABLE17beta1REL_17_STABLE17beta1Also in: Add 64-bit XIDs into PostgreSQL 15 (not this thread)
LLM analysis, 100% surelanded as postedfrom patchset #1213 of 11 reviewers saw the final version

The twophase.c patch in #121 has the same stable patch ID as the commit and credits Maxim, who introduced its epoch reconstruction in #95. The current verdict in thread #45306 distinguishes its earlier direct filename widening and confirms this later submission as the committed source.

  • Jacob Championno review foundtrailerJacob's discussion in #80 concerns SLRU WAL compatibility, before the two-phase filename patch was introduced.
  • Heikki Linnakangassaw the final versiontrailerHeikki proposed FullTransactionId two-phase filenames in #89 and described the resulting series as nearly ready in #97, after which this component's code stayed unchanged.
  • Alexander Korotkovsaw the final versiontrailerKorotkov approved the series in #113, posted #121, and committed the two-phase patch.
  • Japin Lino review foundtrailerJapin's reviews in #24 and #73 address formatting and pg_upgrade, not the later two-phase filename change.
  • Pavel Borisovno review foundtrailerPavel's late technical reviews in #102 and #117 concern SLRU functions and filenames, with no identifiable review of twophase.c.
  • Tom Laneno review foundtrailerTom discusses printf casting and the SLRU inline warning in #11 and #122, without reviewing the two-phase implementation.
  • Peter Eisentrautno review foundtrailerPeter's substantive reviews precede the two-phase patch and concern formatting and SLRU design, notably #26.
  • Andres Freundno review foundtrailerAndres's signedness and UBSAN feedback in #16 and #67 concerns earlier SLRU work.
  • Andrey Borodinno review foundtrailerAndrey reviewed SLRU wraparound and upgrade design in #72, before the two-phase filename patch existed.
  • Dilip Kumarno review foundtrailerDilip's WAL-overhead question in #60 concerns the broader 64-bit XID effort rather than two-phase filenames.
  • Aleksander Alekseevsaw the final versiontrailerAleksander explicitly said the two-phase patch looked good in #96, and its implementation remained unchanged through #121.
Use larger segment file names for pg_notify
28 Nov 2023Alexander Korotkovmaster17beta1REL_19_STABLE17beta1REL_18_STABLE17beta1REL_17_STABLE17beta1Also in: Add 64-bit XIDs into PostgreSQL 15 (not this thread)
LLM analysis, 100% surelanded as postedfrom patchset #1212 of 11 reviewers saw the final version

The notification patch in #121 has the same stable patch ID as the commit, including long filenames, wraparound removal and max_notify_queue_pages; its From header identifies Aleksander. The current verdict in thread #45306 confirms that its earlier attachments lack this implementation and attributes the final patch to this thread.

  • Jacob Championno review foundtrailerJacob reviewed WAL compatibility in #80, before the notification-queue conversion was submitted.
  • Heikki Linnakangassaw an earlier versiontrailerHeikki proposed the notification conversion and configurable limit in #89 and reviewed the series in #97, before its queue-full condition changed from equality to >=.
  • Alexander Korotkovsaw the final versiontrailerKorotkov requested the safer queue-limit comparison in #101, approved the revised series in #113, and committed it.
  • Japin Lino review foundtrailerJapin's concrete feedback concerns earlier formatting and pg_upgrade patches in #24 and #73.
  • Pavel Borisovno review foundtrailerPavel discusses the shared SLRU infrastructure in #102 and #117, without identifiable review of the notification-specific implementation.
  • Tom Laneno review foundtrailerTom's formatting guidance in #6 and #11 predates the notification-queue patch.
  • Peter Eisentrautno review foundtrailerPeter's reviews of the early series, including #26, do not cover the later notification conversion.
  • Andres Freundno review foundtrailerAndres's UBSAN report in #67 concerns SLRU callback arguments rather than notification wraparound removal.
  • Andrey Borodinno review foundtrailerAndrey's review in #72 addresses the earlier general SLRU migration, before the notification-specific patch existed.
  • Dilip Kumarno review foundtrailerDilip's question in #60 concerns future XID-related WAL overhead, not notification queues.
  • Aleksander Alekseevsaw the final versiontrailerAleksander implemented the queue-limit correction in #108 and retained that tested notification patch in #111.
Index SLRUs by 64-bit integers rather than by 32-bit integers
28 Nov 2023Alexander Korotkovmaster17beta1REL_19_STABLE17beta1REL_18_STABLE17beta1REL_17_STABLE17beta1Also in: Add 64-bit XIDs into PostgreSQL 15 (not this thread)
LLM analysis, 100% surelanded as postedfrom patchset #1214 of 11 reviewers saw the final version

The first patch in #121 matches the SLRU internals, callers, WAL descriptions and test interfaces, with only whitespace and comment wrapping differences. Its November 2023 submission supersedes the earlier implementation in thread #45306, whose current verdict also identifies this thread as the final source.

  • Jacob Championsaw an earlier versiontrailerJacob identified the WAL-format compatibility issue reported in #78 and clarified upgrade behavior in #80, before the major simplifications.
  • Heikki Linnakangassaw the final versiontrailerHeikki redirected migration design in #89, caught filename-width limitations in #97, and reviewed the final legacy-name explanation in #116 and #118.
  • Alexander Korotkovsaw the final versiontrailerKorotkov reviewed assertions, inline functions and test coverage in #101, approved the corrected series in #113, and committed the final version.
  • Japin Lisaw an earlier versiontrailerJapin supplied the GetClogDirName cleanup in #73, which was incorporated before the upgrade machinery was subsequently redesigned and removed.
  • Pavel Borisovsaw the final versiontrailerPavel justified inline functions for type checking in #102 and reviewed preservation of legacy filename handling in #117 and #119.
  • Tom Laneno review foundtrailerTom's pre-commit feedback in #6 and #11 concerns the separate XID-formatting work, while his SLRU declaration complaint in #122 came after this commit.
  • Peter Eisentrautsaw an earlier versiontrailerPeter challenged signedness, filename formatting, FileTag width and missing tests in #26, #36 and #38, before substantial redesign.
  • Andres Freundsaw an earlier versiontrailerAndres questioned signed page types in #16 and reported the misaligned int64 callback load in #67, before later architectural changes.
  • Andrey Borodinsaw an earlier versiontrailerAndrey examined whether the series actually removed wraparound and questioned its split and upgrade algorithm in #72.
  • Dilip Kumarsaw an earlier versiontrailerDilip raised the WAL-size implications of the proposed groundwork in #60, before the series was narrowed and reworked.
  • Aleksander Alekseevsaw the final versiontrailerAleksander tested earlier versions in #35, corrected addressing and assertion issues, and endorsed the final filename treatment in #120.

Credits across 11 commits

Author
Aleksander Alekseev5
Alexander Korotkov1 LLM4
Maxim Orlov4
Pavel Borisov4
Michael Paquier2 LLM3
Nikita Glukhov3
Teodor Sigaev3
Yura Sokolov3
Reviewer
Alexander Korotkovcommitter5
Aleksander Alekseev4
Andres Freund3
Andrey Borodin3
Dilip Kumar3
Heikki Linnakangas3
Jacob Champion3
Japin Li3
Pavel Borisov3
Peter Eisentraut3
Tom Lane3
Michael Paquiercommitter1
Noah Misch1
Reported by
Noah Misch3
Alexander Lakhin2
Michael Paquier1
Thomas Munro1
Tom Lane1
Committer
Alexander Korotkov7
Michael Paquier4

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