Dec 2025

4 commits
Update .abi-compliance-history for PrepareToInvalidateCacheTuple().
17 Dec 2025Noah MischREL_17_STABLE17.8REL_16_STABLE16.12REL_15_STABLE15.16REL_14_STABLE14.21
LLM analysis, 99% surewritten from the discussion, no matching patch

Adds only .abi-compliance-history documentation for the PrepareToInvalidateCacheTuple() signature change discussed in #16 and subsequently backported; no indexed attachment contains this bookkeeping change.

Update .abi-compliance-history for CacheInvalidateHeapTupleInplace().
16 Dec 2025Noah MischREL_18_STABLE18.2
LLM analysis, 99% surewritten from the discussion, no matching patch

Adds only .abi-compliance-history documentation for removing CacheInvalidateHeapTupleInplace()'s third argument, the cleanup proposed in #29 and implemented in #30; no matching bookkeeping attachment exists.

Correct comments of "Fix data loss at inplace update after heap_update()".
15 Dec 2025Noah Mischmaster19beta1REL_19_STABLE19beta1
LLM analysis, 100% surepart of a posted patchsetlight changesfrom patchset #301 of 1 reviewer saw the final version

Contains exactly the heap_inplace_lock() deletion explanation and systable_inplace_update_begin() indexed-column comment hunks from #30, split from its invalidation-specific cleanup.

  • Paul A Jungwirthsaw the final versiontrailerHe identified the faulty deletion paragraph in #29 and approved the rewritten comments in #32.
Revisit cosmetics of "For inplace update, send nontransactional invalidations."
15 Dec 2025Noah Mischmaster19beta1REL_19_STABLE19beta1REL_18_STABLE18.2
LLM analysis, 100% surepart of a posted patchsetlight changesfrom patchset #302 of 2 reviewers saw the final version

Takes the invalidation comments, README.tuplock explanation, decode.c cleanup, and unused-parameter removal from #30 unchanged; its two other comment hunks were committed separately in 0839fbe400.

  • Paul A Jungwirthsaw the final versiontrailerHe proposed clearer comments and removing the unused parameter in #29, then explicitly approved the revised comments and signature in #32.
  • Surya Poondlasaw the final versiontrailerPaul reports their joint review in #29, after which only comment revisions and removal of an unused parameter followed.

Apr 2025

1 commit
Assert lack of hazardous buffer locks before possible catalog read.
17 Apr 2025Noah Mischmaster18beta1REL_19_STABLE18beta1REL_18_STABLE18beta1REL_17_STABLE17.8REL_16_STABLE16.12REL_15_STABLE15.16REL_14_STABLE14.21Also in: catcache search while BUFFER_LOCK_EXCLUSIVE on catalog (modified before commit)

Oct - Nov 2024

6 commits
Revert "For inplace update, send nontransactional invalidations."
2 Nov 2024Noah MischREL_17_STABLE17.1REL_16_STABLE16.5REL_15_STABLE15.9REL_14_STABLE14.14REL_13_STABLE13.17REL_12_STABLE12.21reverts 1 commit here
LLM analysis, 100% surewritten from the discussion, no matching patch1 of 1 reviewer saw the final version

Reverts the stable-branch invalidation implementation following Alexander Lakhin's hang report in #19 and the release-risk decision in #22; there is no attached revert patch.

  • Alexander Lakhinsaw the final versionLLM analysis, 95% sureHis reproducer and blocked-backend traces in #19 demonstrated the regression that directly motivated this discussion-derived revert.
Revert "WAL-log inplace update before revealing it to other sessions."
2 Nov 2024Noah MischREL_17_STABLE17.1REL_16_STABLE16.5REL_15_STABLE15.9REL_14_STABLE14.14REL_13_STABLE13.17REL_12_STABLE12.21revert
LLM analysis, 100% surewritten from the discussion, no matching patch

Executes the stable-branch dependency revert proposed in #22, undoing the WAL-before-visibility change so the invalidation change could also be reverted; no attached patch supplies this reversal.

Fix inplace update buffer self-deadlock.
2 Nov 2024Noah Mischmaster18beta1REL_19_STABLE18beta1REL_18_STABLE18beta1
LLM analysis, 100% surelanded as postedfrom patchset #221 of 1 reviewer saw the final version

Applying #22 to the commit's parent reproduces heapam.c, inval.c, and inval.h exactly, including moving invalidation preparation before buffer locking and adding ForgetInplace_Inval().

  • Alexander Lakhinsaw the final versionLLM analysis, 100% sureThe commit explicitly credits Alexander Lakhin as reviewer, while #19 supplies his original reproducer and the available thread contains no separate review of the final attachment.
Move I/O before the index_update_stats() buffer lock region.
2 Nov 2024Noah Mischmaster18beta1REL_19_STABLE18beta1REL_18_STABLE18beta1REL_17_STABLE17.1REL_16_STABLE16.5REL_15_STABLE15.9REL_14_STABLE14.14REL_13_STABLE13.17+1 more
LLM analysis, 100% suremodified before commitlight changesfrom patchset #20

Implements the index_update_stats() reordering from #20 with identical logic; the committer only rewords comments describing possible differences between rel->rd_rel and the tuple being modified.

At end of recovery, reset all sinval-managed caches.
25 Oct 2024Noah MischREL_17_STABLE17.1REL_16_STABLE16.5REL_15_STABLE15.9REL_14_STABLE14.14REL_13_STABLE13.17REL_12_STABLE12.21
LLM analysis, 100% surelanded as postedfrom patchset #160 of 1 reviewer saw the final version

The recovery-reset attachment for PostgreSQL 16 in #16 applies to this commit's parent and reproduces all three changed files exactly, adding SIResetAll() and its StartupXLOG() call.

  • Andres Freundsaw an earlier versionLLM analysis, 83% sureHis review in #5 analyzed WAL compatibility and the cache-reset alternative, which Noah refined into an end-of-recovery reset in #6 before attaching the implementation in #16.
For inplace update, send nontransactional invalidations.
25 Oct 2024Noah Mischmaster18beta1REL_19_STABLE18beta1REL_18_STABLE18beta1REL_17_STABLE17.1REL_16_STABLE16.12REL_15_STABLE15.16REL_14_STABLE14.14REL_13_STABLE13.17+1 more
LLM analysis, 100% surelanded as postedfrom patchset #163 of 5 reviewers saw the final version

Applying the development-branch invalidation attachment from #16 to the commit's parent reproduces every changed file exactly; apparent textual differences are diff-hunk alignment.

  • Paul A Jungwirthsaw the final versiontrailerIn #26, he reports reviewing and testing the repaired stable-branch submission carried by #25, including reproducing the old hang and verifying its absence with the patch.
  • Surya Poondlasaw the final versiontrailerPaul reports their joint review of the repaired backport and associated comment changes in #29, supporting the eventual backport credit.
  • Ilyasov Iansaw the final versiontrailerPaul explicitly reports their joint review and successful hang testing of the repaired backport in #26.
  • Nitin Motianisaw an earlier versiontrailerHe examined locking and callback design in #12 and reported successful testing in #15, before #16 corrected abort cleanup logic.
  • Andres Freundsaw an earlier versiontrailerHis review in #5 questioned removing transactional invalidations and examined WAL compatibility, prompting the later restoration of transactional invalidation.

Jun 2024

1 commit
Remove comment about xl_heap_inplace "AT END OF STRUCT".
28 Jun 2024Noah Mischmaster17beta3REL_19_STABLE17beta3REL_18_STABLE17beta3REL_17_STABLE17beta3REL_16_STABLE16.4REL_15_STABLE15.8REL_14_STABLE14.13REL_13_STABLE13.16+1 more
LLM analysis, 100% surelanded as postedfrom patchset #4

Exactly matches the latest attached one-line removal of the obsolete tuple-data comment from xl_heap_inplace in src/include/access/heapam_xlog.h in #4.

Credits across 12 commits

Author
Noah MischLLM8
Reviewer
Paul A Jungwirth3
surya poondla2
Andres Freund1
Ilyasov Ian1
Nitin Motiani1
Reported by
Paul A Jungwirth2
Alexander Lakhin1
Ilyasov Ian1
surya poondla1
Committer
Noah Misch12

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