Mar 2023

5 commits
amcheck: Generalize one of the recently-added update chain checks.
27 Mar 2023Robert Haasmaster16beta1REL_19_STABLE16beta1REL_18_STABLE16beta1REL_17_STABLE16beta1REL_16_STABLE16beta1
LLM analysis, 100% surepart of a posted patchsetlight changesfrom patchset #960 of 1 reviewer saw the final version

Exactly matches the second patch in #96, excluding index hashes and hunk positions: it moves the HEAP_UPDATED invariant into check_tuple_header() and adjusts the TAP expectations. It lands only that component of the two-patch series.

  • Andres Freundsaw an earlier versionLLM analysis, 88% sureReviewing the landed version of #78, Andres proposed the broader invariant in #84 and endorsed check_tuple_header() as its location in #93, before its implementation in #96.
amcheck: Tighten up validation of redirect line pointers.
27 Mar 2023Robert Haasmaster16beta1REL_19_STABLE16beta1REL_18_STABLE16beta1REL_17_STABLE16beta1REL_16_STABLE16beta1
LLM analysis, 100% surepart of a posted patchsetlight changesfrom patchset #961 of 1 reviewer saw the final version

Exactly matches the first patch in #96, excluding index hashes and hunk positions, tightening redirect-target validation in contrib/amcheck/verify_heapam.c. The commit credits Andres Freund and Robert Haas, with Andres's implementation originating in #85.

  • Robert Haassaw the final versioncommittercommitted the patch
amcheck: Fix verify_heapam for tuples where xmin or xmax is 0.
24 Mar 2023Robert Haasmaster16beta1REL_19_STABLE16beta1REL_18_STABLE16beta1REL_17_STABLE16beta1REL_16_STABLE16beta1REL_15_STABLE15.3REL_14_STABLE14.8
LLM analysis, 100% surelanded as postedfrom patchset #971 of 1 reviewer saw the final version

All three hunks match #97 modulo rebase: invalid xmin reports corruption, and invalid xmax returns before examining the unset status output. The changed title and explanatory commit message do not alter the patch.

  • Andres Freundsaw the final versionLLM analysis, 100% sureAndres reported the Valgrind failure in #86 and agreed with the diagnosis and patch in #98, which the commit explicitly acknowledges.
amcheck: Fix a few bugs in new update chain validation.
23 Mar 2023Robert Haasmaster16beta1REL_19_STABLE16beta1REL_18_STABLE16beta1REL_17_STABLE16beta1REL_16_STABLE16beta1
LLM analysis, 99% surepart of a posted patchsetmoderate changesfrom patchset #851 of 2 reviewers saw the final version

Takes the direct HEAP_HOT_UPDATED checks from #85, omits its redirect changes, and replaces its consumer-side bounds guard with Robert Haas's producer-side check from #89. The commit explicitly attributes its first hunk to Robert and the other two to Andres.

  • Peter Geoghegansaw an earlier versionLLM analysis, 86% surePeter confirmed in #82 that CTIDs may legitimately exceed the truncated line-pointer array and that the HOT-status macro caused a validation problem, before the corrective implementation was posted.
  • Robert Haassaw the final versioncommittercommitted the patch
Fix new test case to work on (some?) big-endian architectures.
22 Mar 2023Tom Lanemaster16beta1REL_19_STABLE16beta1REL_18_STABLE16beta1REL_17_STABLE16beta1REL_16_STABLE16beta1
LLM analysis, 100% surewritten from the discussion, no matching patch

Tom Lane reported mamba's big-endian failure in #80 and committed two corrected bitfield constants in src/bin/pg_amcheck/t/004_verify_heapam.pl without an attached fix patch. Robert's response in #88 came after the commit.

Nov 2022

1 commit
Don't test HEAP_XMAX_INVALID when freezing xmax.
23 Nov 2022Peter Geogheganmaster16beta1REL_19_STABLE16beta1REL_18_STABLE16beta1REL_17_STABLE16beta1REL_16_STABLE16beta1
LLM analysis, 100% surewritten from the discussion, no matching patch1 of 1 reviewer saw the final version

Implements the heap_prepare_freeze_tuple() cleanup discussed in #54, #55, and #56: accept only zero as already-frozen raw xmax, drop HEAP_XMAX_INVALID, and reuse the fetched xid. No attached patch modifies src/backend/access/heap/heapam.c.

  • Andres Freundsaw the final versionLLM analysis, 100% sureAndres proposed testing only !TransactionIdIsValid(xid) and avoiding the redundant raw-xmax fetch in #55, which Peter accepted in #56 and committed.

Credits across 6 commits

Author
Andres FreundLLM2
Robert HaasLLM2
Peter Geoghegan1
Reviewer
Robert Haascommitter2
Committer
Robert Haas4
Peter Geoghegan1
Tom Lane1

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