Oct 2022

1 commit
Under has_wal_read_bug, skip recovery/t/032_relfilenode_reuse.pl.
29 Oct 2022Noah Mischmaster16beta1REL_19_STABLE16beta1REL_18_STABLE16beta1REL_17_STABLE16beta1REL_16_STABLE16beta1REL_15_STABLE15.1
LLM analysis, 99% surewritten from the discussion, no matching patch

The commit extends the has_wal_read_bug suppression established by #28 to src/test/recovery/t/032_relfilenode_reuse.pl, which appears in no attached patchset. Its message cites another kittiwake failure, and its Discussion trailer links the extension to this thread.

May 2022

1 commit
Under has_wal_read_bug, skip contrib/bloom/t/001_wal.pl.
7 May 2022Noah Mischmaster15beta1REL_19_STABLE15beta1REL_18_STABLE15beta1REL_17_STABLE15beta1REL_16_STABLE15beta1REL_15_STABLE15beta1REL_14_STABLE14.3REL_13_STABLE13.7+3 more
LLM analysis, 99% surewritten from the discussion, no matching patch

The commit applies has_wal_read_bug and skip_all to contrib/bloom/t/001_wal.pl after failures on snapper and kittiwake. No attachment changes that file; this is a later extension of the suppression approach discussed in #33.

Feb 2022

1 commit
Use Test::Builder::todo_start(), replacing $::TODO.
10 Feb 2022Noah Mischmaster15beta1REL_19_STABLE15beta1REL_18_STABLE15beta1REL_17_STABLE15beta1REL_16_STABLE15beta1REL_15_STABLE15beta1REL_14_STABLE14.3REL_13_STABLE13.7+3 more
LLM analysis, 100% surewritten from the discussion, no matching patch

The replacement of local $TODO with Test::More->builder->todo_start('filesystem bug') comes directly from Noah Misch's inline proposal in #34. The commit applies it to contrib/amcheck/t/003_cic_2pc.pl and updates the corresponding comment in src/test/recovery/t/027_stream_regress.pl.

Jan 2022

1 commit
On sparc64+ext4, suppress test failures from known WAL read failure.
27 Jan 2022Noah Mischmaster15beta1REL_19_STABLE15beta1REL_18_STABLE15beta1REL_17_STABLE15beta1REL_16_STABLE15beta1REL_15_STABLE15beta1REL_14_STABLE14.2REL_13_STABLE13.6+3 more
LLM analysis, 100% suremodified before commitlight changesfrom patchset #283 of 4 reviewers saw the final version

The commit preserves the has_wal_read_bug detector and recovery-test skip from #28, but changes the amcheck test from skip_all to conditional local $TODO. It also adjusts documentation and explains why the recovery test must still be skipped, as agreed in #33.

  • Thomas Munrosaw an earlier versionLLM analysis, 87% sureIn #19, he identified O_DIRECT alignment failures in the runtime workaround and questioned supporting it, prompting the substantial switch to test suppression.
  • Andres Freundsaw the final versionLLM analysis, 99% sureHe challenged the runtime approach in #20, proposed automatic detection in #25, and supplied conditional $TODO guidance in #32 that the committed version follows.
  • Michael Paquiersaw the final versionLLM analysis, 99% sureIn #29, he advocated TODO marking so the affected tests would continue exercising SPARC and two-phase-commit code, which the amcheck change implements.
  • Tom Lanesaw the final versionLLM analysis, 99% sureIn #30, he accepted TODO marking provided it was restricted to known affected platforms, matching the committed conditional detector.

Nov 2021

1 commit
Report any XLogReadRecord() error in XlogReadTwoPhaseData().
12 Nov 2021Noah Mischmaster15beta1REL_19_STABLE15beta1REL_18_STABLE15beta1REL_17_STABLE15beta1REL_16_STABLE15beta1REL_15_STABLE15beta1REL_14_STABLE14.2REL_13_STABLE13.6+3 more
LLM analysis, 100% surelanded as postedfrom patchset #12 of 2 reviewers saw the final version

The single twophase.c hunk exactly matches #1, including both branches of the enhanced error report. Noah Misch confirmed the push in #5.

  • Michael Paquiersaw the final versionLLM analysis, 100% sureHe explicitly said the patch looked good in #2, and 3354746910 credits his review.
  • Tom Lanesaw the final versionLLM analysis, 100% sure3354746910 explicitly credits his review, although his visible messages in this thread begin after that commit.

Credits across 5 commits

Author
Noah MischLLM2
Committer
Noah Misch5

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