Mar 2022

2 commits
Harden TAP tests that intentionally corrupt page checksums.
25 Mar 2022Tom Lanemaster15beta1REL_19_STABLE15beta1REL_18_STABLE15beta1REL_17_STABLE15beta1REL_16_STABLE15beta1REL_15_STABLE15beta1REL_14_STABLE14.3REL_13_STABLE13.7+2 more
LLM analysis, 100% suremodified before commitmoderate changesfrom patchset #670 of 2 reviewers saw the final version

The checksum-field inversion from #67 becomes PostgreSQL::Test::Cluster::corrupt_page_checksum, used by both pg_checksums and pg_basebackup tests. The commit also adds error details and binary mode, replaces unsafe seek calls in amcheck tests, and incorporates the requests in #68 and #70.

  • Andres Freundsaw an earlier versionLLM analysis, 98% sureIn #68, he endorsed #67 and requested a shared helper covering pg_basebackup, which was implemented afterward alongside further test changes.
  • Dagfinn Ilmari MannsÃ¥kersaw an earlier versionLLM analysis, 98% sureIn #70, he approved #67 subject to including $! in error messages, before the expanded helper-based implementation landed.
Fix possible recovery trouble if TRUNCATE overlaps a checkpoint.
24 Mar 2022Robert Haasmaster15beta1REL_19_STABLE15beta1REL_18_STABLE15beta1REL_17_STABLE15beta1REL_16_STABLE15beta1REL_15_STABLE15beta1REL_14_STABLE14.3REL_13_STABLE13.7+3 more
LLM analysis, 99% suremodified before commitlight changesfrom patchset #361 of 6 reviewers saw the final version

All eleven files' code and comment hunks match #36 after normalizing rebase metadata. The only remaining edit is removal of the proposed commit message's backpatching sentence; the commit explicitly credits source sender Kyotaro Horiguchi as the patch author.

  • Heikki Linnakangassaw an earlier versionLLM analysis, 100% sureIn #17, he identified buffer-scan and I/O-lock problems in #5 and proposed delaying checkpoint completion, prompting the architectural replacement.
  • Andres Freundsaw an earlier versionLLM analysis, 100% sureIn #9 and #18, he reviewed truncation-failure handling and the buffer-retention approach of #5, before the checkpoint-barrier rework.
  • Alvaro Herrerasaw an earlier versionLLM analysis, 99% sureIn #6, he evaluated the failure-handling approach of #5 using an observed storage-disconnection corruption case, and the commit explicitly credits his review.
  • Masahiko Sawadasaw an earlier versionLLM analysis, 100% sureIn #8, he challenged the generic PANIC handling in #5 and explored deferred truncation before the approach was replaced.
  • Tom Lanesaw an earlier versionLLM analysis, 100% sureIn #26 and #28, he reviewed the enum-based #22, objecting to mutually exclusive barriers and the procarray.h dependency before the final bitmap implementation.
  • Robert Haassaw the final versioncommittercommitted the patch

Credits across 2 commits

Author
Kyotaro HoriguchiLLM1
Tom LaneLLM1
Reviewer
Robert Haascommitter1
Committer
Robert Haas1
Tom Lane1

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