May 2022

5 commits
Fix slow animal timeouts in 032_relfilenode_reuse.pl.
13 May 2022Thomas Munromaster15beta1REL_19_STABLE15beta1REL_18_STABLE15beta1REL_17_STABLE15beta1REL_16_STABLE15beta1REL_15_STABLE15beta1
LLM analysis, 100% suremodified before commitlight changesfrom patchset #371 of 1 reviewer saw the final version

The commit preserves all changes in #37 to src/test/recovery/t/032_relfilenode_reuse.pl and adds a three-line explanation of restarting the timeout. Robert requested that comment in #38, and Thomas confirmed adding it and pushing in #39.

  • Robert Haassaw the final versiontrailerRobert accepted the timeout change and requested an explanatory comment in #38, which was the only substantive attachment-to-commit difference.
Add logging for excessive ProcSignalBarrier waits.
11 May 2022Thomas Munromaster15beta1REL_19_STABLE15beta1REL_18_STABLE15beta1REL_17_STABLE15beta1REL_16_STABLE15beta1REL_15_STABLE15beta1
LLM analysis, 100% suremodified before commitlight changesfrom patchset #321 of 1 reviewer saw the final version

Matches the logging attachment in #32, changing its elog call to ereport/errmsg with conventional backend-PID wording and formatting in src/backend/storage/ipc/procsignal.c. Thomas described that edit when announcing the commit and left the separate dummy test-function patch uncommitted #34.

  • Robert Haassaw the final versiontrailerRobert endorsed the logging proposal in #33, after which only the reporting style and wording changed.
Fix race in 032_relfilenode_reuse.pl.
8 May 2022Thomas Munromaster15beta1REL_19_STABLE15beta1REL_18_STABLE15beta1REL_17_STABLE15beta1REL_16_STABLE15beta1REL_15_STABLE15beta1
LLM analysis, 100% surewritten from the discussion, no matching patch

Adds wait_for_catchup() before the standby uses pg_prewarm and corrects a test comment, implementing Thomas's diagnosis of grison's failure in #31. No attachment proposes this repair; #35 explicitly confirms that this commit addressed it.

Fix old-fd issues using global barriers everywhere.
7 May 2022Thomas Munromaster15beta1REL_19_STABLE15beta1REL_18_STABLE15beta1REL_17_STABLE15beta1REL_16_STABLE15beta1REL_15_STABLE15beta1
LLM analysis, 100% surelanded as postedfrom patchset #291 of 2 reviewers saw the final version

The diff exactly matches the global-barrier attachment in #29, including dbcommands.c, tablespace.c, removal of USE_BARRIER_SMGRRELEASE, and the recovery test. Its patch header and commit credit Andres, whose original implementation and test were submitted in #10.

  • Robert Haassaw an earlier versiontrailerRobert reviewed the original fix and test in #11 and #14 and endorsed the revised barriers in #26, before #29 added a new database file-copy redo barrier.
  • Thomas Munrosaw the final versiontrailerThomas examined Andres's original fix in #17, revised and tested it through #29, and committed the final version as announced in #30.
Rethink PROCSIGNAL_BARRIER_SMGRRELEASE.
7 May 2022Thomas Munromaster15beta1REL_19_STABLE15beta1REL_18_STABLE15beta1REL_17_STABLE15beta1REL_16_STABLE15beta1REL_15_STABLE15beta1
LLM analysis, 100% suremodified before commitlight changesfrom patchset #291 of 1 reviewer saw the final version

Matches the smgr-release attachment in #29, with only comment wrapping and the smgrrelease() loop variable changed from int i to ForkNumber forknum. The EXTENSION_DONT_OPEN behavior and preservation of SMgrRelation objects are unchanged.

  • Robert Haassaw the final versiontrailerRobert challenged pointer invalidation in #18, reviewed the counter approach in #21 and #24, and endorsed the final segment-based design in #26, after which this patch received only light edits.

Credits across 5 commits

Author
Thomas MunroLLM3
Andres Freund1
Reviewer
Robert Haas4
Thomas Munro1
Reported by
Andres Freund1
Committer
Thomas Munro5

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