May 2019

1 commit
Probe only 127.0.0.1 when looking for ports on Unix.
8 May 2019Thomas Munromaster12beta1REL_19_STABLE12beta1REL_18_STABLE12beta1REL_17_STABLE12beta1REL_16_STABLE12beta1REL_15_STABLE12beta1REL_14_STABLE12beta1REL_13_STABLE12beta1+4 more
LLM analysis, 100% suremodified before commitlight changesfrom patchset #311 of 1 reviewer saw the final version

The PostgresNode.pm address-list change matches #31 exactly; the committed version further revises the explanatory comment. Thomas confirms that additional comment adjustment in #32.

  • Noah Mischsaw the final versionLLM analysis, 90% sureIn #30 he corrected an earlier version's comment about loopback probing, and #31 explicitly incorporates that feedback; that earlier version has no patchset entry.

Apr 2019

9 commits
Test both 0.0.0.0 and 127.0.0.x addresses to find a usable port.
15 Apr 2019Noah Mischmaster12beta1REL_19_STABLE12beta1REL_18_STABLE12beta1REL_17_STABLE12beta1REL_16_STABLE12beta1REL_15_STABLE12beta1REL_14_STABLE12beta1REL_13_STABLE12beta1+4 more
LLM analysis, 100% surewritten from the discussion, no matching patch

Adds can_bind() and probes both wildcard and concrete loopback addresses in PostgresNode.pm, unlike the wildcard-only implementation in #22. Noah identifies this unattached follow-up as the Windows port-conflict fix in #29.

MSYS: Translate REGRESS_SHLIB to a Windows file name.
14 Apr 2019Noah Mischmaster12beta1REL_19_STABLE12beta1REL_18_STABLE12beta1REL_17_STABLE12beta1REL_16_STABLE12beta1REL_15_STABLE12beta1REL_14_STABLE12beta1REL_13_STABLE12beta1+2 more
LLM analysis, 100% surewritten from the discussion, no matching patch

Adds MSYS-to-Windows path translation for REGRESS_SHLIB in src/test/recovery/t/017_shm.pl, absent from the submitted patches. Noah describes this post-landing repair in #29.

MSYS: Skip src/test/recovery/t/017_shm.pl.
14 Apr 2019Noah MischREL_10_STABLE10.8REL9_6_STABLE9.6.13
LLM analysis, 100% surewritten from the discussion, no matching patch

Reverts the pg_ctl kill fallback on affected older branches and skips 017_shm.pl under MSYS because those branches lack the required signal support. This unattached compatibility fix is explained in #29.

When Perl "kill(9, ...)" fails, try "pg_ctl kill".
13 Apr 2019Noah Mischmaster12beta1REL_19_STABLE12beta1REL_18_STABLE12beta1REL_17_STABLE12beta1REL_16_STABLE12beta1REL_15_STABLE12beta1REL_14_STABLE12beta1REL_13_STABLE12beta1+4 more
LLM analysis, 100% surewritten from the discussion, no matching patch

Changes PostgresNode::kill9() to fall back to pg_ctl when Perl kill() fails, whereas #22 still aborts on that failure. Noah reports the MSYS failure and this repair in #29; no matching fix was attached.

Consistently test for in-use shared memory.
13 Apr 2019Noah Mischmaster12beta1REL_19_STABLE12beta1REL_18_STABLE12beta1REL_17_STABLE12beta1REL_16_STABLE12beta1REL_15_STABLE12beta1REL_14_STABLE12beta1REL_13_STABLE12beta1+6 more
LLM analysis, 100% suremodified before commitlight changesfrom patchset #221 of 4 reviewers saw the final version

Matches the full patch in #22 except for the poll_start() comment, which adds delayed waitpid() as another reason to retry. That clarification follows #25 and Noah's acceptance in #26.

  • Dmitry Dolgovsaw an earlier versionLLM analysis, 99% sureHe tested #9 and reported its failing shared-memory assertion in #10, before the attachment refactor and later behavioral fixes.
  • Kyotaro Horiguchisaw an earlier versionLLM analysis, 100% sureHis code review in #14 prompted the state-name correction and attachment refactor, and the commit explicitly credits his review of earlier versions.
  • Daniel Gustafssonsaw an earlier versionLLM analysis, 100% sureHe tested and reviewed #16 in #17, before the EACCES behavior change and concurrency repairs, and the commit explicitly credits his earlier review.
  • Thomas Munrosaw the final versionLLM analysis, 99% sureIn #25 he corrected the signal-delivery explanation and identified the waitpid() delay, which Noah accepted in #26 and added to the committed comment.
Revert "Consistently test for in-use shared memory."
5 Apr 2019Noah Mischmaster12beta1REL_19_STABLE12beta1REL_18_STABLE12beta1REL_17_STABLE12beta1REL_16_STABLE12beta1REL_15_STABLE12beta1REL_14_STABLE12beta1REL_13_STABLE12beta1+6 morereverts 2 commits here
LLM analysis, 100% surewritten from the discussion, no matching patch

Implements the revert announced in #21 after continued buildfarm failures, removing the initial safety landing and its follow-up repairs. No revert patch was attached.

Make src/test/recovery/t/017_shm.pl safe for concurrent execution.
4 Apr 2019Noah Mischmaster12beta1REL_19_STABLE12beta1REL_18_STABLE12beta1REL_17_STABLE12beta1REL_16_STABLE12beta1REL_15_STABLE12beta1REL_14_STABLE12beta1REL_13_STABLE12beta1+4 more
LLM analysis, 100% surewritten from the discussion, no matching patch

Implements the extra port-holder postmaster proposed in #20 to prevent concurrent 017_shm.pl executions from choosing conflicting shared-memory keys. No matching attachment preceded the commit; #22 contains a later revision of this mechanism.

Consistently test for in-use shared memory.
4 Apr 2019Noah Mischmaster12beta1REL_19_STABLE12beta1REL_18_STABLE12beta1REL_17_STABLE12beta1REL_16_STABLE12beta1REL_15_STABLE12beta1REL_14_STABLE12beta1REL_13_STABLE12beta1+6 more
LLM analysis, 99% surepart of a posted patchsetlight changesfrom patchset #161 of 3 reviewers saw the final version

Commits the shared-memory safety attachment from #16, excluding the separately landed HINT attachment. Relative to that safety patch, it adds shmdt() failure logging requested in #17 and renames the recovery test to 017_shm.pl.

  • Dmitry Dolgovsaw an earlier versionLLM analysis, 99% sureHe tested #9 and reported its failing recovery test in #10, before the subsequent PGSharedMemoryAttach() refactor.
  • Kyotaro Horiguchisaw an earlier versionLLM analysis, 100% sureHis detailed review in #14 led to the attachment-lifetime refactor and clearer state naming in #16, and the commit explicitly credits him.
  • Daniel Gustafssonsaw the final versionLLM analysis, 100% sureHe reported successful check-world testing and requested the test rename and shmdt() diagnostics in #17, with only those light changes before landing.
Update HINT for pre-existing shared memory block.
1 Apr 2019Noah Mischmaster12beta1REL_19_STABLE12beta1REL_18_STABLE12beta1REL_17_STABLE12beta1REL_16_STABLE12beta1REL_15_STABLE12beta1REL_14_STABLE12beta1REL_13_STABLE12beta1+6 more
LLM analysis, 99% surepart of a posted patchsetlight changesfrom patchset #162 of 2 reviewers saw the final version

Exactly matches the standalone HINT attachment last submitted in #16, changing miscinit.c to recommend terminating old server processes. It commits only that attachment, leaving the accompanying safety implementation for a separate commit.

  • Daniel Gustafssonsaw the final versionLLM analysis, 100% sureHe explicitly reviewed and tested both attachments in #17, and the HINT commit also names him as a reviewer.
  • Kyotaro Horiguchisaw the final versionLLM analysis, 100% sureThe commit explicitly credits him as a reviewer, although his visible review in #14 addresses the companion safety implementation rather than the HINT text.

Credits across 10 commits

Author
Noah MischLLM3
Thomas MunroLLM1
Committer
Noah Misch9
Thomas Munro1

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