Mar 2021

4 commits
Rename wait event WalrcvExit to WalReceiverExit.
24 Mar 2021Fujii Masaomaster14beta1REL_19_STABLE14beta1REL_18_STABLE14beta1REL_17_STABLE14beta1REL_16_STABLE14beta1REL_15_STABLE14beta1REL_14_STABLE14beta1
LLM analysis, 100% surelanded as postedfrom patchset #181 of 1 reviewer saw the final version

The four-file rename in #18 has the same stable patch ID as the commit, including the documentation wording and WAIT_EVENT_WAL_RECEIVER_EXIT constant.

  • Thomas Munrosaw the final versiontrailerThomas endorsed the proposed replacement names in #19, and the committed patch is unchanged.
Add condition variable for recovery resume.
12 Mar 2021Thomas Munromaster14beta1REL_19_STABLE14beta1REL_18_STABLE14beta1REL_17_STABLE14beta1REL_16_STABLE14beta1REL_15_STABLE14beta1REL_14_STABLE14beta1
LLM analysis, 100% surepart of a posted patchsetmoderate changesfrom patchset #140 of 2 reviewers saw the final version

Commits the recovery-pause member of #14, adapted to RecoveryPauseState and renamed recoveryNotPausedCV with broadcasts restricted to resume. It also converts the additional sleep loop in RecoveryRequiresIntParameter(), removes explicit sleep preparation, and retains the promotion-path early return.

  • Heikki Linnakangassaw an earlier versionLLM analysis, 87% sureHeikki identified recoveryPausesHere() as a loop whose postmaster-death response would be delayed by #1 in #2, motivating the wait-loop changes that later became this condition-variable patch.
  • Fujii Masaosaw an earlier versionLLM analysis, 99% sureFujii examined the recovery-pause WaitLatch() change in #6, identifying promotion latency and excessive wakeups that Thomas explicitly revisited before adopting condition variables in #13.
Add condition variable for walreceiver shutdown.
12 Mar 2021Thomas Munromaster14beta1REL_19_STABLE14beta1REL_18_STABLE14beta1REL_17_STABLE14beta1REL_16_STABLE14beta1REL_15_STABLE14beta1REL_14_STABLE14beta1
LLM analysis, 100% surepart of a posted patchsetmoderate changesfrom patchset #140 of 1 reviewer saw the final version

Commits the walreceiver condition-variable member of #14, renamed to walRcvStoppedCV and narrowed to broadcast only transitions to STOPPED. Thomas explicitly describes this simplification in #17.

  • Heikki Linnakangassaw an earlier versionLLM analysis, 85% sureHeikki identified ShutdownWalRcv() as an overlooked sleeping loop affected by #1 in #4, prompting the prerequisite fix that later became this condition-variable implementation.
Poll postmaster less frequently in recovery.
12 Mar 2021Thomas Munromaster14beta1REL_19_STABLE14beta1REL_18_STABLE14beta1REL_17_STABLE14beta1REL_16_STABLE14beta1REL_15_STABLE14beta1REL_14_STABLE14beta1Also in: Excessive PostmasterIsAlive calls slow down WAL redo
LLM analysis, 99% surepart of a posted patchsetlight changesfrom patchset #143 of 3 reviewers saw the final version

Commits the third member of #14, with POSTMASTER_POLL_RATE_LIMIT replacing the literal 1024, a named uint32 counter, and revised comments. This matching March 2021 submission is newer than every patchset listed for thread #38497.

  • Heikki Linnakangassaw the final versiontrailerHeikki identified delayed death detection in sleeping loops and requested documentation in #2 and #4; the counter mechanism he reviewed retained only light changes in this commit.
  • Fujii Masaosaw the final versiontrailerFujii examined promotion latency and archive-recovery death detection in #6 and #10, while the polling-rate component subsequently received only light edits.
  • Michael Paquiersaw the final versiontrailerMichael examined the counter implementation in #12 and requested the named rate-limit constant in #16, which the commit adopts.

Sep 2020

1 commit
Allow WaitLatch() to be used without a latch.
23 Sep 2020Thomas Munromaster14beta1REL_19_STABLE14beta1REL_18_STABLE14beta1REL_17_STABLE14beta1REL_16_STABLE14beta1REL_15_STABLE14beta1REL_14_STABLE14beta1
LLM analysis, 100% surepart of a posted patchsetlight changesfrom patchset #71 of 1 reviewer saw the final version

Commits only the latch prerequisite from the two-patch submission #7; that member has an identical stable patch ID to the commit. Thomas confirms its separate application in #9.

  • David Rowleysaw the final versionLLM analysis, 100% sureDavid ran the Windows recovery tests against the latch member of #7 and reported successful results in #8 before it was committed.

Credits across 5 commits

Author
Thomas MunroLLM4
Fujii Masao1
Reviewer
Fujii Masao1
Heikki Linnakangas1
Michael Paquier1
Thomas Munro1
Committer
Thomas Munro4
Fujii Masao1

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