Jan 2021

4 commits
Rethink SQLSTATE code for ERRCODE_IDLE_SESSION_TIMEOUT.
11 Jan 2021Tom Lanemaster14beta1REL_19_STABLE14beta1REL_18_STABLE14beta1REL_17_STABLE14beta1REL_16_STABLE14beta1REL_15_STABLE14beta1REL_14_STABLE14beta1
LLM analysis, 100% surewritten from the discussion, no matching patch

Changes ERRCODE_IDLE_SESSION_TIMEOUT from 08P02 to 57P05 in src/backend/utils/errcodes.txt, implementing Tom Lane's proposal in #51 and rationale in #48. No attached patchset contains this change.

Improve commentary in timeout.c.
7 Jan 2021Tom Lanemaster14beta1REL_19_STABLE14beta1REL_18_STABLE14beta1REL_17_STABLE14beta1REL_16_STABLE14beta1REL_15_STABLE14beta1REL_14_STABLE14beta1
LLM analysis, 99% surewritten from the discussion, no matching patch

Adds only a schedule_alarm() comment explaining why a signal handler overwriting signal_due_at before setitimer() remains safe. The commit describes Tom Lane's post-commit re-reading of the timer work discussed in #45; no attached patchset contains this explanation.

Improve timeout.c's handling of repeated timeout set/cancel.
6 Jan 2021Tom Lanemaster14beta1REL_19_STABLE14beta1REL_18_STABLE14beta1REL_17_STABLE14beta1REL_16_STABLE14beta1REL_15_STABLE14beta1REL_14_STABLE14beta1
LLM analysis, 100% surepart of a posted patchsetmoderate changesfrom patchset #431 of 2 reviewers saw the final version

Commits the timeout.c component of #43, whose patch header credits Thomas Munro and whose original submission is described in #16. Tom Lane retains lazy timer reuse but sets signal_pending before setitimer(), adds overdue-timeout recovery, omits the submitted reschedule_timeouts() reset, and expands the comments.

  • Hayato Kuroda (Fujitsu)saw an earlier versionLLM analysis, 99% sureAnalyzed timer scheduling and interruption recovery in #29 and supplied the fix in #33, before Tom Lane revised the race handling at commit.
  • Tom Lanesaw the final versioncommittercommitted the patch
Add idle_session_timeout.
6 Jan 2021Tom Lanemaster14beta1REL_19_STABLE14beta1REL_18_STABLE14beta1REL_17_STABLE14beta1REL_16_STABLE14beta1REL_15_STABLE14beta1REL_14_STABLE14beta1
LLM analysis, 100% surepart of a posted patchsetmoderate changesfrom patchset #431 of 6 reviewers saw the final version

Commits the idle-session component of the two-patch series in #43, retaining its timeout implementation while revising documentation, variable names, and SQLSTATE to 08P02. Tom Lane also changes MAX_TIMEOUTS to USER_TIMEOUT + 10 in timeout.h, as explained in #45; the timer optimization lands separately.

  • Cary Huangsaw an earlier versionLLM analysis, 100% sureTested the feature and installcheck-world and identified the missing sample configuration entry in #8, before subsequent priority, SQLSTATE, and timeout-capacity changes.
  • Bharath Rupireddysaw an earlier versionLLM analysis, 92% sureAnalyzed cached postgres_fdw connections and requested a warning in #11, which was incorporated into later documentation before additional code changes.
  • Thomas Munrosaw an earlier versionLLM analysis, 99% sureMeasured excessive setitimer() calls with the idle-session feature in #16 and supplied the companion optimization before later changes to the idle-session patch.
  • David G. Johnstonsaw an earlier versionLLM analysis, 96% sureReviewed documentation in #25 and proposed documentation and code-comment corrections in #39, with final SQLSTATE and timeout-capacity edits still to follow.
  • Hayato Kuroda (Fujitsu)saw an earlier versionLLM analysis, 96% sureRequested the middleware warning in #21, questioned timeout priority in #23, and checked code comments in #35, before the committer's SQLSTATE and timeout-capacity changes.
  • Tom Lanesaw the final versioncommittercommitted the patch

Credits across 4 commits

Author
Japin LiLLM1
Thomas MunroLLM1
Reviewer
Tom Lanecommitter2
Committer
Tom Lane4

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