Feb - Mar 2025

4 commits
Fix race condition in TAP test 007_pre_auth
6 Mar 2025Michael Paquiermaster18beta1REL_19_STABLE18beta1REL_18_STABLE18beta1
LLM analysis, 100% surelanded as postedfrom patchset #481 of 1 reviewer saw the final version

The diff exactly combines the two patches in #48, adding wait_event and backend_type restrictions in src/test/authentication/t/007_pre_auth.pl. Michael explicitly chose this simpler fix over the polling-helper proposal in #54.

  • Andres Freundsaw the final versionLLM analysis, 98% sureReported the CI failure in #43 and reviewed the fix's polling loops in #49, before the unchanged fix from #48 was committed.
Split pgstat_bestart() into three different routines
4 Mar 2025Michael Paquiermaster18beta1REL_19_STABLE18beta1REL_18_STABLE18beta1
LLM analysis, 100% suremodified before commitlight changesfrom patchset #401 of 2 reviewers saw the final version

Matches the standalone startup-reporting patch in #40, with comment cleanup and removal of the application_name TODO in backend_status.c. Michael forwarded Jacob's implementation from #39, and the commit explicitly credits Jacob.

  • Michael Paquiersaw the final versiontrailerReviewed initialization, tests, and security-field timing in #9, #32, and #35, prepared #40, and committed the final version.
  • Andres Freundsaw an earlier versiontrailerCriticized repeated initialization and explained its architectural hazards in #22 and #28, prompting the substantial three-phase rewrite in #31.
backport: Extend background_psql() to be able to start asynchronously
19 Feb 2025Andres FreundREL_17_STABLE17.5REL_16_STABLE16.9REL_15_STABLE15.13REL_14_STABLE14.18REL_13_STABLE13.21Also in: BackgroundPsql swallowing errors on windows (not this thread)
LLM analysis, 100% surepart of a posted patchsetlight changesfrom patchset #182 of 2 reviewers saw the final version

Backports the asynchronous-connection component of #18, preserving the wait parameter and public wait_connect() implementation in BackgroundPsql.pm and Cluster.pm with additional constructor documentation and minor prose cleanup. The supplied verdict for thread #51068 confirms that no attachment there introduces this functionality and that its discussion only prompted the backport.

  • Michael Paquiersaw the final versionLLM analysis, 100% surePrepared the asynchronous-connection component for its original landing, adding missing constructor documentation as reported in #20, and that reviewed implementation was preserved in the backport.
  • Andres Freundsaw the final versioncommittercommitted the patch
backport: Improve handling of empty query results in BackgroundPsql
19 Feb 2025Andres FreundREL_17_STABLE17.5REL_16_STABLE16.9REL_15_STABLE15.13REL_14_STABLE14.18REL_13_STABLE13.21Also in: BackgroundPsql swallowing errors on windows (not this thread)
LLM analysis, 100% surepart of a posted patchsetlight changesfrom patchset #183 of 3 reviewers saw the final version

Backports the empty-result component of #18, preserving its optional-newline fix in BackgroundPsql.pm and recovery-test changes with only explanatory comments and formatting edits. The supplied verdict for thread #51068 confirms that its attachments assume this behavior and that its discussion prompted the backport rather than supplied the implementation.

  • Andrew Dunstansaw the final versiontrailerEndorsed the empty-result component in #6, after which that component received only light editorial changes.
  • Michael Paquiersaw the final versiontrailerExamined the removed recovery-test assertions in #19 and prepared the original landing reported in #20, with only light edits carried into the backport.
  • Andres Freundsaw the final versioncommittercommitted the patch

Nov 2024

2 commits
Improve handling of empty query results in BackgroundPsql::query()
7 Nov 2024Michael Paquiermaster18beta1REL_19_STABLE18beta1REL_18_STABLE18beta1
LLM analysis, 100% surepart of a posted patchsetlight changesfrom patchset #182 of 2 reviewers saw the final version

Commits the empty-result prerequisite alone from #18: makes the newline before the query banner optional and converts three recovery-test calls to query_safe(). Adds an explanatory comment and formatting cleanup, with the landing confirmed in #20.

  • Andrew Dunstansaw the final versiontrailerApproved the empty-result component in #6, after which that component received only light editorial changes.
  • Michael Paquiersaw the final versiontrailerReviewed the recovery-test assertion changes in #19 and committed the final patch, reporting completion in #20.
Extend Cluster.pm's background_psql() to be able to start asynchronously
6 Nov 2024Michael Paquiermaster18beta1REL_19_STABLE18beta1REL_18_STABLE18beta1
LLM analysis, 100% surepart of a posted patchsetlight changesfrom patchset #181 of 1 reviewer saw the final version

Commits only the asynchronous BackgroundPsql prerequisite from #18, preserving the wait parameter and public wait_connect() implementation. Adds constructor documentation and minor prose cleanup, as Michael confirms in #20.

  • Michael Paquiersaw the final versioncommittercommitted the patch

Credits across 6 commits

Author
Jacob Champion6
Reviewer
Michael Paquiercommitter4
Andres Freundcommitter3
Andrew Dunstan2
Reported by
Andres Freund1
Co author
Michael Paquier1
Committer
Michael Paquier4
Andres Freund2