May 2022

3 commits
Remove PGDLLIMPORT marker from __pg_log_level
13 May 2022Michael Paquiermaster15beta1REL_19_STABLE15beta1REL_18_STABLE15beta1REL_17_STABLE15beta1REL_16_STABLE15beta1REL_15_STABLE15beta1
LLM analysis, 100% surewritten from the discussion, no matching patch2 of 2 reviewers saw the final version

The one-line change removes PGDLLIMPORT from __pg_log_level in src/include/common/logging.h, reversing that portion of #97. Andres and Tom explained why the frontend-only variable should remain unmarked, and Michael explicitly agreed to remove the marker #99 #100 #101.

  • Andres Freundsaw the final versionLLM analysis, 100% sureIdentified the frontend-only PGDLLIMPORT marking as inappropriate, directly prompting the removal #99.
  • Tom Lanesaw the final versionLLM analysis, 100% sureConfirmed that omitting the frontend marker was intentional, supporting the exact correction committed #100.
Mark a few 'bbsink' related functions / variables static.
12 May 2022Andres Freundmaster15beta1REL_19_STABLE15beta1REL_18_STABLE15beta1REL_17_STABLE15beta1REL_16_STABLE15beta1REL_15_STABLE15beta1
LLM analysis, 99% surewritten from the discussion, no matching patch2 of 2 reviewers saw the final version

No attachment changes the bbsink code; Andres raised the missing static qualifiers directly in #94, and Michael and Robert agreed they could be corrected #95 #96. The commit makes bbsink operation tables and private basebackup_target.c functions static.

  • Michael Paquiersaw the final versionLLM analysis, 97% sureAgreed that the externally visible bbsink operation tables were an error #95.
  • Robert Haassaw the final versionLLM analysis, 94% sureApproved correcting the issues raised in the follow-up discussion, including the bbsink declarations #96.
Add some missing PGDLLIMPORT markings
12 May 2022Michael Paquiermaster15beta1REL_19_STABLE15beta1REL_18_STABLE15beta1REL_17_STABLE15beta1REL_16_STABLE15beta1REL_15_STABLE15beta1
LLM analysis, 100% surelanded as postedfrom patchset #972 of 2 reviewers saw the final version

The complete three-file diff matches #97 exactly apart from abbreviated index hashes: pqsignal.h declarations, the logging.h marker, and the script usage comment. Michael confirmed the landing in #98.

  • Andres Freundsaw the final versionLLM analysis, 98% sureFound the three unmarked signal-mask variables left by the landed header patch, directly motivating this mechanical correction #94.
  • Robert Haassaw the final versionLLM analysis, 91% sureConfirmed why the helper missed the signal declarations and endorsed correcting their style while preserving the header-only scope #96.

Apr 2022

2 commits
Apply PGDLLIMPORT markings broadly.
8 Apr 2022Robert Haasmaster15beta1REL_19_STABLE15beta1REL_18_STABLE15beta1REL_17_STABLE15beta1REL_16_STABLE15beta1REL_15_STABLE15beta1
LLM analysis, 99% surepart of a posted patchsetlight changesfrom patchset #842 of 5 reviewers saw the final version

This lands the header-marking component of the two-patch series #84; the helper lands separately. Differences reflect regenerated markings for intervening WAL, statistics, preload-library and JSON changes, plus formatting, while preserving the same declaration-marking approach.

  • Tom Lanesaw an earlier versionLLM analysis, 95% sureReviewed the original exposure policy and required already-private variables to remain private, a restriction adopted before the broader automated rework #6 #71 #73.
  • Bruce Momjiansaw an earlier versionLLM analysis, 88% sureExplicitly endorsed exporting the GUC variables to match Unix behavior before the patch expanded to all header globals #9.
  • Peter Geoghegansaw an earlier versionLLM analysis, 85% sureSupported the original GUC proposal while expressing reservations about the verbosity of marking every global variable #22.
  • Andres Freundsaw the final versionLLM analysis, 95% sureReviewed the generated markings and questioned frontend and shared-library declarations; subsequent changes to the header component were mechanical additions and rebasing #78.
  • John Naylorsaw the final versionLLM analysis, 88% sureChecked the frontend-macro concern and explained that existing PGDLLIMPORT handling already permitted unconditional markings for common-library variables #82.
Helper script to apply PGDLLIMPORT markings.
8 Apr 2022Robert Haasmaster15beta1REL_19_STABLE15beta1REL_18_STABLE15beta1REL_17_STABLE15beta1REL_16_STABLE15beta1REL_15_STABLE15beta1
LLM analysis, 100% surepart of a posted patchsetlight changesfrom patchset #841 of 3 reviewers saw the final version

This lands only src/tools/mark_pgdllimport.pl from the two-part series #84. Its executable code is unchanged; the added 21-line introductory and copyright comment implements Andrew's review request #85.

  • Andrew Dunstansaw the final versionLLM analysis, 100% sureSaid the script looked fine and requested its introductory comment and copyright notice, both added before commit; the commit message also explicitly credits his review #85.
  • Andres Freundsaw an earlier versionLLM analysis, 86% sureRequested scripting to prevent missing markers and acknowledged the resulting first version, before its declaration-recognition logic was revised #76 #78.
  • Bruce Momjiansaw an earlier versionLLM analysis, 82% sureRequested automatic checking for forgotten markings in the original proposal, a suggestion subsequently implemented through the helper script #48.

Credits across 5 commits

Author
Robert HaasLLM2
Michael PaquierLLM1
Reported by
Andres Freund1
Committer
Michael Paquier2
Robert Haas2
Andres Freund1

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