May 2022

3 commits
Add a new shmem_request_hook hook.
13 May 2022Robert Haasmaster15beta1REL_19_STABLE15beta1REL_18_STABLE15beta1REL_17_STABLE15beta1REL_16_STABLE15beta1REL_15_STABLE15beta1
LLM analysis, 100% surelanded as postedfrom patchset #1282 of 3 reviewers saw the final version

The nine-file hook patch in #128 has the same stable patch ID as the commit. Michael Paquier forwarded it; its header and the commit credit Nathan Bossart and Julien Rouhaud.

  • Tom Lanesaw an earlier versionLLM analysis, 98% sureTom supported enforcing the new request phase and requested FATAL errors in #110, before the shared guard and LWLock enforcement were implemented in #112.
  • Anton A. Melnikovsaw the final versionLLM analysis, 100% sureAnton reported successful Linux and Windows tests and verified correct MaxBackends-dependent allocation in an extended pg_stat_statements in #113, after which only cleanup and documentation changes followed.
  • Robert Haassaw the final versioncommittercommitted the patch
Remove non-functional code for unloading loadable modules.
11 May 2022Robert Haasmaster15beta1REL_19_STABLE15beta1REL_18_STABLE15beta1REL_17_STABLE15beta1REL_16_STABLE15beta1REL_15_STABLE15beta1
LLM analysis, 100% surepart of a posted patchsetlight changesfrom patchset #1243 of 3 reviewers saw the final version

Takes the complete unloading-cleanup component of #124, with an identical stable patch ID, leaving its hook component pending. The retained changes remove disabled unloading code, _PG_fini functions and documentation references.

  • Michael Paquiersaw the final versionLLM analysis, 100% sureMichael caught missing documentation cleanup in #121, approved the revised cleanup and suggested applying it first in #123, and is credited in the commit.
  • Tom Lanesaw the final versionLLM analysis, 95% sureBefore the cleanup attachment existed, Tom recommended deleting the unusable unloading machinery and explained its unresolved hook and custom-GUC problems in #117, which Nathan adopted in #118.
  • Robert Haassaw the final versioncommittercommitted the patch
Fix misleading comments about background worker registration.
6 May 2022Robert Haasmaster15beta1REL_19_STABLE15beta1REL_18_STABLE15beta1REL_17_STABLE15beta1REL_16_STABLE15beta1REL_15_STABLE15beta1
LLM analysis, 100% surepart of a posted patchsetlight changesfrom patchset #1127 of 7 reviewers saw the final version

Exactly matches the comment-only component of #112 by stable patch ID, leaving the hook component uncommitted. It restores the same postmaster.c and postinit.c corrections previously committed and then reverted.

  • Anton A. Melnikovsaw the final versionLLM analysis, 100% sureAnton reviewed and tested the series, specifically questioned the breadth of the GUC comment in #113, and is explicitly credited by the commit.
  • Michael Paquiersaw the final versionLLM analysis, 98% sureMichael approved these same comment corrections in #32 and confirmed reviewing and committing them in #33, before their unchanged restoration.
  • Andres Freundsaw the final versionLLM analysis, 94% sureAndres identified extension-driven GUC changes as the reason initialization must remain late in #6 and #8, directly supplying the corrected comment rationale acknowledged in #9.
  • Greg Sabino Mullanesaw the final versionLLM analysis, 95% sureGreg caught the change/chance typo in the initialization comment in #16 and approved the corrected version in #18; the retained comment component subsequently changed only lightly.
  • Alvaro Herrerasaw the final versionLLM analysis, 82% sureAlvaro endorsed Nathan's analysis and patch in #5, whose correction of obsolete background-worker comments survived with only wording changes.
  • Bharath Rupireddysaw the final versionLLM analysis, 84% sureBharath confirmed the obsolete worker-count dependency and specifically discussed the inaccurate initialization comment in #2, with only comment wording changes following for this component.
  • Robert Haassaw the final versioncommittercommitted the patch

Apr 2022

1 commit
Revert the addition of GetMaxBackends() and related stuff.
12 Apr 2022Robert Haasmaster15beta1REL_19_STABLE15beta1REL_18_STABLE15beta1REL_17_STABLE15beta1REL_16_STABLE15beta1REL_15_STABLE15beta1reverts 4 commits here
LLM analysis, 100% surepart of a posted patchsetlight changesfrom patchset #753 of 3 reviewers saw the final version

Exactly matches the revert component of #75 by stable patch ID; neither early initialization nor the GUC restriction was committed. Robert explicitly reports committing only the revert in #80.

  • Julien Rouhaudsaw the final versionLLM analysis, 98% sureJulien called the revised series sensible in #76, with his remaining ordering and error-message comments applying to the two components omitted from this commit.
  • Michael Paquiersaw the final versionLLM analysis, 91% sureBefore a revert attachment existed, Michael proposed reverting the accessor because it did not solve extension resource sizing and a replacement needed more design work in #60.
  • Robert Haassaw the final versioncommittercommitted the patch

Jan - Feb 2022

4 commits
Fix typo in multixact.c
10 Feb 2022Michael Paquiermaster15beta1REL_19_STABLE15beta1REL_18_STABLE15beta1REL_17_STABLE15beta1REL_16_STABLE15beta1REL_15_STABLE15beta1
LLM analysis, 100% surelanded as postedfrom patchset #501 of 1 reviewer saw the final version

The single multixact.c comment correction in #50 has the same stable patch ID as the commit; Michael confirms applying it in #51.

  • Michael Paquiersaw the final versioncommittercommitted the patch
Reduce more the number of calls to GetMaxBackends()
10 Feb 2022Michael Paquiermaster15beta1REL_19_STABLE15beta1REL_18_STABLE15beta1REL_17_STABLE15beta1REL_16_STABLE15beta1REL_15_STABLE15beta1
LLM analysis, 100% surelanded as postedfrom patchset #472 of 2 reviewers saw the final version

The async.c, lock.c and lockfuncs.c caching changes exactly match #47 by stable patch ID.

  • Nathan Bossartsaw the final versiontrailerNathan agreed with Michael's analysis in #49 and explicitly accepted the attached changes in #50.
  • Robert Haassaw the final versiontrailerRobert explicitly accepted Michael's attached cleanup in #48, explaining that he had noticed similar cases while reviewing the original accessor patch.
Remove MaxBackends variable in favor of GetMaxBackends() function.
8 Feb 2022Robert Haasmaster15beta1REL_19_STABLE15beta1REL_18_STABLE15beta1REL_17_STABLE15beta1REL_16_STABLE15beta1REL_15_STABLE15beta1
LLM analysis, 100% surelanded as postedfrom patchset #421 of 8 reviewers saw the final version

The accessor conversion in #42 has the same stable patch ID as the commit. Robert's later explanation #46 also says he reviewed and committed Nathan's patch without implementing additional changes.

  • Greg Sabino Mullanesaw an earlier versionLLM analysis, 100% sureGreg tested the patch, proposed dropping accessor flags and checked EXEC_BACKEND behavior in #16, #18 and #20, before later caching and inline-helper changes.
  • Michael Paquiersaw an earlier versionLLM analysis, 100% sureMichael identified additional repeated calls and questioned the narrow accessor approach in #30, and the commit explicitly credits his review.
  • Andres Freundsaw an earlier versionLLM analysis, 99% sureAndres challenged early initialization and proposed hiding MaxBackends behind a checked accessor in #6 and #10, prompting the major redesign.
  • Wang, Shenhaosaw an earlier versionLLM analysis, 98% sureShenhao requested caching accessor results outside loops in #21, which Nathan implemented before further caller and inline-helper revisions.
  • Fujii Masaosaw an earlier versionLLM analysis, 96% sureFujii questioned the scope and caller churn in #23, prompting a temporary macro-based revision before the explicit-accessor design returned.
  • Alvaro Herrerasaw an earlier versionLLM analysis, 87% sureAlvaro said Nathan's original proposal was correct in #5, before it was redesigned into a guarded accessor.
  • Bharath Rupireddysaw an earlier versionLLM analysis, 87% sureBharath reviewed the original initialization reorder and raised use-case and EXEC_BACKEND questions in #2, well before the accessor redesign.
  • Robert Haassaw the final versioncommittercommitted the patch
Fix comments about bgworker registration before MaxBackends initialization
29 Jan 2022Michael Paquiermaster15beta1REL_19_STABLE15beta1REL_18_STABLE15beta1REL_17_STABLE15beta1REL_16_STABLE15beta1REL_15_STABLE15beta1
LLM analysis, 100% surelanded as postedfrom patchset #316 of 6 reviewers saw the final version

The postmaster.c and postinit.c comment corrections exactly match #31 by stable patch ID; Michael reports applying them in #33.

  • Andres Freundsaw the final versionLLM analysis, 95% sureAndres explained the extension GUC changes requiring late initialization in #6 and #8, directly motivating the corrected comments acknowledged in #9.
  • Greg Sabino Mullanesaw the final versionLLM analysis, 96% sureGreg corrected the initialization-comment typo in #16 and approved the revised patch in #18, with only light changes to the extracted comments afterward.
  • Robert Haassaw the final versionLLM analysis, 86% sureRobert endorsed the original patch containing the obsolete-comment corrections in #4, and the eventual comment-only component differs only in wording.
  • Alvaro Herrerasaw the final versionLLM analysis, 84% sureAlvaro endorsed Nathan's original analysis and patch in #5, including the obsolete background-worker comments later extracted for this commit.
  • Bharath Rupireddysaw the final versionLLM analysis, 85% sureBharath confirmed the changed worker-count calculation and discussed the inaccurate initialization comment in #2, before its wording was refined.
  • Michael Paquiersaw the final versioncommittercommitted the patch

Credits across 8 commits

Author
Nathan Bossart5 LLM7
Michael PaquierLLM1
Reviewer
Robert Haascommitter6
Michael Paquiercommitter2
Nathan Bossart1
Committer
Robert Haas5
Michael Paquier3

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