Mar 2021

3 commits
Make test_target_session_attrs more robust against connection failure.
3 Mar 2021Tom Lanemaster14beta1REL_19_STABLE14beta1REL_18_STABLE14beta1REL_17_STABLE14beta1REL_16_STABLE14beta1REL_15_STABLE14beta1REL_14_STABLE14beta1
LLM analysis, 100% surewritten from the discussion, no matching patch

The post-commit broken-pipe failure reported in #142 led Tom Lane to pass SHOW port through psql -c with undef stdin in src/test/recovery/t/001_stream_rep.pl; #145 explains the diagnosis and fix. No attached patchset contains this change.

Extend the abilities of libpq's target_session_attrs parameter.
3 Mar 2021Tom Lanemaster14beta1REL_19_STABLE14beta1REL_18_STABLE14beta1REL_17_STABLE14beta1REL_16_STABLE14beta1REL_15_STABLE14beta1REL_14_STABLE14beta1
LLM analysis, 99% surepart of a posted patchsetsignificant changesfrom patchset #1391 of 8 reviewers saw the final version

Lands the libpq, documentation and test portion of #139, with its server GUC change committed separately. Tom substantially rewrites PQconnectPoll, resets cached server properties, and replaces which_primary_host with a complete second host scan, as explained in #141.

  • Laurenz Albesaw an earlier versionLLM analysis, 100% sureTested the early implementation, found its endless reconnect loop in #3, verified the fix in #6, and reviewed the refactored version in #11.
  • Takayuki Tsunakawasaw an earlier versionLLM analysis, 100% sureReviewed initialization, enum compatibility, connection rejection and documentation in #70, #83, #84, #86 and #96, before subsequent redesigns.
  • Peter Smithsaw an earlier versionLLM analysis, 100% sureReviewed and tested the implementation in #112 and #113, identified duplicate code and a wrong jump in #117, and confirmed the revisions in #119.
  • Michael Paquiersaw an earlier versionLLM analysis, 97% sureExamined the fallback sentinel logic in #17 and advocated a simpler explicit state and a read-only mode, both addressed during later development.
  • Alvaro Herrerasaw an earlier versionLLM analysis, 98% sureReviewed the connection state machine and enum placement in #97, prompting restructuring and internal enum placement in later submissions.
  • Daniel Gustafssonsaw an earlier versionLLM analysis, 98% sureIdentified broken recovery tests caused by renamed node variables in #110, which were repaired in #111.
  • Robert Haassaw an earlier versionLLM analysis, 94% sureChallenged the patch's read-write/primary equivalence in #114, leading to the explicitly acknowledged semantic correction in #116.
  • Tom Lanesaw the final versioncommittercommitted the patch
Mark default_transaction_read_only as GUC_REPORT.
2 Mar 2021Tom Lanemaster14beta1REL_19_STABLE14beta1REL_18_STABLE14beta1REL_17_STABLE14beta1REL_16_STABLE14beta1REL_15_STABLE14beta1REL_14_STABLE14beta1
LLM analysis, 100% surepart of a posted patchsetlight changesfrom patchset #1392 of 4 reviewers saw the final version

Extracts the default_transaction_read_only GUC_REPORT flag and its protocol/libpq documentation from #139. The guc.c change is identical; documentation receives minor grammar and formatting corrections.

  • Laurenz Albesaw the final versionLLM analysis, 99% sureExplicitly credited as a reviewer in d16f8c8e41, although no review of this server-side reporting change by him appears in the thread.
  • Takayuki Tsunakawasaw an earlier versionLLM analysis, 98% sureReviewed the earlier transaction_read_only reporting patch and its documentation in #70, #83 and #96, before the reported variable changed.
  • Peter Smithsaw an earlier versionLLM analysis, 97% sureReviewed the transaction_read_only reporting explanation and cached state in #117, before the switch to default_transaction_read_only, and is explicitly credited by the commit.
  • Tom Lanesaw the final versioncommittercommitted the patch

Jan 2021

1 commit
Introduce a new GUC_REPORT setting "in_hot_standby".
5 Jan 2021Tom Lanemaster14beta1REL_19_STABLE14beta1REL_18_STABLE14beta1REL_17_STABLE14beta1REL_16_STABLE14beta1REL_15_STABLE14beta1REL_14_STABLE14beta1
LLM analysis, 99% surepart of a posted patchsetmoderate changesfrom patchset #1311 of 3 reviewers saw the final version

Lands the in_hot_standby portion of #131, which incorporates Tom's implementation from #121, while excluding transaction_read_only reporting. As #132 explains, the commit removes the extra state variable and show-hook side effects and completes the documentation.

  • Laurenz Albesaw the final versionLLM analysis, 99% sureExplicitly credited as a reviewer in bf8a662c9a, although his visible reviews concern the earlier client-side feature rather than this server GUC.
  • Takayuki Tsunakawasaw an earlier versionLLM analysis, 100% sureReviewed recovery-state initialization and notification placement in #86 and requested server-state documentation in #96, before Tom replaced the signaling design.
  • Peter Smithsaw an earlier versionLLM analysis, 99% sureExamined recovery-exit signaling and server-state reporting in #112 and #117, before the end-of-query implementation replaced it.

Nov 2020

1 commit
Avoid spamming the client with multiple ParameterStatus messages.
25 Nov 2020Tom Lanemaster14beta1REL_19_STABLE14beta1REL_18_STABLE14beta1REL_17_STABLE14beta1REL_16_STABLE14beta1REL_15_STABLE14beta1REL_14_STABLE14beta1
LLM analysis, 100% surelanded as postedfrom patchset #127

Applying #127 to the commit's parent reproduces all four committed files exactly, including last_reported storage and deferred ParameterStatus emission. #129 confirms the push.

Sep 2019

1 commit
Restructure libpq code to remove some duplicity
10 Sep 2019Alvaro Herreramaster13beta1REL_19_STABLE13beta1REL_18_STABLE13beta1REL_17_STABLE13beta1REL_16_STABLE13beta1REL_15_STABLE13beta1REL_14_STABLE13beta1REL_13_STABLE13beta1
LLM analysis, 100% surelanded as postedfrom patchset #902 of 2 reviewers saw the final version

Applying #90 to the commit's parent reproduces libpq-fe.h exactly and fe-connect.c modulo whitespace. The attachment credits Hari Babu Kommi, and #91 confirms Alvaro tested and pushed the cleanup.

  • Takayuki Tsunakawasaw the final versiontrailerReviewed the cleanup's enum compatibility and subsequent revisions in #70, #72, #83 and #86, followed only by equivalent control-flow cleanup, rebasing and documentation separation.
  • Álvaro Herrerasaw the final versiontrailerRevised the state transition in #90, manually checked protocol-v2 behavior in #91, and committed the final version.

Credits across 6 commits

Author
Tom LaneLLM2
vignesh CLLM2
Hari Babu Kommi1
Reviewer
Tom Lanecommitter2
Alvaro Herrera1
Tsunakawa, Takayuki1
Committer
Tom Lane5
Alvaro Herrera1

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