Mar 2021

5 commits
Fix new pthread code to respect --disable-thread-safety.
13 Mar 2021Thomas Munromaster14beta1REL_19_STABLE14beta1REL_18_STABLE14beta1REL_17_STABLE14beta1REL_16_STABLE14beta1REL_15_STABLE14beta1REL_14_STABLE14beta1
LLM analysis, 100% surelanded as postedfrom patchset #45

The configure.ac and generated configure changes exactly match #45; git patch-id confirms identical patches.

pgbench: Refactor thread portability support.
10 Mar 2021Thomas Munromaster14beta1REL_19_STABLE14beta1REL_18_STABLE14beta1REL_17_STABLE14beta1REL_16_STABLE14beta1REL_15_STABLE14beta1REL_14_STABLE14beta1
LLM analysis, 100% surepart of a posted patchsetlight changesfrom patchset #40

Lands the thread-portability member of #40, replacing Windows pthread emulation with THREAD_* macros and removing invalid pthread_t assumptions. Relative to that attachment, only platform-description comments change.

pgbench: Improve time logic.
10 Mar 2021Thomas Munromaster14beta1REL_19_STABLE14beta1REL_18_STABLE14beta1REL_17_STABLE14beta1REL_16_STABLE14beta1REL_15_STABLE14beta1REL_14_STABLE14beta1
LLM analysis, 100% surepart of a posted patchsetlight changesfrom patchset #402 of 3 reviewers saw the final version

Lands the timing member of #40, preserving Fabien's integer-microsecond arithmetic and revised connection-time accounting. Differences are formatting, output wording and registration of pg_time_usec_t in the pgindent typedef list.

  • Kyotaro Horiguchino review foundtrailerHis timing recommendations in #3 preceded the timing-refactoring patch, and the supplied thread contains no subsequent review of that code.
  • Hayato Kurodasaw the final versiontrailerReviewed timing-patch documentation and code style in #15 and checked the fixes in #18, after which this member received rebasing and light cleanup.
  • Thomas Munrosaw the final versioncommittercommitted the patch
Add missing pthread_barrier_t.
10 Mar 2021Thomas Munromaster14beta1REL_19_STABLE14beta1REL_18_STABLE14beta1REL_17_STABLE14beta1REL_16_STABLE14beta1REL_15_STABLE14beta1REL_14_STABLE14beta1
LLM analysis, 100% surepart of a posted patchsetlight changesfrom patchset #40

Lands the missing-pthread-barrier prerequisite from #40, retaining its mutex/condition-variable implementation and configure checks. Changes are header/comment cleanup, use of c.h, formatting and a pgindent typedef entry.

pgbench: Synchronize client threads.
10 Mar 2021Thomas Munromaster14beta1REL_19_STABLE14beta1REL_18_STABLE14beta1REL_17_STABLE14beta1REL_16_STABLE14beta1REL_15_STABLE14beta1REL_14_STABLE14beta1
LLM analysis, 99% surepart of a posted patchsetmoderate changesfrom patchset #401 of 5 reviewers saw the final version

Lands the synchronization member of the four-patch series #40, adding barrier-initialization error reporting and adjusting the threadless stub in pgbench.c. The attached patch credits Fabien COELHO and Andres Freund, whose original inline barrier appears in #2.

  • Marina Polyakovasaw an earlier versiontrailerIdentified Windows API-version requirements in #23, prompting #24, before the subsequent portability redesign.
  • Kyotaro Horiguchisaw an earlier versiontrailerResponded to the inline prototype #2 with default-barrier and timing recommendations in #3, before any attached series.
  • Hayato Kurodasaw an earlier versiontrailerFound the connection-failure deadlock in #15, supplied #16 and checked its integration in #18, before the portability redesign and final initialization changes.
  • David Rowleysaw an earlier versiontrailerThomas reported his Windows testing and barrier-macro correction in #40, but the committed initialization error handling was added afterward.
  • Thomas Munrosaw the final versioncommittercommitted the patch

Credits across 5 commits

Author
Thomas MunroLLM3
Fabien COELHO2
Andres Freund1
Reviewer
Hayato Kuroda (Fujitsu)2
Kyotaro Horiguchi2
Thomas Munrocommitter2
David Rowley1
Marina Polyakova1
Reported by
Tom Lane1
Committer
Thomas Munro5

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