Apr 2025

4 commits
Use streaming read I/O in autoprewarm
4 Apr 2025Melanie Plagemanmaster18beta1REL_19_STABLE18beta1REL_18_STABLE18beta1
LLM analysis, 99% surepart of a posted patchsetmoderate changesfrom patchset #410 of 5 reviewers saw the final version

Commits the fourth component of #41, including AutoPrewarmReadStreamData and its typedef registration. The substantive edit replaces READ_STREAM_FULL with READ_STREAM_DEFAULT | READ_STREAM_USE_BATCHING, changing read ramp-up and enabling AIO batching; the patch credits Nazir and Melanie as co-authors.

  • Heikki Linnakangassaw an earlier versiontrailerIn #38, Heikki reviewed the nested-loop series, questioned buffer-cache overshoot, and identified the missing tablespace comparisons before subsequent logic changes.
  • Daniel Gustafssonsaw an earlier versiontrailerDaniel requested comment, interrupt-check, and assertion changes in #40 and approved the posted final series in #42, but the committed stream flags subsequently changed behavior.
  • Andrey M. Borodinsaw an earlier versiontrailerAndrey identified read-ahead overshoot in #2, proposed documenting or limiting it in #6, and supported readiness in #14 before the control-flow rewrites.
  • Kirill Reshkesaw an earlier versiontrailerKirill reported successful testing and requested a bounds assertion in #11, then supported readiness in #13 before the substantial rewrites.
  • Matheus Alcantarasaw an earlier versiontrailerMatheus reproduced performance improvements, tested a larger buffer pool, and questioned callback state in #8, well before the implementation was restructured.
Refactor autoprewarm_database_main() in preparation for read stream
4 Apr 2025Melanie Plagemanmaster18beta1REL_19_STABLE18beta1REL_18_STABLE18beta1
LLM analysis, 100% surepart of a posted patchsetlight changesfrom patchset #411 of 2 reviewers saw the final version

Commits the third component of #41; its contrib/pg_prewarm/autoprewarm.c diff is identical, while the commit message further explains split control between the callback and main loop. Both the attachment and commit credit Nazir and Melanie as co-authors.

  • Heikki Linnakangassaw an earlier versiontrailerHeikki endorsed the clearer refactoring in #38, before tablespace handling and database-boundary checks were revised in #39.
  • Daniel Gustafssonsaw the final versiontrailerDaniel reviewed the refactoring and requested wording fixes in #40, then approved #41 in #42, after which its code was unchanged.
Fix autoprewarm neglect of tablespaces
4 Apr 2025Melanie Plagemanmaster18beta1REL_19_STABLE18beta1REL_18_STABLE18beta1
LLM analysis, 100% surepart of a posted patchsetlight changesfrom patchset #412 of 3 reviewers saw the final version

Exactly matches the first component of #41, adding tablespace checks at all three relation/fork transition sites in contrib/pg_prewarm/autoprewarm.c. It lands only that independent fix from the four-patch series.

  • Heikki Linnakangassaw an earlier versionLLM analysis, 98% sureHeikki identified the missing tablespace comparisons while reviewing #37 in #38, directly prompting this fix, but did not visibly review its later implementation.
  • Daniel Gustafssonsaw the final versionLLM analysis, 88% sureDaniel gave an overall positive review of #39 in #40 and approved the final series in #42, which included the unchanged tablespace fix.
  • Nazir Bilal Yavuzsaw the final versionLLM analysis, 94% sureNazir explicitly approved the final series #41 in #43, including Melanie's independently authored tablespace fix.
Remove misleading read stream asserts in a few users
3 Apr 2025Melanie Plagemanmaster18beta1REL_19_STABLE18beta1REL_18_STABLE18beta1
LLM analysis, 100% surewritten from the discussion, no matching patch1 of 1 reviewer saw the final version

Daniel questioned the redundant exhaustion assertion in #40, and Melanie proposed cleaning up other users in #41. The commit removes those assertions from amcheck, GiST, B-tree, and SP-GiST; none of the attached patches changes those files.

  • Daniel Gustafssonsaw the final versionLLM analysis, 99% sureDaniel's pre-commit review in #40 explained why the redundant assertion misleadingly suggested a race, and Melanie explicitly adopted that reasoning for other users in #41.

Credits across 4 commits

Author
Melanie PlagemanLLM1
Reviewer
Daniel Gustafsson2
Heikki Linnakangas2
Andrey Borodin1
Kirill Reshke1
Matheus Alcantara1
Reported by
Heikki Linnakangas1
Co author
Melanie Plageman2
Nazir Bilal Yavuz2
Committer
Melanie Plageman4

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