Jan 2023

4 commits
Round off random_normal() test results one more decimal place.
10 Jan 2023Tom Lanemaster16beta1REL_19_STABLE16beta1REL_18_STABLE16beta1REL_17_STABLE16beta1REL_16_STABLE16beta1
LLM analysis, 100% surewritten from the discussion, no matching patch

No attachment contains this follow-up change from extra_float_digits = 0 to -1 in random.sql and its expected output. The commit cites buildfarm failures, and #38 identifies AIX rounding differences as the reason.

Remove pg_regress' never-documented "ignore" feature.
10 Jan 2023Tom Lanemaster16beta1REL_19_STABLE16beta1REL_18_STABLE16beta1REL_17_STABLE16beta1REL_16_STABLE16beta1
LLM analysis, 100% surewritten from the discussion, no matching patch1 of 1 reviewer saw the final version

Removes ignore-directive parsing, ignored-failure accounting, and related reporting from src/test/regress/pg_regress.c, which no attached patch changes. Tom proposed this cleanup in #33 and reaffirmed it in #35.

  • Dean Rasheedsaw the final versionLLM analysis, 88% sureDean endorsed the proposed removal of failure ignoring in #34, before the cleanup commit; no attachment contained the pg_regress.c removal.
Upgrade the random.sql regression test.
10 Jan 2023Tom Lanemaster16beta1REL_19_STABLE16beta1REL_18_STABLE16beta1REL_17_STABLE16beta1REL_16_STABLE16beta1
LLM analysis, 100% suremodified before commitlight changesfrom patchset #35

Matches the resulting random.sql and expected-output contents of #35 except for replacing a non-ASCII hyphen in comments, and additionally removes ignore: random from parallel_schedule. The commit credits Tom Lane and Dean Rasheed as authors, so Dean is excluded from suggested reviewers.

Invent random_normal() to provide normally-distributed random numbers.
9 Jan 2023Tom Lanemaster16beta1REL_19_STABLE16beta1REL_18_STABLE16beta1REL_17_STABLE16beta1REL_16_STABLE16beta1
LLM analysis, 100% suremodified before commitlight changesfrom patchset #282 of 7 reviewers saw the final version

The implementation matches #28, with a URL-encoding correction in pg_prng.c, a catalog-version bump, and removal of one regression-test iteration with its corresponding duplicate-count adjustment. The commit credits Paul Ramsey, whose patch Tom revised and posted.

  • David G. Johnstonsaw an earlier versionLLM analysis, 100% sureDavid proposed mean-first arguments and clearer transformation code in #2, adopted before the later argument-handling and generator revisions.
  • Justin Pryzbysaw an earlier versionLLM analysis, 100% sureJustin identified the erroneous make_interval binding in #3, which Paul fixed before subsequent implementation revisions.
  • Michael Paquiersaw an earlier versionLLM analysis, 99% sureMichael reviewed declarations and missing tests in #8, requested cheaper tests in #17, and supplied the concrete header-compilation failure in #25, before Tom's implementation fixes.
  • Mark Dilgersaw an earlier versionLLM analysis, 99% sureMark tested negative standard deviations and infinite parameters and reported their surprising results in #10, before later generator and argument-handling changes.
  • Fabien COELHOsaw an earlier versionLLM analysis, 100% sureFabien reported successful compilation and make check, questioned epsilon truncation, and proposed pgbench reuse in #19, which preceded Tom's generator revisions.
  • Dean Rasheedsaw the final versionLLM analysis, 97% sureDean accepted random_normal() while reviewing shared-state and parallel-restriction semantics in #30, then agreed in #32 to defer broader code and documentation reorganization.
  • Tom Lanesaw the final versioncommittercommitted the patch

Credits across 4 commits

Author
Paul RamseyLLM1
Tom LaneLLM1
Reviewer
Tom Lanecommitter1
Committer
Tom Lane4

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