Jun 2021

3 commits
Remove libpq's use of abort(3) to handle mutex failure cases.
29 Jun 2021Tom Lanemaster15beta1REL_19_STABLE15beta1REL_18_STABLE15beta1REL_17_STABLE15beta1REL_16_STABLE15beta1REL_15_STABLE15beta1
LLM analysis, 100% surewritten from the discussion, no matching patch1 of 1 reviewer saw the final version

The commit implements Tom Lane's proposal in #8 and #10: remove PGTHREAD_ERROR and replace its mutex-failure uses with Assert(false). None of the attachments changes these mutex paths in fe-connect.c, fe-secure-openssl.c, or libpq-int.h.

  • Daniel Gustafssonsaw the final versionLLM analysis, 100% sureDaniel endorsed replacing mutex-failure aborts with assertions and limiting the change to HEAD in #11, before the commit.
Don't use abort(3) in libpq's fe-print.c.
28 Jun 2021Tom Lanemaster15beta1REL_19_STABLE15beta1REL_18_STABLE15beta1REL_17_STABLE15beta1REL_16_STABLE15beta1REL_15_STABLE15beta1REL_14_STABLE14beta3REL_13_STABLE13.4+4 more
LLM analysis, 99% suremodified before commitlight changesfrom patchset #7

The commit preserves #7's fe-print.c changes: graceful OOM returns, boolean do_field(), and centralized allocation cleanup. Tom consolidates three allocation checks and separates another allocation from its condition, leaving the patch's error-handling design intact.

Remove undesirable libpq dependency on stringinfo.c.
26 Jun 2021Tom Lanemaster14beta3REL_19_STABLE14beta3REL_18_STABLE14beta3REL_17_STABLE14beta3REL_16_STABLE14beta3REL_15_STABLE14beta3REL_14_STABLE14beta3
LLM analysis, 100% surewritten from the discussion, no matching patch3 of 3 reviewers saw the final version

No attachment changes parseServiceFile() in fe-connect.c; the commit follows the dependency report in #3, diagnosis in #4, and revert proposal in #5. It implements Tom's plan in #6 to replace StringInfo with a 1024-byte stack buffer while retaining the single cleanup path.

  • Jacob Championsaw the final versionLLM analysis, 96% sureJacob reported libpq's unexpected pg_get_line_buf() and pfree() symbol dependencies in #3, prompting the service-file fix.
  • Michael Paquiersaw the final versionLLM analysis, 99% sureMichael traced the service-file dependency to initStringInfo() calling exit-on-OOM palloc() and identified it as a PostgreSQL 14 issue in #4.
  • Daniel Gustafssonsaw the final versionLLM analysis, 99% sureDaniel proposed reverting the StringInfo-dependent service-file changes because switching to PQExpBuffer was too invasive for PostgreSQL 14 in #5.

Credits across 3 commits

Author
Tom LaneLLM1
Committer
Tom Lane3

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