Aug 2021

3 commits
Rethink regexp engine's backref-related compilation state.
8 Aug 2021Tom Lanemaster15beta1REL_19_STABLE15beta1REL_18_STABLE15beta1REL_17_STABLE15beta1REL_16_STABLE15beta1REL_15_STABLE15beta1REL_14_STABLE14beta3
LLM analysis, 100% surewritten from the discussion, no matching patch

This unattached rework restores subRE-pointer bookkeeping, avoids replacing existing parenthesized-atom endpoints, and records the actual capturing node in src/backend/regex/regcomp.c 00116dee5a. Tom explained that endpoint-only bookkeeping conflicted with his planned REG_NOSUB optimization and announced the replacement in #9.

Fix use-after-free issue in regexp engine.
8 Aug 2021Tom Lanemaster15beta1REL_19_STABLE15beta1REL_18_STABLE15beta1REL_17_STABLE15beta1REL_16_STABLE15beta1REL_15_STABLE15beta1REL_14_STABLE14beta3
LLM analysis, 100% surelanded as postedfrom patchset #8

The commit diff exactly matches #8 apart from Git index-line abbreviation, including the parseqatom() return-value change, freed-node pointer clearing, and both regression-test files cc1868799c.

Make regexp engine's backref-related compilation state more bulletproof.
8 Aug 2021Tom Lanemaster15beta1REL_19_STABLE15beta1REL_18_STABLE15beta1REL_17_STABLE15beta1REL_16_STABLE15beta1REL_15_STABLE15beta1REL_14_STABLE14beta3
LLM analysis, 99% suremodified before commitmoderate changesfrom patchset #7

The commit takes the endpoint-pair bookkeeping from #7, adds explanatory comments, and changes the nested-capture wrapper's endpoints from lp/rp to s/s2 in src/backend/regex/regcomp.c cb76fbd7ec. Its regression test had already landed with cc1868799c, so it is absent from this diff.

Credits across 3 commits

Author
Tom LaneLLM2
Committer
Tom Lane3

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