Mar - Apr 2026

5 commits
Avoid unsafe access to negative index in a TupleDesc.
6 Apr 2026Tom LaneREL_18_STABLE18.4REL_17_STABLE17.10REL_16_STABLE16.14REL_15_STABLE15.18REL_14_STABLE14.23
LLM analysis, 99% surewritten from the discussion, no matching patch

The commit adds only an attno > 0 guard before TupleDescAttr in src/backend/commands/copy.c, implementing the conservative stable-branch approach proposed in #11. Neither attachment contains this guard for COPY; the attached HEAD fix rejects system columns instead.

Disallow system columns in COPY FROM WHERE conditions.
6 Apr 2026Tom Lanemaster19beta1REL_19_STABLE19beta1
LLM analysis, 100% surelanded as postedfrom patchset #111 of 1 reviewer saw the final version

The diff exactly matches #11 across src/backend/commands/copy.c and the copy2 regression files, including system-column rejection, dropped-column handling, and comment cleanups.

  • Peter Eisentrautsaw the final versionLLM analysis, 99% surePeter explicitly agreed with prohibiting system columns in #12, reviewing the proposal attached in #11 before it was committed unchanged.
Add a test for creating an index on a whole-row expression.
24 Mar 2026Robert Haasmaster19beta1REL_19_STABLE19beta1
LLM analysis, 100% surelanded as postedfrom patchset #11 of 1 reviewer saw the final version

The changes to src/test/regress/sql/indexing.sql and its expected output exactly match the whole-row expression index test in the third patch of #1.

  • Tom Lanesaw the final versiontrailerTom endorsed all three patches in #2, including the whole-row index test that landed unchanged.
Bounds-check access to TupleDescAttr with an Assert.
24 Mar 2026Robert Haasmaster19beta1REL_19_STABLE19beta1
LLM analysis, 100% surelanded as postedfrom patchset #11 of 1 reviewer saw the final version

All added and removed lines match the second patch of #1, covering tupdesc.c, tupdesc.h, plperl.c, and pl_exec.c; only rebase-related locations differ.

  • Tom Lanesaw the final versiontrailerTom endorsed the assertion patch in #2, while noting that he had not exhaustively checked every caller and supported limiting it to HEAD.
Prevent spurious "indexes on virtual generated columns are not supported".
24 Mar 2026Robert Haasmaster19beta1REL_19_STABLE19beta1REL_18_STABLE18.4
LLM analysis, 100% surelanded as postedfrom patchset #11 of 1 reviewer saw the final version

Both attno > 0 guards in src/backend/commands/indexcmds.c match the first patch of #1, with only surrounding branch context and line locations differing.

  • Tom Lanesaw the final versiontrailerTom endorsed the fix in #2, confirmed the invalid zero-attribute access using assertions in #4, and reported another affected buildfarm run in #6.

Credits across 5 commits

Author
Robert HaasLLM3
Tom Lane2
Reviewer
Tom Lane3
Reported by
Alexander Lakhin2
Committer
Robert Haas3
Tom Lane2

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