Jun 2021

2 commits
Fix copying data into slots with FDW batching
16 Jun 2021Tomas Vondramaster14beta2REL_19_STABLE14beta2REL_18_STABLE14beta2REL_17_STABLE14beta2REL_16_STABLE14beta2REL_15_STABLE14beta2REL_14_STABLE14beta2
LLM analysis, 100% suremodified before commitlight changesfrom patchset #140 of 1 reviewer saw the final version

Applying #14 to the commit parent reproduces the committed executor code and expected output exactly. The only remaining difference is that the commit keeps the existing 70,000-row INSERT test enabled in contrib/postgres_fdw/sql/postgres_fdw.sql.

  • Alexander Pyhalovsaw an earlier versionLLM analysis, 94% sureAlexander reported duplicate-key failures and supplied the reproducer in #13 that became the regression test, before the executor fix was added in #14.
Optimize creation of slots for FDW bulk inserts
11 Jun 2021Tomas Vondramaster14beta2REL_19_STABLE14beta2REL_18_STABLE14beta2REL_17_STABLE14beta2REL_16_STABLE14beta2REL_15_STABLE14beta2REL_14_STABLE14beta2
LLM analysis, 100% suremodified before commitmoderate changesfrom patchset #100 of 3 reviewers saw the final version

The commit combines both patches from #10, expands comments, and renames ri_BatchInitialized to ri_NumSlotsInitialized. It also changes ExecEndModifyTable's cleanup bound from ri_NumSlots to the initialized-slot count, a functional correction beyond the comment cleanup mentioned in #12.

  • Andres Freundsaw an earlier versionLLM analysis, 96% sureAndres diagnosed repeated descriptor references defeating resource-owner hashing and proposed copying the tuple descriptor in #2, which the implementation adopted before subsequent cleanup-logic changes.
  • Tom Lanesaw an earlier versionLLM analysis, 88% sureTom identified degradation in ResourceArrayAdd as well as ResourceArrayRemove and endorsed using a transient descriptor in #3, before the implementation and later cleanup-logic changes.
  • Bharath Rupireddysaw an earlier versionLLM analysis, 99% sureBharath reported the regression-test crash in #9 and improved test execution time after the correction in #11, but the committed version subsequently changed the cleanup-loop bound.

Credits across 2 commits

Author
Tomas VondraLLM2
Reported by
Alexander Pyhalov1
Committer
Tomas Vondra2

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