Jul 2020

3 commits
Weaken type-OID-matching checks in array_recv and record_recv.
21 Jul 2020Tom Lanemaster14beta1REL_19_STABLE14beta1REL_18_STABLE14beta1REL_17_STABLE14beta1REL_16_STABLE14beta1REL_15_STABLE14beta1REL_14_STABLE14beta1
LLM analysis, 100% surelanded as postedfrom patchset #89

The attachment in #89 and the commit have identical stable patch IDs: all changes in proto.c, arrayfuncs.c, and rowtypes.c match, with only intervening base changes.

Fix construction of updated-columns bitmap in logical replication.
20 Jul 2020Tom Lanemaster14beta1REL_19_STABLE14beta1REL_18_STABLE14beta1REL_17_STABLE14beta1REL_16_STABLE14beta1REL_15_STABLE14beta1REL_14_STABLE14beta1REL_13_STABLE13beta3+3 more
LLM analysis, 100% surewritten from the discussion, no matching patch

Peter Geoghegan reported the invalid read in #90, and Tom Lane identified the missing column mapping in #91; no attached patch implements this fix. The commit corrects updatedCols construction in worker.c, adds array-bound assertions, and extends 003_constraints.pl.

Allow logical replication to transfer data in binary format.
18 Jul 2020Tom Lanemaster14beta1REL_19_STABLE14beta1REL_18_STABLE14beta1REL_17_STABLE14beta1REL_16_STABLE14beta1REL_15_STABLE14beta1REL_14_STABLE14beta1
LLM analysis, 100% suremodified before commitmoderate changesfrom patchset #801 of 4 reviewers saw the final version

The latest matching submission is Dave Cramer's #80, and Tom Lane confirms committing it with adjustments in #89. Edits add pg_dump support and binary-option worker restarts, restore logicalrep_read_update/delete APIs, revise tuple storage and binary-input validation, check typsend, and guard requests to older publishers.

  • Dmitry Dolgovsaw an earlier versionLLM analysis, 99% sureDmitry questioned the relation-ID API changes in #37 and identified the missing GetSubscription binary assignment in #41, both addressed before subsequent tuple-storage and committer changes.
  • Petr Jelineksaw an earlier versionLLM analysis, 99% surePetr reviewed cursor initialization and tuple-storage overhead in #54 and #56, explained why architecture checks were unnecessary in #74, and is explicitly credited as a reviewer by 9de77b5453.
  • Daniel Gustafssonsaw an earlier versionLLM analysis, 99% sureDaniel supplied tests and detailed code feedback in #69, reported testing in #71, and completed cleanup in #77, before moderate committer changes in 9de77b5453.
  • Tom Lanesaw the final versioncommittercommitted the patch

Credits across 3 commits

Author
Dave CramerLLM1
Tom LaneLLM1
Reviewer
Tom Lanecommitter1
Committer
Tom Lane3

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