Apr - May 2019

6 commits
Remove reindex_catalog test from test schedules.
10 May 2019Andres Freundmaster12beta1REL_19_STABLE12beta1REL_18_STABLE12beta1REL_17_STABLE12beta1REL_16_STABLE12beta1REL_15_STABLE12beta1REL_14_STABLE12beta1REL_13_STABLE12beta1+6 more
LLM analysis, 100% surewritten from the discussion, no matching patch

No attachment removes reindex_catalog from the regression schedules; the commit implements the plan in #69 while retaining the test files for manual execution. Andres reports the development-branch landing in #81.

Remove RelationSetIndexList().
3 May 2019Tom Lanemaster12beta1REL_19_STABLE12beta1REL_18_STABLE12beta1REL_17_STABLE12beta1REL_16_STABLE12beta1REL_15_STABLE12beta1REL_14_STABLE12beta1REL_13_STABLE12beta1+1 more
LLM analysis, 85% surepart of a posted patchsetmoderate changesfrom patchset #2

The RelationSetIndexList() body and declaration deletions exactly match the cleanup portion of #2, but the commit additionally removes transaction-end forced-list cleanup and converts rd_indexvalid to bool. Tom Lane implemented this expanded follow-up after explicitly leaving it out of #62; the original CatalogIndexInsert() approach was not adopted.

Fix reindexing of pg_class indexes some more.
2 May 2019Tom Lanemaster12beta1REL_19_STABLE12beta1REL_18_STABLE12beta1REL_17_STABLE12beta1REL_16_STABLE12beta1REL_15_STABLE12beta1REL_14_STABLE12beta1REL_13_STABLE12beta1+6 more
LLM analysis, 100% suremodified before commitlight changesfrom patchset #621 of 1 reviewer saw the final version

The commit preserves #62's index.c and relcache.c changes, adding an index-kind assertion requested in #63 and polishing comments and an error message. The mapped-index update suppression, explicit XID assignment, and invalidation mechanism are unchanged.

  • Andres Freundsaw the final versionLLM analysis, 99% sureAndres challenged the index-hiding implementation in #54 and reviewed #62 in #63, requesting the adopted index-kind assertion and identifying a stale comment.
Run catalog reindexing test from 3dbb317d32 serially, to avoid deadlocks.
1 May 2019Andres Freundmaster12beta1REL_19_STABLE12beta1REL_18_STABLE12beta1REL_17_STABLE12beta1REL_16_STABLE12beta1REL_15_STABLE12beta1REL_14_STABLE12beta1REL_13_STABLE12beta1+6 more
LLM analysis, 100% surewritten from the discussion, no matching patch

No attached patch contains this test split: it moves catalog REINDEX checks from create_index into a separately scheduled reindex_catalog test. This implements the decision in #41 and #42 to retain coverage while investigating cache-clobber failures.

Fix potential assertion failure when reindexing a pg_class index.
30 Apr 2019Andres Freundmaster12beta1REL_19_STABLE12beta1REL_18_STABLE12beta1REL_17_STABLE12beta1REL_16_STABLE12beta1REL_15_STABLE12beta1REL_14_STABLE12beta1REL_13_STABLE12beta1+6 more
LLM analysis, 99% surepart of a posted patchsetmoderate changesfrom patchset #210 of 1 reviewer saw the final version

This extracts the reindex_index() ordering fix and CatalogIndexInsert() HOT assertions from #21, leaving its storage/table-AM changes to the companion commit. The committed version moves setup outside PG_TRY, integrates HOT assertions into the existing index loop, and adds core and logical-decoding regression tests.

  • Tom Lanesaw an earlier versiontrailerTom reviewed #21 in #23, requesting preservation of the non-assert fast path and movement of setup outside PG_TRY; the final assertion-loop implementation and regression tests followed without another posted patch review.
Fix several recently introduced issues around handling new relation forks.
30 Apr 2019Andres Freundmaster12beta1REL_19_STABLE12beta1REL_18_STABLE12beta1REL_17_STABLE12beta1REL_16_STABLE12beta1REL_15_STABLE12beta1REL_14_STABLE12beta1REL_13_STABLE12beta1+1 more
LLM analysis, 100% surepart of a posted patchsetmoderate changesfrom patchset #210 of 1 reviewer saw the final version

This commits the storage/table-AM portion of #21, including explicit RelFileNode arguments, SMgrRelation return values, and moving buffer flushing into the copy implementations. RelationSetNewRelfilenode() was revised before landing, notably changing the mapping update from immediate to deferred; the reindex ordering and HOT assertions landed separately.

  • Tom Lanesaw an earlier versiontrailerTom reviewed the WIP in #23 and opposed moving CommandCounterIncrement() to callers, but the subsequently committed relcache sequencing and deferred mapping update differ from the reviewed attachment.

Credits across 6 commits

Author
Andres Freund2
Tom LaneLLM2
Reviewer
Tom Lane2
Reported by
Michael Paquier2
Committer
Andres Freund4
Tom Lane2

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