Aug 2026

14 commits
Clarify logic in CreateSubscription().
15 Aug 2026Jeff DavismasterunreleasedREL_19_STABLE19beta4
LLM analysis, 100% surelanded as postedfrom patchset #251 of 1 reviewer saw the final version

The subscriptioncmds.c diff exactly matches #25, including the stmt->conninfo condition and both assertions.

  • Tom Lanesaw the final versionLLM analysis, 100% sureAfter reporting the Coverity concern in #24, he explicitly approved the attached cleanup in #26 before it was committed.
When changing owner of a subscription, do not throw an error.
5 Aug 2026Jeff DavismasterunreleasedREL_19_STABLE19beta3
LLM analysis, 100% surepart of a posted patchsetlight changesfrom patchset #152 of 2 reviewers saw the final version

Extracts the owner-change component of #15 unchanged: subscriptioncmds.c removes the USAGE error and changes missing-mapping validation to WARNING, with matching regression tests.

  • Shlok Kyalsaw the final versiontrailerHe tested the patches and confirmed that the ownership issue reported by Hayato Kuroda was fixed in #21.
  • Amit Kapilasaw the final versiontrailerHe reviewed the overall series and endorsed its design and high-level code in #22.
Revert "Validate subscription conninfo on owner change"
5 Aug 2026Jeff DavismasterunreleasedREL_19_STABLE19beta3revert
LLM analysis, 100% surepart of a posted patchsetlight changesfrom patchset #151 of 1 reviewer saw the final version

Exactly matches the revert component of #15, removing owner-change conninfo validation, its documentation, and its tests while retaining the DROP SERVER tests.

  • Amit Kapilasaw the final versiontrailerHe endorsed the overall series after reviewing its design and high-level code in #22.
CREATE SUBSCRIPTION: do not construct conninfo unnecessarily.
5 Aug 2026Jeff DavismasterunreleasedREL_19_STABLE19beta3
LLM analysis, 100% surepart of a posted patchsetlight changesfrom patchset #151 of 1 reviewer saw the final version

Exactly matches the CREATE SUBSCRIPTION component of #15, constructing server-derived conninfo only when opts.connect is true and otherwise checking FDW connection capability.

  • Amit Kapilasaw the final versiontrailerHe endorsed the overall series after reviewing its design and high-level code in #22.
For subscription DDL, demote user mapping checks to WARNING.
5 Aug 2026Jeff DavismasterunreleasedREL_19_STABLE19beta3
LLM analysis, 100% surepart of a posted patchsetlight changesfrom patchset #151 of 1 reviewer saw the final version

Exactly matches the user-mapping component of #15, introducing GetUserMappingExtended() in foreign.c and using WARNING for subscription DDL checks.

  • Amit Kapilasaw the final versiontrailerHe endorsed the overall series after reviewing its design and high-level code in #22.
Always check foreign-server USAGE when resolving subscription conninfo.
5 Aug 2026Jeff DavismasterunreleasedREL_19_STABLE19beta3
LLM analysis, 99% surepart of a posted patchsetlight changesfrom patchset #153 of 4 reviewers saw the final version

Exactly extracts the unconditional foreign-server USAGE-check component of #15. Its Co-authored-by credit and Jeff's explanation in #8 identify Shlok Kyal's contribution.

  • Shlok Kyalsaw the final versiontrailerHe confirmed by testing in #21 that his reported foreign-server permission issue was fixed.
  • Hayato Kuroda (Fujitsu)saw an earlier versiontrailerHe identified the REFRESH PUBLICATION privilege problem and discussed deferred validation in #6, before the consolidated implementation, without identifying a reviewed attachment.
  • Amit Kapilasaw the final versiontrailerHe endorsed the overall series after reviewing its design and high-level code in #22.
  • Jeff Davissaw the final versioncommittercommitted the patch
Be precise about when ALTER SUBSCRIPTION needs conninfo.
5 Aug 2026Jeff DavismasterunreleasedREL_19_STABLE19beta3
LLM analysis, 99% surepart of a posted patchsetmoderate changesfrom patchset #151 of 3 reviewers saw the final version

Extracts the ALTER conninfo-selection component of #15, dropping its ENABLE case and adding ENABLE/DISABLE tests to accommodate the retain_dead_tuples fix. The patch credits Shlok Kyal as co-author, also acknowledged in #8.

  • Hayato Kuroda (Fujitsu)no review foundtrailerAlthough credited in the patch header, his visible reviews in #17 and #20 concern the separate conninfo-loading refactor rather than this ALTER decision logic.
  • Amit Kapilasaw an earlier versiontrailerHe endorsed the posted series in #22, after which the committed component changed ENABLE handling and added regression coverage.
  • Jeff Davissaw the final versioncommittercommitted the patch
Build subscription conninfo after checking that it's enabled.
5 Aug 2026Jeff DavismasterunreleasedREL_19_STABLE19beta3
LLM analysis, 100% surepart of a posted patchsetlight changesfrom patchset #151 of 1 reviewer saw the final version

Exactly matches the worker-ordering component of #15, moving conninfo construction after enabled checks in maybe_reread_subscription() and InitializeLogRepWorker().

  • Amit Kapilasaw the final versiontrailerHe discussed disabled-worker error behavior in #13 and endorsed the updated series in #22.
Remove Subscription conninfo field; generate in caller.
5 Aug 2026Jeff DavismasterunreleasedREL_19_STABLE19beta3
LLM analysis, 100% surepart of a posted patchsetmoderate changesfrom patchset #150 of 2 reviewers saw the final version

Extracts the conninfo refactor from #15, with a local worker.c change that saves old_conninfo, installs the replacement, and then frees the old allocation; other differences are rebase context.

  • Amit Kapilasaw an earlier versiontrailerHis explanatory comment from #13 was adopted and he endorsed the revised series in #22, before the committed allocation-order change.
  • Hayato Kuroda (Fujitsu)saw an earlier versiontrailerHe investigated concurrent catalog invalidation in #17 and resolved his concern in #20, before the committed allocation-order change.
Validate publisher for retain_dead_tuples in the apply worker.
4 Aug 2026Amit KapilamasterunreleasedREL_19_STABLE19beta3
LLM analysis, 100% surelanded as postedfrom patchset #122 of 2 reviewers saw the final version

Matches #12 in subscriptioncmds.c, worker.c, and subscriptioncmds.h; the only diff-text change is comment wrapping.

  • Jeff Davissaw the final versiontrailerHe analyzed the original approach in #10 and explicitly approved the replacement while preferring its ENABLE behavior in #14.
  • Hayato Kurodasaw the final versiontrailerHe tested the replacement on PostgreSQL 19 and master with Jeff's reproducer and reported that it looked good in #16.
postgres_fdw: reject use_scram_passthrough for subscriptions.
3 Aug 2026Jeff DavismasterunreleasedREL_19_STABLE19beta3
LLM analysis, 100% surepart of a posted patchsetlight changesfrom patchset #8

Exactly extracts the SCRAM pass-through rejection component of #8, including postgres_fdw connection checks, TAP coverage, and documentation.

Improve DROP SERVER handling of dependent subscriptions.
3 Aug 2026Jeff DavismasterunreleasedREL_19_STABLE19beta3
LLM analysis, 100% surepart of a posted patchsetlight changesfrom patchset #8

Exactly extracts the DROP SERVER component of #8, generalizing shared-object locks and moving the dependency error after locking; the commit expands its explanatory message.

Fix lock release for role membership grants in DROP OWNED BY.
3 Aug 2026Jeff DavismasterunreleasedREL_19_STABLE19beta3REL_18_STABLE18.6REL_17_STABLE17.11REL_16_STABLE16.15
LLM analysis, 100% surepart of a posted patchsetlight changesfrom patchset #8

Exactly extracts the role-membership lock-release component of #8, including the AuthMemRelationId branch in dependency.c and the new isolation test.

Do not log subscription conninfo.
3 Aug 2026Jeff DavismasterunreleasedREL_19_STABLE19beta3REL_18_STABLE18.6REL_17_STABLE17.11REL_16_STABLE16.15REL_15_STABLE15.19REL_14_STABLE14.24
LLM analysis, 100% surewritten from the discussion, no matching patch

Removes the conninfo DEBUG1 message from worker.c; no attached patch proposes that deletion. The commit attributes it to the initial report, and #23 identifies it as the fix for the logging finding.

Credits across 14 commits

Author
Jeff DavisLLM10
Amit Kapila1
Reviewer
Amit Kapila8
Hayato Kuroda (Fujitsu)4
Jeff Daviscommitter3
Shlok Kyal2
Reported by
Noah Misch10
Tom Lane1
Co author
Shlok Kyal2
Committer
Jeff Davis13
Amit Kapila1

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