model-writtenrun 5324 - may be wrong, links go to the source messages. Commit-by-commit LLM analysis is in the Commits dialog.
- Amit Langote relayed Simon's off-list report that enabling or disabling a statement trigger on a partitioned table incorrectly searched for that trigger on partitions, and proposed execution-time recursion restricted to row triggers while preserving ONLY semantics in #1.
- Zhihong Yu questioned whether new enum values would disturb existing numeric values in #2; Peter Eisentraut explained that this mattered for backpatching in #3, and Amit agreed in #4.
- Dmitry Koval found that the first patch failed to update child triggers when the parent already had the requested state in #6; Amit corrected the recursion condition and added regression coverage in #7, which Dmitry declared ready for a committer in #8.
- Alvaro Herrera replaced the proliferation of recursive enum variants with a boolean in AlterTableCmd in #10, proposing that other ALTER TABLE operations could eventually use the same mechanism.
- Tom Lane raised cached parse-tree mutation concerns in #11, prompting Amit to consider runtime storage in #12; Alvaro established that ATPrepCmd already copies the node in #13, resolving Tom's concern and prompting a documentation request in #14.
- Amit endorsed the boolean approach and suggested naming it recurse in #15; Alvaro incorporated that name and explanatory comments in #16, alongside an illustrative ADD COLUMN conversion explicitly deferred because broader cleanup remained incomplete and backpatching risked ABI changes and destabilization.
- Amit approved the revised trigger fix in #17; Alvaro proposed preserving the existing EnableDisableTrigger() API on stable branches through a separate EnableDisableTriggerNew() entry point in #18, and Amit agreed in #19.
- Alvaro committed the trigger-fix portion of #16 as ec0925c22a, with only comment wrapping and whitespace adjustments, crediting Amit as co-author and Dmitry as reviewer and backpatching through PostgreSQL 11.
- Alvaro reported in #20 that the committed tests exposed an underspecified ORDER BY on the buildfarm and that he had pushed a follow-up fix; Amit acknowledged completion in #21, leaving the broader ALTER TABLE recursion cleanup deferred.