model-writtenrun 6337 - may be wrong, links go to the source messages. Commit-by-commit LLM analysis is in the Commits dialog.
- Kyotaro Horiguchi reported that invalid ALTER TABLE operations produced an uninformative wrong-type error and proposed explicit missing relation-kind combinations for master and PostgreSQL 11–13, also removing unused combinations, in #1.
- Michael Paquier explained that translation required complete error sentences and requested regression coverage in #2; Horiguchi supplied a supplementary test patch covering the live combinations in #4.
- After the work was deferred to the next commitfest in #5, Ahsan Hadi reported passing installcheck-world and meaningful errors for various object types in #7.
- Horiguchi pointed to Peter Eisentraut's broader redesign in #8, and Paquier favored that approach for long-term maintenance while initially questioning the need for backpatches in #10.
- Eisentraut reported that his separate redesign had resolved the PostgreSQL 15 behavior, supported narrower fixes for older branches, and requested retaining apparently unused combinations in #11; Horiguchi supplied revised PostgreSQL 11–14 patches in #12.
- Following Paquier's renewed request for tests in #13, Horiguchi supplied branch-specific fixes and tests in #14, explaining that PostgreSQL 11–12 rejected ATTACH PARTITION earlier during parsing and that PostgreSQL 14 no longer needed the ATT_INDEX case.
- Paquier applied the final submission as 0c83eb2e0e, adding the missing combinations and targeted tests while preserving existing cases; exact attachment equivalence cannot be checked because the final patches were not extracted.
- Paquier confirmed application and justified retaining the currently unreachable older-branch combination against future code changes in #15; Horiguchi acknowledged completion in #16, leaving no further work proposed in this thread.