model-writtenrun 9630 - may be wrong, links go to the source messages. Commit-by-commit LLM analysis is in the Commits dialog.
- Peter Eisentraut proposed a regression test and fix in #1 after finding that updates between differently represented but equal primary-key values, such as float -0 and 0, failed to cascade because an optimization skipped the referential action.
- Andrew Gierth questioned the claimed SQL-standard violation in #2, arguing that the standard's requirement for distinct values might refer to SQL equality rather than distinguishable representations.
- Tom Lane reviewed the patch in #3, saw possible performance and user-expectation benefits, but objected to introducing a comparison rule solely for ON UPDATE CASCADE and raised broader consistency concerns.
- Laurenz Albe supplied the standard's definition of distinctness in #4, Andrew defended using the same equality semantics as the foreign-key constraint in #5, and Peter countered that representation-changing updates must still have observable effects in #6.
- Peter explained in #7 that the patch corrected an optimization rather than redefining foreign-key behavior, accepted broader comparison consistency despite possible extra work, and posted #8 with bytewise comparisons for all primary-key update actions plus conflict resolution.
- Peter committed both patches from #8 together as 1ffa59a85c, with only a regression-test comment adjustment beyond rebasing, and confirmed completion in #9; no outstanding patch work was identified.