model-writtenrun 6307 - may be wrong, links go to the source messages. Commit-by-commit LLM analysis is in the Commits dialog.
- Alexander Lakhin reported that passing NULL to an anycompatiblemultirange function produced a cache lookup failure instead of a useful polymorphic-type error in #1.
- Neil Chen supplied explicit unknown-input checks and an expected-error update in #2, which Alvaro Herrera considered sensible in #3.
- Alexander Korotkov's deeper review found that inference from anycompatiblerange to anycompatiblemultirange and in the reverse direction also failed, requiring broader work than the initial fix #5 #6.
- Alexander Lakhin then reported an unexpected-operator error when comparing a multirange column with an empty range in #9, and Alexander Korotkov identified it as a separate selectivity-estimation bug in #10.
- Alexander Korotkov corrected the selectivity operator switches and added empty-range and empty-multirange tests in #11, committing that patch with only whitespace adjustments as 178ec460db and announcing the push in #12.
- Alexander Korotkov questioned whether enforcement errors could become assertions in #13, but Tom Lane retained the checks and supplied a broader polymorphic-resolution patch based partly on Neil's work in #15, reorganizing type inference and adding regression tests.
- Tom Lane separately proposed letting a uniquely matching polymorphic candidate reach enforcement to produce more informative errors in #17, while Alexander Korotkov questioned whether that could misrepresent the user's intended function in #18.
- Alexander Korotkov approved the reworked patch and its handling of mismatched range/multirange arguments in #19, and Tom Lane committed #15 with a comment clarification as 336ea6e6ff.
- Both bug fixes had landed by the end of the thread, while broader diagnostic improvements remained a proposal: Tom Lane suggested retaining a no-matching-function primary error with richer details in #21, and Alexander Korotkov supported that direction in #22.