model-writtenrun 5182 - may be wrong, links go to the source messages. Commit-by-commit LLM analysis is in the Commits dialog.
- Ranier Vilela proposed correcting a duplicated isnan check in circle_same so that both circles' radii are examined #1.
- Daniel Gustafsson requested a reproducer, then demonstrated that a NaN-radius circle incorrectly compared equal to a finite-radius circle; the patch corrected that result, and he identified PostgreSQL 12 as the oldest version needing the fix #2 #4.
- Julien Rouhaud highlighted earlier unresolved geometric NaN issues, while Daniel Gustafsson checked the intended equality semantics and reported that testing and code inspection supported this fix #5 #6.
- Ranier Vilela reported a changed regression result on Windows, and Daniel Gustafsson explained that the change was platform-independent and correct because NaN and zero radii should not compare equal #7 #8.
- Ranier Vilela submitted a revised patch incorporating the corrected regression output, which Daniel Gustafsson committed unchanged as 8cb2a22bbb, with backpatches through PostgreSQL 12 #9.
- The thread ended with Ranier Vilela thanking Daniel Gustafsson for the commit; the submitted fix was fully landed #10.