tritao · April 23, 2026 14:02
diff --git a/gistfile1.txt b/gistfile1.txt
 - Baseline behavior: a model with adjacent equal planar faces, cylindrical faces, and BSpline faces should refine to the same result as before the PR.
 - No false negatives: equal faces must still merge when their geometric values sit near quantization/grid boundaries.
 - No false positives: faces with the same structural hash/grid key but different actual geometry must not be merged.
 - Mixed bucket case: multiple BSpline faces in the same or neighboring grid cells, where only some are equal, should produce the right equality groups.
 - Plane/cylinder fallback: confirm planes and cylinders still use the old behavior effectively, since their structural hash/grid key now does not optimize them.
 - Performance regression/sanity case: a shape with many same-type faces, especially many BSpline faces, should run faster or at least not worse than the old pairwise path.
	- Baseline behavior: a model with adjacent equal planar faces, cylindrical faces, and BSpline faces should refine to the same result as before the PR.
	- No false negatives: equal faces must still merge when their geometric values sit near quantization/grid boundaries.
	- No false positives: faces with the same structural hash/grid key but different actual geometry must not be merged.
	- Mixed bucket case: multiple BSpline faces in the same or neighboring grid cells, where only some are equal, should produce the right equality groups.
	- Plane/cylinder fallback: confirm planes and cylinders still use the old behavior effectively, since their structural hash/grid key now does not optimize them.
	- Performance regression/sanity case: a shape with many same-type faces, especially many BSpline faces, should run faster or at least not worse than the old pairwise path.
No results found