Open clbarnes opened 4 years ago
This is probably due to randomly selecting the rays for containment checks. As of https://github.com/clbarnes/ncollpyde/commit/0b127e44929c0d0452dc9ee654f4c2f156fc0845 the seed is consistent (but can be made random).
Deterministic rays do not solve this issue.
"Resolved" in that I've added a tolerance to the one benchmark which was failing... Not ideal, so I'll leave this open for now.
This may be resolved by the rebase on parry and the use of parry's enhanced-determinism feature.
https://github.com/clbarnes/ncollpyde/runs/1406818474?check_suite_focus=true
test_ncollpyde_contains
test fails randomly on CI; count of points in the mesh differs by 1.