During development my method of testing (other than running cargo test of course) was to run the demo on a sample OBJ file and compare its output with the original. This obviously isn't scalable and doesn't ensure correctness.
We should have a continuously integrated test suite which ensures correctness of meshopt-rs by comparing its results with the original implementation on multiple input datasets, and a (separate?) benchmarking suite where performance improvements can be tracked (and ideally, compared to C++).
During development my method of testing (other than running
cargo testof course) was to run the demo on a sample OBJ file and compare its output with the original. This obviously isn't scalable and doesn't ensure correctness.We should have a continuously integrated test suite which ensures correctness of
meshopt-rsby comparing its results with the original implementation on multiple input datasets, and a (separate?) benchmarking suite where performance improvements can be tracked (and ideally, compared to C++).