Meshing issues on demo inputs with commit 0d5ebca

[mzucker]

Pulled commit 0d5ebca today and ran the demos:

cargo run --release --bin fidget-cli -- --input models/bear.vm mesh --eval jit --depth 7 --out bear_7.stl
cargo run --release --bin fidget-cli -- --input models/bear.vm mesh --eval jit --depth 8 --out bear_8.stl
cargo run --release --bin fidget-cli -- --input models/bear.vm mesh --eval jit --depth 9 --out bear_9.stl

All three of the files exhibit jagged points in MacOS Preview, as shown here:

When viewed in MeshLab, there are some apparent triangles with incorrect orientations:

These disappear if you select the "Double" option for back-face rendering:

The colonnade demo input also produced "spikies" when run at depth 7:

cargo run --release --bin fidget-cli -- --input models/colonnade.vm mesh --eval jit --depth 7 --out colonnade_7.stl    

The spikies disappear when run at depth 8:

I didn't see any issues in MeshLab with incorrectly wound triangles with the colonnade demo.

Rust information: rustc 1.83.0 (90b35a623 2024-11-26)

System information: Mac OS Sonoma 14.7.2 (23H311) on a 13-inch 2020 MacBook Pro with 2 GHz Quad-Core Intel Core i5 processor.

Please let me know if you want me to upload STL outputs or provide any additional system info.

[mkeeter]

Thanks! I'm hoping this is the same issue as #225, because that's a simpler model to investigate.

[Wulfsta]

@mkeeter is it possible the back facing triangle issue is the same as #198?

[mkeeter]

I think this is improved by #232. Meshing is sensitive to carefully selected tuning parameters, so I've tweaked them a little bit and no longer see the spikes.

[mzucker]

Spikes are indeed gone but there seem to remain some back-facing triangles, at least according to MeshLab:

Meshlab also reports issues with non-manifold edges/faces.

This is bear at depth 7.