Marius Kintel

Out of scope for this PR, but we should really rethink the cache. Having the possibility of the cache returning different geometry types is already the source of some problems.

Yeah, the idea of caching CGAL objects separately was that conversion to PolySet was somewhat lossy, and the same logic applies to Manifold geometry. Conversion performance is less of a...

Yeah, that's why we might want to introduce a concept of "non-backend-specific geometries". One challenge is that "implicit unions" introduce the concept of GeometryList, which essentially yields geometry trees.

I'll think about it for a bit. Since automated tests don't help us much, we have to be a bit careful.

Yeah, that looks reasonable. Something is odd about the original code though. Why isn't this->root set when the PolySet is convex? I'll try to find a good way of running...

Wait, manifold using only triangles doesn't imply that the PolySet is convex..

Thanks for the discussion! Closing in favour of #4996

..before we start a long thread on why the -D argument is weird: Yes, I'd really like to improve this, but without breaking old code :)

Fixed conflicts

Works for me on macOS as of 2024-03-24 Any updates for other platforms?