mesh.copy() segfaults, can't deep copy rtree.index.Index?? about trimesh HOT 3 CLOSED

Wow, the multiple copy operations sure exposes all sorts of exciting segfaults. After I added a unit test which demonstrates this behavior, I noticed the hard faulting was pretty platform specific and intermittent. In addition to the rtree I noticed some instances where the cached convex hull crashed it as well.

Anyway, I updated mesh.copy to only copy the mesh data and not the cache, and that appears to fix it (so maybe I could cache that freaking r-tree).

Thanks for the report!

from trimesh.

Yes, I suspect this is a Heisenbug -- under a deecopy(), somebody somewhere is smashing memory and the failure is totally dependent on context and stack.

Thanks for the quick followup! I'll take a look at your changes and report any problems.

from trimesh.

Heh yeah, caching those ctypes objects is a recipe for random pain. In additional testing + the unit test added in b9ec0b1 I think this is "fixed" though feel free to re-open if you see it again.

from trimesh.