I've rebased on main, added a description for .points
(maybe it should use double backticks for monospace font in documentation instead?) and updated the performance section of the PR description…
.points
register_class
overrides classes with the same bl_idname
I've updated the existing code so that RNA_raw_type_sizeof
returns a size_t
. This should also fix possible numeric overflow in the PyMem_Malloc
calls in bpy_rna.cc#foreach_getset
.
I made a PR for my changes (blender/blender-addons#105123), it doesn't touch the issue of the dual tetrahedron being flipped, it's only for the normals issue.
Flipping polygon normals can be done by reversing the order of the vertices of each face, I've done this for each of the scaled * -1 cases in my fork: https://projects.blender.org/Mysteryem/blender…
I think it would be better to try and find the cause of the problem in the existing code and fix that, than to add a bunch of code to 'fix' things afterwards.
Since the FBX shape key data is sparse, it may sound like a good idea to individually access and set where there is sparse data, but from profiling done back when I updated FBX IO to use…