Simplification leads to incorrect normals on flat surfaces

[traking]

Hi!

I'm using meshopt_simplify(..., target_error = 1e-2, target_index_count = size_t(src_indices.size() * 0.5f)) and getting following results with attached mesh (in .OBJ):

As you can see there are some incorect normals being selected. Is there anything that I can adjust in algorithm in order to fix this behavior?

Thanks!

desk.zip

[zeux]

This is a duplicate of #158; it requires the simplifier to actually compare the normal metric/error during simplification which currently isn't done. Of course you can, as a temporary measure, recompute normals post-simplification to work around it.