Some few issues and sigma2 >= sigma1 assert error #6
Description
Hello again, thank you for fixing the compilation errors. I've been testing it and found some issues:
Coherence:
The thekla atlas example tells it "Produced debug_packer_final.tga" but it does not by default because DEBUG_OUTPUT is 0
Changing DEBUG_OUTPUT to 1 will make a new compilation error:
It is including "nvimage/ImageIO.h" but there's no such file. A fix is to:
thekla_atlas/src/nvmesh/param/AtlasPacker.cpp
Line 842 in e393793
Comment the include line, line 843 and uncomment line 842.
sigma2 >= sigma1
I have attached an object file in this issue, this model will produce sigma2 >= sigma1 assert error, it does come from:
The problem is that t1 s1 and t2 s2 are IND0 (indefinite).. I was trying to track where it gets indefinite, it comes from this:
thekla_atlas/src/nvmath/Solver.cpp
Line 352 in e393793
There's a division by 0 happening in there that does that. I did try override alpha to -0.000001f if it's 0.0f so it could bypass this sigma assert, but it does explode the UVs (the triangles spikes everywhere in texture coordinates). There's also an overload of the same function and equation happening a little above from this one but I'm not sure if it's being used.
I also did build an obj validator to make sure it doesn't have null indices etc, everything is fine, every program loads the model fine.
Sorry if I misspelled something, English is not my main language and my C/C++ knowledge is also very limited.