This looks good to me now. You can commit now.
Changing the threshold to 1e-6 is good, but doesn't fix the main problem. The main problem is that the fraction can also be very close to 1 (in the case where the division by 45 degrees gives a…
Whoops, just realized that my code isn't quite right either. Need a fabs() around the fract_45 and fract_90 when comparing against 1.0f.
My test of this code didn't work. The code in my comment on line 1061 works, though you may want to do something else. I also suggest changing the epsilon to 1e-6 for floats from 1e-7.
This isn't quite good enough. The fract_45 can be something like 0.999999 in the case where angle_snapped is just very slightly less than DEG2RADF(45.0f). Similarly for fract_90. This happened to me when I tested this code with mesh bisect -- the if test failed when the division inside the fractf on line1058 yielded 49.99999679159879.
Change 'being' to 'are' (better English)
I am so glad this landed. Thanks for your great work, Lukasz, and thanks Campbell for your help too.
We should update the Blender Manual to describe how to use this. I just had a look and the…
We think this is ready for commit, but Campbell wants to do one final review and test. Hopefully this will be committed this week.
Thanks Mike93. I am just checking that this means that I am waiting for an update from you before looking at this further. You only need reply if that's an incorrect assumption.
Hi Mike93,
The patch is now unbuildable because of the change in where libraries are. I can fix this if you like, but might be better if you merged a current main into your branch and made it…
Rather than try to modify this pull request to something that would be completely different, I just committed [8549bc94deaa]([url](https://projects.blender.org/blender/blender/commit/8549bc94deaa91…
The problem is that the code assumes that if there is more than one segment then the corner pattern will be M_TRI_FAN (for which there will be an early return) or M_ADJ (where all those mesh_vert()…
