The problem at the moment is that we need to pass info between 2 line art calculations so we know "if the edge is projected onto this exact triangle".

This problem doesn't only affect shadow, but if we solve it properly it should also benefit stuff like:

• Shadow region edge removal.
• Illuminated light contour only.
• Other styles that require 2-time calculation from the light object and the camera.

Current idea is to use index but thread loading has made the it a bit problematic, (Also, the object will cull differently in visibility check stage). Will come up with a efficient way of doing it.

After some struggle today I get the occlusion right for the front/back facing shadow receiving surface. Due to occlusion function precision issue, light contour will sometimes be a bit problematic as they are essentially "overlapping edges" in that context. This needs to have an architectural solution as current shadow is algorithmetically correct, we do not have a way to share "this is the matching edge that we tested in the shadow computation" between two runs of line art. Due to different culling, we might not be able to use a set of loaded geometry again which enables shadow and occlusion calculation in one go, so there needs to be a data passage. This will also enable us to select edges that are in the occluded region differently as stated before.