I have to admit that one-week sprint is too short even for such small project, next sprints will be two-weeks, and marked by even week number.

Currently I stuck a bit with proper shadows drawing, it is really tangled when the spatial structure is not Euclidean, so I need more time to solve these issues. Traditional shadow maps are not working because:

• they have to be layered for different parts of space, which really increases memory expenses;
• stencil buffer have to be separated from depth buffer, which is not supported;
• peter-panning can be solved by switching between backface and frontface culling, but portal rendering messes this technique;
• a lot of lesser issues mixed with these above.

Hope next week I’ll came with a solution both good-looking and efficient.

Interesting: how this problem was solved in Portal by Valve? Or they don’t have portal shadows?