RE: [Algorithms] Dynamic plane shifting BSP traversal

[Neil S. <ne...@r0...>]

> Now the collision check:
> First we shift down A to A' and check the swept circle.
> We found that A' splits X->Y into two parts. (collision 
> normal is A.normal) X'->Y is at the front side, we recurse 
> (so now we at the D node).
> We shift up D to D' and we found that X'->Y is at the back 
> side (solid) so we found a collision. Right? No, there's no 
> collision in this scene!

Shouldn't you be shifting C and D inwards (opposite direction to the
normal)? That's what you seem to be doing for A, but you're doing the
opposite for C and D.

- Neil.