In those lines of the algorythm explanation:
3. If no part of [a1..a2] is visible, mark the cell as not visible and advance to next cell
4. If some part of [a1..a2] is visible, merge it into the shadow queue; mark the cell as visible
Isn't it that the angle pair should be merged into the shadow queue ONLY if it's a blocking cell? Because according this explanation we merge any of angle pair into the shadows. Which is wrong I assume.
--Fireton (talk) 15:24, 16 August 2016 (CEST)