Robustness issue with StraightSkeleton and inexact construction kernel
Looking at the StraightSkeleton for the input reported in ticket #80 (closed) reveals a robustness issue that results in a piece of skeleton falling outside the input polygon, see:
The artifact goes away if the input polygon is first translated to have 0,0 as its lowest-left point and the output skeleton is translated back to its original position, see:
The skeleton computed with "extact construction" kernel is otherwise visually equal to the one computed with the improved robusness (shifting). Running the translation doesn't significantly affect performances.
Maybe this issue should be reported upstream to CGAL..