Not the most elegant fix but should do the trick. The bug was in the way I tracked halfedge ids associated with convex hull's edges — when an opposing edge of a flipped quad also belonged to the convex hull (usually in the very beginning of a triangulation), flipping would invalidate the associated adjacent edge id of that edge.
Closes #11. cc @redblobgames
Not the most elegant fix but should do the trick. The bug was in the way I tracked halfedge ids associated with convex hull's edges — when an opposing edge of a flipped quad also belonged to the convex hull (usually in the very beginning of a triangulation), flipping would invalidate the associated adjacent edge id of that edge.