Comment 10 for bug 1404715

Had to get this working (at least to some extent), so I did a bit of research on the problem. The root cause seems to be that in certain circumstances the edges in the sweep list end up in a wrong order. When after that new edges are inserted they may end up at _very_ wrong positions, and before long the whole list is all messed up.

Unfortunately, I did not have the time to find a real fix, but I worked around the problem by adding some NULL checks to the active_edges function (see the attached workaround patch), which at least makes the SIGSEGV disappear. The result is probably not pixel perfect, as the edges are still in the wrong order, but that's not that big an issue for me.

I am also attaching another test case, reproduced from real live data this time, which demonstrates hopefully a little less pathological case for examination.