by randi » Mon Apr 17, 2006 10:13 pm
I have not looked at the map in that bug report. However, in reference to the map Enjay mentioned, it really does have a thin, misplaced one-dimensional subsector that he was "lucky" enough to land inside of, which I discovered by using Wintex's node builder.
Changing the point on side function shouldn't help that because it is a genuine defect in the map's nodes. And keep in mind that a floating point routine will never be able to duplicate a fixed point version for diagonal lines.