[Algorithms] more plane-poly clipping woes

[Charles B. <cb...@cb...>]

Well, I'm now using "thick" BSP planes as suggested
to me by Angel Popov.  This basically means treating
all your BSP planes as having some thickness; then
any vert which lies in that thickness is treated as
being already on the plane.  When you clip a poly
against a plane, the verts which are "on" the plane
are not modified.  This prevents you from clipping
segments where one point has a distance of 999
and the other has a distance of -0.001

The problem is that the clipper is still not consistent.
That is, if I take a polygon and clip it against a
plane to produce two new polygons, and then test those
against the plane, they should tell me they're on the front
and back sides; instead they can occasionally report that
they intersect the plane, because a vertex produced by
clipping was not inside the thickness of the plane.

The only solution I can think of for this is to have a
larger epsilon for plane-testing than for plane-clipping
(double seems to work), but that concerns me a bit (is
there a way that could lead to inconsistency?).



--------------------------------------
Charles Bloom           www.cbloom.com