[ODE] Opcode TrisMesm collisions

[Erwin de Vries]

> > Is "double-sided collisions" hard to code into OPCODE ? :)
>
> It's already there.
>
> OPCODE only does "collision detection", not "contact generation". I report
> all triangles touched by, say a sphere, regardless of winding - since the
> sphere and triangles would collide anyway, in both CW or CCW mode.
>
> For ray-mesh you even have a "culling" boolean parameter to tell the
system
> to backface cull the faces or not.
>
> So the problem could be in the contact generation part instead, i.e. the
> code contributed by Erwin.

Yea yea blame it on me again. :-)

Pierre's right though. The problem with double sided contacts is that you
need to determine on which side of a triangle the object is. In the case of
a sphere this means that you have to determine on which side the center is.
I guess you can do the same for a box, although it is less obvious. Problem
with this method is that you're essentially reducing the robustness for fast
moving objects by a factor 2. No need to explain that, right? I agree that
it should be an option on a per triangle basis though, because it can be
very useful at times.

Erwin