[ODE] Trimesh-Trimesh and Triangle Normals

[Peter Kyme]

Pierre Terdiman wrote:
>>Yeah, i found this while i was testing ode for our project a few months
>>ago (or a year by now), but the question is why other users seems to dont
>>need inverting the second object normal, like  the trimesh demo, that
>>seems to run fine in Russ pc and others, that actually tested before
>>releasing the trimesh-trimesh code, idont see how this can be related to
>>the pc's arch/OS i.e P4/winXP
> 
> 
> Random guess: isn't it just because the winding of the mesh is different...
> ?
> 

Surely that would just reverse both normals involved in the collision. 
The problem here is that the normal for only one of the triangles (the 
one belonging to geom2) needs to be flipped.