Re[6]: [ODE] coldet

[Pierre Terdiman]

> NB>    Yes, sure Coldet can detect contact points (contacts normal is
> NB>    another case, but I think you can have normals from points on mesh)
>
> NB>    Then how is Opcode and coldet different? Sorry , I can't catch you
> NB>    in this...
>
>  I am sorry. I rather new in it and do not know Opcode properly.
>  I decided that it can only be used to find intersecting triangles
>  not contact points.

And you would be right : I just report intersecting triangles indeed.

But (I think) the point is, slapping a dedicated triangle-triangle routine
on returned pairs is probably best done in the client code, since many apps
will probably need different ways to drive their collision response.

That's what ColDet does, using two different triangle-triangle overlap
codes, depending on the situation. The fast one is the standard routine from
Tomas Möller. That's also what I'm using. The slow one (that computes a
collision point) is named "MyTriTri.c" in ColDet. It would be unfortunate to
use it in the generic case since it's slower. But you can grab it and use it
as well on Opcode's returned pairs, maybe (modulo license troubles).

Anyway I'm not sure it's really that useful. Returning a single collision
point is dubious, and that's not even the problem - the problem is the
penetration depth, at least for ODE. [Now, right, 100% Baraff-like
simulators don't need this.]

Well, various libs, various strategies, see for your particular needs.

Pierre