[ODE] Triangle-box collider contribution

[Alen Ladavac]

> I believe that we don't have a working capsule-triangle case, so I
> think that'd be most welcome.

No problem. It is one of the simplest ones and most tested on our side, so I
guess it will be good.

> Box-triangle was the most noticeable problem and this code seems
> to address all of the glitches I'd hitherto noticed.

I'm glad to hear that. It also makes me feel safer about it being used on
our side as well - one of the best aspects of sharing your code. ;)

> Our cylinder-anything code seems problematic, so that's another
> good candidate (at the moment cylinders are only a contrib, not a
> core primitive;

Well, our cylinder-something colliders seem to work ok on our side so far.
To be honest we had to do some fakes. When it needs to stand up, we
approximate its base with an octagon. But for rolling we have a special case
that can generate 2 contacts quickly. The code looks a bit ugly, but works
well.


> I don't know whether it makes sense for ODE to
> ever promote cylinders to a core primitive though).

Why not? They are a necessity if you want to model barrels and cans, because
they can both stand upwards and roll. Useful for wheels as well, if you can
afford it - spheres are fast but very poor aproximation for wheels.

> How would such a porting happen -- would you require NDA, or
> simply supply a block of code with its interfaces dangling?  If
> the latter, we could at least put the code in a 'porters welcome'
> directory somewhere (i.e. in ODE CVS).

I'll just mail the snippets with some guidelines to the porters, and provide
them with explanations on the math library if anything is unclear.

> Thanks for donating the box-tri collider, and to asko for porting it.
You are wellcome. :)

Cheers,
Alen