[ODE] no box/trimesh collision, Unix only

[Adam D. Moss]

Jeffrey Smith wrote:
>>CVS head (dated 30.12.2003):
>>Everything works good under Windows platform.
> 
> Do all of the following tests work?
> 
> a) bounding box much smaller than triangles in trimesh
> b) bounding box much larger than triangles in trimesh
> c) bounding box contact face straddles an edge between two triangles in
>    trimesh
> 
> I've only had success with "a"; "b" and "c" fail completely.  (I am using the 
> CVS head from two weeks ago).

Here are the two checkin candidates that IMO seem most likely
to be the culprits simply because of what they touched:
2003/08/29 21:26:11 (rewrite of trimesh/box collisions)
and
2003/11/14 19:40:10 (allow AABBs on trimesh)

Note that ODE 0.039 was tagged around 2003/07/11 06:07:56, so
neither of these changes are in 0.039.

The first thing to check would be whether box/trimesh was
okay before/after 2003/08/29.  As a data point, I'm not seeing
any problems with box-trimesh on CVS-HEAD/Linux/x86/gcc3.1.1,
but I only check it with a spin of test_trimesh.exe.

--Adam
-- 
Adam D. Moss   . ,,^^   adam at gimp.org   http://www.foxbox.org/   co:3