[ODE] Box/trimesh collision problem

[Adam D. Moss]

Jon Watte (ODE) wrote:
>> If it's any help, I see box/trimesh issues too.  In fact these
>> are readily apparent, for me, with test_trimesh.exe where
>> some boxes (particularly those that settle near an edge) will
>> behave fine, but others will, after a couple of bounces, fall
>> right through the trimesh.
> 
> That used to work, so if you're seeing issues, something's probably 
> busted. But try the basics of trimesh collision: Are you creating enough 
> contacts? Are you sure you're passing the number of indices, not the 
> number of triangles, to the build function?

This is with an unmodified test_trimesh.cpp.  I assume that it's
doing things correctly (well, not always a safe assumption with
ODE's own test_* programs, I know firsthand!), and a brief glance
at the code shows that it's asking for 40 contacts.

That it is reasonably visible with test_trimesh.exe means either
that it's a problem fairly specific to my build setup, or that
test_trimesh.cpp has bugs and no-one really noticed/cared, or that
trimesh<->{box,ccylinder} (but not sphere) has bugs but most
people are quietly living with them or working around them.

I've just noticed that dropping boxes onto the bunnies in
test_moving_trimesh.exe is also rather hit-and-miss.  I know that
this used to look great -- I think the last time I checked it
was shortly after the switch to the Croteam collider (cvs-reverting
to the older collider is a little bit worse for test_trimesh.exe),
which seemed good, and not much has changed on the trunk since then.
Pretty odd.

Thanks,
--Adam
-- 
Adam D. Moss   -   adam at gimp.org