[ODE] Geom Transforms hate me

[Adam D. Moss]

Russ Smith wrote:
 >>If it's any help, I looked at this myself several months ago and
 >>concluded that the mass offset adjustment for composite objects
 >>in test_boxstack probably wasn't doing the right thing anyway
 >
 > do you know what the bug is?

I'm afraid not -- I ended up building a composite-mass solution
from first principles IIRC pretty much as you describe in the
paragraph below, with an extra reverse-offset again when it comes
to drawing/using the body's position to account for having translated
it away from the relative position it was supposed to have been created
in.  Actually it was a bit of a nightmarish trial-and-error process.

Each mass/geom's rotation is also critical in creating the composite
mass, not just its translation, and the rotation has to be done
in the right frame of refence -- I think that threw me the most.

 > the fact that geometry and mass are completely independent things in ODE
 > certainly makes it annoying to create composites. what the boxstack demo
 > is doing IIRC is (1) computing the total mass and inertia of the composite
 > object by summing up the pieces, (2) finding the offset of the body center
 > of mass to the composite object center of mass, (3) offsetting
 > the geometry and body positions so that the body center of mass is equal
 > to the body point of reference. this last step is needed because ODE
 > bodies *must* have center-of-mass == point-of-reference.

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