[ODE] how to glue bodies together ?

[XFred]

On Wed, 28 Jan 2004 13:32:15 -0500
"Martin C. Martin" <martin at metahuman.org> wrote:

> Can you explain what you mean by "unreliable"?

Sure. Fixed joints only fix position, not rotation.
We tested that :
create 1 box without rotation ,
create another box with 45 degrees of rotation ,
then, fix the 2 together with a fixed joint :
The first step makes the rotated box back to a 0 degree rotation.

Obviously, this is said in the doc, and unfortunately we can't rely on fixed joint because of this behaviour. 
Using a hinge joint instead seems to do the trick...

> 
> > Don't combine the bodies.  Instead, just use one body - combine the
> > geoms, and combine the masses so it acts like both bodies.
> 
> This gets tricky when you have a object that you want to break apart or 
> join together at some point.  While it's possible to create/destroy 
> bodies and geoms when that happens, that's a PITA.  In some 
> applications, the problems of using a fixed joint may be small.
> 
> Now that I think about it, maybe a function to "join" two bodies/geoms, 
> and another one to split them, would be useful?
> 
> - Martin

Our application aims on very accurate collision detection. We already use some kind of geometric approximation by using primitives.
So, computation time is not critical, but accuracy is .
Detecting breaking is not our goal either.

-- 
Fred D.