[Bf-committers] ODE changes

[bf-committers@blender.org]

On Mon, Nov 04, 2002 at 09:30:17AM +0100, Maarten Gribnau wrote:
> While we're on the topic: I have crashes trying to run Blender with old NaN demo
> files using physics. Is that a known problem?

Yes, very unfortunately, the removal of the commercial SOLID library has
led to a situation where it is currently impossible to run any old NaN demo
files with physics :-(

The ODE physics will be a complete new replacement physics system. It is
unclear to me if it is possible to preserve compatibility with the old
impulse-based Sumo/SOLID physics (since ODE is constraint-based and has
different parameters to tune like ERP and CFM, which don't exist in
Sumo).

I've looked at the backwards-compatabilty physics issue briefly, and I think
the best bet is to use SOLID 2.0 (the free predecessor of the commercial SOLID 
version used in pre-GPL Blender) and to try to add the missing functionality
to SOLID 2.0 which should theoretically make it again possible to use Sumo
physics. (I was able to compile Blender with Sumo physics, but the link
step failed because of the missing SOLID library.) The main missing features
are penetration depth computation and raytesting, both of which can be done
with alternative libraries (albeit with some limitations). In particular,
penetration depth can be done with ODE's collision detector (although you
are slightly but not considerably more limited than with SOLID with respect
to available collision shapes), and raytesting is available as an add-on
library to ODE as well as being in Opcode 1.2 I believe.

So basically my idea of how to go forward with the physics issue is to use
ODE for new physics, because it is much more powerful than Sumo (ODE
supports full articulated rigid-body dynamics and can be extended with
user-defined constraints). For backwards compatibility we could extend
essentially rewrite SOLID 2.0 to be interface-compatible with SOLID 3.0
and add the missing features, then re-enable Sumo.

> Are there any test files that
> should run with ODE physics that I can use to test the OSX version? Hans told me
> you created a tank demo file?

I've created two, available at

http://www.xype.net/~nlin

-Norman