[ODE] Stability in ODE

[Timothy J. Wood]

   I just started looking at ODE today and it seems pretty nice.  I've 
got a basic update of the Mac OS X using 'native' window/OpenGL system 
(instead of X11 which few people have on Mac OS X).

   One issue, though -- the stability of the simulation seems somewhat 
questionable.  I'm wondering if this is just an artifact of the choice 
of time step in the test cases, something wrong with the build on Mac 
OS X (a compiler bug, maybe) or maybe ODE itself just isn't that mature.

   I've added an extra test case to my local build, test_stability1.cpp, 
which has two thin boxes shaped into a narrow V.  You can the drop 
various solids down into the notch.  This is a pretty standard 
stability test, but ODE doesn't seem to do so well.

   The test I copied from (boxstack) had a step of 0.05.  With this, an 
axially aligned cube would severely interpenetrate the V, as would a 
sphere (although, less so).  The cylinder just shoot right through the 
V w/o seeming to hit it at all.

   Decreasing the step to 0.005, the cube stops and then falls over -- 
another common failure mode.  It does stop, however.  The sphere stops 
nicely while the cylinder still sinks completely through the V.

   Going down the 0.0005 (!) the cube still falls over and the cylinder 
still goes through the V.

   I've put a copy of my test case at:

http://www.omnigroup.com/~bungi/Library/Web/test_stability1.cpp

   You should be able to drop it into the ODE directory and make simple 
modifications to the makefiles to get it to build.

   Could someone on a 'supported' platform let me know whether this is a 
bug in ODE or my port (or possibly, I suppose, my test case :)

   Thanks!

-tim