[ODE] LCP internal error, s <= 0

[don]

Is there still interest in an example that causes:

    ODE Message 3: LCP internal error, s <= 0

I have release 0.03 which seems to have the fix described by Russell in
Dec 01.  I have also tried the hack posted by brunoh@student.ethz.ch in
Jan 02.

My system is a simple set of 5 boxes and some joints.  There is force
in just one joint.  I don't think there are any contacts, the boxes are
all in one group, and it doesn't touch the ground.  (This will change
after I get things working...).

I have a version that always gets this error, with no external inputs.
The forced joint runs against its stop, things sit for a moment, then
goes bad.  If you wait a while, the whole structure inverts itself in
one dimension.

I did try detecting the stop and setting the velocity and force of the
joint motor to zero, that didn't solve it though.

Don Reid
don_reid (at) attbi.com