[ODE] CVS + OpCode buggy?

[Max Bane]

Hey, all.
I thought I'd start taking a serious look at ODE and downloaded the current 
CVS tree.  I enabled OpCode (which also forced me to set single floating 
point precision) and all compiled fine.  But there's lots of weirdness when I 
try running the various test programs.
For example test_ode.exe ends with "ODE INTERNAL ERROR: A*x = b+w, maximum 
difference = 1.907349e-05 [\n]  Aborted"
And in test_joints.exe every other test is "FAILED".
And worst of all, test_crash.exe segfaults when I try to turn on faststep.

So should I just give up and get the stable release? Or am I missing something 
in my compilation?
-- 
Max Bane
He played the king as if afraid someone else would play the ace.
		-- John Mason Brown, drama critic