The thing is, it works here (my ODE lib is single-precision, so are the thirdparty ones I compiled), because dReal is typedeffed as float here.
The fact that _Hypnos and his double-precision build could only make it work by making it a float, that sounds very much like this really ought to be a float. Can anyone confirm this? If so, that’s one bug less in Panda3D 