Overflow in AABB - This Bastard is Killing Me

[m-ryan]

I have been looking this issue from every angle I can think of, but there seems to be no fix I can find.

I am working on a game where objects that need to be simulated vary greatly in size. They can range from massive airships to small single-person fighter planes. It was only recently that this problem has come up.

Here are the dimensions of the bounding box being generated.

Box from node dimensions: 0.399119 0.087765 0.367182

I don't think this is too small for bullet is it? If it is then I am in serious trouble because bringing the players perspective to a size this small is really the only way I can get the scale difference I need while still having a good simulation on the larger end.

I have also guaranteed that there are no uninitialized values being passed to bullet to the point some would consider me obsessive.

Could someone please list me out as many other cases as they can think of where this error would crop up, because documentation on this issue is extremely lacking, and none of the other forum posts have helped me.

I am using the DBVT broadphase.

[m-ryan]

I believe I discovered the source issue, and it actually was me. There was an uninitialized double I was multiplying a thrust vector by that must have corrupted the entire thing. However, in the process of working out this issue I did discover that most of the classes that inherit from btQuadWord leaves the m_floats array in an unitialized state, and I would like to recommend that this be addressed as soon as possible.