Random assert "Assertion failed: mNodes[nodeID].aabb.contains(newAABB)"

[caxapexac]

This happens sometimes, just by random after ~5-50secs of running code that doesn't use anything except static rigidbodies and collisionbodies. Just character controller that uses collide to find extrusion vector and some static geometry. The more rigidbodies I have the more probability of crash (it crashed after about 5 secs after adding ~30 rigidbodies)

What should I try to check to locate the root of the problem. I mean, this assert fails on broad phase but I don't touch broad phase manually, I only call world.update() then interpolate transforms almost as written in docs...

[DanielChappuis]

Thanks for reporting this. Where is this line exactly in the code? Can I ask you to post here the values of the AABB coordinates for the two variables 'aabb' and 'newAABB' when the assert occurs?

[caxapexac]

Sorry for the late reply, I'm trying to reproduce the issue, wait for the several days please

[Casqade]

@DanielChappuis I also had the same assertion failure under slightly different conditions (steps):

1. RigidBody centered inside CollisionBody (both with colliders)
2. Change RigidBody type to dynamic
3. Call world update
4. A custom EventListener::onContact callback applies force to the RigidBody
5. Assertion failure at DynamicAABBTree.cpp:209

Only after following your comment, I found out that one of the aabb's contained NaN values. These came from step 4: the applied force vector was also NaN (incorrect calculation on my side).

I suggest to introduce something like assert(force.isFinite()) in RigidBody::applyWorldForceAt[Local/World]Position. In case the assertion won't break anything, it would save the debugging time. Even in my simple setup it took me too long to find the culprit.

[DanielChappuis]

Yes that's a good idea. Thanks a lot for your feedback.