motion state translation reset to (0,0,0) in some cases

[gjaegy]

Hi,

just spent my day fixing some border-line case bug, where my motion state transform was being set to identity/(0,0,0) during the step update.

This is what happens:

• create and initialize a motion state with the current position/orientation
• create a new rigid body using that motion state
• before adding the rigid body to the world, call btRigidBody::setWorldTransform() to initialize the rigid body position to be safe
• add rigid body to world
• step the physic simulation (::stepSimulation()) with a timeStep < fixedTimeStep (for instance, 5ms for a fixedTimeStep of 16.66ms)

This will actually call btDiscreteDynamicsWorld::synchronizeSingleMotionState() for that body, and update the motionstate using the value returned by btRigidBody::getInterpolationWorldTransform().
However, since the simulation hasn't actually stepped, and because the rigid body has just been added, btRigidBody::getInterpolationWorldTransform() returns an identity transform, which override the transform stored in the motion state.

My workaround is to call both btRigidBody::setWorldTransform() & btRigidBody::setInterpolationWorldTransform() before adding the rigid body to the world. But that is definitively counter-intuitive.

I hope that makes sense. If not don't hesitate to get back to me.

Cheers
Greg

[erwincoumans]

Thanks for the report. We probably should call setInterpolationWorldTransform when adding the body to the world.