Until `MechanicalWorld::step` is called, `GeometricalWorld::interferences_with_point` uses old collider positions about nphysics HOT 1 OPEN

My workaround of using a timestep of 0.0 ended up not being so good because it would cause dynamic bodies that are temporarily colliding with static bodies to be forced apart, which is not viable behavior for an editor. Therefore I took another look at what made step work for interferences_with_point by removing chunks of MechanicalWorld::step until the editor exhibited the original issue.

After some false positives, I narrowed it to this chunk:

bodies.foreach_mut(&mut |_, b: &mut dyn Body<N>| {
    b.clear_forces();
    b.validate_advancement();
});

I transplanted this into my sync with step function and it works! My theory is that even though I was setting the body and collider position to the new position, the old body position is temporarily cached and used to reset the attached collider position before the broad phase based point interference is run. The call to Body::validate_advancement clears out the old position allowing the broad phase to work as expected with the new collider position. After simplifying my sync function looks like this:

pub fn sync(&mut self) {
    self.mech.maintain(&mut self.geo, &mut self.bodies, &mut self.colliders, &mut self.joints);
    self.bodies.foreach_mut(&mut |_, b| {
        b.validate_advancement();
    });
    self.geo.sync_colliders(&self.bodies, &mut self.colliders);
    self.geo.perform_broad_phase(&self.colliders);
}

I'm not sure if this is a supported use case or if it will continue to work, so it would be nice to get some input from a maintainer, but I'm basically happy with my solution. Thanks for taking a look.

from nphysics.