Feature request: Panic if QueryPipeline fails to be updated

[RedBreadcat]

The following code softlocks within physics.step(); in the loop, after "Step: 9" is printed. The freeze does not occur if self.query_pipeline.update(&self.bodies, &self.colliders); is uncommented.

use rapier2d::{
    na::{Isometry2, Vector2},
    prelude::*,
};

struct Physics {
    pipeline: PhysicsPipeline,
    query_pipeline: QueryPipeline,
    integration_parameters: IntegrationParameters,
    island_manager: IslandManager,
    broad_phase: BroadPhase,
    narrow_phase: NarrowPhase,
    bodies: RigidBodySet,
    colliders: ColliderSet,
    impulse_joints: ImpulseJointSet,
    multibody_joints: MultibodyJointSet,
    ccd_solver: CCDSolver,
}

impl Physics {
    fn step(&mut self) {
        self.pipeline.step(
            &Vector2::default(),
            &self.integration_parameters,
            &mut self.island_manager,
            &mut self.broad_phase,
            &mut self.narrow_phase,
            &mut self.bodies,
            &mut self.colliders,
            &mut self.impulse_joints,
            &mut self.multibody_joints,
            &mut self.ccd_solver,
            Some(&mut self.query_pipeline),
            &(),
            &(),
        );

        //self.query_pipeline.update(&self.bodies, &self.colliders);
    }

    fn build_rb_kinematic_with_collider(
        &mut self,
        pos: Vector2<f32>,
    ) -> (RigidBodyHandle, ColliderHandle) {
        println!("build");
        let rb = Self::build_rigidbody_kinematic_template(pos);
        let rbh = self.bodies.insert(rb);
        let col = Self::build_primary_collider_kinematic_template();
        let colh = self
            .colliders
            .insert_with_parent(col, rbh, &mut self.bodies);

        (rbh, colh)
    }

    fn build_rigidbody_kinematic_template(pos: Vector2<f32>) -> RigidBody {
        RigidBodyBuilder::new(RigidBodyType::KinematicVelocityBased)
            .position(Isometry2::new(pos, 0.0))
            .gravity_scale(0.0)
            .can_sleep(true)
            .build()
    }

    fn build_primary_collider_kinematic_template() -> Collider {
        let shape = SharedShape::ball(1.0);
        let cb = ColliderBuilder::new(shape);

        cb.build()
    }
}

impl Default for Physics {
    fn default() -> Self {
        let pipeline = PhysicsPipeline::new();
        let query_pipeline = QueryPipeline::new();
        let integration_parameters = IntegrationParameters {
            dt: 0.01,
            ..Default::default()
        };
        let island_manager = IslandManager::new();
        let broad_phase = BroadPhase::new();
        let narrow_phase = NarrowPhase::new();
        let bodies = RigidBodySet::new();
        let colliders = ColliderSet::new();
        let impulse_joints = ImpulseJointSet::new();
        let multibody_joints = MultibodyJointSet::new();
        let ccd_solver = CCDSolver::new();

        Self {
            pipeline,
            query_pipeline,
            integration_parameters,
            island_manager,
            broad_phase,
            narrow_phase,
            bodies,
            colliders,
            impulse_joints,
            multibody_joints,
            ccd_solver,
        }
    }
}

fn main() {
    let mut physics = Physics::default();

    let a = Vector::new(1.0, 0.0);
    let b = Vector::new(0.0, 0.0);

    physics.build_rb_kinematic_with_collider(a);

    let mut step = 0;
    loop {
        // 2 are needed
        physics.build_rb_kinematic_with_collider(b);
        physics.build_rb_kinematic_with_collider(b);

        step += 1;
        println!("Step: {}", step);
        physics.step();
    }
}

It was a bit difficult to figure out what was happening, as the freeze only occurred when rigidbodies were in certain positions. (If a == b, I could not trigger a freeze). If possible, it would be nice if the code panicked, or a more elegant solution that made it impossible for one to forget to update the QueryPipeline.

The freeze itself occurred inside this function call: https://github.com/dimforge/rapier/blob/1a4183cc94acc3210e4ae467abbea7d68e51c5ff/src/pipeline/query_pipeline.rs#L349

In more complex code that periodically deleted rigidbodies and colliders, and had dynamic physics objects too, I occasionally panicked inside the closure on the next line, due to the handle's two internal integer components being huge numbers that looked like uninitialized memory.

Thank you for your work!

[sebcrozet]

The qbvh::refit should not freeze (and, more generally, the query pipeline’s update should never fail). This is a bug in the QBVH update code.

[sebcrozet]

I wasn’t able to reproduce the bug with the provided code, but, other similar issues were reported and a fix just got merged in https://github.com/dimforge/parry/pull/185. So I’m closing this issue, but, feel free to open a new issue if this happens again.