Matrix index out of bounds. #310

Johannes0021 · 2025-01-29T20:28:50Z

// Error / Panic --------------------------------------------------------------

parry2d-f64-0.18.0/src/query/clip/clip_aabb_line.rs:141:19:
Matrix index out of bounds.

// What happened ---------------------------------------------------------

I accidentally created a triangle ColliderShape with:

point_a -> Vec2 { data: [40.0, 0.0] }
point_b -> Vec2 { data: [ 0.0, 80.0] } // Same as point_c
point_c -> Vec2 { data: [ 0.0, 80.0] } // Same as point_b

In clip_aabb_line.rs the origin parameter has NAN values -> pub fn clip_aabb_line(...);

In the loop in the else part:
let mut inter_with_near_halfspace = (aabb.mins[i] - origin[i]) * denom; // Results in NaN (origin[i] == NaN)
let mut inter_with_far_halfspace = (aabb.maxs[i] - origin[i]) * denom; // Results in NaN (origin[i] == NaN)

near_diag stays false
near_side stays 0

in line 141 [0 - 1] Matrix index out of bounds.

// Solution ----------------------------------------------------------------

let near = if near_diag {
    (tmin, -dir.normalize(), near_side)
} else {
    let mut normal = Vector::zeros();

    if near_side < 0 {
        normal[(-near_side - 1) as usize] = 1.0;
    } else if near_side > 0 { // <-------------------------------------- Here
        normal[(near_side - 1) as usize] = -1.0;
    }
    else { // <--------------------------------------------------------- Here
        return None;
    }

    (tmin, normal, near_side)
};

let far = if far_diag {
    (tmax, -dir.normalize(), far_side)
} else {
    let mut normal = Vector::zeros();

    if far_side < 0 {
        normal[(-far_side - 1) as usize] = -1.0;
    } else if far_side > 0 { // <--------------------------------------- Here
        normal[(far_side - 1) as usize] = 1.0;
    }
    else { // <--------------------------------------------------------- Here
        return None;
    }

    (tmax, normal, far_side)
};

hakolao · 2025-04-29T21:48:08Z

I'm getting this too, an only panic I've seen in a long time. Would be nice to get this fixed.

Johannes0021 · 2025-05-05T22:30:41Z

Problem:
If a triangle collider has two or all three points the same,
the code divides by zero inside project_local_point_and_get_location.
This makes the result NaN and breaks further calculations.

How to Reproduce:
This example uses Rapier2D for demonstration, but the root cause of the error lies in Parry.

use rapier2d::prelude::*;

fn main() {
    let mut rigid_body_set = RigidBodySet::new();
    let mut collider_set = ColliderSet::new();

    /* Create the ground. */
    let collider = ColliderBuilder::ball(100.0).build();
    collider_set.insert(collider);

    /* Create the bouncing ball. */
    let rigid_body = RigidBodyBuilder::dynamic()
        .translation(vector![0.0, 0.0])
        .build();

    let a = point![40.0,  0.0];
    let b = point![ 0.0, 80.0];
    let c = point![ 0.0, 80.0];

    let collider = ColliderBuilder::triangle(a, b, c).restitution(0.7).build();
    let triangle_body_handle = rigid_body_set.insert(rigid_body);
    collider_set.insert_with_parent(collider, triangle_body_handle, &mut rigid_body_set);

    /* Create other structures necessary for the simulation. */
    let gravity = vector![0.0, -9.81];
    let integration_parameters = IntegrationParameters::default();
    let mut physics_pipeline = PhysicsPipeline::new();
    let mut island_manager = IslandManager::new();
    let mut broad_phase = DefaultBroadPhase::new();
    let mut narrow_phase = NarrowPhase::new();
    let mut impulse_joint_set = ImpulseJointSet::new();
    let mut multibody_joint_set = MultibodyJointSet::new();
    let mut ccd_solver = CCDSolver::new();
    let mut query_pipeline = QueryPipeline::new();
    let physics_hooks = ();
    let event_handler = ();

    /* Run the game loop, stepping the simulation once per frame. */
    for _ in 0..3 {
        physics_pipeline.step(
            &gravity,
            &integration_parameters,
            &mut island_manager,
            &mut broad_phase,
            &mut narrow_phase,
            &mut rigid_body_set,
            &mut collider_set,
            &mut impulse_joint_set,
            &mut multibody_joint_set,
            &mut ccd_solver,
            Some(&mut query_pipeline),
            &physics_hooks,
            &event_handler,
        );

        let triangle_body = &rigid_body_set[triangle_body_handle];
        println!("Triangle altitude: {}", triangle_body.translation().y); // Prints NaN
    }
}

Output:
Triangle altitude: NaN
Triangle altitude: NaN
Triangle altitude: NaN

What Happens:

In "src/query/point/point_triangle.rs" in the fn project_local_point_and_get_location (around line 254), the code does:

let u = (ac_bp - ab_bp) / (ac_bp - ab_bp + ab_cp - ac_cp); // proj on bc = b + bc * u

This resolves in the NaN.

Propagation and Panic:
If a invalid triangle makes u become NaN, later code still uses these NaN values. For example, in clip_aabb_line:

let inter_with_near_halfspace = (aabb.mins[i] - origin[i]) * denom;
let inter_with_far_halfspace = (aabb.maxs[i] - origin[i]) * denom;

Both results are NaN because origin[i] is NaN. The flags near_diag and far_diag stay false, and near_side and far_side remain 0. Eventually the code does:

normal[(near_side - 1) as usize] = 1.0;

This becomes normal[-1], causing a panic: index out of bounds.

Note: I’m not sure how critical this is in practice, since you normally create triangles with valid, distinct points.
I investigated this mainly to pinpoint the root cause of the matrix index out of bounds panic.*

Johannes0021 · 2025-05-05T22:57:22Z

Just to add:
This was the first place I found where a triangle with duplicate points leads to issues, but I’m not sure if it’s the only one.

Johannes0021 added a commit to Johannes0021/parry that referenced this issue Mar 4, 2025

Fix dimforge#310: Matrix index out of bounds.

921832b

Johannes0021 linked a pull request Mar 4, 2025 that will close this issue

Fix #310: Matrix index out of bounds. #319

Open

Vrixyz added the bug Something isn't working label Mar 5, 2025

Johannes0021 added a commit to Johannes0021/parry that referenced this issue Mar 19, 2025

Fix dimforge#310: Matrix index out of bounds. + Clippy

68cb27c

Johannes0021 added a commit to Johannes0021/parry that referenced this issue Mar 19, 2025

Fix dimforge#310: Matrix index out of bounds. + Clippy

c168599

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Uh oh!

Uh oh!

Matrix index out of bounds. #310

Matrix index out of bounds. #310

Johannes0021 commented Jan 29, 2025 •

edited

Loading

hakolao commented Apr 29, 2025

Uh oh!

Johannes0021 commented May 5, 2025 •

edited

Loading

Uh oh!

Johannes0021 commented May 5, 2025

Uh oh!

Uh oh!

Matrix index out of bounds. #310

Matrix index out of bounds. #310

Comments

Johannes0021 commented Jan 29, 2025 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

hakolao commented Apr 29, 2025

Uh oh!

Johannes0021 commented May 5, 2025 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

Johannes0021 commented May 5, 2025

Uh oh!

Johannes0021 commented Jan 29, 2025 •

edited

Loading

Johannes0021 commented May 5, 2025 •

edited

Loading