Collision resolution with contact_point_response

[totebo]

When using contact_point_response, the normals that are returned are sometimes incorrect:

Here is a minimal example of this behaviour: CollisionResolution.zip (12.8 KB)

I’m using the “resolving kinematic collisions” code from this page.

Searching the forum, dmitriy discourages the usage of contact_point_response at all in this post.

The use case I have is that I would like the collisions to work regardless of the size or type of the fixture colliding. Using raycasts is therefore not ideal, because I’d need to know the size of the fixtures to fire them accurately.

Is there really no other way of solving the scenario above? As you might have guessed, this is for a platform game.

[ross.grams]

I wonder if this would be solved if we could enable continuous collision detection on bodies, since then it would sweep the path to find the earliest possible collision, rather than simply evaluating the frame and choosing the shortest vector to resolve the collision, regardless of movement.

[totebo]

That sounds like a great idea to me!

Edit: Could it have something to do with this post?

Testing the example above with a dynamic body produces the same result if the gravity is high enough: