`KinematicCollision3D.get_local_shape()` always returns only the zero index shape

[viksl]

Godot version

4.0 dotnet

System information

Windows 10, Nvidia GTX 1660Ti, Vulkan

Issue description

I'd first like to say I'm not sure if this is a bug or if I'm just understanding the functionality wrong, I've been trying reddit, discord for days, nothing so I decided to come with it here.

I'd like to get the collision shape which collided during KinematicCollision from inside a CharacterBody3D, I found there's get_local_shape(index) method but this method always returns only the 0 index CollisionObject.

If I'm using this wrong I'm sorry, is there some way to use this or something else to get the particular colliding shape?

Test project is in gdscript.

Thank you.

Steps to reproduce

1. Play the project
2. Wait until objects collide
3. Read the output, it prints the first shape only (ChracterBodyShape1)
   The local_shape_at_index prints the shape that should be colliding but it returns only the 0 index CollisionShape3D which is named CharacterBodyShape1, the shape which collided is CharacterBodyShape2
4. Stop the project
5. Move the CharacterBodyShape2 above CharacterBodyShape1 in the sceen tree so it's the first child of the CharacterBody3D node.
6. Play the project
7. Watch the Output again (again only the first shape this time ChracterBodyShape2 since you moved it up)

As you can see now it prints the colliding shape but only because it's the 0 index shape.

Minimal reproduction project

TestGetLocalShape.zip

[adamscott]

Be careful, you're conflating the collision_index of collisions with get_slide_collision_count().
collision_index refers to collision.get_collision_count(), the deepest collision being 0.

[viksl]

@adamscott I've updated the code to reflect that mistake and updated teh MRP in the original post but as I mentioned previously I did check all the indices manually anyway and they all return a wrong shape. Perhaps I'm still doing something wrong? I really can't get to the colliding shape and only the 0 index one.

Thank you for pointing out my mistake but does it change anything considering the above mentioned, please?

EDIT: I've edited the original post and removed stale info considering the changes, problem is sitll the same.

[adamscott]

Yeah, something seem wrong. I'll try to put more time on this soon to confirm the issue.

[Raven38911]

I ran into this issue while working on an own project and also tried the MRP - I can confirm that this is still broken in 4.2.

There is also a comment by @r0401 under the closed (very old) issue #31144, where he reports having the same problem in 4.0.1 and 4.1.

[Belserich]

Still buggy in 4.3 stable

[adamscott]

Thanks @Belserich for the ping ! I decided to really investigate this time and I found out the issue.