Discussion of CharacterBody behavior

[TokageItLab]

Godot version

4.0

System information

Any

Issue description

This issue is a tracker that summarizes some of the issues that occur with the current physics behavior of CharacterBody.

Since the way characters are required to move varies from game to game, the implementation should be as minimalistic as possible.

The problems that occur can be divided into two categories: engine-side problems and user-side problems. However, even for user-side problems, it is possible to provide solutions as add-ons or tutorials. Then, I suggest that the implementation should try to be as uncomplicated as possible so that the user can understand what is going on.

I prepared a sample project that looks like this.

• Move_and_slide_with_snap fails to follow a moving platform if stop_on_slope is set to true #30311
• [Bullet] Inaccurate behaviour of 3D KinematicBody move_and_slide_with_snap on vertical moving platform. #36334
• is_on_floor() is jittering when going up the steep slope as wall from floor #50756
• apply_velocity_on_leave in CharacterBody3D is may not working correctly #52957
• CharacterBody3D causes slipping when going up the slope with slanting to it #52958
• When jumping on a horizontal-moving slope, the landing point will be shifted #53435
• Can't determine if it's stacked or not
  • Result of CharacterBody's gravity calculation on steep slope in air is undesirable #51874
  • Implement priority order of collision solving in StaticBody
  • Implement is_inside_solid_geometry() in CharacterBody

Steps to reproduce

Play scene.

Minimal reproduction project

https://github.com/TokageItLab/3d-platform-test-for-godot4

[pouleyKetchoupp]

For testing purpose, I've made a 3.x version of the same project that can be used to test backported features:
3d-platform-test-for-godot3-main.zip

It's mostly okay, but I had problems converting render meshes from 4.0 to 3.x so I had to recreate them and I took shortcuts so they are not all looking exactly the same. Physics and controls should be setup the same way.

[TokageItLab]

@pouleyKetchoupp I opened two proposals for solving the collision between StaticBody and CharacterBody.
- Implement priority order of collision solving in StaticBody
- Implement is_inside_solid_geometry() in CharacterBody

[TokageItLab]

@pouleyKetchoupp #30311 was closed, so I tested Moving-Slope. It no longer slid down, but I encountered two problems.

• On the slope moving upward, the player speed is reduced (since the character's velocity is strange, I'll look into the problem in the PlayerController as well)

2021-08-11.5.12.41-1.mov

• When the wall is close, the collision detection is very sketchy

2021-08-11.5.12.412-1.mov

[pouleyKetchoupp]

@TokageItLab Thanks for testing and letting me know! I've noticed the one with the slowdown earlier today, I'm going to have a look to both.

[TokageItLab]

@pouleyKetchoupp Currently, some methods such as get_slide_count() and get_floor_velocity() can't found in CharacterBody. Do you know what was the reason for it's removal and are there superseding?

Edited:
I found #51645, nice work @fabriceci !

[TokageItLab]

@pouleyKetchoupp @fabriceci @fire Updated the sample project. The movement on the moving slope is now perfect. Excellent! However, the jittering slope problem still occurs in 3D.

And..., there is a regression where the CharacterBody (Player's toes?) is always buried in the floor. This was due to a change in the CapsuleCollider specification, so I fixed!

[pouleyKetchoupp]

@TokageItLab For the remaining jittering on the slope, I think it makes sense it's still there because #51027 was made just for 2D for now. We'll keep an eye on this case when porting those changes to 3D.

[TokageItLab]

@pouleyKetchoupp @fabriceci In relation to the slope jittering, it reminded me of #51874.