Loading a script that uses it's own class_name creates orphans despite scripts being references

[Ashe]

Godot version:
3.2.3.stable.custom_build (I got it from the AUR)

OS/device including version:
Using GLES3 video driver
OpenGL ES 3.0 Renderer: GeForce GTX TITAN X/PCIe/SSE2
Arch Linux (don't think it's relevant?)

Issue description:
You get a lot of orphan warnings when a script uses it's class name inside itself and it gets loaded without being used. Simply mentioning the class_name anywhere else will bubble up the issue.

ERROR: ~List: Condition "_first != __null" is true.
   At: ./core/self_list.h:112.
ERROR: ~List: Condition "_first != __null" is true.
   At: ./core/self_list.h:112.
WARNING: cleanup: ObjectDB instances leaked at exit (run with --verbose for details).
   At: core/object.cpp:2135.
Leaked instance: GDScriptNativeClass:824
Leaked instance: GDScript:1177 - Resource path: res://broken_class.gd
Hint: Leaked instances typically happen when nodes are removed from the scene tree (with `remove_child()`) but not freed (with `free()` or `queue_free()`).
ERROR: clear: Resources still in use at exit (run with --verbose for details).
   At: core/resource.cpp:477.
Resource still in use: res://broken_class.gd (GDScript)

Steps to reproduce:

# test.gd
extends Node
const SCRIPT_PATH := "res://broken_class.gd";
func _ready():
  var script := load(SCRIPT_PATH);

# broken_class.gd
class_name BrokenClass
extends Node

var variable := false;

func broken_function(BrokenClass: BrokenClass) -> bool:
  return variable == variable;

Minimal reproduction project:
engine_bug.zip

[Ashe]

If this is actually intentional (much like you shouldn't use self.set_variable since it causes a recursive dependency) then maybe consider adding a clause in the static typing docs since I didn't see anywhere that there was a rule about using a classes name inside its own file. There's only one for using two classes that depend on each other (which has an in-editor warning and is fine).

[Calinou]

Duplicate of #16959?

[Ashe]

Duplicate of #16959?

Ah yes, the comments of that bug are certainly related, although the original post seems a little different.

Firstly, in that bug, using 'load' is listed as a work around, but it was the load that triggered the bug for me.

Secondly, that bug is about two scripts loading each other. This is about a script recursively loading itself due to class_name, triggered by a load.

Either way sorry for submitting if it is an issue. I didn't see that other one.

[Calinou]

Either way sorry for submitting if it is an issue. I didn't see that other one.

No worries! I can leave this issue open; it sounds different enough.

[qarmin]

Looks that it may be fixed in Godot 3.2.4(I can't reproduce leak)

[Ashe]

Looks that it may be fixed in Godot 3.2.4(I can't reproduce leak)

That's promising news - I think the AUR is still using 3.2.3 so looking forward to fix these things! I guess I should close unless anyone has anything else to add.

[akien-mga]

The specific test project in the OP seems fixed in 3.2.4 RC 3 indeed.

There are still similar issues which are reproducible such as #16959 (comment) and #21461 though.