Cannot Load Saved Materials

[cybereality]

Godot version

v4.0.beta2.official [f8745f2]

System information

Ubuntu 22.04, AMD RX 6800 XT

Issue description

I can create a material (either standard or ORM) on a mesh. However, if I save this, I cannot load it anywhere else, even on the same mesh. It gives an error: "The selected resource (StandardMaterial3D) does not match any type expected for this property (BaseMaterial3D, ShaderMaterial)". This is a new project, with nothing else but a cube.

Steps to reproduce

Create a MeshInsance.
Give it a material (standard or orm). It can be on the mesh or the override.
Save the material.
Click on the material drop-down and choose load or quick load.
Choose the same material.
Watch the error.

Minimal reproduction project

MaterialBug.zip

[manueldun]

Does this count for hacktoberfest? If it does i am going to give this a try

[Calinou]

It's strange that this happens, as ORMMaterial3D and StandardMaterial3D both inherit from BaseMaterial3D.

Does this count for hacktoberfest? If it does i am going to give this a try

Yes, any issue counts.

[Zylann]

I had an existing scene containing various mesh instances with a StandardMaterial3D assigned to them. Some of them inexplicably lost that material, and I can't assign it back because of this error. Despite that, some mesh instances still have the material assigned just fine.

Interestingly, after I re-save several times the scene:

• If I use Quick Load, it shows the error.
• If I use Load, it also shows the error.
• If I use New, it doesnt appear to do anything, but reloading the scene shows a new material assigned.
• If I drag and drop from the FileSystem dock, it assigns the material, but any mesh I hovered while moving over the 3D viewport prior to reaching the inspector, loose their material.

[Calinou]

If you can compile the engine from source, you could look into bisecting the regression to greatly speed up troubleshooting.

Assuming this issue doesn't occur in 4.0.beta1:

• 4.0.beta1's commit hash (good) is 4ba934b.
• 4.0.beta2's commit hash (bad) is f8745f2.

[manueldun]

If you can compile the engine from source, you could look into bisecting the regression to greatly speed up troubleshooting.

Assuming this issue doesn't occur in 4.0.beta1:

• 4.0.beta1's commit hash (good) is 4ba934b.
• 4.0.beta2's commit hash (bad) is f8745f2.

I will start bisecting tomorrow morning to get an idea of how it should be fixed. I just found the code where it shows the warning. Thanks.

[fracteed]

Just wanted to confirm that I also have this issue in beta2, that didn't exist in beta1. Meshes that already had materials are fine and the material can be edited, but another material can no longer be applied.

[clayjohn]

For posterity, this bug was introduced by #62413 and will be closed by #66795