Infer array length from custom array if no vertex array is provided #150
Conversation
While the `RenderingServer::mesh_create_surface_data_from_arrays` method does support vertexless meshes (see godotengine#62046 and godotengine#83446), it enforces that the size of custom arrays is dependent on the size of the vertex array. This effectively means that custom arrays cannot be used in vertexless meshes. This commit changes the way the array length is computed so that if no vertex array is provided, its length will be inferred from the custom arrays, if provided. It therefore adds support for custom arrays in vertexless meshes.
Co-authored-by: A Thousand Ships (she/her) <[email protected]>
…et_vertex_to_custom_array_length_factor`
| if (array_len == 0) { | ||
| Vector<Variant> custom_array = p_arrays[RS::ARRAY_CUSTOM0 + i]; | ||
| uint32_t factor_reciprocal = _get_vertex_to_custom_array_length_factor(format, RS::ARRAY_CUSTOM0 + i); | ||
| array_len = custom_array.size() / factor_reciprocal; |
There was a problem hiding this comment.
I know it's probably not a problem, but is there any chance custom_array.size() could return 0?
There was a problem hiding this comment.
Well, ARRAY_CUSTOM0 can be any size, but if it’s 0, then the code should just behave the same way it does without my changes.
There was a problem hiding this comment.
The thing is, in C++, dividing by 0 is undefined. So there is a chance that doing this could introduce garbage into the variable that can screw things up. Having a check before the division to make sure that it's not 0 and bail if it is will prevent the garbage from happening.
There was a problem hiding this comment.
I’m not dividing by custom_array.size(), though. custom_array.size() is the dividend in this case. I’m dividing by factor_reciprocal which shouldn’t ever be 0.
There was a problem hiding this comment.
I used the wrong verbage, but the problem still remains. It's undefined behavior in C++, which could cause garbage values, even if the dividend is 0.
The code and the idea itself is good, I just feel like having a custom_array.size() != 0 check before dividing would fix a potential problem source from undefined behavior and the dark magic that is the C++ compilers.
There was a problem hiding this comment.
I used the wrong verbage, but the problem still remains. It's undefined behavior in C++, which could cause garbage values, even if the dividend is 0.
The code and the idea itself is good, I just feel like having a
custom_array.size() != 0check before dividing would fix a potential problem source from undefined behavior and the dark magic that is the C++ compilers.
Dividing 0 by any integer other than 0 is well defined behaviour in C++ or any other language for that matter.
Checking the dividend for 0 would be pointless.
There was a problem hiding this comment.
After educating myself farther, I was wrong. I'm sorry about that.
|
This needs to be squashed. |
I am unsure if PRs that haven’t been merged into Godot are already being considered, but if yes, then this is the Godot PR:
godotengine/godot#93549