bind group buffer size error (update to 0.12)

[dhardy]

Could I ask for a little help please? I'm attempting to update wgpu (from 0.11, currently to 0.12):

Caused by:
    In a RenderPass
      note: encoder = `render`
    In a draw command, indexed:false indirect:false
      note: render pipeline = `SS render_pipeline`
    Buffer is bound with size 12 where the shader expects 16 in group[0] compact index 1

The pipeline in question:

        // shared bind group:
        let bgl_common = device.create_bind_group_layout(&wgpu::BindGroupLayoutDescriptor {
            label: Some("common bind group layout"),
            entries: &[
                wgpu::BindGroupLayoutEntry {
                    binding: 0,
                    visibility: wgpu::ShaderStages::VERTEX,
                    ty: wgpu::BindingType::Buffer {
                        ty: wgpu::BufferBindingType::Uniform,
                        has_dynamic_offset: false,
                        min_binding_size: None, // TODO
                    },
                    count: None,
                },
                wgpu::BindGroupLayoutEntry {
                    binding: 1,
                    visibility: wgpu::ShaderStages::FRAGMENT,
                    ty: wgpu::BindingType::Buffer {
                        ty: wgpu::BufferBindingType::Uniform,
                        has_dynamic_offset: false,
                        min_binding_size: None, // TODO
                    },
                    count: None,
                },
            ],
        });

        let pipeline_layout = device.create_pipeline_layout(&wgpu::PipelineLayoutDescriptor {
            label: Some("SS pipeline_layout"),
            bind_group_layouts: &[bgl_common],
            push_constant_ranges: &[],
        });

        let render_pipeline = device.create_render_pipeline(&wgpu::RenderPipelineDescriptor {
            label: Some("SS render_pipeline"),
            layout: Some(&pipeline_layout),
            vertex: wgpu::VertexState {
                module: &shaders.vert_shaded_square,
                entry_point: "main",
                buffers: &[wgpu::VertexBufferLayout {
                    array_stride: size_of::<Vertex>() as wgpu::BufferAddress,
                    step_mode: wgpu::VertexStepMode::Vertex,
                    attributes: &wgpu::vertex_attr_array![0 => Float32x2, 1 => Float32x4, 2 => Float32x2],
                }],
            },
            primitive: wgpu::PrimitiveState {
                topology: wgpu::PrimitiveTopology::TriangleList,
                strip_index_format: None,
                front_face: wgpu::FrontFace::Cw,
                cull_mode: Some(wgpu::Face::Back), // not required
                unclipped_depth: false,
                polygon_mode: wgpu::PolygonMode::Fill,
                conservative: false,
            },
            depth_stencil: None,
            multisample: Default::default(),
            fragment: Some(wgpu::FragmentState {
                module: &shaders.frag_shaded_square,
                entry_point: "main",
                targets: &[wgpu::ColorTargetState {
                    format: super::RENDER_TEX_FORMAT,
                    blend: Some(wgpu::BlendState::ALPHA_BLENDING),
                    write_mask: wgpu::ColorWrites::ALL,
                }],
            }),
            multiview: None,
        });

The first binding in bg_common is type Scale = [f32; 4];. The second, the one causing the error, is a [f32; 3]. Usage in shader:

layout(set = 0, binding = 1) uniform FragCommon {
    vec3 lightNorm;
};

Filling in the min_binding_size (to 16 and 12 respectively) results in a different error:

Caused by:
    In Device::create_render_pipeline
      note: label = `SS render_pipeline`
    error matching FRAGMENT shader requirements against the pipeline
    shader global ResourceBinding { group: 0, binding: 1 } is not available in the layout pipeline layout
    buffer structure size 16, added to one element of an unbound array, if it's the last field, ended up greater than the given `min_binding_size`

Full code: use https://github.com/kas-gui/kas/tree/work (kas-gui/kas@6a484d6 and parent commit). Try running e.g. cargo run --example gallery.

[dhardy]

Well, I "solved" the issue. Apparently this binding is now considered to have size 16:

layout(set = 0, binding = 1) uniform FragCommon {
    vec3 lightNorm;
};

Padding the input buffer to 16 bytes fixes my build, but I wonder if this is a bug in shader translation. That said, the naga translates the GLSL to this WGSL, which is still 12 bytes:

struct FragCommon {
    lightNorm: vec3<f32>,
}