material properties only work with direct assignment (=) but not with compound-assignment (e.g.: *=)

[gibssa]

Describe the bug
clipSpaceTransform won't compile at all in matc for a vertexDomain:device shader (docs, previous discussion).

To Reproduce

1. git clone [email protected]:google/filament.git
2. ./build.sh release matc
3. Minimal shader test_direct_pixels.mat:

material {
    name: TestDirectPixels,
    parameters: [
    ],
    requires : [
        uv0
    ],
    blending : transparent,
    vertexDomain : device,
    shadingModel : unlit
}

vertex {
    void materialVertex(inout MaterialVertexInputs material) {
        material.clipSpaceTransform *= 0.3;
    }
}

fragment {
    void material(inout MaterialInputs material) {
        prepareMaterial(material);
        vec2 uv = getUV0();
        material.baseColor = float4(0.0, 0.1, 0.4, 0.4);
    }
}

5. ./out/cmake-release/tools/matc/matc --api=metal --platform=mobile --output=./test_direct_pixels.filamat ./test_direct_pixels.mat
6. matc compilation fails:

➜  filament git:(main) ✗ ./out/cmake-release/tools/matc/matc --api=metal --platform=mobile --output=./test_direct_pixels.filamat ./test_direct_pixels.mat
ERROR: Unable to parse vertex shader
ERROR: test_direct_pixels.mat:15: 'clipSpaceTransform' : no such field in structure 'material'
ERROR: test_direct_pixels.mat:15: '' : compilation terminated
ERROR: 2 compilation errors.  No code generated.


Could not compile material ./test_direct_pixels.mat

Expected behavior
matc compilation should work, though it may have broken here?

Screenshots
N/A

Logs
(see Step 6 above)

Desktop (please complete the following information):

• OS: macOS
• GPU: MacBook Pro Nov 2023
• Backend: Metal

Smartphone (please complete the following information):

• Device: N/A (offline matc)
• OS: N/A (offline matc)

Additional context
I'm looking to create a 2D quad, with 256 instances, and do efficient vertex transforms on the GPU.

Thanks as always team!

[pixelflinger]

note that

test_filamat --gtest_filter=\*.StaticCodeAnalyzerDirectAssignVertex

passes

[pixelflinger]

ah this looks like a bug in how we detect when a property is used. Currently you can't use *= you have to use =. It shouldn't be a problem because it's set to identity by default.

[gibssa]

ah this looks like a bug in how we detect when a property is used. Currently you can't use *= you have to use =. It shouldn't be a problem because it's set to identity by default.

Works! Thank you for the triage.