DirectX is required to compile to SPIR-V

[sutbult]

I built SDL_shadercross locally using this command:

cmake -B build -DSDLSHADERCROSS_DXC=OFF -DSDLSHADERCROSS_INSTALL=ON -DSDLSHADERCROSS_SHARED=ON

As you can see, I disabled DXC as I in this early stage of development of the project I am working on, I am only interested in compiling my program on Linux and running SDL_GPU with the Vulkan backend.

I have this function:

SDL_GPUShader* compileShader(SDL_GPUDevice* device) {
    SDL_ShaderCross_HLSL_Info info = {0};
    SDL_ShaderCross_GraphicsShaderMetadata metadata = {0};

    SDL_GPUShader *vertexShader = SDL_ShaderCross_CompileGraphicsShaderFromHLSL(
        device,
        &info,
        &metadata);

    if(vertexShader == NULL) {
        fprintf(
            stderr,
            "SDL_ShaderCross_CompileGraphicsShaderFromHLSL failed: %s\n",
            SDL_GetError());
        return NULL;
    }

    return vertexShader;
}

When the function is called, I get this output:

SDL_ShaderCross_CompileGraphicsShaderFromHLSL failed: Shadercross was not built with DXC support, cannot compile using DXC!

I would expect something like "Your source is null" instead.

I also tried this:

shadercross triangle.vert.hlsl -o triangle.vert.spv

Then I got this result:

ERROR: Failed to compile SPIR-V From HLSL: Shadercross was not built with DXC support, cannot compile using DXC!

So it seems like the command line tool has the same problem.

[thatcosmonaut]

Not really sure how you expect to compile HLSL to SPIR-V without DXC, the compiler that compiles HLSL to SPIR-V.

[sutbult]

Okay, it just seems kind of counter-intuitive to require something related to DirectX to compile shader code for Vulkan.

[thatcosmonaut]

You could always use GLSL and a compiler like glslc to produce SPIR-V. You aren't required to use HLSL or SDL_shadercross.

[sutbult]

You could always use GLSL and a compiler like glslc to produce SPIR-V. You aren't required to use HLSL or SDL_shadercross.

That's a good point. However, I solved it by running the command-line tool through Docker.