Also test all shader types and be a bit more careful to make shaders
reasonably valid in these tests.
Details
Diff Detail
- Repository
- rG LLVM Github Monorepo
Event Timeline
clang/test/SemaHLSL/shader_type_attr.hlsl | ||
---|---|---|
30 | I don't think we should expect this error, given that there is no previous declaration for doubledUp(). Maybe something else like %0 and %1 are incompatible attributes? |
clang/test/SemaHLSL/shader_type_attr.hlsl | ||
---|---|---|
30 | I don't necessarily disagree, but if we're going to change that I think it should be in a different change than this one. |
clang/test/SemaHLSL/shader_type_attr.hlsl | ||
---|---|---|
30 | I suspect this change is just to prevent the presence of the compute annotation from triggering the missing numthreads error. | |
61 | Oh interesting! DXC doesn't actually support this syntax, but I think it is worth supporting in Clang. |
clang/test/SemaHLSL/entry_shader.hlsl | ||
---|---|---|
1 | Another case of trying to keep things to one error - anyhit doesn't support the numthreads attribute | |
clang/test/SemaHLSL/shader_type_attr.hlsl | ||
30 | That's correct - just trying to keep to testing the conflicting attribute error and not have a spurious numthreads as well. | |
61 | Hm, it might be better to stick to the syntax dxc handles for these tests at least. I'll update that. Sure, checking both attributes seems like a good idea. |
Why was this changed to mesh instead of "anyhit" ?