bader wrote:

> > > The spirv-tools feature/substitution setup is carried over from Tooling's 
> > > lit.local.cfg into a new lit.local.cfg in clang-linker-wrapper/, since 
> > > the parent clang/test/lit.cfg.py does not set up that feature. The dead 
> > > spirv-tools block is removed from Tooling/lit.local.cfg now that no tests 
> > > there depend on it.
> > 
> > 
> > Why do OffloadTools tests depend on spirv-tools? I expect only SPIR-V 
> > backend tests need those.
> 
> @bader ,
> 
> ```
> # executed command: clang-linker-wrapper 
> --host-triple=x86_64-unknown-linux-gnu --linker-path=/usr/bin/ld 
> openmp-spirv.cpp.tmp.host.o -o 
> clang-linker-wrapper/Output/openmp-spirv.cpp.tmp
> 
> # .---command stderr------------
> # | clang: error: cannot find SPIR-V Tools binary 'spirv-link', which is 
> required for SPIR-V compilation. It can be obtained from your system package 
> manager or from KhronosGroup/SPIRV-Tools on GitHub
> # | /localdisk2/yplyakhi/llvm-project/build/bin/clang-linker-wrapper: error: 
> 'clang' failed
> # `-----------------------------
> # error: command failed with exit status: 1
> ```
> 
> this is my understanding: `clang-linker-wrapper` runs `clang 
> --target=spirv64-intel`, and the driver selects SPIRV::Linker for the link 
> step, which resolves to `spirv-link`.

Shouldn't we run the test in "dry-run" mode to focus on testing 
`clang-linker-wrapper` functionality?

https://github.com/llvm/llvm-project/pull/198434
_______________________________________________
cfe-commits mailing list
[email protected]
https://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits

Reply via email to