[clang-linker-wrapper][NFC] Remove dead code

[bader]

InputFiles variable defined in the function is not used. If statements below define local scope variables with the same name.

[AlexeySachkov]

This seems to be a community code: https://github.com/llvm/llvm-project/blame/main/clang/tools/clang-linker-wrapper/ClangLinkerWrapper.cpp#L939

I would suggest to drop our own copy within HasSYCLOffloadKind instead to reduce divergence with the upstream.

[bader]

"HasNonSYCLOffloadKinds branch" builds similar vector differently. I think this is the reason to sink InputFiles initialization into if-conditions.

[AlexeySachkov]

"HasNonSYCLOffloadKinds branch" builds similar vector differently. I think this is the reason to sink InputFiles initialization into if-conditions.

It modifies InputFiles, but the code to initialize seems to be absolutely the same. In that case it is not clear to me as to why we can't re-use the community code here.

If your concern is that we have some other if and other scopes below which also define InputFiles and shadow the original variable, then I would say we should fix those. Shadowing variables is not a good idea

[bader]

It modifies InputFiles, but the code to initialize seems to be absolutely the same. In that case it is not clear to me as to why we can't re-use the community code here.

They are different. Please compare:
https://github.com/intel/llvm/blob/sycl/clang/tools/clang-linker-wrapper/ClangLinkerWrapper.cpp#L2376-L2383 and https://github.com/intel/llvm/blob/sycl/clang/tools/clang-linker-wrapper/ClangLinkerWrapper.cpp#L2464-L2475.

Shadowing variables is not a good idea

That's the reason I remove it. If we go with your proposal, we still have to use shadowing due to the difference I mentioned above.

[AlexeySachkov]

That's the reason I remove it. If we go with your proposal, we still have to use shadowing due to the difference I mentioned above.

Isn't better to fix shadowing in our own code which was added on top of the upstream instead of dropping the upstream code?

[bader]

If you have any suggestions, I'm all ears.

[AlexeySachkov]

If you have any suggestions, I'm all ears.

#16926 is an alternative proposal. I would love to see even more unification, but I can't quickly propose anything as it requires more investigation

[bader]

Your proposal doesn't work. I left a comment in #16926.

[asudarsa]

Good catch Alexey. I think this variable was missed as the code evolved over time.
Also, there will be considerable cleanup if/when we fully move to using clang-sycl-linker.

Thanks