CUDA nightly unused argument -fsycl-device-lib-jit-link on AWS

[sarnex]

Describe the bug

Seems we use werror there and see this:

********************
FAIL: SYCL :: DeviceLib/std_complex_math_test.cpp (577 of 2385)
******************** TEST 'SYCL :: DeviceLib/std_complex_math_test.cpp' FAILED ********************
Exit Code: 1

Command Output (stdout):
--
# RUN: at line 2
/__w/llvm/llvm/toolchain/bin//clang++  -Werror -fsycl -fsycl-targets=nvptx64-nvidia-cuda  /__w/llvm/llvm/llvm/sycl/test-e2e/DeviceLib/std_complex_math_test.cpp -fno-fast-math -o /__w/llvm/llvm/build-e2e/DeviceLib/Output/std_complex_math_test.cpp.tmp1.out
# executed command: /__w/llvm/llvm/toolchain/bin//clang++ -Werror -fsycl -fsycl-targets=nvptx64-nvidia-cuda /__w/llvm/llvm/llvm/sycl/test-e2e/DeviceLib/std_complex_math_test.cpp -fno-fast-math -o /__w/llvm/llvm/build-e2e/DeviceLib/Output/std_complex_math_test.cpp.tmp1.out
# note: command had no output on stdout or stderr
# RUN: at line 3
env SYCL_UR_CUDA_ENABLE_IMAGE_SUPPORT=1 ONEAPI_DEVICE_SELECTOR=cuda:gpu  /__w/llvm/llvm/build-e2e/DeviceLib/Output/std_complex_math_test.cpp.tmp1.out
# executed command: env SYCL_UR_CUDA_ENABLE_IMAGE_SUPPORT=1 ONEAPI_DEVICE_SELECTOR=cuda:gpu /__w/llvm/llvm/build-e2e/DeviceLib/Output/std_complex_math_test.cpp.tmp1.out
# .---command stdout------------
# | Pass
# `-----------------------------
# RUN: at line 5
/__w/llvm/llvm/toolchain/bin//clang++  -Werror -fsycl -fsycl-targets=nvptx64-nvidia-cuda  /__w/llvm/llvm/llvm/sycl/test-e2e/DeviceLib/std_complex_math_test.cpp -fsycl-device-lib-jit-link -fno-fast-math -o /__w/llvm/llvm/build-e2e/DeviceLib/Output/std_complex_math_test.cpp.tmp2.out
# executed command: /__w/llvm/llvm/toolchain/bin//clang++ -Werror -fsycl -fsycl-targets=nvptx64-nvidia-cuda /__w/llvm/llvm/llvm/sycl/test-e2e/DeviceLib/std_complex_math_test.cpp -fsycl-device-lib-jit-link -fno-fast-math -o /__w/llvm/llvm/build-e2e/DeviceLib/Output/std_complex_math_test.cpp.tmp2.out
# .---command stderr------------
# | clang++: error: argument unused during compilation: '-fsycl-device-lib-jit-link' [-Werror,-Wunused-command-line-argument]
# `-----------------------------
# error: command failed with exit status: 1

--

********************
FAIL: SYCL :: DeviceLib/std_complex_math_fp64_test.cpp (578 of 2385)
******************** TEST 'SYCL :: DeviceLib/std_complex_math_fp64_test.cpp' FAILED ********************
Exit Code: 1

Command Output (stdout):
--
# RUN: at line 4
/__w/llvm/llvm/toolchain/bin//clang++  -Werror -fsycl -fsycl-targets=nvptx64-nvidia-cuda  /__w/llvm/llvm/llvm/sycl/test-e2e/DeviceLib/std_complex_math_fp64_test.cpp -o /__w/llvm/llvm/build-e2e/DeviceLib/Output/std_complex_math_fp64_test.cpp.tmp1.out
# executed command: /__w/llvm/llvm/toolchain/bin//clang++ -Werror -fsycl -fsycl-targets=nvptx64-nvidia-cuda /__w/llvm/llvm/llvm/sycl/test-e2e/DeviceLib/std_complex_math_fp64_test.cpp -o /__w/llvm/llvm/build-e2e/DeviceLib/Output/std_complex_math_fp64_test.cpp.tmp1.out
# note: command had no output on stdout or stderr
# RUN: at line 5
env SYCL_UR_CUDA_ENABLE_IMAGE_SUPPORT=1 ONEAPI_DEVICE_SELECTOR=cuda:gpu  /__w/llvm/llvm/build-e2e/DeviceLib/Output/std_complex_math_fp64_test.cpp.tmp1.out
# executed command: env SYCL_UR_CUDA_ENABLE_IMAGE_SUPPORT=1 ONEAPI_DEVICE_SELECTOR=cuda:gpu /__w/llvm/llvm/build-e2e/DeviceLib/Output/std_complex_math_fp64_test.cpp.tmp1.out
# .---command stdout------------
# | Pass
# `-----------------------------
# RUN: at line 7
/__w/llvm/llvm/toolchain/bin//clang++  -Werror -fsycl -fsycl-targets=nvptx64-nvidia-cuda  /__w/llvm/llvm/llvm/sycl/test-e2e/DeviceLib/std_complex_math_fp64_test.cpp -fsycl-device-lib-jit-link -o /__w/llvm/llvm/build-e2e/DeviceLib/Output/std_complex_math_fp64_test.cpp.tmp2.out
# executed command: /__w/llvm/llvm/toolchain/bin//clang++ -Werror -fsycl -fsycl-targets=nvptx64-nvidia-cuda /__w/llvm/llvm/llvm/sycl/test-e2e/DeviceLib/std_complex_math_fp64_test.cpp -fsycl-device-lib-jit-link -o /__w/llvm/llvm/build-e2e/DeviceLib/Output/std_complex_math_fp64_test.cpp.tmp2.out
# .---command stderr------------
# | clang++: error: argument unused during compilation: '-fsycl-device-lib-jit-link' [-Werror,-Wunused-command-line-argument]
# `-----------------------------
# error: command failed with exit status: 1

https://github.com/intel/llvm/actions/runs/15338458359/job/43160339204

I don't know if we should just remove the flag or if it's supposed to be used.

To reproduce

No response

Environment

No response

Additional context

No response

[aelovikov-intel]

+ @intel/llvm-reviewers-cuda , can you look into this? Either fix or at least disable the test somehow to make the job green in Nightly...

[npmiller]

Created a quick workaround patch here: #18985

Other tests using this flag are already using -Wno-error=unused-command-line-argument, so it seems fair.

But this flag has no effect on Nvidia/AMD compilation, it's only used for SPIR, from the clang driver:

          bool UseJitLink =
              IsSPIR &&
              Args.hasFlag(options::OPT_fsycl_device_lib_jit_link,
                           options::OPT_fno_sycl_device_lib_jit_link, false);

So the warning is correct, and it doesn't really have anything to do with CUDA.