[BUG]: cmake installation fails on (lack of) llvm

[VictorEijkhout]

Is this a duplicate?

• I confirmed there appear to be no duplicate issues for this bug and that I agree to the Code of Conduct

Type of Bug

Silent Failure

Component

Infrastructure

Describe the bug

-- Found CUDAToolkit: /home1/apps/nvidia/Linux_aarch64/24.7/cuda/12.5/targets/sbsa-linux/include;/opt/apps/nvidia_math/12.4/include (found version "12.5.82")
-- Performing Test CMAKE_HAVE_LIBC_PTHREAD
-- Performing Test CMAKE_HAVE_LIBC_PTHREAD - Success
-- Found Threads: TRUE
-- Enabled CUDA architectures:
-- Linker detection: unknown
CMake Warning at libcudacxx/cmake/AddLLVM.cmake:1702 (message):
  LLVM_EXTERNAL_LIT set to LLVM_DEFAULT_EXTERNAL_LIT-NOTFOUND, but the path
  does not exist.
Call Stack (most recent call first):
  libcudacxx/cmake/AddLLVM.cmake:1747 (get_llvm_lit_path)
  libcudacxx/cmake/AddLLVM.cmake:1792 (add_lit_target)
  libcudacxx/test/libcudacxx/CMakeLists.txt:116 (add_lit_testsuite)


CMake Error at libcudacxx/test/libcudacxx/CMakeLists.txt:120 (find_program):
  Could not find libcudacxx_LIT using the following names: lit

Does this mean I can only install with llvm compiler? I'm using nvidia compiler.

How to Reproduce

0. Download 2.8.0 release
1. run cmake

Expected behavior

cmake should succeed.

Reproduction link

No response

Operating System

No response

nvidia-smi output

No response

NVCC version

No response

[miscco]

That seems to be a lapse because our tests require lit the LLVM Infrastructure Test client.

@alliepiper we should definitely fix that. Until then, you should be able to just do a pip install lit