Arch Linux User Repository

Builds fine here, might be something with your system.

-- The CXX compiler identification is GNU 15.2.1 -- The CUDA compiler identification is NVIDIA 13.0.88 with host compiler GNU 15.2.1

namcap says nvidia-utils package might be needed, you could try if it helps

cuda_memtest E: Dependency nvidia-utils detected and not included (libraries ['usr/lib/libnvidia-ml.so.1'] needed in files ['usr/bin/cuda_memtest'])

Does not build any more. At build()

Compiler: /opt/cuda/bin/nvcc
/usr/bin/g++-14: No such file or directory

So maybe the problem is with cuda package not depending on g++-14. Cuda is version 13.0 and g++ is version 15.2.1 in Arch now.

@revel: Well, since they learn programming from NVIDIA, one big code blob is probably by design /s

@dreieck This package was not meant to track a git repo, I did it out of desperation because the original package was outdated and unmaintained and the fork doesn't bother to make releases. Hopefully this will change.

nvcc is part of cuda package, should work for you.

✗ nvcc -V
nvcc: NVIDIA (R) Cuda compiler driver
Copyright (c) 2005-2023 NVIDIA Corporation
Built on Fri_Nov__3_17:16:49_PDT_2023
Cuda compilation tools, release 12.3, V12.3.103
Build cuda_12.3.r12.3/compiler.33492891_0
✗ which nvcc
/opt/cuda/bin/nvcc
✗ pacman -Qo /opt/cuda/bin/nvcc
/opt/cuda/bin/nvcc is owned by cuda 12.3.1-2

Fails to build for me due to missing nvcc (cuda version 12.3.1-2 is installed):

CMake Error at /usr/share/cmake/Modules/Internal/CMakeCUDAFindToolkit.cmake:104 (message):
  Failed to find nvcc.

  Compiler requires the CUDA toolkit.  Please set the CUDAToolkit_ROOT
  variable.
Call Stack (most recent call first):
  /usr/share/cmake/Modules/CMakeDetermineCUDACompiler.cmake:89 (cmake_cuda_find_toolkit)
  CMakeLists.txt:43 (enable_language)

Maybe another missing makedependency?

Regards!

You download the latest state from a git repository.

So you must:

• Rename your package to cuda_memtest-git (i.e. with the -git suffix),
• add cuda_memtest to the provides and conflicts arrays.

(For renaming, create newly named package and submit a merge request for the old package.)

Regards and thanks for maintaining!

@MartinX3: Possibly your AUR helper doesn't install cuda dependency early enough. Try installing it first and then this package. If that's not it, try to install cuda-tools and please report here if that helps.

@revel: There are a few issues with the package:

1. The package description is somewhat misleading since the OpenCL part (ocl_memtest) is not maintained and possibly even broken.

2. There's no explanation as to why the patch in prepare() is commented out by default?

3. Similarly there are two generic lines in package(), intertwined with two OpenCL-related, but it is again not clear why would they be commented out by default?

   install -d "$pkgdir/usr/bin/" install -d "$pkgdir/usr/share/$pkgname/"

4. As has been mentioned before the standard package name would be cuda_memtest-git given the source versioning.

Sadly it is still broken for me.

==> Beginne build()...
-- The CXX compiler identification is GNU 13.1.1
-- Detecting CXX compiler ABI info
-- Detecting CXX compiler ABI info - done
-- Check for working CXX compiler: /usr/lib/ccache/bin/c++ - skipped
-- Detecting CXX compile features
-- Detecting CXX compile features - done
CMake Error at /usr/share/cmake/Modules/CMakeDetermineCUDACompiler.cmake:180 (message):
  Failed to find nvcc.

  Compiler requires the CUDA toolkit.  Please set the CUDAToolkit_ROOT
  variable.
Call Stack (most recent call first):
  CMakeLists.txt:43 (enable_language)

Done, thanks @dwheeler