pip install fails without cuda

[Tim Berberich]

Hi Nathanaël,

calling 'pip install shtns' fails in building wheel on a check for cuda:

checking for gcc option to support OpenMP... -fopenmp
-L/opt/cuda/lib64 -L/opt/cuda/lib64/stubs
checking for cudaMalloc in -lcudart... no
configure: error: cudart library not found. Try adding LDFLAGS="-L/usr/local/cuda/lib64"
error: command './configure' failed with exit code 1
[end of output]

note: This error originates from a subprocess, and is likely not a problem with pip.
ERROR: Failed building wheel for shtns
Failed to build shtns
ERROR: Could not build wheels for shtns, which is required to install pyproject.toml-based projects

I don’t have cuda installed so failing to find it is expected. Is there an option to tell the installation process to skip the cuda check since I just want to use the CPU version of shtns.

The same error is thrown when using the installation via your git directly, i.e.:
./configure --enable-python
make
make install

‌

[Nathanaël Schaeffer]

Mmm, this is strange. This should only happen if you have the CUDA_PATH environment variable set.
To explicitely disable cuda, try:

CUDA_PATH= pip install shtns

or

unset CUDA_PATH
pip install shtns

‌

[Tim Berberich]

Thanks for the help, this worked!
It seems it was a problem on my side withe a leftover environment variable.