Using pyenv breaks PythonCall

[doorisajar]

Affects: PythonCall

Describe the bug
After setting up a local Python version using pyenv, creating a virtual environment, and setting up PythonCall to use that venv instead of Conda, PythonCall can't be loaded:

julia> using PythonCall
┌ Warning: Python library "/Users/rdoake/.pyenv/versions/3.9.13/lib/python3.9/config-3.9-darwin/libpython3.9.a" could not be opened.
└ @ PythonCall.C ~/.julia/packages/PythonCall/1f5yE/src/cpython/context.jl:118
┌ Warning: Python library "/Users/rdoake/.pyenv/versions/3.9.13/lib/libpython3.9.a" could not be opened.
└ @ PythonCall.C ~/.julia/packages/PythonCall/1f5yE/src/cpython/context.jl:118
ERROR: InitError: ArgumentError: NULL library handle

The files referenced in the error message do exist.

Steps to reproduce

• Create a directory & shell into it
• pyenv local x.x # I used 3.9 to disambiguate from my default system Python, which is 3.10
• create a venv at .venv using the tool of your choice (I tested with Poetry and venv)
• launch Julia
• ] activate .
• > ENV["JULIA_CONDAPKG_BACKEND"]="Null" # don't use Conda
• > ENV["JULIA_PYTHONCALL_EXE"]=".venv/bin/python3" # point PythonCall to the venv python
• ] add PythonCall
• > using PythonCall

My system

• MacOS 13.4
• Julia 1.9.0
• PythonCall 0.9.13

Additional context
If I instead use poetry to specify a Python version for the virtual environment, and then do the above steps without using pyenv, then PythonCall works as expected.

[cjdoris]

How did you install the version of Python you are trying to use? From the logs you've sent, it's a statically linked build of Python which is unfortunately not supported by PythonCall (because it needs to dynamically link to libpython). Often this is the case for Python bundled in system package managers.

[doorisajar]

I manage installed Python versions using pyenv.

[cjdoris]

Ah, I didn't realise pyenv builds Python for you. Looks like you can get it to build the shared library here: https://github.com/pyenv/pyenv/blob/master/plugins/python-build/README.md#building-with---enable-shared

[doorisajar]

Just realized I forgot to circle back to this -- that workaround does enable PythonCall.jl to work with a pyenv-installed Python version. I can keep doing that for the time being. Thanks!