JuliaPy
Support free-threaded Python [3.14]
It's no longer marked experimental, as of 3.14 now in rc2.
https://py-free-threading.github.io/running-gil-disabled/
The free-threaded Python executable will always have a python3.13t alias (for Python 3.13); whether python, python3 and/or python3.13 point at the free-threaded executable depends on the installation method used (see the install guide for more details).
I believe it's off by default in 3.13, but should also be supportable there.
It's a question if JuliaCall should install and use those t versions. I'm not sure if it means any changes for JuliaCall, other than pointing to the right runtime.
Cool, I didn't know it was non-experimental now.
It would certainly make PythonCall simpler if it could assume there was no GIL. However since the GIL gets re-enabled if you load a module which requires the GIL then for now at least we need to keep the GIL handling.
If the ABI is the same for free-threaded and ordinary python then PythonCall should just work with it.
Is a free-threaded build available on conda-forge or similar?
It's possible we could have a bit of special functionality. For instance we could have a way to specify we want a free-threaded build. And we could have a preference which forces free-threaded mode (PYTHON_GIL=0) and turns our GIL handling into no-ops.
To answer my own question, yes there are some meta-packages to specify which flavour of python to use.
https://anaconda.org/conda-forge/python-freethreading
So someone can already just add python-freethreading to their CondaPkg.toml to get such a build.
Would be good to run the test suite on it.