numpy defaults to free-threaded version on conda-forge

My apologies if this is a wrong venue. Is this a conda-forge problem? When I run mamba create -n foo meson-python numpy 'python>=3.13' the free-threaded python is selected: + python 3.13.0 h6355ac2_0_cp313t conda-forge Cached + numpy 2.1.3 py313hb01392b_0 conda-forge Cached When python spec is moved to the front, I do get the "usual" version. However, when I build a python extension with conda-build, I do not control the order. The extension is created for the free-threaded version. Here is my requirements section in recipe/meta.yaml: requirements: build: - {{ compiler('c') }} - {{ compiler('fortran') }} host: - python >=3.13 - python_abi =3.13.* *_cp313 # needed to prevent selection of freethreaded python - meson-python - numpy >=2.0 - pip run: - python >=3.13 - numpy >=2.0 I did manage to find a workaround, which is the python_abi line. Am I doing it right?

Yes, that does look like a conda-forge issue. I think it's this one: https://github.com/conda-forge/python-feedstock/issues/738 On Tue, Nov 26, 2024 at 6:06 PM george trojan via NumPy-Discussion < numpy-discussion@python.org> wrote:
My apologies if this is a wrong venue. Is this a conda-forge problem? When I run
mamba create -n foo meson-python numpy 'python>=3.13'
the free-threaded python is selected:
+ python 3.13.0 h6355ac2_0_cp313t conda-forge Cached + numpy 2.1.3 py313hb01392b_0 conda-forge Cached
When python spec is moved to the front, I do get the "usual" version. However, when I build a python extension with conda-build, I do not control the order. The extension is created for the free-threaded version. Here is my requirements section in recipe/meta.yaml:
requirements: build: - {{ compiler('c') }} - {{ compiler('fortran') }} host: - python >=3.13 - python_abi =3.13.* *_cp313 # needed to prevent selection of freethreaded python - meson-python - numpy >=2.0 - pip run: - python >=3.13 - numpy >=2.0
I did manage to find a workaround, which is the python_abi line. Am I doing it right?
_______________________________________________ NumPy-Discussion mailing list -- numpy-discussion@python.org To unsubscribe send an email to numpy-discussion-leave@python.org https://mail.python.org/mailman3/lists/numpy-discussion.python.org/ Member address: ralf.gommers@googlemail.com

It looks similar as https://github.com/conda-forge/python-feedstock/issues/738, but the issue shows up only when installing numpy. mamba create -n pybug python mamba activate pybug mamba update --all all work fine. Then mamba install cython is OK mamba install numpy wants to change python to free-threaded. mamba --version mamba 1.5.10 conda 24.9.2 Le mar. 26 nov. 2024, à 17 h 44, Ralf Gommers <ralf.gommers@gmail.com> a écrit :
Yes, that does look like a conda-forge issue. I think it's this one: https://github.com/conda-forge/python-feedstock/issues/738
On Tue, Nov 26, 2024 at 6:06 PM george trojan via NumPy-Discussion < numpy-discussion@python.org> wrote:
My apologies if this is a wrong venue. Is this a conda-forge problem? When I run
mamba create -n foo meson-python numpy 'python>=3.13'
the free-threaded python is selected:
+ python 3.13.0 h6355ac2_0_cp313t conda-forge Cached + numpy 2.1.3 py313hb01392b_0 conda-forge Cached
When python spec is moved to the front, I do get the "usual" version. However, when I build a python extension with conda-build, I do not control the order. The extension is created for the free-threaded version. Here is my requirements section in recipe/meta.yaml:
requirements: build: - {{ compiler('c') }} - {{ compiler('fortran') }} host: - python >=3.13 - python_abi =3.13.* *_cp313 # needed to prevent selection of freethreaded python - meson-python - numpy >=2.0 - pip run: - python >=3.13 - numpy >=2.0
I did manage to find a workaround, which is the python_abi line. Am I doing it right?
_______________________________________________ NumPy-Discussion mailing list -- numpy-discussion@python.org To unsubscribe send an email to numpy-discussion-leave@python.org https://mail.python.org/mailman3/lists/numpy-discussion.python.org/ Member address: ralf.gommers@googlemail.com
participants (2)
-
george trojan
-
Ralf Gommers