✔️ Linting Passed

All linting checks passed. Your pull request is in excellent shape! ☀️

_{Generated for commit: ac431ca. Link to the linter CI: here}

Looking at the failing CI test, I see at https://dev.azure.com/scikit-learn/scikit-learn/_build/results?buildId=77349&view=logs&jobId=85dc54f1-b746-54c6-aaa8-29d3ad6201c9&j=85dc54f1-b746-54c6-aaa8-29d3ad6201c9&t=5f5d7705-2bc0-5dba-fcc6-e1cf6a70293a

Starting: Select python version to run CI python scripts
==============================================================================
Task         : Use Python version
Description  : Use the specified version of Python from the tool cache, optionally adding it to the PATH
Version      : 0.248.1
Author       : Microsoft Corporation
Help         : https://docs.microsoft.com/azure/devops/pipelines/tasks/tool/use-python-version
==============================================================================
Found tool in cache: Python 3.9.23 x64
Finishing: Select python version to run CI python scripts

but here https://scikit-learn.org/stable/install.html it says that 1.7 requires Python 3.10 or later

not sure if this is related to my CI issue though...

I think the python version reported in the step that you linked to is a red herring. In our CI we always (I think) setup a conda environment in which all the work is done. So the python version shown in the CI step is more like the "system Python", not what is actually used. For example for this job the conda environment is based on https://github.com/scikit-learn/scikit-learn/blob/main/build_tools/azure/pymin_conda_forge_openblas_min_dependencies_environment.yml which specifies python 3.10

I wonder if the explanation for why this job fails and the others dont is the version of meson-python (0.16 for the minimum deps job, 0.18 in pymin_conda_forge_openblas_ubuntu_2204). Have you tried locally with meson-python 0.16? And/or is there something about this "new" style of specifying metadata in the meson-python change log?

If we can confirm it is due to meson-python, I would then summon @lesteve to see if we can bump the version up or not.

@betatim I've tried locally with different meson-python versions, and that seems to be the issue:

For 0.18.0, 0.17.1, installation succeeds; for 0.17.0, 0.16.0 it doesn't. Can't figure out from changelog a likely reason for why, though: https://mesonbuild.com/meson-python/changelog.html

Show-versions output:

System:
    python: 3.11.11 (main, Dec 10 2024, 14:15:35) [Clang 16.0.0 (clang-1600.0.26.4)]
executable: /Users/richard/.pyenv/versions/sklearn/bin/python
   machine: macOS-14.7.6-x86_64-i386-64bit

Python dependencies:
      sklearn: 1.8.dev0
          pip: 24.0
   setuptools: 65.5.0
        numpy: 1.26.4
        scipy: 1.15.3
       Cython: 3.1.2
       pandas: None
   matplotlib: None
       joblib: 1.5.1
threadpoolctl: 3.6.0

Built with OpenMP: True

threadpoolctl info:
       user_api: blas
   internal_api: openblas
    num_threads: 2
         prefix: libopenblas
       filepath: /Users/richard/.pyenv/versions/3.11.11/envs/sklearn/lib/python3.11/site-packages/numpy/.dylibs/libopenblas64_.0.dylib
        version: 0.3.23.dev
threading_layer: pthreads
   architecture: Haswell

       user_api: openmp
   internal_api: openmp
    num_threads: 4
         prefix: libomp
       filepath: /usr/local/Cellar/libomp/20.1.7/lib/libomp.dylib
        version: None

Let's go ahead and bump up the minimum version in

After updating the minimum version in that file you'll have to regenerate the lockfiles with build_tools/update_environments_and_lock_files.py

This looks like this PR broke the wheel build, see #31590 (comment) for more details.