build/pkgs/numpy 2.0.1, scipy 1.13.1, cmake 3.30.2, gsl 2.8, meson 1.5.1, meson_python 0.16.0, ninja_build 1.12.1, openblas 0.3.28, pybind11 2.13.1, pythran 0.16.1

[mkoeppe]

• https://numpy.org/doc/stable/release/2.0.0-notes.html

• https://scipy.github.io/devdocs/release/1.13.1-notes.html

• Fixes using numpy 2.0 #38242

📝 Checklist

• The title is concise and informative.
• The description explains in detail what this PR is about.
• I have linked a relevant issue or discussion.
• I have created tests covering the changes.
• I have updated the documentation and checked the documentation preview.

⌛ Dependencies

• Depends on provide compatibility with numpy 2.0 #38250 (merged here)
• Depends on build/pkgs/setuptools_scm: Upgrade to 8.1.0, fix dependency typing_extensions, change some packages to wheel #38226 (merged here)

[github-actions]

Documentation preview for this PR (built with commit 16eb5e4; changes) is ready! 🎉
This preview will update shortly after each push to this PR.

[mkoeppe]

Tests at https://github.com/mkoeppe/sage/actions/runs/10059877764

[kwankyu]

I get

sage -t --warn-long 69.9 --random-seed=274666681782489637879204037307995713165 src/sage/graphs/generic_graph.py
**********************************************************************
File "src/sage/graphs/generic_graph.py", line 15996, in sage.graphs.generic_graph.GenericGraph.?
Failed example:
    (graphs.FruchtGraph()).clustering_coeff(weight=True)                  # needs networkx
Expected:
    {0: 0.3333333333333333, 1: 0.3333333333333333, 2: 0,
     3: 0.3333333333333333, 4: 0.3333333333333333,
     5: 0.3333333333333333, 6: 0.3333333333333333,
     7: 0.3333333333333333, 8: 0, 9: 0.3333333333333333,
     10: 0.3333333333333333, 11: 0}
Got:
    {0: np.float64(0.3333333333333333),
     1: np.float64(0.3333333333333333),
     2: 0,
     3: np.float64(0.3333333333333333),
     4: np.float64(0.3333333333333333),
     5: np.float64(0.3333333333333333),
     6: np.float64(0.3333333333333333),
     7: np.float64(0.3333333333333333),
     8: 0,
     9: np.float64(0.3333333333333333),
     10: np.float64(0.3333333333333333),
     11: 0}
**********************************************************************
File "src/sage/graphs/generic_graph.py", line 16006, in sage.graphs.generic_graph.GenericGraph.?
Failed example:
    (graphs.FruchtGraph()).clustering_coeff(nodes=[0,1,2],                # needs networkx
                                            weight=True)
Expected:
    {0: 0.3333333333333333, 1: 0.3333333333333333, 2: 0}
Got:
    {0: np.float64(0.3333333333333333), 1: np.float64(0.3333333333333333), 2: 0}
**********************************************************************
1 item had failures:
   2 of 1269 in sage.graphs.generic_graph.GenericGraph.?
    [3915 tests, 2 failures, 23.88 s]
----------------------------------------------------------------------
sage -t --warn-long 69.9 --random-seed=274666681782489637879204037307995713165 src/sage/graphs/generic_graph.py  # 2 doctests failed
----------------------------------------------------------------------

[kwankyu]

@tobiasdiez scipy is a dependency of the distribution package sagemath-standard. As explained in the comment, scipy imposes certain constrains on cython. So it is reasonable to reflect those constraints into src/pyproject.toml. No?

I can now answer to this. After Tobias's PR #36982, src/pyproject.toml is the source of the version constraints on some dependencies (in this case, cython) of sage-the-distribution. So yes.

This may impose a version constraint stricter than necessary on the dependency of sagemath-standard. However, according to Matthias: #37902 (comment), it is dubious whether sagemath-standard gets real harm from the stricter version constraint.

Anyway there is no mechanism in sage that can be used to maintain two sets of version constraints, one for sage-the-distribution and one for modularized distributions.

[kwankyu]

Working well. LGTM.

[mkoeppe]

The demand by Tobias Diez (https://groups.google.com/g/sage-devel/c/sVcJiJdUxRs/m/oEntIDE9AgAJ), to deviate from the long-standing practice of synchronized version constraints, are out of scope of this upgrade PR. Setting to "disputed".

[kwankyu]

+1 from me.