spkg bumps for Python 3.14

[dimpase]

needed for building Sage with Python 3.14, as tracked on #40890

📝 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

[github-actions]

Documentation preview for this PR (built with commit eb347d3; changes) is ready! 🎉
This preview will update shortly after each push to this PR.

[dimpase]

we are not 3.14-compatible yet.
How about 3.13?

[enriqueartal]

I deleted my post since I find the patch in issue #40890. In fedora 42, sage (dev and stable) works with system python 3.13.7.
Now a virtual system is trying to build it in fedora 43 with this PR, the above patch and also #40882 and #40888, but it may take a while.

[enriqueartal]

I have to repeat make several times because it stops without an error but there is no package to blame.

[enriqueartal]

Running ./bootstrap produces some complains about dashes and underscore and it seems it affects hatch-vcs

[dimpase]

I have to repeat make several times because it stops without an error but there is no package to blame.

this is probably because you forgot to merge #40882 (unbreak the current beta) before testing

[dimpase]

Running ./bootstrap produces some complains about dashes and underscore and it seems it affects hatch-vcs

./bootstrap:89: installing /home/dp/sage/src/doc/en/reference/spkg/*.rst
Traceback (most recent call last):
  File "/home/dp/sage/build/bin/sage-package", line 39, in <module>
    run()
    ~~~^^
  File "/home/dp/sage/build/bin/../sage_bootstrap/cmdline.py", line 520, in run
    app.dependencies(*args.package_class, types=types, format=args.format)
    ~~~~~~~~~~~~~~~~^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/home/dp/sage/build/bin/../sage_bootstrap/app.py", line 183, in dependencies
    elif format == 'rst' and Package(dep).has_file('SPKG.rst'):
                             ~~~~~~~^^^^^
  File "/home/dp/sage/build/bin/../sage_bootstrap/package.py", line 53, in __new__
    self.__init__(package_name)
    ~~~~~~~~~~~~~^^^^^^^^^^^^^^
  File "/home/dp/sage/build/bin/../sage_bootstrap/package.py", line 79, in __init__
    raise ValueError('package names use underscores, not dashes, got {0}'.format(package_name))
ValueError: package names use underscores, not dashes, got hatch-fancy-pypi-readme
Traceback (most recent call last):
  File "/home/dp/sage/build/bin/sage-package", line 39, in <module>
    run()
    ~~~^^
  File "/home/dp/sage/build/bin/../sage_bootstrap/cmdline.py", line 520, in run
    app.dependencies(*args.package_class, types=types, format=args.format)
    ~~~~~~~~~~~~~~~~^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/home/dp/sage/build/bin/../sage_bootstrap/app.py", line 183, in dependencies
    elif format == 'rst' and Package(dep).has_file('SPKG.rst'):
                             ~~~~~~~^^^^^
  File "/home/dp/sage/build/bin/../sage_bootstrap/package.py", line 53, in __new__
    self.__init__(package_name)
    ~~~~~~~~~~~~~^^^^^^^^^^^^^^
  File "/home/dp/sage/build/bin/../sage_bootstrap/package.py", line 79, in __init__
    raise ValueError('package names use underscores, not dashes, got {0}'.format(package_name))
ValueError: package names use underscores, not dashes, got setuptools-scm

[enriqueartal]

Exactly. Enabling system-packages one can go on.

[dimpase]

also names of packages in distro/ for the newly introduced spkgs have to be adjusted. E.g. for gentoo there should be dev-python/hatch-vcs

[enriqueartal]

sage_setup complains about 3.14.0

[dimpase]

The mess with underscores etc is very unpleasant. Sage's spkgs cannot have more than 1 underscore in their names - I've recently seen that with roman_numerals (it had to be renamed from roman_numerals_py)

[dimpase]

sage_setup complains about 3.14.0

#40890 (comment)

[enriqueartal]

sage_setup complains about 3.14.0

#40890 (comment)

I missed this; applied.

[dimpase]

diff --git a/build/pkgs/argon2_cffi/dependencies b/build/pkgs/argon2_cffi/dependencies
index 81c83adedaf..8b0d10fd255 100644
--- a/build/pkgs/argon2_cffi/dependencies
+++ b/build/pkgs/argon2_cffi/dependencies
@@ -1,4 +1,4 @@
-argon2_cffi_bindings | $(PYTHON_TOOLCHAIN) flit_core $(PYTHON) hatchling hatch-vcs hatch-fancy-pypi-readme
+argon2_cffi_bindings | $(PYTHON_TOOLCHAIN) flit_core $(PYTHON) hatchling hatch_vcs hatch_readme
 
 ----------
 All lines of this file are ignored except the first.
diff --git a/build/pkgs/hatch_vcs/dependencies b/build/pkgs/hatch_vcs/dependencies
index 93937564726..210ca14d94e 100644
--- a/build/pkgs/hatch_vcs/dependencies
+++ b/build/pkgs/hatch_vcs/dependencies
@@ -1 +1 @@
-setuptools-scm hatchling | pip $(PYTHON)
+setuptools_scm hatchling | pip $(PYTHON)

fixes the bootstrap for me

[enriqueartal]

As I said I bypassed this enabling system-packages but this is useful to avoid the problem, thanks.
Now it fails with sagelib because of atexit, (even applying your proposed patch). Do you need the log?

[dimpase]

As I said I bypassed this enabling system-packages but this is useful to avoid the problem, thanks. Now it fails with sagelib because of atexit, (even applying your proposed patch). Do you need the log?

Thanks, and no, we understand what's going on, Basically most of atexit.pyx is already obsolete on Python 3.13,
see python/cpython#114279

So we should work on refactoring/replacing it, see #40890 for more details

[dimpase]

Could you please review this on Python <3.14 ? We need these in anyway.

[dimpase]

@enriqueartal and review #40888 too, please

[enriqueartal]

I am testing in Fedora 42 with python 3.13. There is an error with argon2_cffi. It is not built because of hatch_readme. In config.log it claims that system package will be used. In distros/fedora.txt it looks for python3-hatchling which is not the right one and moreover it is not installed. The same happens in config.log after putting the right fedora package (not installed). With ./configure --with-system-hatch_readme=no, it is OK. I hope it will finish shortly for sagelib.

[enriqueartal]

Besides the above problems, everything works.

[dimpase]

OK, indeed, build/pkgs/hatch_readme/distros/ need adjusting. I'll push an update

[dimpase]

I've added and corrected data on packages, in particular on Fedora. Can you have another look whether it works with system packages hatch* ?

[enriqueartal]

Do you mean to use enable-system-packages?

[dimpase]

yes

[dimpase]

perhaps there must be version constraints set, to avoid cryptic errors stemming ftom incompatible packages used

[enriqueartal]

When building ipykernel it fails:

ModuleNotFoundError: No module named 'jupyter_core'

Maybe another confusion between _ and -?

[dimpase]

On Thu, Oct 02, 2025 at 11:57:55AM -0700, Enrique Manuel Artal Bartolo wrote: enriqueartal left a comment (sagemath/sage#40892) When building `ipykernel` it fails: ``` ModuleNotFoundError: No module named 'jupyter_core' ``` Maybe another confusion between _ and -?

please a bit more context, perhaps config.log, to understand where each affected spkg comes from

[enriqueartal]

This is config.log and ipykernel-6.27.1.log. I hope it may help.

[dimpase]

This is config.log and ipykernel-6.27.1.log. I hope it may help.

this is what I also saw when I tried this with some homebrew Python packages, so this should be fixed by the last commit.
(more precisely, the error was about incompatibility with another packages, not hatchling (that was already updated))
Could you check?

[cxzhong]

You also need to bump flit_core and info. Because the old flit_core is not fit in Python 3.14 and the old info os not fit for gcc 15.

[dimpase]

You also need to bump flit_core and info. Because the old flit_core is not fit in Python 3.14 and the old info os not fit for gcc 15.

flit-core is already bumped here. I'm less worried about info, as a sane system setup will have (tex)info installed, and so it can be used. Also, there is no relation between info and Python, so this should be another PR to make.

[dimpase]

@cxzhong - I bumped info in #41034, please review

[cxzhong]

Then scipy, ipython, notebook, pip, wheel, pyproject-api, pytest which are important. and we can remove tomli, from python 3.11, we do not need this library. It is a python standard library