src/bin/sage-cython: Repurpose as PEP 420 fixer

[mkoeppe]

We ship it with the distribution sagemath-environment.

This is needed for building the modularized distributions of the Sage library using meson-python.

With meson >= 1.4.0 (mesonbuild/meson#12674), it then suffices to set CYTHON=sage-cython.

With earlier versions of meson, one can use a native file

[binaries]
cython = 'sage-cython'

... as done in #37012 (54c95d6) and #35095.

📝 Checklist

• The title is concise, informative, and self-explanatory.
• 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 accordingly.

⌛ Dependencies

[tornaria]

Please don't un-deprecate this. Instead it should be removed in favor of using an unpatched version of cython.

Right now, compiling cython files that use sage modules is broken because of namespace packages, and patching cython here and in sage_setup gives the misleading impression that it works.

This is the opposite of "fitting in the python/cython ecosystem".

[mkoeppe]

Note that it is a "temporary workaround".

Anyone report the namespace problem to Cython upstream yet?

[tornaria]

Temporary hack. And when/if cython fixes it we still have to ship the hack since distros are slow updating cython, etc.

Let's stick to stable features, not the bleeding edge.

[mkoeppe]

@tornaria Kindly limit the use of fighting words such as "hack".

Let's stick to stable features, not the bleeding edge.

We have been successfully using Cython with namespace packages ever since 2021.

If anything, it is that our porting effort to Cython 3 is incomplete.

• Use Cython 3 implicit namespace package support #36228 tracks a remaining task, that's all.

[mkoeppe]

@tornaria Is there a dispute here in the sense of the "disputed PRs" process? https://groups.google.com/g/sage-devel/c/IgBYUJl33SQ/m/MV6LXkqRCgAJ

[jhpalmieri]

I'm posting to record a vote of -1 on behalf of Tobias Diez.

[orlitzky]

It's hard to coherently vote on all of these conflicting proposals at once. I lean towards re-adding the init files in the short term, but if the consensus is that we shouldn't do it, then I don't object to this. On the other hand, if I vote to approve this AND to re-add the init files, then we may wind up with two approved but conflicting branches.

[mkoeppe]

This PR here only adds a useful command and does not conflict with anything.

[mkoeppe]

This PR is prepared exactly how PRs should be prepared. It is not a test of ideological consistency.

It simply a useful tool for the build of where there are no __init__.py files -- even if the "re-add init files" PRs is merged. I'm using it in #35095 to build the modularized distributions using meson-python.

[orlitzky]

It simply a useful tool for the build of where there are no __init__.py files -- even if the "re-add init files" PRs is merged. I'm using it in #35095 to build the modularized distributions using meson-python.

OK that's helpful to know.

[mkoeppe]

@tornaria Please check #37009 (comment)

[tornaria]

@tornaria Please check #37009 (comment)

-1 vote from me here.

[github-actions]

Documentation preview for this PR (built with commit 1d21aa2; changes) is ready! 🎉
This preview will update shortly after each push to this PR.

[mkoeppe]

@tornaria Would you be able to express in technical terms what your concern about this PR is, if any?

I am using it in a follow-up for building the modularized distributions with meson-python; it has no effect on building the monolithic Sage library.

[tornaria]

#37009 (comment)

[mkoeppe]

There's nothing there, Gonzalo, just declarations.

[mkoeppe]

@roed314 @saraedum I think the "disputed" tag should be removed; a review has not even been started.