Specify use_sources=True in cython compilation

[user202729]

In my testing, this allows cython compilation of code with from sage.[...] cimport [...] to work when built with meson. Ability to compile cython code when use legacy build system is unaffected.

Actually without this change cython compilation works fine on conda & meson CI anyway (probably because of some coincidence, I didn't inspect)

📝 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 6478f43; changes) is ready! 🎉
This preview will update shortly after each push to this PR.

[dimpase]

has this fixed any failing tests?

[user202729]

for some reason (probably some path coincidence or whatever) it works fine on CI either way. Maybe I should figure out some way to test it on CI.

[dimpase]

Please post an explicit example of what fails for you, and the specifics of your Cython setup.

[user202729]

Okay here it is: if --editable install is used with meson then it fails.

e.g.

https://github.com/sagemath/sage/actions/runs/12913561812/job/36011095007#step:10:14938

2025-01-22T17:30:30.3441153Z **********************************************************************
2025-01-22T17:30:30.3441906Z File "src/sage/arith/long.pxd", line 119, in sage.arith.long.pxd.integer_check_long
2025-01-22T17:30:30.3486008Z Failed example:
2025-01-22T17:30:30.3486410Z     cython('''
2025-01-22T17:30:30.3486776Z     from sage.arith.long cimport *
2025-01-22T17:30:30.3487271Z     from sage.rings.integer cimport smallInteger

[...]

2025-01-22T17:30:30.3550836Z     RuntimeError: Error compiling Cython file:
2025-01-22T17:30:30.3551358Z     ------------------------------------------------------------
2025-01-22T17:30:30.3551848Z     ...
2025-01-22T17:30:30.3552031Z 
2025-01-22T17:30:30.3552192Z     from sage.arith.long cimport *
2025-01-22T17:30:30.3552601Z     ^
2025-01-22T17:30:30.3552970Z     ------------------------------------------------------------
2025-01-22T17:30:30.3553339Z 
2025-01-22T17:30:30.3553979Z     _tmp_sage_3enyaq4d_tmp_1nsgnqax_pyx_0.pyx:2:0: 'sage/arith/long.pxd' not found

Side note, do you think it's desirable to test all combinations (Python 3.11, 12 × editable: true, false) each pull request? Or we could cut down to only test e.g. {Python 3.11 non-editable, Python 3.12 editable} on pull request and test the full matrix in pushes to branch.

[user202729]

There's a bunch of additional things that fail with --editable apparently.

Maybe we can just merge the first commit in and enable --editable on CI later? (or should they be put as known CI failure to be fixed gradually?)