Dependencies: Restrict Python to strictly below 3.13

[agoscinski]

Recstrict Python version ito strictly below 3.13 until we have resolved the dependency issues in PR #6600.

[codecov[bot]]

Codecov Report

All modified and coverable lines are covered by tests :white_check_mark:

Project coverage is 77.87%. Comparing base (ef60b66) to head (067a482). Report is 132 commits behind head on main.

Additional details and impacted files

```diff @@ Coverage Diff @@ ## main #6603 +/- ## ========================================== + Coverage 77.51% 77.87% +0.37% ========================================== Files 560 567 +7 Lines 41444 42120 +676 ========================================== + Hits 32120 32796 +676 Misses 9324 9324 ```

:umbrella: View full report in Codecov by Sentry.
:loudspeaker: Have feedback on the report? Share it here.

[sphuber]

Not saying we shouldn't do this, but is this really necessary? Our own code is compatible with Python 3.13. It is just that certain dependencies aren't. Adding this limitation is not going to do much then, right? Especially not if we are not going to release it soon. By the time we would release it we will probably be compatible with 3.13 anyway and we would have to remove it again. So far we have never imposed an explicit limit and it has never caused issues. I am also not aware that this is recommended as a best practice by the PyPA.

[danielhollas]

Yeah, I think this is actually discouraged, to the point where for example uv ignores maximum python version limits when installing dependencies (I think).

[agoscinski]

Yeah, I think this is actually discouraged, to the point where for example uv ignores maximum python version limits when installing dependencies (I think).

Read the same https://github.com/astral-sh/uv/issues/8374#issuecomment-2424246009

There is a huge discussion about it, but I don't see any result comig from it https://discuss.python.org/t/requires-python-upper-limits/12663/80 The metadata field is used by some like pyreads https://pyreadiness.org/3.13/ but the problem that you install Python 3.13 because it is the newest and then AiiDA crashes remains to me. I don't see a simple solution for this.

[agoscinski]

Since we agreed for the release to not include it, this can be closed. If this comes up again as a general discussion one can open an issue.

[danielhollas]

The main issue is, even if you did include it in the release it would be kind of pointless, since the resolver would simply backtrack to the previous version that did not include the constraint. :-/

Dne st 6. 11. 2024 6:38 uživatel Alexander Goscinski < @.***> napsal:

Since we agreed for the release to not include it, this can be closed. If this comes up again as a general discussion one can open an issue.

— Reply to this email directly, view it on GitHub https://github.com/aiidateam/aiida-core/pull/6603#issuecomment-2458827935, or unsubscribe https://github.com/notifications/unsubscribe-auth/ACIY64PMORNW2GL22ROR5JLZ7G2PLAVCNFSM6AAAAABREU3V42VHI2DSMVQWIX3LMV43OSLTON2WKQ3PNVWWK3TUHMZDINJYHAZDOOJTGU . You are receiving this because you commented.Message ID: @.***>

[sphuber]

The main issue is, even if you did include it in the release it would be kind of pointless, since the resolver would simply backtrack to the previous version that did not include the constraint. :-/

:point_up: That is exactly it.

[sphuber]

Another interesting and well laid out point of view just for reference: https://iscinumpy.dev/post/bound-version-constraints/