[PULP-807] Add changes required for Django 5

[pedro-psb]

Compatibility findings

• django-guid adds support to django 5:
  • https://github.com/snok/django-guid/releases/tag/3.4.0
• import string new location:
  • https://docs.djangoproject.com/en/5.2/ref/utils/#django.utils.module_loading.import_string
  • previously imported from storages
• shell auto-import new behavior:
  • https://docs.djangoproject.com/en/5.2/howto/custom-shell/
  • Now a pulpcore-manaer shell call auto-imports models and other stuff (like shell_plus). The problem is that it prints a message to stdout which breaks our trick of using it to get values from the database in the tests.
  • An alternative is to globally disable, which is shown in the link.
• django-filter added oficial suport for django 5.2 in their latest release (25.1):
  • https://github.com/carltongibson/django-filter/blob/main/CHANGES.rst#version-251-2025-02-14
  • lowerbound were 23.2, which doesnt work with django 5.2.
  • did a simple test with 24.2 and it passes a small subset of the previously failing tests. Nevertheless it doesnt oficially suport 5.2

[mdellweg]

shell auto-import new behavior:

* https://docs.djangoproject.com/en/5.2/howto/custom-shell/

* Now a `pulpcore-manaer shell` call auto-imports models and other stuff (like shell_plus). The problem is that it prints a message to stdout which breaks our trick of using it to get values from the database in the tests.

* An alternative is to globally disable, which is shown in the link.

We may be using that "trick" in more places... Can't they make the imports silent?

[pedro-psb]

The docs show how to disable it globally. I'll see how that goes

[mdellweg]

Please don't change the version requirements as if they were following semver. They are not...

[mdellweg]

"Other breaking changes are rare." What I hear is: "We make no guarantees about changes tied to specific types of version changes." But also "You can probably expect 1 or 2 releases per year." (not like we need to touch this line every other week.)

I would argue we are better off to keep the "<=" upper bound.

Having said that, i think what we agreed to try now is: Not touching any LB here, and still allowing to up the django version UB to 5.2.

And now a very wild idea: Given that these changes are really minimal, how do you feel about backporting this version bump?

[pedro-psb]

Having said that, i think what we agreed to try now is: Not touching any LB here, and still allowing to up the django version UB to 5.2.

I see, that was not clear to me, but it sounds fine.

And now a very wild idea: Given that these changes are really minimal, how do you feel about backporting this version bump?

We can try and see what happens. You mean all of it? The django 5 upperbound, the shell command override, and the other bumps?

[mdellweg]

Basically if we can make upgrading Django less of a hassle for everybody, we all win...
For that I'd say this PR should only bump upper bounds of requirements (Probably really only Django) and contain this other two small adjustments you made (that seem to be compatible with Django4 anyway.)

And then we can think about backporting it.

[mdellweg]

That's perfect. Thank you!
It reminds me that we still haven't looked up all dependencies versioning philosophies.

[pedro-psb]

Lowerbound fails under its set of dependencies. I'm trying to isolate what exactly locally.
And just a note, probably its known already: our unit tests are a bit fragile in terms of leaking. If we use pytest-random-order we'll get different set of failures at each run.

[pedro-psb]

That was intriguing... Thanks for the suggestion!