Conversation
bfb58ac to
0ac9dba
Compare
|
@abravalheri Any reservations before I proceed with merging? |
|
Well that's annoying. The test is failing with: But the test failure report doesn't indicate where the error exists and instead shows a traceback about how it came to report the failure (unhelpful, pytest-ruff). The test doesn't fail if I run ruff in standalone mode. I tested on main and the failure isn't here, so I guess I'll bisect the diff and see if I can figure out where the failure is. |
|
Oh, I was wrong - the ruff error is visible in the traceback, and the issue is in line 617. |
Removed "ugh!" comment because it's not at all clear if it's even relevant, and even if it was, it's not actionable.
|
Please update the documentation at https://setuptools.pypa.io/en/stable/deprecated/distutils/apiref.html. |
…/setuptools#4872 commit_hash:289a58f64149a1262c18bbaaf0317085890490c3
…ypa/setuptools#4872 commit_hash:8b8b245ace1fb5e5b188f42ba289071f0ea0e5c2
…ypa/setuptools#4872 commit_hash:8b8b245ace1fb5e5b188f42ba289071f0ea0e5c2
Thanks for catching this. Docs updated in pypa/distutils@a1c0ebc14, which will merge here in due course. |
|
@mdavidsaver tags don't have to be linear. In fact, they're arbitrary strings, they don't even have to represent a version number. It's a backport so it's normal that a lower number (patch to a previous major version) comes after a bigger number (actual latest). Also read: #5142 (comment) |
|
I completely understand the "can" aspect of git tags. I was asking about the "should" aspect. I have expressed my surprise on #5142 (comment) . |
|
Reporting as requested in the release notes: this change broke gobject-introspection. |
I was gonna suggest leaving in an unused parameter, but thinking about it more, maybe this is a rare case where a loud error is better. Script doesn't work vs script that used to dry run now actually makes changes on disk. |
Remove dry_run from our code and work around lack of pkg_resources Related to pypa/setuptools#4872 fixes #34
No, it's not. We use setuptools 69.0.3. We specify the exact version to use in the requirements file. Now this breaks our testing because I don't know how this change is affecting v69 but it does somehow. Brilliant! |
|
|
@jaraco FYI |
|
It looks like |
You can restict setuptools version used during the build using the build constraints: It can also be included in the project's own BTW, there's a PR in pytest-dependency to fix this issue |
-> used in setup.py ref.: pypa/setuptools#4872
Dry run functionality has been removed from setuptools v81.0.0+. pypa/setuptools#4872 pypa/distutils#335

Summary of changes
At first, I considered deprecating this functionality. Then I realized that this functionality is exposed almost exclusively through a parameter to
setup.py, which is already deprecated. Moreover, it seems unlikely that anyone would use this functionality in today's modern usage. I personally haven't used it even going back to the most legacy usage.Removing these changes from Setuptools should unblock the failing tests in pypa/distutils#335, and in fact, maybe they should be merged together for easier rollback if needed. [Update] The changes from pypa/distutils#335 are included in this merge.
Ref pypa/distutils#334
Pull Request Checklist
newsfragments/.(See documentation for details)