Path.relative_to() walk_up support

[cjntaylor]

Adding support for the walk_up keyword argument introduced in Path.relative_to() in python 3.12

[agronholm]

What's all the socket stuff, and why did you reformat .pre-commit-config.yaml?

[cjntaylor]

Apologies, I realized I messed up my change to split the socket test modifications out into a separate commit. That's now fixed, so you can see just the changes that were made to that test file

[agronholm]

However, it did make the formatting more consistent across file and closer to "standard" YAML formatting. If you prefer, I can revert these changes - just let me know.

Could you elaborate on where this "standard" YAML formatting comes from? As far as I can tell, my formatting is consistent with the YAML 1.2 standard.

[uSpike]

Hi, I just wanted to add that I think it'd be good to open a separate issue/PR for changes unrelated to the title of this PR. I think the original changes are great and I think it's a lot less cognitive load on reviewers and agronholm to review when done separately.

Cheers!

[cjntaylor]

Hi, I just wanted to add that I think it'd be good to open a separate issue/PR for changes unrelated to the title of this PR

The only reason I haven't is because they're indirectly related; the full test suite (which includes new tests relevant to this PR that were specifically requested) won't pass without the other test changes. At least locally. For some inexplicable reason it doesn't break on the macOS CI instances, so maybe that's good enough?

I'm happy to do so though. I'll have to wrap my head around making this PR contingent on the PR you're requesting be added (it'll need to be merged first), but, other than that, I think it's fine

Sorry for the confusion. From the beginning I've been indicating that I was unsure of how to handle the other changes w.r.t. this PR so I'm glad someone has weighed in.

Thanks!

[agronholm]

The changes to .pre-commit-config.yaml are largely unncessary. There seems to be a mypy bug that causes it to behave differently when only a portion of the files are being scanned, as opposed to the whole code base. When that happens, I check with pre-commit run -a to see if it's a real problem, and if it doesn't show up in that scan, I just disable the git hooks for that commit. The workaround would be to force mypy to always check all the files with the following config:

        args: [src, tests]
        pass_filenames: false

[cjntaylor]

@uSpike - PR's have been split into #659, #660 and #661, with 659's pre-commit changes determined as irrelevant and removed.

@agronholm - Thanks for your patience in all of this, and apologies for over-complicating things. I've completely isolated the changes into separate PRs so you can consider them independently, and made sure that each one is as minimal a change as possible.

Thank you for the workaround. I've closed #659 accordingly.

Please note that this PR will now fail to run the full test suite on macOS >= 13. Running test_fileio.py in isolation appears to work correctly, and the CI does pass

[agronholm]

LGTM, thanks!