gh-138432: zoneinfo: improve error message for PathLike relative paths

[tungol]

This corrects a small inconsistency with the error handling of zoneinfo.reset_tzpath. A sequence of os.PathLike instances is documented as valid input, but the error pathway for relative paths doesn't handle these correctly. This pull request corrects this, and adds a relevant test case.

• Issue: zoneinfo.reset_tzpath raises TypeError instead of ValueError if passed a sequence with a relative Pathlike #138432

[StanFromIreland]

Please add a news entry.

[tungol]

Done.

[serhiy-storchaka]

I suggest to call os.fspath() for all items before assigning them to TZPATH (and before checking that they are absolute paths). We can also add a check that they all are now strings (bytes object will cause error later in find_tzfile() and available_timezones()).

[tungol]

I suggest to call os.fspath() for all items before assigning them to TZPATH (and before checking that they are absolute paths).

We can do that, but I don't think it's really necessary. Checking for absolute status is done with os.path.isabs, which already handles os.PathLike objects correctly. Within our code, TZPATH is consumed by find_tzfile, which uses os.path.join to produce a string, so it's again automatically handled.

I found this issue while working on a program to apply typeshed stubs to the stdlib and then analyze with mypy, so I'm relatively confident that this is the only place in (python-source) stdlib where os.Pathlike values for TZPATH were overlooked. (Typeshed correctly annotates TZPATH as tuple[str | PathLike[str], ...])

This leaves the possibility of third-party consumers of TZPATH being surprised by encountering an os.PathLike object instead of a string. If we want to change TZPATH to be restricted to str only on that basis, that's probably reasonable, but I think that's probably best left for a different MR.

[serhiy-storchaka]

It will help to check that if it is a path-like object, it is PathLike[str], not PathLike[Any]. bytes and PathLike[bytes] in TZPATH will cause runtime error in os.path.join().

Anyway, if os.fspath() is already implicitly called in os.path.isabs() and os.path.join(), why not call it explicitly and save the result?

[tungol]

@serhiy-storchaka Is this about what you were thinking?

[serhiy-storchaka]

Thank you for your response @tungol. Yes, this is what I suggested.

Few more nitpicks.

[serhiy-storchaka]

LGTM. 👍

Thank you for your contribution.

[miss-islington-app]

Thanks @tungol for the PR, and @serhiy-storchaka for merging it 🌮🎉.. I'm working now to backport this PR to: 3.13.
🐍🍒⛏🤖

[miss-islington-app]

Thanks @tungol for the PR, and @serhiy-storchaka for merging it 🌮🎉.. I'm working now to backport this PR to: 3.14.
🐍🍒⛏🤖

[miss-islington-app]

Sorry, @tungol and @serhiy-storchaka, I could not cleanly backport this to 3.13 due to a conflict.
Please backport using cherry_picker on command line.

cherry_picker 859aecc33b82f45e5b7ae30138d28f2a2f33a575 3.13

[bedevere-app]

GH-138777 is a backport of this pull request to the 3.14 branch.

[bedevere-app]

GH-138778 is a backport of this pull request to the 3.13 branch.