GH-45295: [Python][CI] Make download_tzdata_on_windows more robust and use tzdata package for tzinfo database on Windows for ORC

[amoeba]

Rationale for this change

We have two Windows issues and this PR is addressing both:

1. PyArrow's download_tzdata_on_windows can fail due to TLS issues in certain CI environments.
2. The Python wheel test infrastructure needs a tzinfo database for ORC and the automation fetching that started failing because the URL was made invalid upstream.

These two issues are being solved in one PR simply because they appeared together during the 19.0.1 release process but they're separate.

What changes are included in this PR?

1. Makes download_tzdata_on_windows more robust to TLS errors by attempting to use requests if it's available and falling back to urllib otherwise.
2. Switches our Windows wheel test infrastructure to grab a tzinfo database from the tzdata package on PyPi instead of from a mirror URL. This should be much more stable for us over time.

Are these changes tested?

Yes.

Are there any user-facing changes?

No.

• GitHub Issue: [Python][CI] test_download_tzdata_on_windows fails on Windows wheels due to CERTIFICATE_VERIFY_FAILED #45295

[github-actions]

⚠️ GitHub issue #45295 has been automatically assigned in GitHub to PR creator.

[amoeba]

@github-actions crossbow submit wheel-windows-cp39-amd64

[github-actions]

Unable to match any tasks for `wheel-windows-cp39-amd64`
The Archery job run can be found at: https://github.com/apache/arrow/actions/runs/13149682761

[amoeba]

@kou is there a way to test this PR?

[kou]

@github-actions crossbow submit wheel-windows-cp39-amd64

[github-actions]

Unable to match any tasks for `wheel-windows-cp39-amd64`
The Archery job run can be found at: https://github.com/apache/arrow/actions/runs/13149993032

[kou]

@github-actions crossbow submit wheel-windows-*

[github-actions]

Revision: 29467e2

Submitted crossbow builds: ursacomputing/crossbow @ actions-435df0c2a2

Task	Status
wheel-windows-cp310-cp310-amd64
wheel-windows-cp311-cp311-amd64
wheel-windows-cp312-cp312-amd64
wheel-windows-cp313-cp313-amd64
wheel-windows-cp313-cp313t-amd64
wheel-windows-cp39-cp39-amd64

[amoeba]

Thanks for getting the jobs running, I'll check on them tomorrow.

[raulcd]

This seems to be working and is back to the original issue failing on a specific test on Arrow to download tzdata:

    def test_download_tzdata_on_windows():
        tzdata_path = os.path.expandvars(r"%USERPROFILE%\Downloads\tzdata")
    
        # Download timezone database and remove data in case it already exists
        if (os.path.exists(tzdata_path)):
            shutil.rmtree(tzdata_path)
>       download_tzdata_on_windows()
...
with urlopen('https://data.iana.org/time-zones/tzdata-latest.tar.gz') as response:

Is the function even necessary?
I understand is a util to download tzdata but I am not sure we want to provide that utility if users can just use importlib.resources to get the tzdata one.
Should we just remove the utility function?
@AlenkaF @jorisvandenbossche thoughts?

[github-actions]

Revision: 1d715f9

Submitted crossbow builds: ursacomputing/crossbow @ actions-4b3b781c2e

Task	Status
wheel-windows-cp310-cp310-amd64
wheel-windows-cp311-cp311-amd64
wheel-windows-cp312-cp312-amd64
wheel-windows-cp313-cp313-amd64
wheel-windows-cp313-cp313t-amd64
wheel-windows-cp39-cp39-amd64

[amoeba]

@github-actions crossbow submit wheel-windows-*

[github-actions]

Revision: 84ab16e

Submitted crossbow builds: ursacomputing/crossbow @ actions-715daf418e

Task	Status
wheel-windows-cp310-cp310-amd64
wheel-windows-cp311-cp311-amd64
wheel-windows-cp312-cp312-amd64
wheel-windows-cp313-cp313-amd64
wheel-windows-cp313-cp313t-amd64
wheel-windows-cp39-cp39-amd64

[raulcd]

Thanks @kou and @amoeba ! This is a great find! The failure on appveyor is related and it's related to the concern I have with this change. I think we should fall back to use urllib in case of ImportError

[amoeba]

@github-actions crossbow submit wheel-windows-*

[github-actions]

Revision: 459376c

Submitted crossbow builds: ursacomputing/crossbow @ actions-b4f931ecfe

Task	Status
wheel-windows-cp310-cp310-amd64
wheel-windows-cp311-cp311-amd64
wheel-windows-cp312-cp312-amd64
wheel-windows-cp313-cp313-amd64
wheel-windows-cp313-cp313t-amd64
wheel-windows-cp39-cp39-amd64

[amoeba]

@github-actions crossbow submit wheel-windows-*

[amoeba]

Last round of crossbow job failures look to be due to an issue with GitHub, I'm seeing it on other PRs too. Re-ran jobs and will keep an eye out.

@raulcd whenever you have a moment, could you please look at my recent change to util.py?

[github-actions]

Revision: 459376c

Submitted crossbow builds: ursacomputing/crossbow @ actions-c246a1f639

Task	Status
wheel-windows-cp310-cp310-amd64
wheel-windows-cp311-cp311-amd64
wheel-windows-cp312-cp312-amd64
wheel-windows-cp313-cp313-amd64
wheel-windows-cp313-cp313t-amd64
wheel-windows-cp39-cp39-amd64

[jorisvandenbossche]

@amoeba thanks for looking into this!

Apologies for the slow reply if I could have prevented some wasted time figuring this out, but indeed as you summarize in #45425 (comment), the way we understood it in the past is that we unfortunately cannot use the tzdata as shipped by the Python package. This would require some changes to support that format upstream in https://github.com/HowardHinnant/date.
See #31472 for the issue about it.

@amoeba could you update the title and top comment description of the PR, as I think it no longer reflects the change entirely?

[jorisvandenbossche]

I also see in the wheel build logs the following message (https://github.com/ursacomputing/crossbow/actions/runs/13191501020/job/36825167067#step:9:1122):

SKIPPED [1] Python310\lib\site-packages\pyarrow\tests\test_compute.py:2301: Timezone database is not installed on Windows

so wondering what is going wrong (or are we not downloading the tzdata files in the wheel builds?)

[amoeba]

Thanks for the review @jorisvandenbossche. Good catch on the message in CI. I'm not sure yet but will investigate.