ValueError: SHA256 hash of downloaded file (ascent.dat) does not match the known hash

[pllim]

Hello. We started getting this error when trying to use this file for a test downstream. Is this expected? How do I make it accept the new SHA? Thanks!

Example log: https://github.com/astropy/astropy/actions/runs/3735545217/jobs/6338926910

ValueError: SHA256 hash of downloaded file (ascent.dat) does not match the known hash: ...
Deleted download for safety.
The downloaded file may have been corrupted or the known hash may be outdated.

        else:
>           scipy_data = scipy.datasets.ascent()

astropy/io/fits/tests/test_image.py:1202: 
_ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ 
scipy/datasets/_fetchers.py:71: in ascent
    fname = fetch_data("ascent.dat")
scipy/datasets/_fetchers.py:31: in fetch_data
    return data_fetcher.fetch(dataset_name)
pooch/core.py:587: in fetch
    stream_download(
pooch/core.py:773: in stream_download
    hash_matches(tmp, known_hash, strict=True, source=str(fname.name))

[tupui]

cc @AnirudhDagar

[AnirudhDagar]

Hi @pllim, thanks for the report and sorry for the late reply. I'm on vacation at the moment (typing from my mobile). Can I take a look at this by EOD if not very urgent?

QQ: Are you using the latest SciPy from main branch, or is it the 1.10rc1 in the CI? This should go away with 1.10rc2 and the latest main branch.

Thanks for the ping @tupui!

[pllim]

Hello and thanks for reply. I ended up just directly downloading and unpicking the file myself. We only need it for a single test using the data, not really using scipy functionality.

If no one else has encountered this error, then feel free to close. Happy holidays!

[AnirudhDagar]

As i said it should not happen with 1.10 rc2 release or main branch, but, let's leave this open for a couple more weeks, just in case someone faces the same.

Ideally, we will never change the data files (their hash) without some kind of deprecation. Since this scipy.datasets sub-module was in dev (waiting to be released in 1.10), we changed ascent data file without raising any warnings, which probably broke AstroPy.

Thanks again!