test_precision errors if LANG is set to de_DE.UTF-8

[olebole]

Description

With my native locale, test_precision.py produces an error that does not appear when LANG=C. This seems to be a problem in astropy itself (not in the test).

Expected behavior

The tests should run independently of a certain locale.

Actual behavior

___________ ERROR collecting time/tests/test_precision.py ________________________
/usr/lib/python3/dist-packages/astropy/time/tests/test_precision.py:92: in <module>
    leap_second_table = iers.LeapSeconds.from_iers_leap_seconds()
/usr/lib/python3/dist-packages/astropy/utils/iers/iers.py:1061: in from_iers_leap_seconds
    return cls._read_leap_seconds(
/usr/lib/python3/dist-packages/astropy/utils/iers/iers.py:1034: in _read_leap_seconds
    expires = Time.strptime(match.groups()[0], '%d %B %Y',
/usr/lib/python3/dist-packages/astropy/time/core.py:1643: in strptime
    tt, fraction = _strptime._strptime(to_string(time), format_string)
/usr/lib/python3/dist-packages/astropy/extern/_strptime.py:350: in _strptime
    raise ValueError("time data %r does not match format %r" %
E   ValueError: time data '28 June 2021' does not match format '%d %B %Y'

Steps to Reproduce

Install python3-astropy from current Debian testing and run

LANG=de_DE.UTF-8 python3 -m pytest --pyargs astropy

System Details

• Linux-5.10.0-2-amd64-x86_64-with-glibc2.31
• Python 3.9.1+ (default, Feb 5 2021, 13:46:56) [GCC 10.2.1 20210110]
• Numpy 1.19.4
• astropy 4.2
• Scipy 1.6.0
• Matplotlib 3.3.4

[saimn]

I think this was fixed by #11062, can you check with master ?

[olebole]

I can't check this for the moment, since when building the package, I always run in the pyerfa problem (#11305). However, I am wondering why this is not in 4.2? IMO it should be backported.

[pllim]

why this is not in 4.2?

What isn't? If you mean #11062, it will be when the next 4.0.x and 4.2.x are released.

[olebole]

@pllim

If you mean #11062, it will be when the next 4.0.x and 4.2.x are released.

Ah, I didn't know that. How does one find that out?

[pllim]

@olebole , its milestone is set for 4.0.5, which isn't out yet. Although given the current maintainer availability, there is unlikely to be any release before 4.3. Usually a major release is also accompanied by bug fix releases of older branches.

[mhvk]

Indeed, I think this particular problem is fixed - the other will hopefully be fixed this weekend (and would also need backporting...).

[mhvk]

I'll close this now - please do re-open if you find the problem also on master!