BUG: pandas/tests/arrays/floating/test_construction.py:test_floating_array_disallows_float16 fails on de_CH locale

[burnpanck]

Pandas version checks

• I have checked that this issue has not already been reported.

• I have confirmed this bug exists on the latest version of pandas.

• I have confirmed this bug exists on the master branch of pandas.

Reproducible Example

pytest pandas/tests/arrays/floating/test_construction.py

Issue Description

Running the current test-suite with the current locale set to de_CH fails due to the that specific test marking itself as xfail unless the locale is en_US. However, the test actually passes.

Expected Behavior

The full test-suite passes under any system locale - specifically also de_CH.

Installed Versions

INSTALLED VERSIONS ------------------ commit : 3a4821e python : 3.9.8.final.0 python-bits : 64 OS : Darwin OS-release : 21.1.0 Version : Darwin Kernel Version 21.1.0: Wed Oct 13 17:33:01 PDT 2021; root:xnu-8019.41.5~1/RELEASE_ARM64_T6000 machine : arm64 processor : arm byteorder : little LC_ALL : de_CH.UTF-8 LANG : en_GB.UTF-8 LOCALE : de_CH.UTF-8

pandas : 1.4.0.dev0+1446.g3a4821e055
numpy : 1.20.3
pytz : 2021.3
dateutil : 2.8.2
pip : 21.3.1
setuptools : 58.1.0
Cython : 0.29.26
pytest : 6.2.5
hypothesis : 6.31.6
sphinx : None
blosc : 1.10.6
feather : None
xlsxwriter : None
lxml.etree : 4.7.1
html5lib : None
pymysql : None
psycopg2 : None
jinja2 : None
IPython : 7.30.1
pandas_datareader: None
bs4 : 4.10.0
bottleneck : None
fsspec : 2021.11.1
fastparquet : 0.7.2
gcsfs : None
matplotlib : None
numexpr : 2.8.1
odfpy : None
openpyxl : 3.0.9
pandas_gbq : None
pyarrow : 6.0.1
pyxlsb : None
s3fs : 2021.11.1
scipy : 1.7.3
sqlalchemy : 1.4.28
tables : 3.6.1
tabulate : None
xarray : 0.20.2
xlrd : None
xlwt : None
numba : None

[burnpanck]

The issue probably started with merging PR #44715 - there is a bit of discussion on this topic over there.

[mroeschke]

Unfortunately we are unable to account for all locals in the test suite since it's optimize to pass the CI so closing as wont fix.

[burnpanck]

Of course it's not feasible to write accurate tests for all locales, but that doesn't mean the test have to xfail! Why not just allow failures if the locale is not en_US? Basically, the way the tests are setup here does actually say "we know the outcome of this test for every locale; success for en_US, and failure for any other". That is directly against what you say, namely that you don't intend to make claims about other locales. Having tests fail wrongly on contributor's machines just discourages us from testing our contributions in the first place.
Now, the issue here is pretty old, and I'm not currently contributing on any issue, so it may not be relevant anymore, but IMHO, completely ignoring any other locale than the CI-blessed one is not good practice - especially, because this issue here is likely easily fixable in the sense of just disabling locale-dependent tests rather than xfailing them.