`EncodingWarning` raised in `cacheprovider`

[AA-Turner]

Sphinx recently enabled the EncodingWarning feature of Python 3.10+ in tests.¹²

PyTest reports three EncodingWarnings in _pytest/cacheprovider.py, output below:

.tox/du18/lib/python3.10/site-packages/_pytest/cacheprovider.py:199
  /home/runner/work/sphinx/sphinx/.tox/du18/lib/python3.10/site-packages/_pytest/cacheprovider.py:199: EncodingWarning: 'encoding' argument not specified
    readme_path.write_text(README_CONTENT)

.tox/du18/lib/python3.10/site-packages/_pytest/cacheprovider.py:189
.tox/du18/lib/python3.10/site-packages/_pytest/cacheprovider.py:189
  /home/runner/work/sphinx/sphinx/.tox/du18/lib/python3.10/site-packages/_pytest/cacheprovider.py:189: EncodingWarning: 'encoding' argument not specified
    f = path.open("w")

.tox/du18/lib/python3.10/site-packages/_pytest/cacheprovider.py:160
  /home/runner/work/sphinx/sphinx/.tox/du18/lib/python3.10/site-packages/_pytest/cacheprovider.py:160: EncodingWarning: 'encoding' argument not specified
    with path.open("r") as f:

The remedy is to add an explicit encoding="..." argument, most likely "utf-8".

Reproducer is python -X warn_default_encoding -m pytest ... with a path to a test suite.

Test session preamble:

============================= test session starts ==============================
platform linux -- Python 3.10.4, pytest-7.1.2, pluggy-1.0.0 -- /home/runner/work/sphinx/sphinx/.tox/du18/bin/python
cachedir: .tox/du18/.pytest_cache
libraries: Sphinx-5.0.0+/7970e6c, docutils-0.18.1
base tempdir: /tmp/pytest-of-runner/pytest-0
rootdir: /home/runner/work/sphinx/sphinx, configfile: setup.cfg, testpaths: tests
plugins: cov-3.0.0

A

Footnotes

1. https://docs.python.org/3/library/io.html#io-encoding-warning ↩

2. https://docs.python.org/3/using/cmdline.html?highlight=warn_default_encoding#cmdoption-X ↩

[bluetech]

Thanks for reporting. Indeed the encoding should be utf-8. The README one is definitely UTF-8; the json ones currently don't pass ensure_ascii=False, so they're ASCII, which is a subset of UTF-8. Along the way I'd suggest also passing ensure_ascii=False to the json.dumps call, because why uglify the output if it's not needed?

[peymanslh]

Hi,
Can I work on this issue?

[AA-Turner]

Please! You just need to change the three calls in my orignal report to add an encoding="utf-8" argument, and probably also add ensure_ascii=False as @bluetech noted.

A