"27 years from now" tests fail on platforms with 32-bit `time_t` (y2k38 problem)

[mgorny]

What did you do?

tox -e py311 on a 32-bit x86 system.

What did you expect to happen?

Tests passing (possibly by skipping these tests on OverflowError).

What actually happened?

.pkg: install_requires> python -I -m pip install hatch-vcs hatchling
.pkg: _optional_hooks> python /usr/lib/python3.11/site-packages/pyproject_api/_backend.py True hatchling.build
.pkg: get_requires_for_build_sdist> python /usr/lib/python3.11/site-packages/pyproject_api/_backend.py True hatchling.build
.pkg: build_sdist> python /usr/lib/python3.11/site-packages/pyproject_api/_backend.py True hatchling.build
py311: install_package_deps> python -I -m pip install freezegun pytest pytest-cov
py311: install_package> python -I -m pip install --force-reinstall --no-deps /home/mgorny/humanize/.tox/.tmp/package/1/humanize-4.9.1.dev38.tar.gz
py311: commands[0]> .tox/py311/bin/python -m pytest --cov humanize --cov tests --cov-report html --cov-report term --cov-report xml
========================================================= test session starts =========================================================
platform linux -- Python 3.11.7, pytest-8.0.2, pluggy-1.4.0
cachedir: .tox/py311/.pytest_cache
rootdir: /home/mgorny/humanize
configfile: pyproject.toml
plugins: cov-4.1.0
collected 572 items                                                                                                                   

tests/test_filesize.py ....................                                                                                     [  3%]
tests/test_i18n.py sssssssssssssssssssssss.....                                                                                 [  8%]
tests/test_number.py .......................................................................................................... [ 26%]
..........................................................................................................                      [ 45%]
tests/test_time.py ............................................................................................................ [ 64%]
............................................................................................................................... [ 86%]
..F......................F...................................................                                                   [100%]

============================================================== FAILURES ===============================================================
______________________________________ test_naturaltime_timezone[test_input19-27 years from now] ______________________________________

test_input = FakeDatetime(2047, 6, 20, 0, 0, tzinfo=datetime.timezone.utc), expected = '27 years from now'

    @freeze_time("2020-02-02")
    @pytest.mark.parametrize(
        "test_input, expected",
        [
            (NOW_UTC, "now"),
            (NOW_UTC - dt.timedelta(seconds=1), "a second ago"),
            (NOW_UTC - dt.timedelta(seconds=30), "30 seconds ago"),
            (NOW_UTC - dt.timedelta(minutes=1, seconds=30), "a minute ago"),
            (NOW_UTC - dt.timedelta(minutes=2), "2 minutes ago"),
            (NOW_UTC - dt.timedelta(hours=1, minutes=30, seconds=30), "an hour ago"),
            (NOW_UTC - dt.timedelta(hours=23, minutes=50, seconds=50), "23 hours ago"),
            (NOW_UTC - dt.timedelta(days=1), "a day ago"),
            (NOW_UTC - dt.timedelta(days=500), "1 year, 4 months ago"),
            (NOW_UTC - dt.timedelta(days=365 * 2 + 35), "2 years ago"),
            (NOW_UTC + dt.timedelta(seconds=1), "a second from now"),
            (NOW_UTC + dt.timedelta(seconds=30), "30 seconds from now"),
            (NOW_UTC + dt.timedelta(minutes=1, seconds=30), "a minute from now"),
            (NOW_UTC + dt.timedelta(minutes=2), "2 minutes from now"),
            (NOW_UTC + dt.timedelta(hours=1, minutes=30, seconds=30), "an hour from now"),
            (NOW_UTC + dt.timedelta(hours=23, minutes=50, seconds=50), "23 hours from now"),
            (NOW_UTC + dt.timedelta(days=1), "a day from now"),
            (NOW_UTC + dt.timedelta(days=500), "1 year, 4 months from now"),
            (NOW_UTC + dt.timedelta(days=365 * 2 + 35), "2 years from now"),
            # regression tests for bugs in post-release humanize
            (NOW_UTC + dt.timedelta(days=10000), "27 years from now"),
            (NOW_UTC - dt.timedelta(days=365 + 35), "1 year, 1 month ago"),
            (NOW_UTC - dt.timedelta(days=365 * 2 + 65), "2 years ago"),
            (NOW_UTC - dt.timedelta(days=365 + 4), "1 year, 4 days ago"),
        ],
    )
    def test_naturaltime_timezone(test_input: dt.datetime, expected: str) -> None:
>       assert humanize.naturaltime(test_input) == expected

tests/test_time.py:449: 
_ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _
.tox/py311/lib/python3.11/site-packages/humanize/time.py:249: in naturaltime
    value = _convert_aware_datetime(value)
.tox/py311/lib/python3.11/site-packages/humanize/time.py:275: in _convert_aware_datetime
    value = dt.datetime.fromtimestamp(value.timestamp())
_ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _

cls = <class 'freezegun.api.FakeDatetime'>, t = 2444601600.0, tz = None

    @classmethod
    def fromtimestamp(cls, t, tz=None):
        if tz is None:
>           return real_datetime.fromtimestamp(
                    t, tz=dateutil.tz.tzoffset("freezegun", cls._tz_offset())
                ).replace(tzinfo=None)
E           OverflowError: timestamp out of range for platform time_t

.tox/py311/lib/python3.11/site-packages/freezegun/api.py:375: OverflowError
___________________________________ test_naturaltime_timezone_when[test_input19-27 years from now] ____________________________________

test_input = FakeDatetime(2047, 6, 20, 0, 0, tzinfo=datetime.timezone.utc), expected = '27 years from now'

    @freeze_time("2020-02-02")
    @pytest.mark.parametrize(
        "test_input, expected",
        [
            (NOW_UTC, "now"),
            (NOW_UTC - dt.timedelta(seconds=1), "a second ago"),
            (NOW_UTC - dt.timedelta(seconds=30), "30 seconds ago"),
            (NOW_UTC - dt.timedelta(minutes=1, seconds=30), "a minute ago"),
            (NOW_UTC - dt.timedelta(minutes=2), "2 minutes ago"),
            (NOW_UTC - dt.timedelta(hours=1, minutes=30, seconds=30), "an hour ago"),
            (NOW_UTC - dt.timedelta(hours=23, minutes=50, seconds=50), "23 hours ago"),
            (NOW_UTC - dt.timedelta(days=1), "a day ago"),
            (NOW_UTC - dt.timedelta(days=500), "1 year, 4 months ago"),
            (NOW_UTC - dt.timedelta(days=365 * 2 + 35), "2 years ago"),
            (NOW_UTC + dt.timedelta(seconds=1), "a second from now"),
            (NOW_UTC + dt.timedelta(seconds=30), "30 seconds from now"),
            (NOW_UTC + dt.timedelta(minutes=1, seconds=30), "a minute from now"),
            (NOW_UTC + dt.timedelta(minutes=2), "2 minutes from now"),
            (NOW_UTC + dt.timedelta(hours=1, minutes=30, seconds=30), "an hour from now"),
            (NOW_UTC + dt.timedelta(hours=23, minutes=50, seconds=50), "23 hours from now"),
            (NOW_UTC + dt.timedelta(days=1), "a day from now"),
            (NOW_UTC + dt.timedelta(days=500), "1 year, 4 months from now"),
            (NOW_UTC + dt.timedelta(days=365 * 2 + 35), "2 years from now"),
            # regression tests for bugs in post-release humanize
            (NOW_UTC + dt.timedelta(days=10000), "27 years from now"),
            (NOW_UTC - dt.timedelta(days=365 + 35), "1 year, 1 month ago"),
            (NOW_UTC - dt.timedelta(days=365 * 2 + 65), "2 years ago"),
            (NOW_UTC - dt.timedelta(days=365 + 4), "1 year, 4 days ago"),
        ],
    )
    def test_naturaltime_timezone_when(test_input: dt.datetime, expected: str) -> None:
>       assert humanize.naturaltime(test_input, when=NOW_UTC) == expected

tests/test_time.py:483: 
_ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _
.tox/py311/lib/python3.11/site-packages/humanize/time.py:249: in naturaltime
    value = _convert_aware_datetime(value)
.tox/py311/lib/python3.11/site-packages/humanize/time.py:275: in _convert_aware_datetime
    value = dt.datetime.fromtimestamp(value.timestamp())
_ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _

cls = <class 'freezegun.api.FakeDatetime'>, t = 2444601600.0, tz = None

    @classmethod
    def fromtimestamp(cls, t, tz=None):
        if tz is None:
>           return real_datetime.fromtimestamp(
                    t, tz=dateutil.tz.tzoffset("freezegun", cls._tz_offset())
                ).replace(tzinfo=None)
E           OverflowError: timestamp out of range for platform time_t

.tox/py311/lib/python3.11/site-packages/freezegun/api.py:375: OverflowError

---------- coverage: platform linux, python 3.11.7-final-0 -----------
Name                                                           Stmts   Miss  Cover
----------------------------------------------------------------------------------
.tox/py311/lib/python3.11/site-packages/humanize/__init__.py       8      0   100%
.tox/py311/lib/python3.11/site-packages/humanize/filesize.py      24      0   100%
.tox/py311/lib/python3.11/site-packages/humanize/i18n.py          56      4    93%
.tox/py311/lib/python3.11/site-packages/humanize/number.py       165      7    96%
.tox/py311/lib/python3.11/site-packages/humanize/time.py         224      0   100%
tests/__init__.py                                                  0      0   100%
tests/test_filesize.py                                            10      0   100%
tests/test_i18n.py                                               106     19    82%
tests/test_number.py                                              32      0   100%
tests/test_time.py                                               124      0   100%
----------------------------------------------------------------------------------
TOTAL                                                            749     30    96%
Coverage HTML written to dir htmlcov
Coverage XML written to file coverage.xml

======================================================= short test summary info =======================================================
FAILED tests/test_time.py::test_naturaltime_timezone[test_input19-27 years from now] - OverflowError: timestamp out of range for platform time_t
FAILED tests/test_time.py::test_naturaltime_timezone_when[test_input19-27 years from now] - OverflowError: timestamp out of range for platform time_t
============================================== 2 failed, 547 passed, 23 skipped in 4.97s ==============================================
py311: exit 1 (5.42 seconds) /home/mgorny/humanize> .tox/py311/bin/python -m pytest --cov humanize --cov tests --cov-report html --cov-report term --cov-report xml pid=37634
.pkg: _exit> python /usr/lib/python3.11/site-packages/pyproject_api/_backend.py True hatchling.build
  py311: FAIL code 1 (21.29=setup[15.87]+cmd[5.42] seconds)
  evaluation failed :( (21.38 seconds)

What versions are you using?

• OS: Gentoo Linux x86
• Python: 3.11.7
• Humanize: 218a86e

[hugovk]

Thanks for the report!

.tox/py311/lib/python3.11/site-packages/freezegun/api.py:375: OverflowError

Looks like the exception is coming from the https://github.com/spulec/freezegun test library, so might be worth reporting something there.

Anyway, skipping sounds reasonable in this case, would you like to create a PR? I don't have a setup to test it.

[mgorny]

Thanks for the report!

.tox/py311/lib/python3.11/site-packages/freezegun/api.py:375: OverflowError

Looks like the exception is coming from the https://github.com/spulec/freezegun test library, so might be worth reporting something there.

It's basically coming from Python, when given a date that's outside time_t range for the system. I don't really see what freezegun could do there.

Anyway, skipping sounds reasonable in this case, would you like to create a PR? I don't have a setup to test it.

Sure, I will file one in a minute.

[mgorny]

Filed #177. In the end, I've figured out it's cleaner to change the base date not to trigger overflows in the first place.