⚠️ Nightly upstream-dev CI failed ⚠️

[github-actions]

Workflow Run URL

Python 3.12 Test Summary

xarray/tests/test_backends.py::TestDask::test_save_mfdataset_compute_false_roundtrip: RuntimeError: NetCDF: HDF error
xarray/tests/test_cftimeindex.py::test_distant_cftime_datetime_sub_cftimeindex[365_day]: Failed: DID NOT RAISE <class 'ValueError'>
xarray/tests/test_cftimeindex.py::test_distant_cftime_datetime_sub_cftimeindex[360_day]: Failed: DID NOT RAISE <class 'ValueError'>
xarray/tests/test_cftimeindex.py::test_distant_cftime_datetime_sub_cftimeindex[julian]: Failed: DID NOT RAISE <class 'ValueError'>
xarray/tests/test_cftimeindex.py::test_distant_cftime_datetime_sub_cftimeindex[all_leap]: Failed: DID NOT RAISE <class 'ValueError'>
xarray/tests/test_cftimeindex.py::test_distant_cftime_datetime_sub_cftimeindex[366_day]: Failed: DID NOT RAISE <class 'ValueError'>
xarray/tests/test_cftimeindex.py::test_distant_cftime_datetime_sub_cftimeindex[gregorian]: Failed: DID NOT RAISE <class 'ValueError'>
xarray/tests/test_cftimeindex.py::test_distant_cftime_datetime_sub_cftimeindex[proleptic_gregorian]: Failed: DID NOT RAISE <class 'ValueError'>
xarray/tests/test_coding_times.py::test_cf_timedelta[1D-days-numbers0]: AssertionError: assert dtype('<m8[us]') == dtype('<m8[ns]')
 +  where dtype('<m8[us]') = np.timedelta64(86400000000,'us').dtype
 +  and   dtype('<m8[ns]') = array(86400000000000, dtype='timedelta64[ns]').dtype
xarray/tests/test_coding_times.py::test_cf_timedelta[timedeltas1-days-numbers1]: AssertionError: assert dtype('<m8[us]') == dtype('<m8[ns]')
 +  where dtype('<m8[us]') = array([ 86400000000, 172800000000, 259200000000], dtype='timedelta64[us]').dtype
 +  and   dtype('<m8[ns]') = array([ 86400000000000, 172800000000000, 259200000000000],\n      dtype='timedelta64[ns]').dtype
xarray/tests/test_coding_times.py::test_cf_timedelta[1h-hours-numbers2]: AssertionError: assert dtype('<m8[us]') == dtype('<m8[ns]')
 +  where dtype('<m8[us]') = np.timedelta64(3600000000,'us').dtype
 +  and   dtype('<m8[ns]') = array(3600000000000, dtype='timedelta64[ns]').dtype
xarray/tests/test_coding_times.py::test_cf_timedelta[1ms-milliseconds-numbers3]: AssertionError: assert dtype('<m8[us]') == dtype('<m8[ns]')
 +  where dtype('<m8[us]') = np.timedelta64(1000,'us').dtype
 +  and   dtype('<m8[ns]') = array(1000000, dtype='timedelta64[ns]').dtype
xarray/tests/test_coding_times.py::test_cf_timedelta[1us-microseconds-numbers4]: AssertionError: assert dtype('<m8[us]') == dtype('<m8[ns]')
 +  where dtype('<m8[us]') = np.timedelta64(1,'us').dtype
 +  and   dtype('<m8[ns]') = array(1000, dtype='timedelta64[ns]').dtype
xarray/tests/test_coding_times.py::test_cf_timedelta[timedeltas7-hours-numbers7]: AssertionError: assert dtype('<m8[us]') == dtype('<m8[ns]')
 +  where dtype('<m8[us]') = array([1800000000, 3600000000], dtype='timedelta64[us]').dtype
 +  and   dtype('<m8[ns]') = array([1800000000000, 3600000000000], dtype='timedelta64[ns]').dtype
xarray/tests/test_coding_times.py::test_cf_timedelta[timedeltas9-days-numbers9]: AssertionError: assert dtype('<m8[s]') == dtype('<m8[ns]')
 +  where dtype('<m8[s]') = array(['NaT', 'NaT'], dtype='timedelta64[s]').dtype
 +  and   dtype('<m8[ns]') = array(['NaT', 'NaT'], dtype='timedelta64[ns]').dtype
xarray/tests/test_coding_times.py::test_cf_timedelta_2d: AssertionError: assert dtype('<m8[us]') == dtype('<m8[ns]')
 +  where dtype('<m8[us]') = array([[ 86400000000, 172800000000, 259200000000]],\n      dtype='timedelta64[us]').dtype
 +  and   dtype('<m8[ns]') = array([[ 86400000000000, 172800000000000, 259200000000000]],\n      dtype='timedelta64[ns]').dtype
xarray/tests/test_groupby.py::test_groupby_bins_datetime_mean: AssertionError: assert dtype('<M8[us]') == dtype('<M8[ns]')
 +  where dtype('<M8[us]') = <xarray.DataArray 'time' (index_bins: 148)> Size: 1kB\narray(['2020-01-01T03:00:00.000000', '2020-01-01T08:00:00.000000...     dtype='datetime64[us]')\nCoordinates:\n  * index_bins  (index_bins) interval[int64, right] 2kB (0, 5] ... (735, 740].dtype
 +    where <xarray.DataArray 'time' (index_bins: 148)> Size: 1kB\narray(['2020-01-01T03:00:00.000000', '2020-01-01T08:00:00.000000...     dtype='datetime64[us]')\nCoordinates:\n  * index_bins  (index_bins) interval[int64, right] 2kB (0, 5] ... (735, 740] = <xarray.Dataset> Size: 5kB\nDimensions:     (index_bins: 148)\nCoordinates:\n  * index_bins  (index_bins) interval[int64,...64[us] 1kB 2020-01-01T03:00:00 ... 2020-...\n    float       (index_bins) float64 1kB 0.004032 0.01075 ... 0.9852 0.9919.time
 +  and   dtype('<M8[ns]') = <class 'numpy.dtype'>('datetime64[ns]')
 +    where <class 'numpy.dtype'> = np.dtype
xarray/tests/test_groupby.py::test_groupby_bins_mean_time_series: AssertionError: assert dtype('<M8[us]') == dtype('<M8[ns]')
 +  where dtype('<M8[us]') = <xarray.DataArray 'time' (trial_bins: 5)> Size: 40B\narray(['2024-01-01T15:00:03.333333', '2024-01-01T15:00:16.666666',..., dtype='datetime64[us]')\nCoordinates:\n  * trial_bins  (trial_bins) interval[float64, right] 80B (-0.009, 1.8] ... (....dtype
 +    where <xarray.DataArray 'time' (trial_bins: 5)> Size: 40B\narray(['2024-01-01T15:00:03.333333', '2024-01-01T15:00:16.666666',..., dtype='datetime64[us]')\nCoordinates:\n  * trial_bins  (trial_bins) interval[float64, right] 80B (-0.009, 1.8] ... (... = <xarray.Dataset> Size: 160B\nDimensions:      (trial_bins: 5)\nCoordinates:\n  * trial_bins   (trial_bins) interval[float...s) float64 40B 5.0 25.0 45.0 65.0 85.0\n    time         (trial_bins) datetime64[us] 40B 2024-01-01T15:00:03.333333 .....time
 +  and   dtype('<M8[ns]') = <class 'numpy.dtype'>('datetime64[ns]')
 +    where <class 'numpy.dtype'> = np.dtype
xarray/tests/test_interp.py::test_datetime[x_new2-expected2]: AssertionError: Left and right DataArray objects are not close
Differing values:
L
    array([nan, nan])
R
    array([0.5, 1.5])
xarray/tests/test_interp.py::test_datetime[x_new3-expected3]: AssertionError: Left and right DataArray objects are not close
Differing values:
L
    array([nan, nan, nan])
R
    array([0.5, 1.5, nan])
xarray/tests/test_interp.py::test_datetime[x_new4-0.5]: AssertionError: Left and right DataArray objects are not close
Differing values:
L
    array([nan])
R
    array([0.5])
xarray/tests/test_interp.py::test_datetime_single_string: AssertionError: Left and right DataArray objects are not close
Differing values:
L
    array(nan)
R
    array(0.5)
xarray/tests/test_plot.py::TestDatasetScatterPlots::test_datetime_hue: TypeError: ufunc 'isfinite' not supported for the input types, and the inputs could not be safely coerced to any supported types according to the casting rule ''safe''
xarray/tests/test_variable.py::test_as_compatible_data_writeable[data0]: pandas.errors.Pandas4Warning: Copy-on-Write can no longer be disabled, setting to False has no impact. This option will be removed in pandas 4.0.
xarray/tests/test_variable.py::test_as_compatible_data_writeable[data1]: pandas.errors.Pandas4Warning: Copy-on-Write can no longer be disabled, setting to False has no impact. This option will be removed in pandas 4.0.
xarray/tests/test_variable.py::TestVariable::test_index_0d_timedelta64: AssertionError: assert dtype('<m8[us]') == 'timedelta64[ns]'
 +  where dtype('<m8[us]') = np.timedelta64(3600000000,'us').dtype
xarray/tests/test_variable.py::TestVariable::test_datetime64_conversion[values0-ns]: AssertionError: assert dtype('<M8[us]') == dtype('<M8[ns]')
 +  where dtype('<M8[us]') = <xarray.Variable (t: 3)> Size: 24B\narray(['1970-01-01T00:00:00.000000', '1970-01-02T00:00:00.000000',\n       '1970-01-03T00:00:00.000000'], dtype='datetime64[us]').dtype
 +  and   dtype('<M8[ns]') = <class 'numpy.dtype'>('datetime64[ns]')
 +    where <class 'numpy.dtype'> = np.dtype
xarray/tests/test_variable.py::TestVariable::test_datetime64_conversion[values1-ns]: AssertionError: assert dtype('<M8[us]') == dtype('<M8[ns]')
 +  where dtype('<M8[us]') = <xarray.Variable (t: 3)> Size: 24B\narray(['1970-01-01T00:00:00.000000', '1970-01-02T00:00:00.000000',\n       '1970-01-03T00:00:00.000000'], dtype='datetime64[us]').dtype
 +  and   dtype('<M8[ns]') = <class 'numpy.dtype'>('datetime64[ns]')
 +    where <class 'numpy.dtype'> = np.dtype
xarray/tests/test_variable.py::TestVariable::test_timedelta64_conversion[values5-ns]: AssertionError: assert dtype('<m8[us]') == dtype('<m8[ns]')
 +  where dtype('<m8[us]') = <xarray.Variable (t: 3)> Size: 24B\narray([           0,  86400000000, 172800000000], dtype='timedelta64[us]').dtype
 +  and   dtype('<m8[ns]') = <class 'numpy.dtype'>('timedelta64[ns]')
 +    where <class 'numpy.dtype'> = np.dtype
xarray/tests/test_variable.py::TestVariableWithDask::test_index_0d_timedelta64: AssertionError: assert dtype('<m8[us]') == 'timedelta64[ns]'
 +  where dtype('<m8[us]') = np.timedelta64(3600000000,'us').dtype
xarray/tests/test_variable.py::TestVariableWithDask::test_datetime64_conversion[values0-ns]: AssertionError: assert dtype('<M8[us]') == dtype('<M8[ns]')
 +  where dtype('<M8[us]') = <xarray.Variable (t: 3)> Size: 24B\ndask.array<array, shape=(3,), dtype=datetime64[us], chunksize=(3,), chunktype=numpy.ndarray>.dtype
 +  and   dtype('<M8[ns]') = <class 'numpy.dtype'>('datetime64[ns]')
 +    where <class 'numpy.dtype'> = np.dtype
xarray/tests/test_variable.py::TestVariableWithDask::test_datetime64_conversion[values1-ns]: AssertionError: assert dtype('<M8[us]') == dtype('<M8[ns]')
 +  where dtype('<M8[us]') = <xarray.Variable (t: 3)> Size: 24B\ndask.array<array, shape=(3,), dtype=datetime64[us], chunksize=(3,), chunktype=numpy.ndarray>.dtype
 +  and   dtype('<M8[ns]') = <class 'numpy.dtype'>('datetime64[ns]')
 +    where <class 'numpy.dtype'> = np.dtype
xarray/tests/test_variable.py::TestVariableWithDask::test_timedelta64_conversion[values5-ns]: AssertionError: assert dtype('<m8[us]') == dtype('<m8[ns]')
 +  where dtype('<m8[us]') = <xarray.Variable (t: 3)> Size: 24B\ndask.array<array, shape=(3,), dtype=timedelta64[us], chunksize=(3,), chunktype=numpy.ndarray>.dtype
 +  and   dtype('<m8[ns]') = <class 'numpy.dtype'>('timedelta64[ns]')
 +    where <class 'numpy.dtype'> = np.dtype
xarray/tests/test_variable.py::TestVariable::test_timedelta64_conversion_scalar[values3-ns]: AssertionError: assert dtype('<m8[us]') == dtype('<m8[ns]')
 +  where dtype('<m8[us]') = <xarray.Variable ()> Size: 8B\narray(86400000000, dtype='timedelta64[us]').dtype
 +  and   dtype('<m8[ns]') = <class 'numpy.dtype'>('timedelta64[ns]')
 +    where <class 'numpy.dtype'> = np.dtype
xarray/tests/test_variable.py::TestVariable::test_timedelta64_conversion_scalar[values4-ns]: AssertionError: assert dtype('<m8[us]') == dtype('<m8[ns]')
 +  where dtype('<m8[us]') = <xarray.Variable ()> Size: 8B\narray(86400000000, dtype='timedelta64[us]').dtype
 +  and   dtype('<m8[ns]') = <class 'numpy.dtype'>('timedelta64[ns]')
 +    where <class 'numpy.dtype'> = np.dtype
xarray/tests/test_variable.py::TestVariable::test_0d_timedelta[values0-ns]: AssertionError: assert dtype('<m8[us]') == dtype('<m8[ns]')
 +  where dtype('<m8[us]') = <xarray.Variable ()> Size: 8B\narray(1000000, dtype='timedelta64[us]').dtype
 +  and   dtype('<m8[ns]') = <class 'numpy.dtype'>('timedelta64[ns]')
 +    where <class 'numpy.dtype'> = np.dtype
xarray/tests/test_variable.py::TestIndexVariable::test_timedelta64_conversion[values5-ns]: AssertionError: assert dtype('<m8[us]') == dtype('<m8[ns]')
 +  where dtype('<m8[us]') = <xarray.IndexVariable 't' (t: 3)> Size: 24B\narray([           0,  86400000000, 172800000000], dtype='timedelta64[us]').dtype
 +  and   dtype('<m8[ns]') = <class 'numpy.dtype'>('timedelta64[ns]')
 +    where <class 'numpy.dtype'> = np.dtype
xarray/tests/test_variable.py::test_datetime_conversion[DatetimeIndex(['2000-01-01'], dtype='datetime64[us]', freq='D')-ns]: AssertionError: assert dtype('<M8[us]') == dtype('<M8[ns]')
 +  where dtype('<M8[us]') = <xarray.Variable (time: 1)> Size: 8B\narray(['2000-01-01T00:00:00.000000'], dtype='datetime64[us]').dtype
 +  and   dtype('<M8[ns]') = <class 'numpy.dtype'>('datetime64[ns]')
 +    where <class 'numpy.dtype'> = np.dtype
xarray/tests/test_variable.py::test_datetime_conversion[DatetimeIndex(['2000-01-01 00:00:00-05:00'], dtype='datetime64[us, America/New_York]', freq='D')-ns]: AssertionError: assert datetime64[us, America/New_York] == datetime64[ns, America/New_York]
 +  where datetime64[us, America/New_York] = DatetimeIndex(['2000-01-01 00:00:00-05:00'], dtype='datetime64[us, America/New_York]', freq='D').dtype
 +    where DatetimeIndex(['2000-01-01 00:00:00-05:00'], dtype='datetime64[us, America/New_York]', freq='D') = PandasIndexingAdapter(array=DatetimeIndex(['2000-01-01 00:00:00-05:00'], dtype='datetime64[us, America/New_York]', freq='D'), dtype=datetime64[us, America/New_York]).array
 +      where PandasIndexingAdapter(array=DatetimeIndex(['2000-01-01 00:00:00-05:00'], dtype='datetime64[us, America/New_York]', freq='D'), dtype=datetime64[us, America/New_York]) = <xarray.Variable (time: 1)> Size: 8B\n<DatetimeArray>\n['2000-01-01 00:00:00-05:00']\nLength: 1, dtype: datetime64[us, America/New_York]._data
 +  and   datetime64[ns, America/New_York] = <class 'pandas.DatetimeTZDtype'>('ns', <DstTzInfo 'America/New_York' LMT-1 day, 19:04:00 STD>)
 +    where <class 'pandas.DatetimeTZDtype'> = pd.DatetimeTZDtype
 +    and   <DstTzInfo 'America/New_York' LMT-1 day, 19:04:00 STD> = <function timezone at 0x7fa74d4aa160>('America/New_York')
 +      where <function timezone at 0x7fa74d4aa160> = pytz.timezone
xarray/tests/test_variable.py::test_datetime_conversion[0   2000-01-01 00:00:00-05:00\ndtype: datetime64[us, America/New_York]-ns]: AssertionError: assert dtype('<M8[us]') == dtype('<M8[ns]')
 +  where dtype('<M8[us]') = <xarray.Variable (time: 1)> Size: 8B\narray(['2000-01-01T05:00:00.000000'], dtype='datetime64[us]').dtype
 +  and   dtype('<M8[ns]') = <class 'numpy.dtype'>('datetime64[ns]')
 +    where <class 'numpy.dtype'> = np.dtype
xarray/tests/test_variable.py::test_timedelta_conversion[1 day, 0:00:00-ns]: AssertionError: assert dtype('<m8[us]') == dtype('<m8[ns]')
 +  where dtype('<m8[us]') = <xarray.Variable ()> Size: 8B\narray(86400000000, dtype='timedelta64[us]').dtype
 +  and   dtype('<m8[ns]') = <class 'numpy.dtype'>('timedelta64[ns]')
 +    where <class 'numpy.dtype'> = np.dtype
xarray/tests/test_variable.py::test_timedelta_conversion[[datetime.timedelta(days=1)]-ns]: AssertionError: assert dtype('<m8[us]') == dtype('<m8[ns]')
 +  where dtype('<m8[us]') = <xarray.Variable (time: 1)> Size: 8B\narray([86400000000], dtype='timedelta64[us]').dtype
 +  and   dtype('<m8[ns]') = <class 'numpy.dtype'>('timedelta64[ns]')
 +    where <class 'numpy.dtype'> = np.dtype
xarray/tests/test_variable.py::TestIndexVariable::test_index_0d_timedelta64: AssertionError: assert dtype('<m8[us]') == 'timedelta64[ns]'
 +  where dtype('<m8[us]') = np.timedelta64(3600000000,'us').dtype
xarray/tests/test_variable.py::TestIndexVariable::test_datetime64_conversion[values0-ns]: AssertionError: assert dtype('<M8[us]') == dtype('<M8[ns]')
 +  where dtype('<M8[us]') = <xarray.IndexVariable 't' (t: 3)> Size: 24B\narray(['1970-01-01T00:00:00.000000', '1970-01-02T00:00:00.000000',\n       '1970-01-03T00:00:00.000000'], dtype='datetime64[us]').dtype
 +  and   dtype('<M8[ns]') = <class 'numpy.dtype'>('datetime64[ns]')
 +    where <class 'numpy.dtype'> = np.dtype
xarray/tests/test_variable.py::TestIndexVariable::test_datetime64_conversion[values1-ns]: AssertionError: assert dtype('<M8[us]') == dtype('<M8[ns]')
 +  where dtype('<M8[us]') = <xarray.IndexVariable 't' (t: 3)> Size: 24B\narray(['1970-01-01T00:00:00.000000', '1970-01-02T00:00:00.000000',\n       '1970-01-03T00:00:00.000000'], dtype='datetime64[us]').dtype
 +  and   dtype('<M8[ns]') = <class 'numpy.dtype'>('datetime64[ns]')
 +    where <class 'numpy.dtype'> = np.dtype

[kmuehlbauer]

Probably due to pandas-dev/pandas#63196 (3.0.0.dev0+2768.g945385d09c), checking.

Already failing at 3.0.0.dev0+2761.g1895c38229, with slightly differing errors.

Last know working version (in CI):

• https://github.com/pydata/xarray/actions/runs/19776226404/job/56669121237
• xarray: 2025.11.1.dev9+g52283c6b7 (Adapt to change in behavior of pandas.Timestamp constructor #10944)
• pandas: 3.0.0.dev0+2754.g1b5b02c08b (BUG: Fix json_normalize 'meta' validation pandas-dev/pandas#63136)
• numpy: 2.4.0.dev0+git20251125.2c2a0d3

First known failing version (in CI):

• https://github.com/pydata/xarray/actions/runs/19791259793/job/56704550186
• xarray: 2025.11.1.dev9+g52283c6b7
• pandas: 3.0.0.dev0+2755.g809d9b72a0 (API: date_range, timedelta_range infer unit from start/end/freq pandas-dev/pandas#63146)
• numpy: 2.4.0.dev0+git20251125.2c2a0d3

There have been several changes on the pandas-side since the above commit changing time handling, which might change behaviour:

• API: date_range, timedelta_range infer unit from start/end/freq pandas-dev/pandas#63146
• ENH: resolution inference for array_to_timedelta64 pandas-dev/pandas#63018
• API: microsecond resolution for Timedelta strings pandas-dev/pandas#63196
• BUG: plotting with non-nano TimedeltaIndex pandas-dev/pandas#63237
• BUG: pytables with non-nano timedelta64 pandas-dev/pandas#63239
• API: Timedelta constructor from keywords give microseconds pandas-dev/pandas#63216

I'm not sure on the best way forward, as it seems this is still worked on in pandas.

[spencerkclark]

Thanks @kmuehlbauer—I agree maybe worth waiting a few days to let the dust settle. From a quick glance at least some of the test failures seem like they might be fairly innocuous, i.e. they may only require us being a little more careful in explicitly specifying units when setting up the tests.