Unit Test Results

See test report for an extended history of previous test failures. This is useful for diagnosing flaky tests.

     21 files  +     2       21 suites  +2   5h 13m 39s ⏱️ + 9m 28s
 18 282 tests ±     0   17 010 ✅ ±     0   1 272 💤 ±    0  0 ❌ ±0 
294 733 runs  +13 460  252 433 ✅ +11 892  42 300 💤 +1 568  0 ❌ ±0 

Results for commit a642b8d. ± Comparison against base commit 08f5daa.

♻️ This comment has been updated with latest results.

CI failures are unrelated.
Please review and merge #12224 first.

I see this failure that looks pertinent when I try to run the dask unit tests locally on my ARM Mac on this PR branch with free-threaded 3.14:

FAILED dask/array/tests/test_array_core.py::test_store_locks - assert 49 == 50
 +  where 49 = <dask.array.tests.test_array_core.CounterLock object at 0x24b5bd37950>.acquire_count
 +  and   50 = <dask.array.tests.test_array_core.CounterLock object at 0x24b5bd37950>.release_count

I also see a few other test failures that look unrelated:

FAILED dask/array/tests/test_array_core.py::test_arithmetic - DeprecationWarning: numpy.fix is deprecated. Use numpy.trunc instead, which is faster and follows the Array API standard.
FAILED dask/dataframe/dask_expr/tests/test_collection.py::test_repartition_partition_size - AssertionError: assert 5 == 4
 +  where 5 = Dask DataFrame Structure:\n                   x      y\nnpartitions=5              \n0              int64  int64\n20               ...    ...\n...              ...    ...\n80               ...    ...\n99               ...    ...\nDask Name: repartition, 2 expressions\nExpr=Repartition(frame=df, partition_size='1kb').npartitions
FAILED dask/dataframe/io/tests/test_parquet.py::test_arrow_partitioning - pandas.errors.Pandas4Warning: Constructing a Categorical with a dtype and values containing non-null entries not in that dtype's categories is deprecated and will raise in a future version.
FAILED dask/dataframe/io/tests/test_parquet.py::test_roundtrip_partitioned_pyarrow_dataset[pyarrow] - KeyError: 2

I got impatient and killed the tests in dask/dataframe/tests/test_categorical.py so I didn't actually run the full test suite.

FAILED dask/array/tests/test_array_core.py::test_store_locks - assert 49 == 50

• where 49 = <dask.array.tests.test_array_core.CounterLock object at 0x24b5bd37950>.acquire_count
• and 50 = <dask.array.tests.test_array_core.CounterLock object at 0x24b5bd37950>.release_count

Are you on numpy git tip? Do you get the same failures with 2.4.1? What about the nightly wheels?

FAILED dask/array/tests/test_array_core.py::test_arithmetic - DeprecationWarning: numpy.fix is deprecated. Use numpy.trunc instead, which is faster and follows the Array API standard.

Thanks; I think the reason CI didn't spot this yet is that numpy nightly wheels are lagging a bit behind git tip?
#12231

FAILED dask/dataframe/dask_expr/tests/test_collection.py::test_repartition_partition_size - AssertionError: assert 5 == 4
where 5 = Dask DataFrame Structure:\n x y\nnpartitions=5 \n0 int64 int64\n20 ... ...\n... ... ...\n80 ... ...\n99 ... ...\nDask Name: repartition, 2 expressions\nExpr=Repartition(frame=df, partition_size='1kb').npartitions
FAILED dask/dataframe/io/tests/test_parquet.py::test_arrow_partitioning - pandas.errors.Pandas4Warning: Constructing a Categorical with a dtype and values containing non-null entries not in that dtype's categories is deprecated and will raise in a future version.
FAILED dask/dataframe/io/tests/test_parquet.py::test_roundtrip_partitioned_pyarrow_dataset[pyarrow] - KeyError: 2

Any idea why these don't show up on 3.14 with pandas nightly? Similar lag issue as above?
#12225

Are you on numpy git tip? Do you get the same failures with 2.4.1? What about the nightly wheels?

Yup, NumPy is built from source. The failure seems to be flaky, it only happened the first time I ran the tests not the second time after I installed optional dependencies. Now that I'm trying just that one test repeatedly, it seems to fail about half of the times I try to run it. I can trigger it with the 2.4.1 NumPy wheel from PyPI.

FWIW, I see no crashes in Pandas. Maybe try uncommenting Pandas to see if any of the CI runners hit the crash?

Thanks; I think the reason CI didn't spot this yet is that numpy nightly wheels are lagging a bit behind git tip?
#12231

Yeah, the deprecation just got merged today so no surprise there.

Any idea why these don't show up on 3.14 with pandas nightly? Similar lag issue as above?

No idea, sorry.

@jacobtomlinson @jsignell does either of you have bandwidth to offer review?
(note that this is blocked by and incorporates #12224)

With pandas 3.0 final I'm getting these failures

FAILED dask/dataframe/dask_expr/tests/test_collection.py::test_repartition_partition_size - AssertionError: assert 5 == 4
FAILED dask/tests/test_tokenize.py::test_tokenize_range_index - assert 16 >= 32

I'd rather investigate them in a follow-up to avoid scope creep.