Set strict xfail option

dask/array/tests/test_array_core.py

@@ -240,7 +240,6 @@ def test_Array_computation():
     LooseVersion(np.__version__) < "1.14.0",
     reason="NumPy doesn't have `np.linalg._umath_linalg` yet",
 )
-@pytest.mark.xfail(reason="Protect from `np.linalg._umath_linalg.inv` breaking")
 def test_Array_numpy_gufunc_call__array_ufunc__01():
     x = da.random.normal(size=(3, 10, 10), chunks=(2, 10, 10))
     nx = x.compute()

dask/array/tests/test_array_function.py

@@ -89,9 +89,6 @@ def test_array_function_sparse(func):
 
 
 @pytest.mark.skipif(missing_arrfunc_cond, reason=missing_arrfunc_reason)
-@pytest.mark.xfail(
-    reason="requires sparse support for __array_function__", strict=False
-)
 def test_array_function_sparse_tensordot():
     sparse = pytest.importorskip("sparse")
     x = np.random.random((2, 3, 4))

dask/array/tests/test_array_utils.py

@@ -3,19 +3,13 @@
 
 import dask.array as da
 from dask.array.utils import meta_from_array, assert_eq
-from dask.array.numpy_compat import _numpy_117
 
 asarrays = [np.asarray]
 
 try:
     import sparse
 
-    asarrays.append(
-        pytest.param(
-            sparse.COO.from_numpy,
-            marks=pytest.mark.xfail(_numpy_117, reason="sparse-257"),
-        )
-    )
+    asarrays.append(sparse.COO.from_numpy)
 except ImportError:
     pass
 

dask/array/tests/test_creation.py

@@ -600,7 +600,7 @@ def test_tile_array_reps(shape, chunks, reps):
         da.tile(d, reps)
 
 
-skip_stat_length = pytest.mark.xfail(_numpy_117, reason="numpy-14061", strict=True)
+skip_stat_length = pytest.mark.xfail(_numpy_117, reason="numpy-14061")
 
 
 @pytest.mark.parametrize(

dask/array/tests/test_gufunc.py

@@ -166,9 +166,6 @@ def foo(x, bar):
     assert_eq(ret, np.array(1.0, dtype="f"))
 
 
-@pytest.mark.xfail(
-    reason="Currently np.einsum doesn't seem to broadcast correctly for this case"
-)
 def test_apply_gufunc_02():
     def outer_product(x, y):
         return np.einsum("...i,...j->...ij", x, y)

dask/bytes/tests/test_http.py

@@ -166,7 +166,7 @@ def test_parquet():
     assert df.columns.tolist() == ["n_nationkey", "n_name", "n_regionkey", "n_comment"]
 
 
-@pytest.mark.xfail(reason="https://github.com/dask/dask/issues/3696")
+@pytest.mark.xfail(reason="https://github.com/dask/dask/issues/3696", strict=False)
 @pytest.mark.network
 def test_bag():
     # This test pulls from different hosts

dask/dataframe/io/tests/test_io.py

@@ -268,13 +268,12 @@ def test_from_pandas_small():
             assert_eq(s, ds)
 
 
-@pytest.mark.xfail(reason="")
-def test_from_pandas_npartitions_is_accurate():
+@pytest.mark.parametrize("n", [1, 2, 4, 5])
+def test_from_pandas_npartitions_is_accurate(n):
     df = pd.DataFrame(
         {"x": [1, 2, 3, 4, 5, 6], "y": list("abdabd")}, index=[10, 20, 30, 40, 50, 60]
     )
-    for n in [1, 2, 4, 5]:
-        assert dd.from_pandas(df, npartitions=n).npartitions == n
+    assert dd.from_pandas(df, npartitions=n).npartitions <= n
 
 
 def test_from_pandas_series():

dask/dataframe/tests/test_accessors.py

@@ -241,4 +241,4 @@ def test_str_accessor_expand_more_columns():
     s = pd.Series(["a b c", "aa bb cc", "aaa bbb ccc"])
     ds = dd.from_pandas(s, npartitions=2)
 
-    s.str.split(n=10, expand=True).compute()
+    ds.str.split(n=10, expand=True).compute()

dask/dataframe/tests/test_dataframe.py

@@ -27,13 +27,7 @@
     total_mem_usage,
 )
 from dask.dataframe import methods
-from dask.dataframe.utils import (
-    assert_eq,
-    make_meta,
-    assert_max_deps,
-    PANDAS_VERSION,
-    PANDAS_GT_0250,
-)
+from dask.dataframe.utils import assert_eq, make_meta, assert_max_deps, PANDAS_VERSION
 
 dsk = {
     ("x", 0): pd.DataFrame({"a": [1, 2, 3], "b": [4, 5, 6]}, index=[0, 1, 3]),
@@ -403,7 +397,6 @@ def test_describe(include, exclude, percentiles, subset):
             )
 
 
-@pytest.mark.xfail(PANDAS_GT_0250, reason="Pandas change.")
 def test_describe_empty():
     df_none = pd.DataFrame({"A": [None, None]})
     ddf_none = dd.from_pandas(df_none, 2)
@@ -413,16 +406,9 @@ def test_describe_empty():
 
     # Pandas have different dtypes for resulting describe dataframe if there are only
     # None-values, pre-compute dask df to bypass _meta check
-    if PANDAS_GT_0250:
-        # https://github.com/pandas-dev/pandas/issues/27183
-        # may be fixed for pandas RC. If so, remove
-        expected = df_none.describe()
-        result = ddf_none.describe(percentiles_method="dask").compute()
-        assert_eq(expected, result)
-    else:
-        assert_eq(
-            df_none.describe(), ddf_none.describe(percentiles_method="dask").compute()
-        )
+    assert_eq(
+        df_none.describe(), ddf_none.describe(percentiles_method="dask").compute()
+    )
 
     with pytest.raises(ValueError):
         ddf_len0.describe(percentiles_method="dask").compute()
@@ -434,7 +420,6 @@ def test_describe_empty():
         ddf_nocols.describe(percentiles_method="dask").compute()
 
 
-@pytest.mark.xfail(PANDAS_GT_0250, reason="Pandas change.")
 def test_describe_empty_tdigest():
     pytest.importorskip("crick")
 

dask/dataframe/tests/test_ufunc.py

@@ -10,7 +10,6 @@
 import dask.array as da
 import dask.dataframe as dd
 from dask.dataframe.utils import assert_eq
-from dask.array.numpy_compat import _numpy_117
 
 
 _BASE_UFUNCS = [
@@ -155,8 +154,7 @@ def test_ufunc(pandas_input, ufunc):
         pytest.param("imag", marks=pytest.mark.filterwarnings("ignore::FutureWarning")),
         "angle",
         "fix",
-        # Possible NumPy / pandas bug: https://github.com/numpy/numpy/issues/13894
-        pytest.param("i0", marks=pytest.mark.xfail(_numpy_117, reason="NumPy-13894")),
+        "i0",
         "sinc",
         "nan_to_num",
     ],

dask/tests/test_distributed.py

@@ -121,8 +121,20 @@ def test_to_hdf_distributed(loop):
             test_to_hdf()
 
 
-@pytest.mark.xfail(reason="HDF not multi-process safe")
-@pytest.mark.parametrize("npartitions", [1, 4, 10])
+@pytest.mark.parametrize(
+    "npartitions",
+    [
+        1,
+        pytest.param(
+            4,
+            marks=pytest.mark.xfail(reason="HDF not multi-process safe", strict=False),
+        ),
+        pytest.param(
+            10,
+            marks=pytest.mark.xfail(reason="HDF not multi-process safe", strict=False),
+        ),
+    ],
+)
 def test_to_hdf_scheduler_distributed(npartitions, loop):
     from ..dataframe.io.tests.test_hdf import test_to_hdf_schedulers
 

dask/tests/test_dot.py

@@ -128,7 +128,9 @@ def test_aliases():
         pytest.param(
             "jpeg",
             Image,
-            marks=pytest.mark.xfail(reason="jpeg not always supported in dot"),
+            marks=pytest.mark.xfail(
+                reason="jpeg not always supported in dot", strict=False
+            ),
         ),
         ("dot", type(None)),
         ("pdf", type(None)),
@@ -159,7 +161,9 @@ def test_dot_graph(tmpdir, format, typ):
         pytest.param(
             "jpeg",
             Image,
-            marks=pytest.mark.xfail(reason="jpeg not always supported in dot"),
+            marks=pytest.mark.xfail(
+                reason="jpeg not always supported in dot", strict=False
+            ),
         ),
         ("dot", type(None)),
         ("pdf", type(None)),

dask/tests/test_order.py

@@ -36,7 +36,6 @@ def test_ordering_keeps_groups_together(abcde):
     assert abs(o[(a, 1)] - o[(a, 3)]) == 1
 
 
-@pytest.mark.xfail(reason="Can't please 'em all")
 def test_avoid_broker_nodes(abcde):
     r"""
 
@@ -173,7 +172,7 @@ def test_avoid_upwards_branching_complex(abcde):
     assert o[(c, 1)] < o[(b, 1)]
 
 
-@pytest.mark.xfail(reason="this case is ambiguous")
+@pytest.mark.xfail(reason="this case is ambiguous", strict=False)
 def test_deep_bases_win_over_dependents(abcde):
     r"""
     It's not clear who should run first, e or d

dask/tests/test_threaded.py

@@ -138,6 +138,7 @@ def test_f():
 @pytest.mark.xfail(
     "xdist" in sys.modules,
     reason="This test fails intermittently when using pytest-xdist (maybe)",
+    strict=False,
 )
 def test_interrupt():
     # Python 2 and windows 2 & 3 both implement `queue.get` using polling,

setup.cfg

@@ -43,6 +43,7 @@ filterwarnings =
     error:::pandas[.*]
     error:::numpy[.*]
     error:::distributed[.*]]
+xfail_strict=true
 
 [metadata]
 license_files = LICENSE.txt