Don't fail when np.intc != np.int32

np.intc (C int), np.int_ (C long) and np.longlong (C long long)
are always distinct type objects, but only two of them are
actually different sizes; np.int32 and np.int64 are aliases
https://sources.debian.org/src/numpy/1:1.18.4-1/numpy/core/_type_aliases.py/#L110

Author: Rebecca N. Palmer <[email protected]>
Bug: pandas-dev/pandas#31856
Forwarded: not-needed pandas-dev/pandas@1660057


Gbp-Pq: Name test_promote_32bit.patch

Author: Debian Science Team

pandas/tests/dtypes/cast/test_promote.py

@@ -96,19 +96,10 @@ def _check_promote(dtype, fill_value, expected_dtype, exp_val_for_scalar=None):
 
 def _assert_match(result_fill_value, expected_fill_value):
     # GH#23982/25425 require the same type in addition to equality/NA-ness
-    res_type = type(result_fill_value)
-    ex_type = type(expected_fill_value)
-    if res_type.__name__ == "uint64":
-        # No idea why, but these (sometimes) do not compare as equal
-        assert ex_type.__name__ == "uint64"
-    elif res_type.__name__ == "ulonglong":
-        # On some builds we get this instead of np.uint64
-        # Note: cant check res_type.dtype.itemsize directly on numpy 1.18
-        assert res_type(0).itemsize == 8
-        assert ex_type == res_type or ex_type == np.uint64
-    else:
-        # On some builds, type comparison fails, e.g. np.int32 != np.int32
-        assert res_type == ex_type or res_type.__name__ == ex_type.__name__
+    # use np.dtype to treat "different" scalar types
+    # with the same data layout as equal
+    # e.g. np.intc and np.int32, np.ulonglong and np.uint64
+    assert np.dtype(type(result_fill_value)) == np.dtype(type(expected_fill_value))
 
     match_value = result_fill_value == expected_fill_value