IntelPython / dpnp / 15054821589

81%

push

github

Apply w/a for expected result in spacing test only on non-Intel NumPy (#2453)

The stock NumPy returns incorrect result for `numpy.spacing(-0.0)`,
while Intel NumPy returns the expected one (the same as dpnp also).
It might be the case when dpnp test is running with the stock NumPy.
Thus the PR add a conditional check to update expected result in
`TestSpacing::test_zeros` in case when non-Intel NumPy is used.

Previously the same logic was based on NumPy version, assuming Intel
NumPy has always `version < "2.0.0"`, but it is not the case anymore.

4772 of 9764 branches covered (48.87%)

Branch coverage included in aggregate %.

18091 of 22118 relevant lines covered (81.79%)

19352.41 hits per line