[BUG]: Numpy 2.4.0 breaks 6 cuda.core tests

[mdboom]

Is this a duplicate?

• I confirmed there appear to be no duplicate issues for this bug and that I agree to the Code of Conduct

Type of Bug

Something else

Component

cuda.core

Describe the bug

With Numpy 2.4.0, the cuda.core test suite fails on the following tests:

FAILED cuda_core/tests/test_utils.py::TestViewCPU::test_args_viewable_as_strided_memory_cpu[in_arr4] - assert (1,) is None
FAILED cuda_core/tests/test_utils.py::TestViewCPU::test_strided_memory_view_cpu_init[in_arr0] - assert (1,) is None
FAILED cuda_core/tests/test_utils.py::TestViewCPU::test_strided_memory_view_cpu[in_arr0] - assert (1,) is None
FAILED cuda_core/tests/test_utils.py::TestViewCPU::test_args_viewable_as_strided_memory_cpu[in_arr0] - assert (1,) is None
FAILED cuda_core/tests/test_utils.py::TestViewCPU::test_strided_memory_view_cpu_init[in_arr4] - assert (1,) is None
FAILED cuda_core/tests/test_utils.py::TestViewCPU::test_strided_memory_view_cpu[in_arr4] - assert (1,) is None

How to Reproduce

Run the cuda.core tests with Numpy 2.4.0 installed. (Numpy 2.3.5 works).

Expected behavior

Tests pass.

My hunch is that Numpy previously set strides to None when the data was C contiguous, but now it is setting it always. I don't see a Numpy CHANGELOG entry for this, but strides is now readonly, fwiw.

Operating System

All Python versions 3.11 and greater in CI. This tracks, since 3.10 is not supported by Numpy 2.4.0, so there we get an earlier version installed.

nvidia-smi output

No response