rok opened a new issue, #55458: URL: https://github.com/apache/arrow/issues/55458
### Describe the bug, including details regarding any error messages, version, and platform. A few tests use NumPy but have no `numpy` mark. They belong to groups that are off by default, so the no-NumPy CI job (`conda-python-3.13-no-numpy`) never runs them, and the missing mark goes unnoticed. `large_memory` tests only run on the ARM64 macOS job, and CUDA tests only run on GPU runners; both have NumPy installed. Affected tests (these modules set `np = None` when NumPy is missing): - `test_convert_builtin.py::test_auto_chunking_list_like` (`large_memory`, uses `np.ones`) - `test_convert_builtin.py::test_array_from_pylist_data_overflow` (`large_memory`, uses `np.zeros`) - `test_convert_builtin.py::test_array_from_pylist_offset_overflow` (`large_memory`, `slow`, uses `np.zeros`) - `test_array.py::test_non_cpu_array` (needs CUDA, uses `np.arange`) With NumPy uninstalled, running these with `--enable-large_memory` (or on a CUDA machine) fails with `AttributeError: 'NoneType' object has no attribute ...` instead of skipping. The fix is to add `@pytest.mark.numpy` to these four tests. Found while reviewing https://github.com/apache/arrow/pull/55264. ### Component(s) Python -- This is an automated message from the Apache Git Service. To respond to the message, please log on to GitHub and use the URL above to go to the specific comment. To unsubscribe, e-mail: [email protected] For queries about this service, please contact Infrastructure at: [email protected]
