Describe the issue:

Calling PyDataMem_GetHandler() results in a segmentation fault.

Reproduce the code example:

See repo https://github.com/lorentzenchr/repo_numpy_mem_handler as a minimal reproducer.

The gist of it is that the C function

static inline void
c_get_current_handler()
{
    PyObject* handler = PyDataMem_GetHandler();
};

is wrapped by Cython and then called for in a Python test. This results in the segfault.

Error message:

Extension modules: numpy._core._multiarray_umath, numpy.linalg._umath_linalg, numpy_mem_handler (total: 3)
zsh: segmentation fault  pytest tests/test_numpy_mem_handler.py

Python and NumPy Versions:

Python 3.13.5
"numpy==2.3.0"
"Cython==3.1.3",
"pytest",
"meson==1.8.3",
"meson-python>=0.16.0",
"ninja==1.13.0",

Runtime Environment:

WARNING: threadpoolctl not found in system! Install it by pip install threadpoolctl. Once installed, try np.show_runtime again for more detailed build information
[{'numpy_version': '2.3.0',
'python': '3.13.5 (main, Jun 11 2025, 15:36:57) [Clang 17.0.0 '
'(clang-1700.0.13.3)]',
'uname': uname_result(system='Darwin', node='MacBookPro.fritz.box', release='24.6.0', version='Darwin Kernel Version 24.6.0: Mon Jul 14 11:28:17 PDT 2025; root:xnu-11417.140.69~1/RELEASE_X86_64', machine='x86_64')},
{'simd_extensions': {'baseline': ['SSE', 'SSE2', 'SSE3'],
'found': ['SSSE3',
'SSE41',
'POPCNT',
'SSE42',
'AVX',
'F16C',
'FMA3',
'AVX2'],
'not_found': ['AVX512F',
'AVX512CD',
'AVX512_KNL',
'AVX512_SKX',
'AVX512_CLX',
'AVX512_CNL',
'AVX512_ICL']}}]

Context for the issue:

In scikit-learn PR#29846 I wanted to play with aligned memory allocations for the histograms of gradient boosting. Interestingly, the segfault shows up on different platforms/setups in CD.