BUG: Internal copyswap/copyswapn calls segfault for new-style user DTypes #30795

Describe the issue:

Several NumPy functions internally call copyswap/copyswapn through PyDataType_GetArrFuncs(descr)->copyswap without NULL checks. For new-style user DTypes registered via PyArrayInitDTypeMeta_FromSpec, these function pointers default to NULL, causing segfaults.

Comment at dtypemeta.c says:

With the reference to the conversation happened with @seberg in numpy/numpy-quaddtype#13 the points discussed are

• copyswap should not be become a first-class new DType API
• Byte-Swapping should be should be handled via casting and byteswap itself is not important as a first-class operation

Possible direction

• For call sites that only copy (swap=0), replace copyswap with memcpy or the strided cast (np.place use it just for copy)
• For actual byte-swapping, these should go through the casting. A "byteswapped view" + cast to native byte order can be the equivalent.
  • PyArray_DescrNewByteorder() support for new style DTypes, or DTypes supporting a byteorder parameter (eg new_dtype(byteorder='>'))

Immediate fix

Add NULL guards at call sites to convert segfaults into clear errors (TypeError or similar), independent of the longer-term casting-based approach.

Reproduce the code example:

import numpy_quaddtype as npq

# Segfault 1: scalar byteswap
npq.QuadPrecision(1).byteswap()

# Segfault 2: np.place
import numpy as np
from numpy_quaddtype import QuadPrecDType
a = np.zeros((18, 1), dtype=QuadPrecDType())
m = np.zeros((18, 1), dtype=np.bool)
m[3:-3] = True
v = np.ones(12, dtype=QuadPrecDType())
np.place(a, m, v)

Error message:

Segmentation fault (core dumped)

Python and NumPy Versions:

2.5.0.dev0+git20251208.7eec5e2
3.12.11 | packaged by Anaconda, Inc. | (main, Jun  5 2025, 13:09:17) [GCC 11.2.0]

Runtime Environment:

[{'numpy_version': '2.5.0.dev0+git20251208.7eec5e2',
  'python': '3.12.11 | packaged by Anaconda, Inc. | (main, Jun  5 2025, '
            '13:09:17) [GCC 11.2.0]',
  'uname': uname_result(system='Linux', node='', release='6.8.0-1044-azure', version='#50~22.04.1-Ubuntu SMP Wed Dec  3 15:13:22 UTC 2025', machine='x86_64')},
 {'simd_extensions': {'baseline': ['X86_V2'],
                      'found': ['X86_V3'],
                      'not_found': ['X86_V4', 'AVX512_ICL']}},
 {'ignore_floating_point_errors_in_matmul': False},
 {'architecture': 'Zen',
  'filepath': '/home/miniconda3/envs/quaddtype/lib/libopenblasp-r0.3.29.so',
  'internal_api': 'openblas',
  'num_threads': 96,
  'prefix': 'libopenblas',
  'threading_layer': 'pthreads',
  'user_api': 'blas',
  'version': '0.3.29'}]

How does this issue affect you or how did you find it:

No response