v2.4.5
📦 numpyView on GitHub →
✨ 1 features🐛 11 fixes🔧 14 symbols
Summary
NumPy 2.4.5 is a patch release focused on bug fixes, typing improvements, and infrastructure maintenance. It resolves issues related to memory safety, type checking, and specific platform builds.
Migration Steps
- If you were relying on the old `np.dtype()` signature used by old pickles, note that the deprecation for it has been undone.
✨ New Features
- Return rank 0 for empty matrices in matrix_rank.
🐛 Bug Fixes
- Fix `np.shape` assignability issue for python lists.
- f2py map complex_long_double to NPY_CLONGDOUBLE.
- Fix heap buffer overflow in timedelta to string casts.
- Fix matvec/vecmat in-place aliasing (out=input produces incorrect results).
- Fix incorrect temp elision for new-style (NEP 43) user-defined dtypes.
- Prevent deadlock due to downstream importing NumPy in dlopen contexts.
- Fix segfault in nditer.multi_index when __getitem__ raises an exception.
- Fix memory leak in np.zeros when fill-zero loop raises an exception.
- np.einsum() fails with a 0-dimensional out argument and non-scalar input.
- Fix signed overflow issue in npy_gcd for INT_MIN on s390x architecture.
- Avoid Undefined Behavior in [safe][add,sub,mul] helpers.