Issues

Issues

• BUG: np.bincount overflows its output length for 2**63 - 1 and writes out of bounds

  numpy/numpy#33305 · opened on Nov 12, 2026

• Modelling part of NumPy in Lean 4 found 55 bugs, 26 unreported. Would these be useful?

  numpy/numpy#33299 · opened on Nov 12, 2026
  4

• BUG: sort(stable=False) changes the count of negative zeros

  numpy/numpy#33286 · opened on Nov 11, 2026

• BUG: np.roots and np.poly do not dispatch __array_function__

  numpy/numpy#33276 · opened on Nov 10, 2026
  2

• ENH: let user DTypes convert Python scalars using the resolved descriptor

  numpy/numpy#33270 · opened on Nov 10, 2026
  1

• BUG: ndarray.any and ndarray.all take their second positional argument as dtype, not out

  numpy/numpy#33267 · opened on Nov 10, 2026
  6

• PERF: short-circuit integer gcd reductions once the result reaches 1

  numpy/numpy#33252 · opened on Nov 8, 2026
  5

• BUG: Excessive temporary memory use in string casts and string operations

  numpy/numpy#33240 · opened on Nov 6, 2026 · 2.6.0 Release

• DOC: Move “NumPy for MATLAB users” out of the user guide

  numpy/numpy#33233 · opened on Nov 6, 2026
  5

• BUG: f2py treats a backslash as an escape inside character literals

  numpy/numpy#33220 · opened on Nov 5, 2026
  2

• BUG: Inconsistent comparison of float between masked array and array

  numpy/numpy#33219 · opened on Nov 4, 2026
  1

• ENH: allow user-DType promoters that only apply to an explicitly requested output DType

  numpy/numpy#33208 · opened on Nov 3, 2026
  3

• BUG: nanargmin / nanargmax do not skip NaT on datetime64

  numpy/numpy#33207 · opened on Nov 3, 2026
  2

• BUG: write_array_header_1_0/2_0 writes NumPy scalars into the header, which read_array_header_* cannot read

  numpy/numpy#33206 · opened on Nov 3, 2026
  4

• BUG: np.bincount with float weights returns integers for empty input

  numpy/numpy#33203 · opened on Nov 2, 2026
  1

• BUG: np.gradient loses precision with integer timestamps

  numpy/numpy#33173 · opened on Oct 29, 2026
  1

• BUG: TestAccuracy.test_validate_transcendentals fails in CI with Ubuntu 26.04

  numpy/numpy#33172 · opened on Oct 29, 2026
  7

• BUG: np.linalg.qr corrupts the heap on macOS 26 arm64 with Accelerate (dgeqrf writes past its own scratch buffer)

  numpy/numpy#33161 · opened on Oct 28, 2026
  5

• BUG: np.polyfit(..., cov=True) returns negative variances for a full-rank fit

  numpy/numpy#33154 · opened on Oct 27, 2026
  3

• BUG: flatiter[True] reads and writes past the end of the array buffer

  numpy/numpy#33081 · opened on Oct 18, 2026
  1

• BUG: np.repeat overflows the result size and writes out of bounds

  numpy/numpy#33080 · opened on Oct 18, 2026
  1

• PERF: dtype.name access overhead

  numpy/numpy#33051 · opened on Oct 15, 2026
  3

• BUG: np.cov copies its whole input to float64 and loses precision at large offsets

  numpy/numpy#33022 · opened on Oct 14, 2026
  8

• BUG: Legacy NPY_DISABLE_CPU_FEATURES names are ignored without a visible warning

  numpy/numpy#33016 · opened on Oct 13, 2026
  4

• ENH: Improve CPU performance of float16 pointwise ufuncs

  numpy/numpy#32970 · opened on Oct 10, 2026
  1