NaN inverse raises unexpected ZeroDivisionError

[Ranfir]

Reporting a bug

• I have tried using the latest released version of Numba (most recent is
  visible in the change log (https://github.com/numba/numba/blob/main/CHANGE_LOG).
• I have included a self contained code sample to reproduce the problem.
  i.e. it's possible to run as 'python bug.py'.

from numba import njit
import numpy as np

@njit
def foo(x):
    return x**-1

foo(np.array([np.nan])) # array([nan])
foo(np.array([np.nan], np.float32)) # ZeroDivisionError: division by zero

Tested under python 3.10.9. This behavior appears in 0.57.1, but is not present in 0.56.4. Setting fastmath=False per #8889 doesn't change this behavior.

[stuartarchibald]

Thanks for the report. Unfortunately this doesn't seem to reproduce for me, I've tested it on a couple of systems with different configurations. There are some issues with pow as noted in #7012, it might be another issue in the implementation that moving to LLVM 14 has highlighted.

[github-actions]

This issue is marked as stale as it has had no activity in the past 30 days. Please close this issue if no further response or action is needed. Otherwise, please respond with any updates and confirm that this issue still needs to be addressed.