The f2c stuff is generated and should be left alone.

The f2c stuff is generated and should be left alone.

Okay, reverted.

This error seems erroneous to me:

  E       AssertionError: 
  E       Arrays are not equal
  E       
  E       Mismatched elements: 2 / 10 (20%)
  E       Max absolute difference among violations: 1208103775
  E       Max relative difference among violations: 32749.
  E        ACTUAL: array([[1208103775,      32750,          2,          3,          4],
  E              [         5,          6,          7,          8,          9]],
  E             dtype=int32)
  E        DESIRED: array([[0, 1, 2, 3, 4],
  E              [5, 6, 7, 8, 9]])
  
  POINTER    = <built-in function POINTER>
  a          = array([[ -217879408,       32745, -1231388384,       32745,           4],
         [          5,           6,           7,          48,           0]],
        dtype=int32)

  numpy/tests/test_ctypeslib.py:218: AssertionError

That does not seem related to this PR. I expect re-running the tests would make them pass again. (Sorry if I am wrong. This is just my guess.)

If that IS the case, perhaps we should open a bug ticket for intermittently failing CI.

You can ignore the PyPy failure, it is a known bug in PyPy.

Changing the error messages is technically a backward compatibility issue - can you add a release note for the off chance someone is using the error message in a test like the numpy tests were?

I have added a release note. I tried to word it in keeping with other release notes. But feel free to holler if you'd like it worded differently.

There was a spelling error in the release note 🙃 so I fixed that and made it a little less wordy. Will merge shortly.

Actually I think @tylerjereddy needs to give it another pass because of branch protection

Actually I think @tylerjereddy needs to give it another pass because of branch protection

I think I found one more case that should be reverted, but I've dismissed the blocking review apart from that.

Thanks @john-science!