[Numpy-discussion] Failures in test_recfunctions.py

[Pierre GM]

Interesting.
The tests pass on my machine
OS X,
Python version 2.5.4 (r254:67916, Dec 29 2008, 17:02:44) [GCC 4.0.1  
(Apple Inc. build 5488)]
nose version 0.10.4

For
>   File
> "/home/nwagner/local/lib64/python2.6/site-packages/numpy/lib/tests/ 
> test_recfunctions.py",
> line 34, in test_zip_descr
>     np.dtype([('', '<i4'), ('', '<i4')]))

I guess I can change '<i4' to int, which should work.

For:
>
> ======================================================================
> FAIL: Test the ignoremask option of find_duplicates
> ----------------------------------------------------------------------
> Traceback (most recent call last):
>   File
> "/home/nwagner/local/lib64/python2.6/site-packages/numpy/lib/tests/ 
> test_recfunctions.py",
> line 186, in test_find_duplicates_ignoremask
>     assert_equal(test[-1], control)
>
> (mismatch 33.3333333333%)
>  x: array([0, 1, 3, 4, 2, 6])
>  y: array([0, 1, 3, 4, 6, 2])
>

there's obviously a machine-dependent element somewhere. I'd blame  
argsort: the last 2 indices that are switched correspond to the masked  
elements in the input of the test. Note that the result is basically  
correct.
I should have access to a linux box, I'll see what I can do.