Re: 01/01: gnu: python-numpy-bootstrap: Reduce matrix size in failing test.

[Federico Beffa <beffa@ieee.org>]

On Mon, Dec 15, 2014 at 9:41 AM, Federico Beffa <beffa@ieee.org> wrote:
> On Sun, Dec 14, 2014 at 11:54 PM, Ludovic Courtès <ludo@gnu.org> wrote:
>> Federico Beffa <beffa@fbengineering.ch> skribis:
>>
>>> +        (alist-cons-before
>>> +         'check 'fix-failing-tests
>>> +         (lambda _
>>> +           (substitute* (find-files "numpy/linalg/tests"
>>> +                                    "test_regression\\.py")
>>> +             (("x = np.eye(1000, 66)")
>>> +              "x = np.eye(10, 66)")))
>>
>> Could you add a comment here explaining that this is reducing the input
>> matrix size (?) and why it matters?  Otherwise it may seem a bit too
>> magical.  :-)
>>
>> Also, this kind of change would be a good candidate for ‘snippet’,
>> because it’s a “static” change.
>
> OK, I need to look into those :-)
>
> Actually the test is still failing and I do not understand why on
> hydra it is failing while on my machine (x86_64) it passes with both
> python versions.  So, at least for the moment, I would propose to
> remove the above phase and disable the tests.
>
> WDYT?

One more thing: the function causing the test to fail is implemented
making use of ATLAS. However, the log does not show ATLAS being built
before numpy (as I would expect given the #:substitutable? #f
argument). If the machines to which hydra distributes the jobs are not
homogeneous, then it could be that the used ATLAS is not working
properly (maybe different ISA extensions?). If that is the case, then
why is the #:substitutable argument not honored?

What machines is the build farm made of? Do they all have the same
version of OS and kernel?

Regards,
Fede