[Numpy-discussion] tests fails when run $nosetests numpy, but not from python prompt

Ralf Gommers ralf.gommers at googlemail.com
Mon Oct 25 11:12:39 EDT 2010


On Mon, Oct 25, 2010 at 5:40 PM, Friedrich Romstedt
<friedrichromstedt at gmail.com> wrote:
> 2010/10/25 Vincent Davis <vincent at vincentdavis.net>:
>> On Sun, Oct 24, 2010 at 6:29 PM,  <josef.pktd at gmail.com> wrote:
>>> On Sun, Oct 24, 2010 at 8:17 PM, Ralf Gommers
>>> <ralf.gommers at googlemail.com> wrote:
>>>> On Sun, Oct 24, 2010 at 2:44 AM, Vincent Davis <vincent at vincentdavis.net> wrote:
>>>>> python2.7 10.5, osx 10.6, numpy 1.5.1rc1
>>>>> Test pass when run from python prompt (see bottom) but fail when run
>>>>> from $nosetests numpy.
>>>>
>>>> That is expected, see
>>>> http://article.gmane.org/gmane.comp.python.numeric.general/40781.
>>>
>>> That only applies to the first group of tests,
>>> On the commandline he has 19 additional test compared to numpy.test()
>>> and failures like
>>>
>>> /tmp/tmpOjoP4S/_test_ext_module_5403.so: mach-o, but wrong architecture
>>>
>>> Are there tests in numpy.test() that are skipped for some other reason
>>> (e.g. decorated as slow) ?
>>

You're right, 13 extra distutils tests labeled as slow. These are
failures on OS X due to #1399.

Ralf


>> A little more looking into this I think it might be a osx issue. I
>> can't seem to find it now but nosetests runs the following
>> /Library/Frameworks/Python.framework/Versions/2.7/Resources/Python.app
>> I am guessing that Python.app loads the wrong python (python2.7 or
>> python2.7-32)  or something like that. It might be that this is more
>> of a nose python2.7 issue.
>
> 105osxpython:numpy Friedrich$ head
> /Library/Frameworks/Python.framework/Versions/2.7/bin/nosetests
> #!/Library/Frameworks/Python.framework/Versions/2.7/Resources/Python.app/Contents/MacOS/Python
>
> Same applies to the 2.6 & 2.5 :file:`nosetests`.  At least for the
> 2.7, the MacOS/Python called has different size than the bin/python2.7
> cmd line executable.  So it's simply something completely different.
> Still it makes the errors not explained.
>
> I remember Ralf's post cited by himself.
>
> FYI, the MacOS/Python executable has all three archs.
>
> We came to the conclusion, so simple not use nosetests, and to leave
> this mysterious issue to later, I will document it on the numpy build
> doc web site.
>
> Friedrich
> _______________________________________________
> NumPy-Discussion mailing list
> NumPy-Discussion at scipy.org
> http://mail.scipy.org/mailman/listinfo/numpy-discussion
>



More information about the NumPy-Discussion mailing list