[Debian-med-packaging] Problems in python-biom-format tests for several architectures

[Andreas Tille]

Tim, could you please comment on this?  If not I'll go for upgrading
python-biom-format to the latest upstream version.

Kind regards

       Andreas (who has missed you last weekend on the sprint)

On Fri, Dec 11, 2015 at 08:36:40AM +0100, Andreas Tille wrote:
> Hi Tim,
> 
> currently qiime is in danger to be removed from testing since
> python-biom-format does not migrate to testing due to some test failures
> on several architectures (i386, armel, armhf, mips, mipsel, powerpc,
> ppc64).  Most 64 bit architectures (but not all - see ppc64) are
> passing.  In i386 the test failures are looking like:
> 
> 
> dh_auto_test
> I: pybuild base:184: cd /«BUILDDIR»/python-biom-format-2.1.4+dfsg/.pybuild/pythonX.Y_2.7/build; python2.7 -m nose tests
> ....................F........F.F..............................................................................................................................................................................................................S.............
> ======================================================================
> FAIL: Test an invalid HDF5 table
> ----------------------------------------------------------------------
> Traceback (most recent call last):
>   File "/usr/lib/python2.7/dist-packages/numpy/testing/decorators.py", line 146, in skipper_func
>     return f(*args, **kwargs)
>   File "/«BUILDDIR»/python-biom-format-2.1.4+dfsg/.pybuild/pythonX.Y_2.7/build/tests/test_commands/test_table_validator.py", line 95, in test_invalid_hdf5
>     self.assertEqual(obs, exp)
> AssertionError: {'valid_table': False, 'report_lines': ["'shape' values do not appear to be inte [truncated]... != {'valid_table': False, 'report_lines': ["Missing attribute: 'creation-date'"]}
> + {'report_lines': ["Missing attribute: 'creation-date'"], 'valid_table': False}
> - {'report_lines': ["'shape' values do not appear to be integers",
> -                   'nnz is not an integer!',
> -                   "Missing attribute: 'creation-date'"],
> -  'valid_table': False}
> 
> ======================================================================
> FAIL: Test a valid HDF5 table
> ----------------------------------------------------------------------
> Traceback (most recent call last):
>   File "/usr/lib/python2.7/dist-packages/numpy/testing/decorators.py", line 146, in skipper_func
>     return f(*args, **kwargs)
>   File "/«BUILDDIR»/python-biom-format-2.1.4+dfsg/.pybuild/pythonX.Y_2.7/build/tests/test_commands/test_table_validator.py", line 79, in test_valid_hdf5
>     self.assertEqual(obs, exp)
> AssertionError: {'valid_table': False, 'report_lines': ["'shape' values do not appear to be inte [truncated]... != {'valid_table': True, 'report_lines': []}
> + {'report_lines': [], 'valid_table': True}
> - {'report_lines': ["'shape' values do not appear to be integers",
> -                   'nnz is not an integer!'],
> -  'valid_table': False}
> 
> ...
> 
> 
> 
> The problem is now:  We are intentionally not packaging the latest
> upstream (see Tim's message in the latest bug log[1]) and before I'll
> check out this for the test case issue (and forward the issue to
> upstream) I wonder whether somebody could confirm that we actually
> could use the latest upstream as well.  Finally we will not manage
> to stick to the old version for ever.
> 
> Kind regards
> 
>        Andreas.
> 
> 
> [1] https://bugs.debian.org/cgi-bin/bugreport.cgi?bug=802867#15
> 
> -- 
> http://fam-tille.de

-- 
http://fam-tille.de