Bug#799195: python-scientific: FTBFS during rebuild against netcdf 4.4.0

[Niels Thykier]

Source: python-scientific
Version: 2.9.4-3
Severity: serious

Hi

The python-scientific package FTBFS during a rebuild against netcdf
4.4.0 (a transition).  Though I cannot tell if the issue is directly
related to netcdf or some other change in unstable since the upload.

The relevant part of the log appears to be:

"""
changing mode of build/scripts-2.7/task_manager from 664 to 775
changing mode of build/scripts-2.7/bsp_virtual from 664 to 775
rm -f Src/MPI/*.o
cd Src/MPI && PYTHONPATH=$(echo /«PKGBUILDDIR»/build/lib.*-2.7) USE_MPICH=yes   /usr/bin/python  compile.py
Traceback (most recent call last):
  File "compile.py", line 8, in <module>
    from Scientific import N
  File "/«PKGBUILDDIR»/build/lib.linux-x86_64-2.7/Scientific/N.py", line 6, in <module>
    from numpy.oldnumeric import *
ImportError: No module named oldnumeric
make[1]: *** [override_dh_auto_build] Error 1
debian/rules:27: recipe for target 'override_dh_auto_build' failed
make[1]: Leaving directory '/«PKGBUILDDIR»'
make: *** [build-arch] Error 2
"""

The issue affected all builds of the package except powerpc (no clue
why powerpc got free of it).  Anyway, a timely fix would be
appreciated so we can finish the netcdf transition without having to
remove python-scientific from testing.

Thanks,
~Niels