[SciPy-user] pynetcdf crashing on Python 2.5

[Robert Kern]

Mary Haley wrote:
> Robert, sorry, you're right, this isn't a NumPy issue, but it
> manifested itself when I tried to build NumPy. I built Python 2.5 (I
> thought successfully), and then when I tried to build NumPy 1.0.3 I
> got the error:
> 
>     ImportError: No module named _md5
> 
> I googled this left and right, and it turns out that when I built
> Python 2.5, some module (hashlib?) didn't get built, possibly because
> it's expecting some particular version of SSL, and I happen to have a
> different version. Python 2.5 is working for me, but I can't build
> NumPy because it depends on "_md5" which I believe is dependent
> on "hashlib".
> 
> I'm still trying to figure thins out. I would be great if I could 
> have NumPy bypass this somehow, because I don't think there's
> a fix for this on the Python 2.5 end.

Ah, okay. We do use the md5 module in part of the build process. See
numpy/core/code_generators/genapi.py . If you just need to get something built,
you can modify that file to use a different hash. You could write your own that
does something extremely simple like keep the strings in a dict mapping to
numbers that you increment.

However, your Python installation is broken. hashlib is a standard module that
should exist in all Python installations.

-- 
Robert Kern

"I have come to believe that the whole world is an enigma, a harmless enigma
 that is made terrible by our own mad attempt to interpret it as though it had
 an underlying truth."
  -- Umberto Eco