[SciPy-dev] Building numpy/scipy on RHEL5 OR CentOS5

Michael Hearne mhearne@usgs....
Wed Jun 25 08:34:05 CDT 2008

My apologies to everyone - I figured out the problem.  I was performing 
the build procedure as myself, after setting UMFPACK="None", but using 
sudo to install.  It didn't occur to me until just now that running as 
root meant that my environment variable wouldn't be visible in the root 
shell.  Duh.

That being said, can I make a suggestion that setup.py for scipy take an 
option that can turn UMFPACK off, instead of the user having to remember 
to do it in the shell, and causing him or herself the same sort of 
confusion I just caused?

It would also be great to have a less convoluted build procedure for 
BLAS/LAPACK, but I'll take one thing at a time.

Also, since Phil seems not to have to set this variable, is there 
something I can provide developers that would help them understand why 
my RHEL5 system seems to require that I set it?

Thanks for everyone's help,

Mike Hearne

Phil Austin wrote:
> Michael Hearne wrote:
>> Phil et. al. - I believe that you and probably others do not have 
>> this problem, but I continue to have it.  I just downloaded the most 
>> recent svn version of scipy and re-attempted the build and install.  
>> The build was successful, but the install complained about ufmpack.
> Just to be sure I understand -- when you say the build was successful,
> does that mean that during the build you see:
>>> UserWarning:
>>>     UMFPACK sparse solver 
>>> (http://www.cise.ufl.edu/research/sparse/umfpack/)
>>>     not found. Directories to search for the libraries can be 
>>> specified in the
>>>     numpy/distutils/site.cfg file (section [umfpack]) or by setting
>>>     the UMFPACK environment variable.
>>>   warnings.warn(self.notfounderror.__doc__)

Michael Hearne
(303) 273-8620
USGS National Earthquake Information Center
1711 Illinois St. Golden CO 80401
Senior Software Engineer
Synergetics, Inc.

More information about the Scipy-dev mailing list