[SciPy-dev] [matplotlib-devel] Preliminary patch for the (new) scipy

Travis Oliphant oliphant at ee.byu.edu
Mon Nov 21 13:00:11 CST 2005


Darren Dale wrote:

>On Monday 21 November 2005 1:12 pm, Travis Oliphant wrote:
>  
>
>>Darren Dale wrote:
>>    
>>
>>>Sorry, I've obviously overlooked something, but maybe I have also
>>>discovered a bug. I've just removed atlas/blas/lapack, updated scipy_core
>>>from svn, removed my build and site-packages/scipy directories, and tried
>>>to build scipy_core. Here is a warning message I get toward the end of
>>>the build and install processes:
>>>      
>>>
>>This is definitely a configuration issue.  For some reason, the system
>>is picking up BLAS information and therefore trying to compile _dotblas
>>for you.   However, it is failing because the libblas file is not found.
>>
>>Can you show us the output of the start of the build (where SYSTEM INFO
>>things are printed).  Or just attach a complete record of the build.
>>This might help track down what the configuration issue is.
>>    
>>
>
>This was due to an oversight on my end. I used gentoo's package manager to 
>completely remove atlas/blas/lapack, and afterwards verified that every 
>instance of /usr/lib/libatlas.* had been removed. I did not check that 
>libblas.* and liblapack.* had been removed, but I should have. Unfortunately, 
>a few broken soft links remained, which scipy found and tried to build 
>against. After removing those links, I was able to build scipy_core. How 
>embarrassing. I'm sorry for the noise. 
>  
>
Not a problem.  In fact, the system_info functions could be made more 
robust by actually trying to link a simple program against what it finds 
to make sure they are actually valid.  So, your example is not 
completely irrelevant.

-Travis




More information about the Scipy-dev mailing list