build problems (Re: [SciPy-dev] Stupid question.)

Prabhu Ramachandran prabhu at
Wed Feb 20 12:58:24 CST 2002

>>>>> "eric" == eric  <eric at> writes:

    >> I guess it serves me right to keep updating from CVS and
    >> trying. :(

    eric> This was added when the test suites were upgraded.  The CVS
    eric> version has the corrct files there.  I'm not sure why update
    eric> didn't synchronize your sandbox correctly.

I haven't done a clean build in a while.  So I cleaned out the build
dir and my older install.  I ran into this:

gcc -shared build/temp.linux-i686-2.1/fortranobject.o build/temp.linux-i686-2.1/flapackmodule.o -L/home/peterson/opt/lib/atlas -Lbuild/temp.linux-i686-2.1 -llapack -lcblas -lf77blas -latlas -lg2c -lc_misc -lcephes -lgist -o build/lib.linux-i686-2.1/scipy/linalg/
/usr/bin/ld: cannot find -llapack

Ummm. scipy_distutils/ has this :

library_path = ['/home/peterson/opt/lib/atlas']

    if sys.platform  == 'win32':
        if not library_path:
            atlas_library_dirs = library_path
        blas_libraries = ['f77blas', 'cblas', 'atlas', 'g2c']
        lapack_libraries = ['lapack'] + blas_libraries 
        if not library_path:
            atlas_library_dirs = unix_atlas_directory(sys.platform)
            atlas_library_dirs = library_path

This will clearly fail by default unless its Pearu's machine.  I
propose to change this so that it works.  I'll commit the changes to

There is another trivial bug in stats/ which I fixed.

Even after that the entire ga module fails to import the stats module
properly and gives me lots of problems.  I simply hacked those lines
out in my install to get things to build.

Rats! Someone modified everything just before I commited.  I'll just
delete my versions re-update and reinstall.  Sigh.

I think its a good idea if we keep each other informed of who is
currently working on what areas and if we are to keep of certain


More information about the Scipy-dev mailing list