[SciPy-user] import failure using g77
Pearu Peterson
pearu at cens.ioc.ee
Sat Oct 19 14:36:35 EDT 2002
On Sat, 19 Oct 2002, Jeffrey B. Layton wrote:
> Hey!
>
> I finally listened to Pearu and built SciPy with g77!
> However, I built g77 from gcc-3.1.1 that I installed
> locally in my home directory (this could be the cause
> of the problem).
(Yes, you might need to fix this using LD_LIBRARY_PATH, see below.)
> I built ATLAS, LAPACK, etc. with gcc-3.1.1.
> I then built/installed SciPy using the command:
>
> /usr/bin/python2.2 setup.py build build_flib --fcompiler=Gnu install
>
> (I had previously installed fftw, Numeric, and f2py).
> Everything good so far.
> So, I start up python2.2 and try to import scipy. Here's
> what I got:
>
> [laytonj at home3 scipy]$ /usr/bin/python2.2
> Python 2.2 (#1, Apr 12 2002, 15:29:57)
> [GCC 2.96 20000731 (Red Hat Linux 7.2 2.96-109)] on linux2
> Type "help", "copyright", "credits" or "license" for more information.
> >>> import scipy
> Traceback (most recent call last):
> File "<stdin>", line 1, in ?
> File "/usr/lib/python2.2/site-packages/scipy/__init__.py", line 30, in ?
> import Numeric
> File "/usr/lib/python2.2/site-packages/Numeric/Numeric.py", line 133, in ?
> import multiarray
> ImportError: libgcc_s.so.1: cannot open shared object file: No such file
> or directory
> >>>
>
>
> Could this be because I built SciPy with gcc-3.1.1 and
> python2.2 was built with gcc-2.96?
It shouldn't be. I am succesfully using extension modules built with
gcc-3.1.1 in python2.2 that is built with gcc-2.95.4.
> Or could this be a bum gcc build? Perhaps a problem with my library
> directories?
Could be. libgcc_s.so should be in /path/to/lib/gcc-lib/i386-linux/3.1.1.
If it exists there, a quick fix would be defining an environment variable
LD_LIBRARY_PATH, e.g.
export LD_LIBRARY_PATH=$LD_LIBRARY_PATH:/path/to/lib/gcc-lib/i386-linux/3.1.1
but read on.
Also note that the import problem above is not due to Scipy but comes from
Numeric import. I bet, direct 'import Numeric' will also fail. May be
rebuilding Numeric will fix the problem.
HTH,
Pearu
More information about the SciPy-User
mailing list