[SciPy-user] how to import umfpack?

[Robert Cimrman]

Robin wrote:
> On Feb 12, 2008 10:15 AM, Robert Cimrman <cimrman3 at ntc.zcu.cz> wrote:
>> ldd -r produces (no _gfortran_* undefined symbols, only the usual Py*...):
>>          linux-gate.so.1 =>  (0xffffe000)
>>          libumfpack.so.0 => /usr/lib/libumfpack.so.0 (0xb7df6000)
>>          libamd.so.0 => /usr/lib/libamd.so.0 (0xb7de5000)
>>          libblas.so.0 => /usr/lib/libblas.so.0 (0xb7dc9000)
>>          libgfortran.so.1 =>
>> /usr/lib/gcc/i486-pc-linux-gnu/4.1.2/libgfortran.so.1 (0xb7d4e000)
>>          libm.so.6 => /lib/libm.so.6 (0xb7d27000)
>>          libgcc_s.so.1 =>
>> /usr/lib/gcc/i486-pc-linux-gnu/4.1.2/libgcc_s.so.1 (0xb7d1b000)
>>          libc.so.6 => /lib/libc.so.6 (0xb7bea000)
>>          libatlas.so.0 => /usr/lib/libatlas.so.0 (0xb781a000)
>>          /lib/ld-linux.so.2 (0x80000000)
>>
>>
>> Otherwise all you do seems fine, I have no other clue what is the problem.
> 
> So it seems there are a lot of dependencies missing from my __umfpack.so:
> robince at bob64:/usr/lib/python2.5/site-packages/scipy/splinalg/dsolve/umfpack$
> ldd __umfpack.so
>         libpthread.so.0 => /lib/libpthread.so.0 (0x00002b2cc544c000)
>         libc.so.6 => /lib/libc.so.6 (0x00002b2cc5667000)
>         /lib64/ld-linux-x86-64.so.2 (0x0000555555554000)
> 
> (no umfpack, amd, atlas or gfortran)
> And yet compilation and linking seemed to complete without errors. Is
> there any way I can get more detail about the build process of the
> umfpack module, perhaps try it by hand, or try to debug it further
> (more verbose output seeing all the commands run to build it etc.)?

well, with shared libraries you get complains only when a symbol is 
actually needed (i.e. when you import). Is it possible force linking the 
libgfortran somehow? I do not know enough about the scipy build system 
to give advice here.

r.