[SciPy-User] Scipy build failure

[David Cournapeau]

Johann Rohwer wrote:
> I've tried to send this previously with a 17kb logfile attached, but 
> the message doesn't seem to make it to the list. Here it is without 
> the attachment:
>
> I'm getting the following build failure in 
> scipy/sparse/linalg/dsolve/SuperLU:
>
> gcc: scipy/sparse/linalg/dsolve/SuperLU/SRC/util.c
> scipy/sparse/linalg/dsolve/SuperLU/SRC/util.c: In function 
> 'superlu_abort_and_exit':
> scipy/sparse/linalg/dsolve/SuperLU/SRC/util.c:31: error: format not a 
> string literal and no format arguments
> scipy/sparse/linalg/dsolve/SuperLU/SRC/util.c: In function 
> 'superlu_abort_and_exit':
> scipy/sparse/linalg/dsolve/SuperLU/SRC/util.c:31: error: format not a 
> string literal and no format arguments
> error: Command "gcc -pthread -fno-strict-aliasing -DNDEBUG -O2 -g -
> pipe -Wformat -Werror=format-security -Wp,-D_FORTIFY_SOURCE=2 -
> fexceptions -fstack-protector --param=ssp-buffer-size=4 -g -fPIC -
> DUSE_VENDOR_BLAS=1 -I/usr/lib64/python2.6/site-
> packages/numpy/core/include -c 
> scipy/sparse/linalg/dsolve/SuperLU/SRC/util.c -o build/temp.linux-
> x86_64-2.6/scipy/sparse/linalg/dsolve/SuperLU/SRC/util.o" failed with 
> exit status 1
>   

The problem is most likely the flags set up by your distribution (I
guess this is RHEL from the security enhanced flags), or more exactly,
the flags show dubious code in superlu.

I applied a quick fix that seems to work when I use your flags - can you
check it up ? (it is r5892).

cheers,

David