[Pkg-scicomp-devel] suitesparse -- create links to /usr/include?

Christophe Prud'homme prudhomm at debian.org
Sun Sep 7 13:51:57 UTC 2008


Ondrej,

we had a discussion with the author of suitesparse, the author of the
suitesparse interface in Octave and the person in charge of the fedora
octave/suitesparse package and we agreed in using /usr/include/suitesparse
as the place of the headers.
With the compiler if you use -I/usr/include/suitesparse in the compiler
command line then you don't have to change your code, just add
/usr/include/suitesparse as a possible location for the headers.
your solution may not work if you have a umfpack header including other
umfpack headers.

what is the problem with that ? I do that for my codes, it works also great
with petsc and trilinos.

Best regards
C.

On Sun, Sep 7, 2008 at 3:14 PM, Ondrej Certik <ondrej at certik.cz> wrote:

> Hi,
>
> Debian is different to other distributions in that umfpack is not in
> /usr/include, but in /usr/include/suitesparse.
> How about creating symlinks to /usr/include ? So that we don't have to
> patch all software that uses umfpack, see for example the patch below
> that is applied in python-scipy. Basically everytime I need to get
> some program using umfpack to work in Debian, I need to fix this.
>
> Is there some convention where include headers should live? For
> example Python headers are in /usr/include/python2.5 too, but every
> project is used to that. Umfpack dependents doesn't seem to be used to
> find umfpack in /usr/include/suitesparse.
>
> Ondrej
>
> Index: python-scipy-0.6.0/scipy/linsolve/umfpack/umfpack.i
> ===================================================================
> --- python-scipy-0.6.0.orig/scipy/linsolve/umfpack/umfpack.i    2007-12-03
> 15:20:37.827719117 +0100
> +++ python-scipy-0.6.0/scipy/linsolve/umfpack/umfpack.i 2007-12-03
> 15:20:51.328488481 +0100
> @@ -10,13 +10,13 @@
>  */
>
>  %{
> -#include <umfpack.h>
> +#include <suitesparse/umfpack.h>
>  #include "numpy/arrayobject.h"
>  %}
>
>  %feature("autodoc", "1");
>
> -#include <umfpack.h>
> +#include <suitesparse/umfpack.h>
>
>  %init %{
>     import_array();
> @@ -189,18 +189,18 @@
>     double Info [ANY]
>  };
>
> -%include <umfpack.h>
> -%include <umfpack_solve.h>
> -%include <umfpack_defaults.h>
> -%include <umfpack_triplet_to_col.h>
> -%include <umfpack_col_to_triplet.h>
> -%include <umfpack_transpose.h>
> -%include <umfpack_scale.h>
> -
> -%include <umfpack_report_symbolic.h>
> -%include <umfpack_report_numeric.h>
> -%include <umfpack_report_info.h>
> -%include <umfpack_report_control.h>
> +%include <suitesparse/umfpack.h>
> +%include <suitesparse/umfpack_solve.h>
> +%include <suitesparse/umfpack_defaults.h>
> +%include <suitesparse/umfpack_triplet_to_col.h>
> +%include <suitesparse/umfpack_col_to_triplet.h>
> +%include <suitesparse/umfpack_transpose.h>
> +%include <suitesparse/umfpack_scale.h>
> +
> +%include <suitesparse/umfpack_report_symbolic.h>
> +%include <suitesparse/umfpack_report_numeric.h>
> +%include <suitesparse/umfpack_report_info.h>
> +%include <suitesparse/umfpack_report_control.h>
>
>  /*
>   The order is important below!
> @@ -212,8 +212,8 @@
>     void **Numeric
>  }
>
> -%include <umfpack_symbolic.h>
> -%include <umfpack_numeric.h>
> +%include <suitesparse/umfpack_symbolic.h>
> +%include <suitesparse/umfpack_numeric.h>
>
>
>  OPAQUE_ARGINOUT( void * )
> @@ -222,8 +222,8 @@
>     void **Numeric
>  }
>
> -%include <umfpack_free_symbolic.h>
> -%include <umfpack_free_numeric.h>
> +%include <suitesparse/umfpack_free_symbolic.h>
> +%include <suitesparse/umfpack_free_numeric.h>
>
>
>
> @@ -245,7 +245,7 @@
>     long *n_col,
>     long *nz_udiag
>  };
> -%include <umfpack_get_lunz.h>
> +%include <suitesparse/umfpack_get_lunz.h>
>
>
>  ARRAY_IN( double, double, DOUBLE )
> @@ -269,6 +269,6 @@
>     int Q [ ]
>  };
>  %apply int  *OUTPUT { int *do_recip};
> -%include <umfpack_get_numeric.h>
> +%include <suitesparse/umfpack_get_numeric.h>
>
>  #endif
>
> _______________________________________________
> Pkg-scicomp-devel mailing list
> Pkg-scicomp-devel at lists.alioth.debian.org
> http://lists.alioth.debian.org/mailman/listinfo/pkg-scicomp-devel
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://lists.alioth.debian.org/pipermail/pkg-scicomp-devel/attachments/20080907/b9df4647/attachment.htm 


More information about the Pkg-scicomp-devel mailing list