[Pkg-scicomp-devel] freefem++ in Debian

[Drew Parsons]

On Tue, 2008-01-29 at 17:18 +0100, Christophe Prud'homme wrote:
> there are still some licensing issues that I haven't had time to look
> at/fix
> The package is ready otherwise.
> 


Thanks for your efforts in preparing it.  There's a small packaging
matter, it just needs a Build-Depends: automake-1.9.

While learning how to use it, I've met an error which doesn't make sense
to me.  Using only 23 grid points per box edge (for testing), I can get
a solution running.  But the moment I try to use a finer grid (even just
24 points per edge), I get the message 

UMFPACK V5.1.0 (May 31, 2007): WARNING: matrix is singular

umfpack_di_numeric failed
  current line = 150
Exec error : umfpack_di_numeric failed
   -- number :1
 catch an erreur in  solve  =>  set  sol = 0 !!!!!!! 
Exec error : umfpack_di_numeric failed
   -- number :1
 err code 7



It doesn't make sense that the matrix would suddenly go singular just by
switching from 23 to 24 grid points per side.  I thought it could be a
bug  in v2.20 that's since been fixed in 2.23, so I'm preparing a deb
from the latest upstream source.  I tested your initial packaging for
2.20 first, and found it specifies aclocal-1.9.  So you'll want to add a
Build-Depends: automake1.9 to cover that.  Does it really need
aclocal-1.9 or could you remove those version definitions from
debian/rules?

Thanks,
Drew


> On Jan 29, 2008 8:40 AM, Drew Parsons <dparsons at debian.org> wrote:
>         Hi Christophe,
>         
>         I notice the changelog for freefem++ (I got a deb from
>         http://people.debian.org/~prudhomm/debian/) says licence
>         problems are
>         now fixed, though it's still not in the Debian archive.
>         
>         Does that mean freefem++ is now ready to let into the archive
>         or do any
>         other problems remain?
>         
>         I'm keen to use it, and I've had some trouble with freefem
>         which I
>         imagine might have been fixed in freefem++.
>         
>         Thanks,
>         Drew
>         
>