Re: porting arapck code to octave

[David Bateman]

David Bateman wrote:
> John W. Eaton wrote:
> > | Why it doesn't work for you is a good question. Maybe there's a buffer
> > | overrun that in my case doesn't do any damage by chance? (Just a wild
> > | guess).
> >
> > Both programs seem to work correctly for me.  I also thought it might
> > be a buffer overrun that was only showing up on some systems (32- vs
> > 64-bit pointers, for example) so I tried it on 32- and 64-bit sytems
> > and it worked on both.  I also tried running it with valgrind and
> > there were no complaints for the call to dssimp.  I'm using Octave
> > 3.0.1 (64-bit system) and Octave 3.0.0 (32-bit system) on a Debian
> > system with arpack from the Debian package.
> >   
> ok, rinse and repeat...
>
> I retried with various compilation and linking options and firstly it 
> appears that if I link to the static library everything works fine. 
> However if I have FFLAGS of "-fPIC -O2" and link arpack with
>
> g++ -shared -Wl,-soname -Wl,libarpack.so -o libarpack.so  SRC/*.o 
> UTIL/*.o -llapack -lblas
>
> as I would have thought was necessary, I have problems.. In fact I 
> can't seem to find any shared library version that I build that does 
> work....
>
> D.

Ok, it seems that if I use the version of arpack from

http://mathema.tician.de/dl/software/arpack-autotools

that contains a few bug fixes and a build with libtool then I can get 
eigs to work correctly... The reason I was building ARPACK myself in the 
first place was that the version that seems to bee distributed with 
debian isn't built with gfortran  with the version in testing.. Will 
this still be the case in future versions of arpack on debian?

Regards
David

--
David Bateman                                address@hidden
35 rue Gambetta                              +33 1 46 04 02 18 (Home)
92100 Boulogne-Billancourt FRANCE            +33 6 72 01 06 33 (Mob)