Re: eigs and ARPACK

[Ben Abbott]

On Jan 3, 2009, at 2:39 PM, Thomas Treichl <address@hidden>  
wrote:

> David Bateman schrieb:
> > Thomas Treichl wrote:
> > > That's it right now, I hope I could give some feedback...
> > >
> > > Best regards,
> > >
> > >  Thomas
> > Thomas thanks for the report, yes it did help even if I'm late in  
> > replying. I hope to have a patch soon, if I can figure out my  
> > linking issue.
>
> Well, I'm even more slow ;) The ARPACK beast made some problems on  
> my Mac but finally I found out that all the problems disappear if I  
> just use g95 to create libarpack.dylib (instead of f2c). The usage  
> of g95 for libarpack.dylib now makes me thinking about some more  
> other things...
>
> At the moment I still have the problem that linking of  
> liboctave.dylib fails because of undefined symbols:
>
>  ld: Undefined symbols:
>  _dnaupd_
>  _dneupd_
>  _dsaupd_
>  _dseupd_
>  _znaupd_
>  _zneupd_
>  /usr/libexec/gcc/i686-apple-darwin8/4.0.1/libtool:
>    internal link edit command failed
>
> that's why I added "$(ARPACK_LIBS)" to "LINK_DEPS" of "Makefile.in"  
> in the directory of liboctave. I cheated, and also compiled the  
> latest Octave sources on my virtual Debian GNU/Linux box. That  
> system doesn't need to know anything about $(ARPACK_LIBS) for  
> liboctave.so but just compiles fine without that information - can  
> you help me to understand where GNU/Linux gets it's information  
> about these functions from? Is my hack the right thing that should  
> be done or am I missing something?
>
> Best regards,
>
>  Thomas

Any chance you tried gfortran?

Ben