Re: what to do about dependencies?

[Tatsuro MATSUOKA]

Hello

--- On Sat, 2012/1/7, Jordi Gutiérrez Hermoso wrote:

> On 6 January 2012 11:36, John W. Eaton <address@hidden> wrote:
> > People often complain that building Octave is too complicated.
> 
> I don't know if building itself is complicated any more than it is for
> any other program using the GNU build system. I think the biggest
> complaint is that it takes too long. I wish we could improve this. I
> remember a couple of years ago, building Octave didn't take so long,
> and I don't think it takes longer now because our codebase is so
> large. Admittedly, it may be gcc's fault. The times I've tried
> building with clang, it seemed faster (but sadly, I couldn't succeed
> with a clang build). I seem to recall that libtool was a big culprit
> in slowing down the build, so perhaps this really can't be helped
> unless we improved libtool itself.
> 
> > For the future, I think we should consider including at least the
> > required dependencies (GNU Readline, PCRE, BLAS+LAPACK (ATLAS?)) and
> > all numerical library dependencies (ARPACK, FFTW3, GLPK, Qhull,
> > QRUPDATE, and SuiteSparse) with Octave.
> 
> I don't like this idea, if you mean having a giant tarball with all
> that extraneous stuff inside it. I wouldn't mind having a separate
> suggested dependencies tarball with everything in it, perhaps even
> with some sort of script that attempts to build and install everything
> inside that tarball... but I think the real problem is that most of
> our external dependencies aren't very well suited to being built
> outside of a Unix-like environment (I'm still thinking Windows is the
> biggest problem).

 (GNU Readline, PCRE, BLAS+LAPACK (ATLAS?)) 
(ARPACK, FFTW3, GLPK, Qhull,QRUPDATE, and SuiteSparse) 
Except fftw, building dll file is really tricky or require the local patch on 
windows.
I think that it might not be realistic solutions at least for the windows build.

Regards

Tatsuro