Re: Octave-3.6.0 testing binary built on MinGW

[Philip Nienhuis]

Ben Abbott wrote:
> On Jan 17, 2012, at 5:30 PM, Philip Nienhuis wrote:
>
> > Ben Abbott wrote:
> > > On Jan 16, 2012, at 5:13 PM, Philip Nienhuis wrote:
> > >
> > > > Hi Tatsuro:
> > > >
> > > > Tatsuro MATSUOKA wrote:
> > > >
> > > > > Hello
> > > > >
> > > > > I have built the octave-3.6.0 and archived the installed directory trees.
> > > > > http://www.tatsuromatsuoka.com/octave/Eng/Win/
> > > > >
> > > > > Please see Readme_octave-3.6.0.txt for details.
> > > > >
> > > > > Note that this release is prepared for the testing purpose for experienced 
> > > > > users.
> > > >
> > > > (shouldn't follow-up be directed to address@hidden, rather than address@hidden)
> > > >
> > > > In a spare hour I've made a preliminary "win32 distribution" using the steps 
> > > > outlined in the attached text, based on your build environment. In hindsight I already 
> > > > spotted some room for improvement in this procedure, but anyway this works too.
> > > > Maybe we can add this text to the wiki later on.
> > > >
> > > > Having done those steps, I'm able to install at least the java&   io packages 
> > > > (using -forge option). They work OK AFAICS.
> > > >
> > > > Installing the linear-algebra package doesn't work yet, due to a problem with gsvd.oct 
> > > > ("dggsvd_" and "zggsvd_ not found").
> > > > I had the same issues with 3.4.2; but Nitzan seemed to do better there.
> > > >
> > > > Anyway, I think your testing binary works good.
> > > >
> > > > (And this week I'll try to prepare a new io pkg (bug fix release)
> > > >
> > > > Philip
> > >
> > > Philip,
> > >
> > > I've been wanting some practice with the wiki, and started working on it a few 
> > > days ago.
> > >
> > > I hope you don't mind that I jumped in and created wiki page and populated it 
> > > with your notes.
> > >
> > >         http://www.octave.org/wiki/index.php?title=Installing_Win32_Distribution
> >
> > No problems, very good!
> >
> > I noted that you've put it under "Installing ..." (though I perceive my text rather as part of "building" 
> > instructions); while in the "Building" section instructions are given ("Octave for Windows") that to me look 
> > more like "installing".
> >
> > I lack time for sorting this out but hopefully someone will fix this 
> > inconsistency....
> >
> > Thanks,
> >
> > Philip
>
> I had thought you were describing how to install Octave for MinGW (using 
> Tatsuro's libs and such).
>
> Did I get that wrong ?

More or less, Ben. Sorry for being too terse :-)

Tatsuro supplied a compiled Octave-3.6.0/MinGW for Windows, consisting 
of just the Octave files (say, that what is installed if you enter "make 
install" after compiling). Just by itself, it can't be made to run.
To be able to run it, one needs either
(1) Tatsu's dependency libs + development environment (MinGW) installed 
in the proper place outside this Octave-3.6.0 directory,
   or
(2) embed the compiled Octave-3.6.0 in a "self-contained" MinGW/msys 
environment and add the dependency libs & support files etc. in place. 
Plus adapt the necessary configuration files. Plus do a lot more 
fiddling. All of this is actually a job for an Octave-MinGW maintainer.

The latter (2) is what my little note is about.

In contrast, under "Building" a description is given of how to merely 
unzip Nitzan's complete Octave "distros" (i.e., 7-zipped archives of 
what you would get after (2) above) into place and there's even 
Benjamin's old 3.2.4 installer (which merely requires a few mouse clicks 
to get a complete Octave installation).

Clear enough now?

P.