Re: [OctDev] 2006.01.28 fails with octave 2.9.4

[<pki...@co...>]

 -------------- Original message ----------------------
From: Quentin Spencer <qsp...@ie...>
> Orion Poplawski wrote:
> 
> >
> > Get lots of:
> >
> > make[1]: Entering directory 
> > `/builddir/build/BUILD/octave-forge-2006.01.28/FIXES'
> > mkoctfile -DHAVE_OCTAVE_29 -v -DALLBITS -DHAVE_ND_ARRAYS rand.cc
> > rm -f randn.octlink
> > /builddir/build/BUILD/octave-forge-2006.01.28/admin/octlink.sh 
> > rand.oct randn.octlink
> > make[1]: 
> > /builddir/build/BUILD/octave-forge-2006.01.28/admin/octlink.sh: 
> > Command not foun
> > d
> > make[1]: *** [randn.octlink] Error 127
> > rm -f rande.octlink
> >
> > Looks like a change to configure.base was checked in to use 
> > octlink.sh, but that octlink.sh was never checked in?
> >
> 
> I got it to build fine with 2.1.72, but I'm seeing this error as well 
> with 2.9.x, and haven't looked any further into it. This error has been 
> reported a couple of times by others in the last week or so. Can someone 
> who understands what is going on explain this (or fix it)?

Sorry, my fault.

Octave 2.9.x has a new autoload feature to tell the interpreter to look
for multiple functions in the same .oct file.  I believe the syntax is
something like:

    autoload('full/path/to/octfile.oct','name');

On 2.1.72, octave-forge should be using "ln -s" as usual.  

On 2.9.4 it should be using "octlink.sh basename.oct name.octlink" 
to create 'name.octlink' which contains:

    autoload(which('basename'),'name');

Before install, mkpkgadd concatenates all the octlink files to the end
of PKG_ADD.

I don't have octlink.sh available remotely, and it will be a few days
yet before I'm home.  I think something like the following should
work.  Be sure to 'chmod a+x admin/octlink.sh' after creating yet.
Please let me know.

admin/octlink.sh:
#!/bin/sh

BASE=`echo $1 | sed -e's/.oct$//'`
NAME=`echo $2 | sed -e's/.octlink$//'`
echo "autoload(which('$BASE'),'$NAME')" > $2

    - Paul