Re: [forge] linear-algebra next release

[Marco Atzeri]

On 6/2/2018 11:47 PM, PhilipNienhuis wrote:
> marco atzeri-2 wrote
> > Hi
> > as current version 2.2.2 does not build on Octave 4.4.0,
> > while repository code does, I uploaded the patches
> > to clean all build warnings and bump description and NEWS
> > to 2.2.3 at
> >
> > http://savannah.gnu.org/patch/?9646
> >
> > I have also prepared the source package and the
> > relative octave-generate_html documentation
> > using last 0.3.1 release.
> >
> > I will open a ticket for package release
> > if the patches are accepted.
> >
> > Regards
> > Marco
>
> Thanks for this work, Marco.
>
> Now that you mention, there's a pending patch report about some typos in the
> NEWS:
> https://savannah.gnu.org/patch/?func=detailitem&item_id=8643

Noted
> Another more important issue is that linear-algebra's binary modules
> gsvd.oct and pgmres.oct shadow functions in core Octave that are newer and
> better.

Only pgmres.oct is built, gsvd.oct not.
It seems the current makefile is broken anyway at least on my cygwin.
Likely after

changeset:   189:6e1dfc22132a
user:        Carnë Draug <address@hidden>
date:        Sat Jun 18 00:47:37 2016 +0100
summary:     Remove autotools cruft from the very old times of Octave Forge.

This also the reason why I do not see any warning of any
shadow functions

> For my own builds I simply delete the entire src/ subdir in the
> linear-algebra package and as far as I'm concerned that could be done for a
> new release as well; in that case linear-algebra would be an .m-file only
> package.
> But for the sake of compatibility with older Octave releases it would be
> nice to automatically detect these functions in core and if missing (older
> octave) add them using autotools, along the lines of deg2rad.m/rad2deg.m in
> the mapping package.

First approach could be better but pgmres.oct does not report any shadow
so a deeper function check is needed anyway.

I will look on that

> That solution (AFAIK invented by Carnë) does add complexity.
> Anyway I think it is up to you to decide.
>
> BTW yet another thing is that there's much better funm.m by one of (TTBOMK)
> Marco Caliari's students that is largely finished. That would be a (very)
> "nice to have" for the linear-algebra package.

any link ?

> Philip

Thanks
Marco