On Mon, 7 Jun 2004 11:25:00 +0800, Sheldon Gill <[EMAIL PROTECTED]> wrote: 
> 
> I have a fix for GNUstep-base needing to source GNUstep.sh. It hasn't been
> committed to CVS yet but it seems to be coming. My solution also allows for a
> somewhat more flexible directory layout.
> 
> I added documentation a while ago to NSTimeZone regarding it's current
> behaviour so as to highlight issues for packagers.
> 

I would like to see this GNUstep.sh fix.  The current issue with
GNUstep.sh is reliably making sure that everybody's shell sources it,
which is a problem when Xterms and  many other terminals default to
not being login shells (and thus don't source /etc/profile).  If you
have a solid solution to this that would be wonderful :)

> 
> This is something being worked on, too. For the moment, if you want a global
> default on installation you've got to set it at compile time. It's the way it
> goes for now.
> 

Well I guess that will have to do for the time being.  I do think the
eventual solution will involve a global defaults domain, however. 
I've discussed this problem with a few people and a couple suggested
the idea of creating a default GNUstep directory in /etc/skel so that
new accounts automatically get a skeleton set of defaults; however,
changing them on a global level after the fact obviously is not
reflected across those accounts.

> > For my fellow packagers, I highly suggest you check out my tree of
> > PKGBUILDS.  You will see some of the hacks I have done to get things
> > packaged properly and hopefully will not have to duplicate my effort.
> 
> Could you write up the hacks you've done and why you've done them that way?
> It would make it easier for others to review, rather than having to check out
> the tree and go through the source to find out whats what. It would also give
> the why which is often not apparent from sources.
> 

I will, when I find the time, as that's not a bad idea.  The ones I
want to draw specific attention to, in any case, are the PKGBUILD and
install file for gnustep-base-cvs, where I use a really dirty hack to
get the documentation to build as part of the package and sort out the
timezone issue pretty effectively.  Some of the things I've done in
gworkspace-cvs are useful for showing how to fix broken packages which
insist on having parts of them already installed (I understand Enrico
has already addressed this issue so I will remove those changes in my
next release of gworkspace packages)

> It could serve as a good starting point for a Packager's Manual as well.

I'd like to see one of these, and I'm gonna talk to Gürkan and Alex
(Perez) about putting one together :)
 
> Regards,
> Sheldon
> 

It was great hearing your feedback.  Thanks for your time, and my regards,

Michael Baehr

Reply via email to