FC1 ghostscript update requires gdk-pixbuf and gtk+?

[Andreas Mueller]

Ralf Corsepius wrote:
> On Wed, 2004-08-18 at 14:34, Andreas Mueller wrote:
> > Ralf Corsepius wrote:
> > > On Wed, 2004-08-18 at 11:28, Tim Waugh wrote:
> > > > I think this is something I'm going to have to let the Fedora
> > > > Legacy project address if need be -- but I do wish that I
> > > > hadn't started down this path in the first place.
> > > >
> > > > Sorry for the mess.
> > >
> > > Which mess?
> > >
> > > ghostscript/FC2 requires gtk2 and gdk_pixbuf2, ghostscript/FC1
> > > requires their gtk1 counterparts, where is the problem?
> >
> > Think of a small text-only server with hylafax. Hylafax needs
> > ghostscript. And ghostscript pulls in urw-fonts, urw-fonts
> > needs /usr/X11R6/bin/mkfontscale (provided by XFree86-font-utils)
> > -> libXfont.so.1 (provided by XFree86-libs) -> freetype. This is
> > just *one* path of dependencies. The effect is that I don't have a
> > text-only server, now I have a lot of X stuff that I don't want.
>
> That's what you already had *before* this new rpm.

Correct, and I didn't like it before.

> The new rpm added gdk-pixbuf and gtk+. Yes, this adds some more
> packages and wastes more disk space, but ... is it really important?
> Most of the packages required by gdk-pixbuf/gtk+ already are required
> elsewhere, so, though it is not nice, this should not be an actual
> problem.

For me it is a problem. You can build a machine without a graphics card, 
but need to have XFree86-Mesa-libGL installed if you want to use 
ghostscript.

(@Michael)
There is a seperate package for gsx, called ghostscript-gtk, but the 
only file in this package is gsx itself. I see no problem that *this* 
package depends on gtk and gdk-pixbuf, but why ghostscript?

> >  And the next
> > logical step is to install gtk+ and gdk-pixbuf?
>
> No, definitely not - As I said above, it isn't nice.
>
> My point is elsewhere: ghostscript for FC2 already depends on gtk2
> (which comprises gdk-pixbuf-2).
>
> So if you consider the ghostscript update pulling-in gtk+ to be a
> packaging regression, then this regression had happened before the
> FC1/update package and also is present in FC2.
>
> => either there is a general packaging bug in both FC1/updates and
> FC2+, and packaging regression that needs to be addressed, or these
> dependencies are the nominal ghostscript dependencies you have got to
> learn to live with.

I don't think that it is a ghostscript dependency. At least the 
ghostscript ./configure can be instructed to not use X11 at all, but I 
didn't look at Red Hat's patches, the ghostscript.spec is really ugly.

I think I have to consider other distributions for text-only servers.

Regards,
Andreas