Re: [pygtk] Installation problems win32

[Cedric Gustin <cedric gustin swing be>]

At 03:34 AM 6/17/2004, Joel Becker wrote:
> On Wed, Jun 16, 2004 at 02:14:11PM -0400, John Ehresman wrote:
> > I think this should continue to be an option, but I'd like to see a
> > distutils / py2exe type solution so the 1st 3 steps can be eliminated.
> > I'm not talking about eliminating pygtk / gtk distributions entirely,
> > but rather making it easy for applications to easily include and use a
> > private copy of the libraries.

you can py2exe your pygtk application but including the all GTK+ runtime 
with your application is overkill IMHO (the minimal distribution does not 
only include a few DLLs but also some configurations files (in the etc 
subdir) and locales.


>         Welcome to DLL Hell.  No Thanks.

This discussion is not limited to the Windows world

See the post by Havoc on his blog

http://log.ometer.com/2004-06.html#16

and the discussion on OSNews

http://207.44.212.20/comment.php?news_id=7399

Personally, I'm in favor of a GTK+ framework installer (à la .Net or Java) 
that only contains the GTK+ runtime as the basic component, with additional 
components (libraries, headers and development tools: libxml2, libglade, 
glade, gtkmm,...) available as extra packages (ideally through a net 
install package manager, something like the cygwin setup). Applications 
(gimp, gaim,...) would then each have their own installer that checks for 
the presence of this GTK+ framework. Same for pygtk, through its 
distutils-generated installer.

This of course implies that this GTK+ distribution is stable enough so that 
developers don't feel the need of embedding their own "patched" gtk+ 
runtime in their application installer. This is not a trivial request. For 
example, gtk+-2.4.3 is not yet available in binary form for win32 on Tor's 
page (http://www.gimp.org/~tml/gimp/win32/downloads.html). However, the 
gladewin32 project (http://gladewin32.sourceforge.net) has a all-in-one 
installer that claims that the "Gtk+ runtime is updated to 2.4.3 with 
unofficial bug fixes". Can we rely on this one and what does "unofficial 
bug fixes" mean ?

Cedric