I've fixed the gtk2 wrt to being parallel-installable on biarch systems like ppc/ppc64. The changes involved moving the arch-dependent config files to another directory and adding -32/-64 suffixes to the installed binaries. This means in particular that packages which install input modules or pixbuf loaders need changes to their postinstall and postuninstall hooks. To make this easier, the gtk2 package now installs two scripts, update-gdk-pixbuf-loaders and update-gtk-immodules, which expect to be given the host triplet and figure out the binary to call and the config file location based on that.
So, if one of your packages installs an input module or pixbuf loader, please update the post hooks accordingly. You can look at librsvg2 or im-sdk for examples.
Regards, Matthias
On Fri, 25 Jun 2004 23:14:41 -0400, Matthias Clasen mclasen@redhat.com wrote:
So, if one of your packages installs an input module or pixbuf loader, please update the post hooks accordingly. You can look at librsvg2 or im-sdk for examples.
Have there been any changes to the pkg-config scripts?
--Chris
Am Sa, den 26.06.2004 schrieb Chris Blizzard um 09:25:
On Fri, 25 Jun 2004 23:14:41 -0400, Matthias Clasen mclasen@redhat.com wrote:
So, if one of your packages installs an input module or pixbuf loader, please update the post hooks accordingly. You can look at librsvg2 or im-sdk for examples.
Have there been any changes to the pkg-config scripts?
--Chris
No
Matthias,
Hmm, seems to be a problem for me here...
gtk2 provides /etc/gtk-2.0/i386-redhat-linux-gnu
update-gtk-immodules %{_host} when run from the %post sees that my machine is an i686, and attempts to do stuff in /etc/gtk-2.0/i686- redhat-linux-gnu which doesn't exist because gtk2 doesn't provide it since gtk2 is an i386 package. So we need to use %{arch} and some other variables rather than %{_host} to make sure we match the gtk2 RPM architecture.
Dan
On Sat, 2004-06-26 at 11:24 -0400, Matthias Clasen wrote:
Am Sa, den 26.06.2004 schrieb Chris Blizzard um 09:25:
On Fri, 25 Jun 2004 23:14:41 -0400, Matthias Clasen mclasen@redhat.com wrote:
So, if one of your packages installs an input module or pixbuf loader, please update the post hooks accordingly. You can look at librsvg2 or im-sdk for examples.
Have there been any changes to the pkg-config scripts?
--Chris
No
Using:
%{_bindir}/update-gtk-immodules %{_arch}-%{_host_vendor}-%{_host_os}
where previously:
%{_bindir}/update-gtk-immodules %{_host}
was used seems to work for gnome-utils at least, and seems somewhat more "correct" WRT what gtk2-2.4.1-3 does. Perhaps im-sdk and librsvg2 need to be updated in the same fashion as well?
Dan
On Sat, 2004-06-26 at 11:24 -0400, Matthias Clasen wrote:
Am Sa, den 26.06.2004 schrieb Chris Blizzard um 09:25:
On Fri, 25 Jun 2004 23:14:41 -0400, Matthias Clasen mclasen@redhat.com wrote:
So, if one of your packages installs an input module or pixbuf loader, please update the post hooks accordingly. You can look at librsvg2 or im-sdk for examples.
Have there been any changes to the pkg-config scripts?
--Chris
No
On Fri, 2004-06-25 at 21:14, Matthias Clasen wrote:
I've fixed the gtk2 wrt to being parallel-installable on biarch systems like ppc/ppc64. The changes involved moving the arch-dependent config files to another directory and adding -32/-64 suffixes to the installed binaries. This means in particular that packages which install input modules or pixbuf loaders need changes to their postinstall and postuninstall hooks. To make this easier, the gtk2 package now installs two scripts, update-gdk-pixbuf-loaders and update-gtk-immodules, which expect to be given the host triplet and figure out the binary to call and the config file location based on that.
So, if one of your packages installs an input module or pixbuf loader, please update the post hooks accordingly. You can look at librsvg2 or im-sdk for examples.
On a side note so that people at RH doesnt think I just grouch. Thankyou for letting people/developers know about the change.
Regards, Matthias
On Fri, 25 Jun 2004 23:14:41 -0400, Matthias Clasen wrote:
This means in particular that packages which install input modules or pixbuf loaders need changes to their postinstall and postuninstall hooks. To make this easier, the gtk2 package now installs two scripts, update-gdk-pixbuf-loaders and update-gtk-immodules, which expect to be given the host triplet and figure out the binary to call and the config file location based on that.
Shouldn't these changes be made to GTK upstream?
thanks -mike
On Mon, 2004-06-28 at 19:25, Mike Hearn wrote:
On Fri, 25 Jun 2004 23:14:41 -0400, Matthias Clasen wrote:
This means in particular that packages which install input modules or pixbuf loaders need changes to their postinstall and postuninstall hooks. To make this easier, the gtk2 package now installs two scripts, update-gdk-pixbuf-loaders and update-gtk-immodules, which expect to be given the host triplet and figure out the binary to call and the config file location based on that.
Shouldn't these changes be made to GTK upstream?
This is a big old hack not acceptable upstream.
Havoc
devel@lists.stg.fedoraproject.org
-
Chris Blizzard -
Dan Williams -
Havoc Pennington -
Matthias Clasen -
Mike Hearn -
Stephen Smoogen