Since there is so much that's going to be new in Fedora 15 compared to
Fedora 14 (GNOME 3, GTK+-3, systemd, NetworkManager-0.9, and so on), the
Red Hat desktop team wanted to get an early start on running the new
stuff even before Fedora 14 is out.
So, I spent a few hours last night and this morning upgrading my primary
laptop from Fedora 13 to Rawhide. Generally worked better than I might
have expected... without too much work I got to a system that seems
basically usable.
One big caveat is that the result of a F14 => rawhide upgrade is a messy
hybrid -
$ rpm -qa | grep fc14 | wc -l
631
$ rpm -qa | grep fc15 | wc -l
701
It's not nearly a pure F15 system yet.
So, I thought I should dump out notes of the procedure I used and a few
snags I hit to help anybody else on the same path.
Upgrade procedure
=================
Step 1: Used preupgrade to upgrade from f13 to f14
Worked apparently perfectly, very slick experience
(Only real wart was that for encrypted root Anaconda has a
"Enter passphrase to decrypt" [ ] This is a global passphrase"
dialog, and the "global passphrase" thing is pretty hard to
understand. "Use this passphrase for all partitions" ? Or just
try it on the other partitions and ask me then if it
didn't work?)
Step 2: Installed fedora-release-rawhide, disabled f14, enabled rawhide
Step 3: Switched to a VT and ran 'yum update --skip-broken'
(I actually first tried 'telinit 3' and that didn't work right - the
GNOME session was left up, VT's were gone, and the system hung when
I then tried to reboot. Wonder if something got broken with the
upstart => systemd => upstart adventure>)
I had to remove a handful of various packages to get this to work,
but I think they were all local builds or stuff I installed from 3rd
party repositories.
Step 4: reboot
Things were mostly working at this point,
Fixing the result
=================
* I was left without a link from /usr/bin/emacs-<version>
to /usr/bin/emacs, so I had to create that manually.
* gtk3-engines didn't get installed on the upgrade, so I had
to install that manually.
* The GTK+ 3 module for ibus seems to be more or less broken.
I had to remove ibus to get gtk3 applications to run without
crashing. (Setting /desktop/gnome/interface/gtk-im-module
to gtk-im-context-simple might have work, but didn't
immediately seem to for some reason, and I needed to get to
a less crashy state.)
* Accessibility modules were producing a ton of bad looking spew
and dconf-editor had a gail/treeview related crash. I couldn't
figure out if there is a way of turning off accessibility
entirely through the new control center, so I turned off
/desktop/gnome/interface/accessibility. Strangely I had to
log out and log in before GTK+ apps stopped loading the
accessibility modules. g-s-d bug?
* All outgoing ssh was hanging - turned out that gnome-keyring-prompt
was hanging weirdly inside g_once_enter_init(); when I finally
saw stderr, it was obviously a need for a rebuild because of
struct size changes in gtk3. So, fired off a rebuild of
gnome-keyring in koji.
Also noted:
'killall gnome-settings-daemon' kills nm-applet, and leaves it in a
state where it doesn't recover even after restarting g-s-d, until
you log out and log back in. Weird.