Bug#512327: upgrade-reports: mostly successful etch->lenny upgrade
Package: upgrade-reports
Severity: normal
Hello,
I have recently performed an upgrade from an etch[1] machine.
Here are the results of, and comments about that upgrade:
The upgrade went, generally, pretty smooth, although there were some
hiccups. The system rebooted (in the end) properly in lenny with the
etch kernel and I haven't upgraded yet the kernel.
Problems:
1) some icons went missing from the panel of my gnome desktop - gq and
old gaim icon - I know both of these packages are not part of lenny
2) both / and /home filesystems were found containing errors(?) after
the reboot into the new lenny system (same kernel). Both were ext3 on
top of LVM2 devices (probably LVM is irrelevant)
3) aptitude kept on:
a) blocking after finishing operations with dpkg (apparently when
writing extended information) - merged bugs 479438, 504063,
511708 (the last one was my report)
b) forgetting recommends which I didn't want (probably as a
consequence of the previous issue)
4) release-notes recommends MODULES=dep when using lilo while
update-initramfs didn't work with dep and worked only with
MODULES=most - this could lead to problems and unbootable systems
after upgrade
5) debian menu is gone
6) zim regression - doesn't minimize to tray any more - a plug-in
needs to be activated
7) during the boot, now there are some errors saying that dmidecode is
not present in the system. These come from the unconditional run of
some dmidecode commands in hotkey-setup (which is no longer
installed).
I don't consider 1) such a big problem, especially since the packages
were removed from the archive and gq had the "?" icon form etch, while
gaim was using some icon which was part of the old gaim package which
was replaced with a fake one (according to the version naming).
2) kind of worried me and lead to some mandatory manual interaction from
the user after reboot (for the / partition check).
There were probably 128 complaints (I made a picture with my phone) of
this form:
/dev/mapper/<name>: Group descriptor N marked uninitialized without
feature set.
N was (probably) from 1 to 127, or so it looks like in the pictures I
took.
I am worried that such checks and manual interaction might break remote
upgrades.
So to recap, after the "aptitude dist-upgrade" was completed, I
rebooted. The / filesystem (ext3 on top of LVM2) was found to contain
those issues, I think it required the root password (but I am not 100%
sure) and a check. After that step was completed, an automatic reboot
occurred. At this reboot, the /home partition was found to have the same
problem at which point I managed to take a picture of the fsck process.
3) a) and b) are probably related, although I am unsure. a) is dealt
with in another BR, while b) is probably either related to a) occurring
at a bad time, or some new bug "keeps bringing back a refused recommended
package". I am unsure what to do about b).
4) Since I have lilo (because I use LVM and I am not using grub2), I
followed the recommendation to: use MODULES=dep in
/etc/initramfs-tools/initramfs.conf, but later, during the upgrade
mkinitramfs complained about this and refused to upgrade forcing me to
revert back to MODULES=most and risking an unbootable system:
..
update-initramfs: Generating /boot/initrd.img-2.6.18-6-686
mkinitramfs: missing hda1 root /dev/mapper/hdapool-rootfilesystem /sys entry
mkinitramfs: workaround is MODULES=most
mkinitramfs: Error please report the bug
update-initramfs: failed for /boot/initrd.img-2.6.18-6-686
..
I tried reverting to the MODULES=dep after reboot (luckily, it worked
with etch's kernel), but I got the same error.
This is a real problem for the upgrade, since, unable to apply the lilo
workaround, the system might become unbootable after upgrade.
5) Since there are some applications which are NOT accessible via
GNOME's menu, but were accessible through the Debian menu, I consider
this issue an important and problematic regression.
6) The default behaviour changed from etch to lenny and now defaults to
closing when pressing the X button, while in etch it minimized to tray.
7) the init script of hotkey-setup starts like this:
twix:~/etch-lenny# head /etc/init.d/hotkey-setup
#!/bin/bash
manufacturer=`dmidecode --string system-manufacturer`
name=`dmidecode --string system-product-name`
version=`dmidecode --string system-version`
SAVED_STATE=/var/run/hotkey-setup
THINKPAD_LOCKFILE=$SAVED_STATE.thinkpad-keys
# This is here because it needs to be executed both if we have a
It is obvious why this error occurs, since dmidecode is unconditionally
ran.
[1] there were some official backports and some local ones.
-- System Information:
Debian Release: 5.0
APT prefers testing
APT policy: (500, 'testing')
Architecture: i386 (i686)
Kernel: Linux 2.6.18-6-686 (SMP w/2 CPU cores)
Locale: LANG=ro_RO.UTF-8, LC_CTYPE=ro_RO.UTF-8 (charmap=UTF-8)
Shell: /bin/sh linked to /bin/bash
Reply to: