I've recently replaced my laptop's HDD as the S.M.A.R.T. data suggested a potential failure. I took the opportunity to try out beta4 of debian-installer and to install a pure wheezy system. Prior to this I've been running a mostly wheezy system with some bits from sid or experimental.

I've got quite a few notes about bugs I've hit in the installer, most of which I still need to file, but perhaps the most problematic appears to be known as #661789. Unless I'm reading this wrong, it would appear this show-stopper bug makes booting on an EFI system impossible without some workarounds, it's been a known problem for 9 months, and d-i beta4's images opt for EFI booting where possible. After the 8th install run I must admit I started getting angry that someone was wasting my time. Perhaps this is covered in some install notes somewhere. Anyway, I got an install working eventually (with legacy BIOS boot mode). and have been running wheezy for the last week.

I've just now had a string of problems which were a fairly typical experience for me running a mixed system. Whenever I hit a problem, particularly in the GNOME3 stack, my first response was to check "am I running the latest version of the packages?" — inevitably not — and so I got around my problems via a reboot or something similarly drastic; updated the packages and left things. Until the next time.

Just now, the following things happened:

  • I resumed the laptop and typed my password into GNOME's screensaver prompt, which then hung. I switched to a text console and tried to log in, but also hung, before timing out. I got in as root (glad I set up a root password) and had a look around, it seems kerberos made it's way into my PAM configuration and all users with uids ≥ 1000 were failing. I have libpam-krb5 installed but Kerberos unconfigured. I'm not sure why this has only bitten now. I'm also fairly sure kerberos is an optional PAM module and pam_unix should still work.

  • Killing gnome-screensaver gets my to my "desktop" but gnome-shell appears to have hung: it isn't drawing anything and all I can see is my wallpaper. Killing the shell process sadly doesn't cure this: I can see the (undecorated) edge of an open window now, but the shell isn't coming back. It seems the session manager doesn't respawn such subprocesses when they die.

  • Give up, restart gdm3, lose my open windows/data, log in afresh. For some reason, there's no networking panel applet visible. Wifi is not up. If I go to the GNOME control panel and visit "Networking", I'm told "The system network services are not compatible with this version". (They were earlier today, with no package changes…). Poking around in a terminal, I see the wifi controller is software disabled. I don't have rfkill installed and can't immediately see how to unblock it via poking around in /sys (probably possible) so I'm left with no choice but to reboot.

I've experienced all of the above (except for the PAM thing) more than once over the last 12 months or so since I blogged about moving to GNOME3. Now I'm running a pure wheezy system, at this point in the freeze the relevant packages are likely to be more stable. So, not running the latest version is no longer a reason to avoid fixing the problem. Time for me to start proper analysis!

Debian is now in deep freeze. I've heard it said a few times that we are coming to the end of that process. Based on my experiences, in my humble opinion, we've got some work to do before we're release worthy.


Comments

comment 1
I’ve a new gnome-shell version in store with a fix for many reported hangs. It should land in sid in a few days.
Comment by Np237
comment 2
I'm preparing a grub2 upload at the moment which should fix #661789. Sorry for the delay.
Comment by Colin Watson