linaro-alip no gdm login panel after logout

[Tom Gall]

linaro-alip 930 snapshot, 930 hwpack for omap on a Rev C4 Beagle Board

After logout, gdm should put up a panel allowing one to login again. This does not happen. Rather a panel is displayed with a picture of a computer and the text linaro-m under it but nothing else.

If from the serial cons I create yet another account, then the panel will change and a list of accounts that can be selected for login will be displayed.

If you log out yet again then same problem occurs

[Tom Gall (tom-gall)]

I think I have this figured out. The PostSession script in /etc/gdm is slightly incorrect.

[Jamie Bennett (jamiebennett)]

Just merged a patch for this, please test again.

[Tom Gall (tom-gall)]

I've gotten to the bottom of this. This bug is just the tip of the iceberg of larger problems...

In tracking why this wasn't working I discovered that both /usr/lib/policykit-1/polkitd and /usr/sbin/console-kit-daemon where not running. This affects other services on the system in interesting ways, all of them bad.

I managed to track down via d-feet, /var/log/daemon.log and then via dbus-send --system --dest=org.freedesktop.ConsoleKit --print-reply /org/freedesktop/ConsoleKit/Manager org.freedesktop.ConsoleKit.Manager.CanRestart

that the issue is with /lib/dbus-1.0/dbus-daemon-launch-helper

The group id is not properly set to messagebus. From images found from the daily snapshot to ones that I've built locally via live helper I've seen values 106, 125 so there doesn't seem to be a set pattern.

Once the group is set to to messagebus and the binary is re-setuid, the system rights itself.

[Tom Gall (tom-gall)]

When looking at the linaro-m-alip tarball from 1007 unpacked into a directory

http://pastebin.ubuntu.com/508296/

so from that pastebin

107 is the messagebus GID in the tarball

106 is the messagebus GID on the system

and ls -lA lib/dbus-1.0/dbus-daemon-launch-helper clearly shows that the tarball GID for the binary is 106 so when the tarball is packaged it's wrong.

[Steve Langasek (vorlon)]

As discussed on IRC, the problem is with how we're unpacking the tarball - you have to use --numeric-owner when unpacking if you want the tarball to keep the numeric uids the same, otherwise the uids will be mapped to the system users of the same name. So this needs to be fixed in linaro-media-create - reassigning.

[Tom Gall (tom-gall)]

fixed in commit 135 on lp:~tom-gall/linaro-image-tools/fix-miss-tar

[Martin Pitt (pitti)]

Accepted linaro-image-tools into maverick-proposed, the package will build now and be available in a few hours. Please test and give feedback here. See https://wiki.ubuntu.com/Testing/EnableProposed for documentation how to enable and use -proposed. Thank you in advance!

[Launchpad Janitor (janitor)]

This bug was fixed in the package linaro-image-tools - 0.3

---------------
linaro-image-tools (0.3) maverick-proposed; urgency=low

  [ Steve Langasek ]
  * no longer using the 'omap' uboot flavor support in live-helper; instead,
    run the mkimage commands directly for omap.
  * fix the paths used for looking up x-loader and u-boot binaries in the
    tarball; these don't belong in /boot at all, pull them from /usr/lib.
  * only enforce the presence of commands we're actually using; this lets
    us automatically pull in btrfs-tools when btrfs is specified.
  * add sudo to a couple more commands in linaro-media-create, so the script
    itself can run as non-root as intended.

  [ James Westby ]
  * Add a linaro-hwpack-create script for the creation of hardware packs.

  [ Steve Langasek ]
  * Fix lookup of linaro-hwpack-install to work correctly when l-m-c is on
    the path. Closes: #652669.

  [ Loïc Minier ]
  * Drop spurious single quote character in boot_args_options for beagle|igep;
    thanks Peter Maydell; LP: #667400.
  * Set LC_ALL=C when entering the chroot as it has no locales. LP: #651905.

  [ Tom Gall ]
  * Use --numeric-owner argument to tar when unpacking, so that files get
    the right owner on boot of the image. LP: #652121.

  [ Dave Martin ]
  * Implement proper cleanup handling on signal / exit, so that we don't
    leave clutter behind on error. This also fixes the message at the end of
    a successful install about being unable to unmount proc. LP: #616385,
    #660196, #667394, #643601, #627383.
 -- Steve Langasek <email address hidden> Tue, 09 Nov 2010 12:50:26 -0800