Bug#775065: dpkg --configure -a results in "Hangup" after first package

[Guillem Jover]

Control: reassign -1 systemd
Control: severity -1 serious

Hi!

On Sun, 2015-01-11 at 09:03:34 +0000, Chris Carr wrote:
> On 11/01/2015 05:13, Guillem Jover wrote:
> >Please run «dpkg -D77777 --configure -a 2>&1 | tee dpkg.log», and attach
> >the output. We can check from there
> 
> It is attached, but it terminates before the output on the screen
> terminated. The last output on the screen was
> 
> D000001: process queue pkg libsane:i386 queue.len 56 progress 10, try 1
> D000040: checking dependencies of libsane:i386 (- <none>)
> D000400:   checking group
> [... a few more D000400:  lines ...]
> D000400:   checking possibility -> adduHangup

Ok then this looks like someone is either forcibly hanging up the
terminal or sending a SIGHUP to dpkg? Which indicates that whoever
is managing it is messing this up, my assumption is either systemd
or plymouth. I've reassigned to the former and also CCed the latter,
so they can investigate further, and reassign if necessary.

> >Hmm, if you are on an emergency shell, I guess the system is not
> >properly booted, are all partitions mounted? Otherwise there's no
> >wonder things are not being configured correctly.
> >
> >If that's the case, then this is probably a problem with your init
> >system or thereabouts.

> Yes I think that's exactly the case, though as you can see from the forum
> thread I dare not hypothesise that these problems have arisen because this
> update has introduced systemd.
> 
> My best guess (and I am not an expert by any means) is that there is some
> memory corruption or overflow in the init system which is polluting
> everything else, including dpkg. I do not understand what "task (plymouth)"
> is or why it keeps being blocked, and I do not know which process is sending
> "Hangup" to dpkg.
> 
> On mounting, the root filesystem mounts ok (/dev/dm-2) and contains all the
> linux directories including /usr, /home, /usr/local etc. etc. The
> filesystems which fail to mount are all Windows ones, which are obviously
> irrelevant at this stage.

Ok. Maybe as a wild guess, try booting w/o the splash kernel option if
you've got it enabled (see bug #602331)? If that fixes the issue then
it would need to be reassigned to plymouth.

Thanks,
Guillem