[buildd-tools-devel] Bug#608054: Bug#608054: schroot: exiting session from one terminal kills others

[Julien Cristau]

On Wed, Dec 29, 2010 at 16:08:48 +0000, Roger Leigh wrote:

> On Sun, Dec 26, 2010 at 07:14:54PM +0100, Julien Cristau wrote:
> > Package: schroot
> > Version: 1.4.16-1
> > Severity: important
> 
> If reproducible, this should probably be severity serious.
> 
> > when i run schroot from two terminals, exiting one kills the other.
> > This seems to be a regression from earlier versions, and is bloody
> > annoying.
> 
> I've had another report of this occuring (not sure if it was from
> yourself), but I wasn't able to reproduce it at the time.
> 
I don't think it was from me.

> I assume this is due to the 15killprocs setup script?  If you
> add an "exit 0" to the top of this script, does it still occur?
> 
exit 0 at the top of 15killprocs fixes it.

> To better understand what's going on here, I really need some more
> information about why this is happening, i.e. which commands you
> run in each terminal and when.
> 
xterm 1:                xterm 2:
schroot
                        schroot
                        ^d
^d

schroot in xterm 2 stays suspended until I exit the chroot from xterm 1,
or, if I don't do that quickly enough, xterm 1 gets:
[1]+  Stopped                 schroot
$ fg
schroot
E: Child terminated by signal 'Killed'
$

> Note that if you're running commands in the /same session/ on both
> then ending the session will naturally kill all running processes
> in that session; this is intended and expected.  If you're running
> commands in /separate sessions/ and ending one kills the other,
> then this is most certainly a major bug.
> 
They're separate sessions, sorry.

> 15killprocs works by looking at /proc/"$pid"/root in each process
> under /proc, and it kills all those under the chroot mount location
> (which should be specific to a schroot session).  If I run schroot
> with "--vebose --debug=notice" I do see it try to kill the processes
> in the other session now, which is really strange.  The mount
> location is definitely different, but it's specifically killing
> processes in the other session.  For reasons I'm not yet sure about
> the SESSION_ID and CHROOT_MOUNT_LOCATION session names differ; it's
> possibly picking up the other session details, but I find that
> somewhat unbelievable.
> 
Cheers,
Julien
-------------- next part --------------
A non-text attachment was scrubbed...
Name: not available
Type: application/pgp-signature
Size: 836 bytes
Desc: Digital signature
URL: <http://lists.alioth.debian.org/pipermail/buildd-tools-devel/attachments/20101229/d0940016/attachment.pgp>