Configuration of upstart package triggers reboot

[f3d]

Binary package hint: upstart

Description: Ubuntu karmic (development branch)
Release: 9.10
Upstart version : 0.6.3-7

I've used a karmic alternative CD (guess it was the last alpha) to try it out on a old laptop. I can't complete installation of the packages, since everytime upstart should be configured a reboot is triggered.

Message is :

Can not write log, openpty() failed (/dev/pts not mounted ?)
Setting up upstart (0.6.3-7) ...
Terminated.

[f3d (f3d)]

There is probably a problem with the kernel or the udev version on the CD.

I was able to finish the installation by booting with a Sysrcd CD I had around.

[Philipp Kern (pkern)]

This was no fun when I tried to recover from cryptsetup breakage yesterday. The telinit in upstart's postinst causes the machine to reboot under Ubuntu's debian-installer, when run from its rescue prompt. So either ubuntu-installer should guard itself against this usage of "telinit u" to re-exec init or upstart should not invoke it unconditionally. I never experienced reboots during dpkg's --configure run before.

[Scott James Remnant (Canonical) (canonical-scott)]

I'm not sure why "telinit u" would trigger a reboot, unless the system actually crashed.

But I did have an idea that we should do that right before remounting the root filesystem, rather than on upgrade - that way we wouldn't lose the state

[Launchpad Janitor (janitor)]

This bug was fixed in the package upstart - 0.6.3-8

---------------
upstart (0.6.3-8) karmic; urgency=low

  * Rather than calling "telinit u" after upgrade, which will lose state,
    have the umountroot initscript take care of it for us by setting a
    flag. LP: #441796.
  * Don't lose the original default runlevel if /etc/inittab exists without
    an initdefault line. LP: #405847.
  * Fix "unhandled error" in shutdown when unable to change runlevel,
    e.g. due to previous Ubiquity bug. LP: #426332.
  * Merge change from trunk that makes it possible to build Upstart using
    a previously built copy of nih-dbus-tool, especially useful when
    cross-compiling. LP: #426740.
  * Merge change from libnih to store our assertion messages in the
    glibc __abort_msg symbol so apport can pick them up. LP: #429411.
  * Merge change from libnih to fix compilation issue with eglibc due
    to changed alphasort() prototype.

 -- Scott James Remnant <email address hidden> Wed, 14 Oct 2009 05:34:13 +0100