systemd for 11.10 ?

John Johansen john.johansen at canonical.com
Thu May 12 07:28:41 UTC 2011


On 05/11/2011 06:12 PM, Phillip Susi wrote:
> On 5/11/2011 10:44 AM, Chow Loong Jin wrote:
>> On the other hand, you can't possibly hope to convince anyone that a persistent
>> screen session requiring a specialized init task is a feature, not a bug.
> 
> It doesn't require a specialized anything.
> 
> A persistent ANYTHING when transitioning to runlevel s is a bug.  As for stopping gdm, it comes down to what does that action mean?  I've always thought of it has meaning stop gdm and everything that has resulted from it.  If you don't want that task to have that meaning, then I believe systemd leaves you the option of configuring it not to, but with upstart you have no choice: it CAN NOT make sure everything resulting from gdm has actually been stopped.
> 
it can't currently, it is possible to fix this, it is not in it self an argument to replace upstart

>> Let's also not forget old SysV-style /etc/init.d/* scripts that may have been
>> started from a graphical terminal, which will inevitably go down with
>> gdm/$display_manager based on what you're proposing. Or are we supposed to break
>> every init script not ported to systemd until the transition to systemd is
>> complete? This would have some serious repercussions on the syncability of
>> packages with init scripts from Debian.
> 
> In other words, any daemons that you manually start instead of using initctl.  I actually have always felt that the whole rc script system is fundamentally broken and the jobs should be listed in inittab so that init knows about them and can restart them if needed, rather than have some script fork off a process that init has no idea exists.
> 
whether you like it or not its what we currently have, and is the expected behavior

> They also would not be "broken" just because if you happen to run it to manually start a daemon, and if you stop gdm, and if systemd is configured to fully stop the whole process tree, then that daemon would also be stopped.
> 
actually I think not having multiple session in many instances is the broken behavior.





More information about the ubuntu-devel mailing list