[lxc-devel] [PATCH 0/2] Some small fixes post 0.9rc1 for 0.9final

[Serge Hallyn]

Quoting Christian Seiler (christian at iwakd.de):
> Hi there,
> 
> I found two little bugs in lxc-start and lxc-shutdown and provided some
> trivial patches that should go in after rc1.
> 
> lxc-start: If pinning the rootfs doesn't work (for example, if the
> rootfs doesn't exist and readlink() that is used to get the parent
> directory fails) it will not remove the network interfaces, causing
> subsequent starts to also fail.
> 
> lxc-shutdown: Probably due to set -e, lxc-shutdown with -t didn't work
> properly anymore, the attached patch fixes that.
> 
> (By the way, is there any reason why lxc-shutdown doesn't use lxc-wait
> and does this strange own timeout logic with a 1s polling loop? In case
> there is no real reason, I propose changing it after 0.9 final and use
> my bugfix for now, just to be sure not to introduce some strange
> regressions.)

Yes - because lxc-wait uses the monitor, meaning only one thing can
lxc-wait on a container at a time.  (This should be fixed at some
point).  This way at least lxc-shutdown can do its thing while the
impatient admin can still use lxc-wait.

-serge