Re: upgrades must not change the installed init system [was: Re: Cinnamon environment now available in testing]
On 09/09/2014 17:01, Russ Allbery wrote:
> Vincent Danjean <vdanjean.ml@free.fr> writes:
>> I agree with your analysis. However, how do you think we can ask the
>> user ? We can have a debconf question. However, whatever the answer is,
>> we must not return an error (i.e. aborting the upgrade). It is really a
>> pain to recover when this occurs.
>
> The original plan was to have the question owned by some package that
> could then switch the init symlink from one implementation to another.
> That way, no abort is required. I'm not sure if that survived contact
> with reality, though, in the sense that I'm not sure how implementable it
> is.
The different init systems would have to be co-installable for this to
work, however they are not.
For the planned fallback in grub to offer sysvinit on systemd systems, a
second copy of sysvinit's /sbin/init was included in a non-conflicting
package in a different location, but systemd's implementation of halt,
poweroff, ... would still be used (as these work with sysvinit's init as
well).
Ansgar
Reply to: