Re: runlevels

[Gordon Matzigkeit <gord@trick.fig.org>]

>>>>> Marcus Brinkmann writes:

 MB> If I understood everything correctly, and if the sysvinit program
 MB> is informed on each runlevel change, this solution is fine with
 MB> me. How hard is it to implement?

Somebody should take a look at the Linux sysvinit package, and see how
hard it would be to conditionalize all the terminal-handling (and
probably lastlog) parts.  If it looks easy, then talk to its upstream
maintainer, and see if they'll accept our changes as a configure
option.

Then the only remaining details are adding an option to shut down to
the rescue shell, and a few tiny porting fixes.

I made the steps in my previous mail too complex: you would't need to
change /hurd/init at all, because nobody else would be calling libc's
reboot() or functions directly, anyway.  You also don't need to make
telinit work by RPC... (I think) init just uses Unix signals for
communication.

I do think that we should keep sysvinit as /sbin/init, and rename
/hurd/init, so that people aren't confused by the two init programs
(accidentally running one instead of the other).  Thomas?

So as far as the actual effort needed to make this work, it should be
pretty minimal: sysvinit won't need any deep magic in order to work,
since it'll only be parsing inittab, running programs, and waiting for
signals.

-- 
 Gordon Matzigkeit <gord@fig.org>  //\ I'm a FIG (http://www.fig.org/)
Committed to freedom and diversity \// I use GNU (http://www.gnu.org/)