[BusyBox] LSB compliant sysvinit applet.

[David Seikel]

On Mon, 7 Mar 2005 09:16:11 -0500 Rob Landley <rob at landley.net> wrote:

> Applets in other directories?  (Are you talking the standalone shell
> code?)

What I'm talking about is the possible directories that busybox applets
can be "in", where the links end up after a busybox --install. 
Currently they are limited to /, /bin, /sbin, /usr/bin, and /usr/sbin. 
I have written the following applets that don't fit into this scheme -

/etc/init.d/rcS - from the default inittab

/usr/share/udhcpc/default.script - ifup calls udhcpc with the default
script.

/lib/lsb/init-functions
/usr/lib/lsb/install_initd
/usr/lib/lsb/remove_initd - LSB insists on these locations.

/etc/init.d/* - init "scripts" implemeted as busybox applets, again LSB
insists.

I currently use links and wrapper shell scripts to get these in the
right place, while keeping the applets themselves in /sbin and
/usr/sbin.  It would be a lot cleaner if busybox supported them
directly.  Probably too much for me to ask for though.

> As for hard wiring in a different inittab...  Have to think about the
> issues when I have more time...

<snip>

> The runlevel switching code is roughly the same logic as the full
> inittab re-reading code.  The tricky bit we don't currently do is
> shutting down the old processes we spawned.  (And what you really want
> is to preserve the processes that are still there at the same
> runlevel.  Don't kill and respawn _everything_ just what needs it.) 
> The logic to do that's not too hard, once you've thought through the
> problem.  Might want to make it a config option, though...

My applets take care of all the runlevel switching already.  Hopefully I
can dump the code someplace public today and you can all see what I am
talking about.