On Sat, 2007-11-10 at 22:57 +0100, Matthijs Kooijman wrote:
> I'm currently working on patch busybox udhcpc to support --env, which seems
> like the sensible solution to that problem.
Why not use the patch I made for Gentoo?
http://sources.gentoo.org/viewcvs.py/gentoo-x86/net-misc/udhcp/files/udhcp-0.9.9_pre20041216-env.patch?view=markup
>
> > If you need some input on how to use runlevels from baselayout-2 with
> > init from bb, I have described my setup in the thread mentioned above.
> > Otherwise I have an absolute minimal system with only uclibc, busybox
> > and baselayout-2 (and of course the kernel :-) ), which is altogether
> > around 3MB.
> I did the same (only added dropbear) and ended up at around 8MB. Did you
> disable a lot of busybox options to get so small?
>
> > Concerning creation of dirs you mentioned. I never had the problem,
> > using standard runlevel startup (sysinit, boot and default) for booting.
> Yeah, that's what I see now as well.
>
> > On the other hand, once I tried to start one script inside of chroot
> > during building of the rootfs and then I got messages about missing
> > dirs, too. But since I do not normally start scripts during building,
> > this is not really a problem for me.
> It stopped being a problem for me too, but the rc program could be slightly
> modified to make it a bit more robust. I won't have the time to prepare a
> patch, though.
>
Attached is a patch to correct this. Ugly, but will do for the time
being.
Thanks
Roy
Index: src/runscript.c
===================================================================
--- src/runscript.c (revision 3090)
+++ src/runscript.c (working copy)
@@ -584,6 +584,9 @@
else if (state & RC_SERVICE_INACTIVE && ! background)
ewarnx ("WARNING: %s has already started, but is inactive", applet);
+ /* Ugly hack to ensure our neede dirs are created */
+ rc_deptree_update_needed ();
+
if (! rc_service_mark (service, RC_SERVICE_STARTING))
eerrorx ("ERROR: %s has been started by something else", applet);
