[Aboriginal] Darn signal bug fixed.

[Rob Landley]

For those of you who don't follow my blog: i fixed the signal bug that
was preventing the M4 build from completing in lfs-bootstrap.  On i686
it build up to udev (the next-to-last package), before dyning because
udev is a horrible pile of crap tied to specific kernel versions and the
udev out of LFS 6.7 won't build against linux 3.0 kernel headers.  (And
yet, it uses autoconf to test whether stuff specified by posix is
available in the environment.  See "horrible pile of crap", above.)

So I need to do an update to LFS 6.8.

Also, in case I haven't documented the decision anywhere: the reason
setup-chroot is in the aboriginal linux root filesystem instead of the
common control image code is that the problem it addresses is a problem
with the system images: the root filesystem is read only, and thus can't
be modified directly, hence the script to make a writeable copy of it
under /home suitable for chrooting into and modifying.

This is intentional, by the way: I used to make ext2 images, and
switched to squashfs because you don't have to fsck them after each run.
 Because they're read only, you can play with them all you like without
accidentally damaging anything or needing to clean something up.  You
can also launch three builds from the same /dev/hda (using different
/dev/hdb files for each qemu instance, of course).

(There's a balance between subtle nuances you miss and never see the
possibilities of, and forcing people to understand buckets of details
before they can do anything.  It's a lesser of two evils type situation:
never really thinking about why that bit's that way, vs HAVING to think
about why that bit's that way.)

Rob