lfs-book/TODO

These are the TODO items for the next LFS release (3.0):

* Update binutils section in Appendix A:
  I have successfully compiled binutils, but the files created doesn't
  match the LFS documentation :
  I get 3 extra files : gasp, gprof and readelf. I don't get nlmconv

* Try to merge
  ft:06:respawn:/sbin/sulogin
    and
  su:S1:respawn:/sbin/sulogin
    into
  id:S016:respawn:/sbin/sulogin

* Consider ftp://ftp.win.tue.nl/pub/linux-local/utils/kbd/
  to replace console-tools/console-data

* Try new procps install:
	make OPT="$CFLAGS" INCDIRS='-I/usr/include -I.' XSCPT='' install	
* Since flex sits in ftp.gnu.org/non-gnu/flex it suggest the main
  download site for flex is elsewhere.

* When installing Glibc in chapter 6 glibc-build/login/pt_chown isn't
  installed (we'll copy manually for now) and the mtrace program isn't
  build. Find out why.

* Rewrite all of the mailinglist info (book and website) and give more
  information how you can set to things like digest, digest2, vacation
  and other useful modes.

* Incorporate LFS FAQ in the book

* Add dedicated package version number entities (to avoid having to make
  changes to 2-7 files to just change the version number of a package).

* Full dependency list. This list isn't a list of "package a depends on
  package b", but a list of "package a depends on b and c from
  package d"

* Bring the book up to speed with the current FHS specs.

* Mention security patches that can be applied to packages.

* At the end of the book suggest the user create the /etc/lfs-<version>
  file and give a few reasons why the user would want
  such a file (example: it's easy to forget after a while which LFS version
  you run; it makes debugging easier for us knowing which LFS version a user
  is running).

* Explain how changing runlevels works. Use the file written by Simon
  Perreault at http://download.linuxfromscratch.org/misc/runlevels.txt

* Don't run lilo from inside chroot anymore. This causes severe problems
  sometimes including LILO not bootstrapping properly.

* Bring back the PPC book (recreate from scratch using current intel
  book) using Jesse McCrosky's notes at
  http://download.linuxfromscratch.org/misc/ppc-notes.txt
  and the patch to the 'patch' package at:
  http://download.linuxfromscratch.org/misc/ppc-patch.patch

* Add descriptions what the patches do we use (like the console-tools and gzip
  patches)

* Try out Slackware's MAKEDEV script and see if it's any better than the
  currently used one. A copy can be found at
  http://download.linuxfromscratch.org/misc/Slackware-MAKEDEV

* Sed'ing the MAKEDEV script gives us hda[1-20], hdb[1-20] and so forth.
  Mention that all of these can't actually be used due to kernel
  limitations (IDE goes to 16 max for example).

* Type in chapter 2 - how to install software: Last paragraph: 
	s/There is on exception/There is one exception/

* Rewrite chapter 2 a little bit. Add that export LFS=/mnt/lfs could be
  added to root's .bash_profile/.bashrc (outside chroot) to make sure
  that between build sessions the variable is still set.

* Add to chapter 2 how to use .gz files (not just .tar.gz)

* Mention LFS should be installed as user root

* Single user mode has been reported not to work properly. Shutdown
  doesn't seem to unmount file systems and the shell doesn't come up
  properly when 'telinit S'.

* Add netkit-base and net-tools to Appendix A.

* Use /etc/HOSTNAME for setting the hostname. This is a bit more
  compliant with other distributions and therefore less confusing to
  people.

* Suggest using this command to strip debug symbols after you finish
  chapter 6: find $LFS -type f -exec strip --strip-debug '{}' ';'

* Modify the killproc function in the functions script. Right now it
  works roughly as follows: killl, wait 2 secs, check if pid's are gone.
  If not, kill with -KILL (aka -9), wait 2 secs, checkif pid's are gone.
  If no, print error message.

  Change this into: kill, don't wait 2 secs but check for pid's right
  away. If pid's are still there, then wait 2 seconds and kill -KILL,
  wait, if pid's still there, print error. Most daemons will exit
  immediately and the 2 second wait for every daemon slows things down
  unnessary.