LiveCD or No LiveCD?

Alexander E. Patrakov patrakov at
Mon Feb 25 09:49:21 PST 2008

Jeremy Huntwork wrote:
> I wonder if we could adapt jhalfs to do the work for us?

Jhalfs works well only when there are only a few (ideally, zero) deviations from 
the book. In the current CD, there are the following deviations in the LFS part 
(the list is probably incomplete):

* is used to set global optimization flags, the architecture, to 
remove the obsolete /usr/man directory, and to suppress static libraries in 
order to save space.

* some more perl modules are built in Chapter 5, to support Markdown, which is a 
perl script and must be run outside of the chroot (and thus has its interpreter 
as /tools/bin/perl). For the same reason, zlib, cdrtools, and zisofstools are 
built at the end of Chapter 5.

* gcc pass1 and binutils pass1 have to be compiled without cflags set for 
everything else in, because the host compiler may be too old for 
understanding advanced CFLAGS.

* some other switches (no-ops for 32-bit x86) are added in order to accomodate 
the 64-bit book.

* in some places, I have to add --disable-shared or --enable-static to override 
the defaults from

* gpm is compiled before ncurses in order to support the use of console mouse in 
ncurses-only applications such as Dialog (try clicking through the net-setup 
dialog with your mouse on the Linux text console--this is impossible if the 
ncurses package is compiled without gpm).

* grub is compiled at the very end--this is an artifact of the reiser4 patch 
that was present in the old LiveCD versions. And it has to override unsuitable 

* glibc has a patch to add the ru_RU.CP1251 locale important for Russians that 
want bug-for-bug compatibility with Windows, and several additional locales

* kbd has many additional third-party keymaps

* lfs-bootscripts are patched

* some packages (udev, module-init-tools) come with non-default configuration files

* ncurses doesn't build the C++ binding, because it is used by nothing on the CD

IMHO, that's already too much for jhalfs to be useful. Some work should be 
started on merging the useful bits back to LFS, and we should decide how to 
handle the rest of the deviations.

Alexander E. Patrakov

More information about the livecd mailing list