[PLUG] Linux from scratch is interesting...

[Jeme A Brelin]

[Michael, can you adjust your linewrap so that it either does not wrap
lines or wraps them at <80 columns.  Those long lines make reading a pain
and I think some are as likely to skip the message as try to decipher it.]

On Tue, 8 Jul 2003, robinsoq  wrote:
> Why is it so important not to compile utilities with dynamic binding
> before chrooting into the LFS system?

As Vincent Damewood pointed out, this is because you'd be linking against
libraries or versions of libraries that just won't exist in your chrooted
system.

I know the LFS book says that you should rebuild everything dynamic and
that's a good idea, but I also like to keep copies of the static binaries
in /sbin just in case /lib is damaged or inaccessible at some point.

> I am wondering how to adapt this system as my real goal is to make this
> work over nfs and boot off of network.  Maybe I only have to modify my
> approach to the init scripts?

You'll also have to build with the most generic options for your processor
family in order to maintain binary compatibility.

> One curiosity I'm running into is with configure.  Does it detect what
> platform your on affecting the compilation result?

Yes.

> I ask as I'm compiling on a Pentium III but want this LFS system to work
> on a true 386 also.

Right.  There are environment variables that can control the options make
passes to gcc.  The cross-compiling hint will help.

> My LFS system is being built in an extra partition on my host box.  I'm
> wondering if the host box should have this system in it's fstab and
> integrity check the partition for it at boot?  Will my network clients
> fsck my LFS system?

I've never built LFS with a filesystem that requires fsck.

> Anyone know why diffutils, a special utility for decrypting function
> labels produced from C++ code where overloading of functions is
> possible, and other various tools are needed in a base Linux system?

diffutils is the GNU diff and patch programs for noting the difference
between files and generating and applying patches.  It's absolutely vital
when compiling software from multiple sources.

<URL: http://www.gnu.org/software/diffutils/diffutils.html >

> Bash I can see, but why do I need gawk?

gawk is the GNU implementation of awk.  It's used in shell scripts
throughout a system.  You'll also find that the LFS book will include
short scripts for patching or modifying a particular configure script or
Makefile and those scripts may well need awk.

And, quite frankly, a unix system without awk just isn't a unix system.

> Anyone have an example of using strings from the binutils package?

I've used strings to look for undocumented command options and all kinds
of odd things.  It can be used to identify some trojans, too.

> Why not distribute patched code instead of providing a patch?

One of the main points, for me, of build a system from scratch is knowing
exactly how it differs from the generic source as created by the package
maintainers.  I download the source from the maintianer's site and check
it against any available signature keys and then apply patches myself.

People used to ask me WHY I went to these pains to keep track of the
software on my computer.  I don't know that I have an answer that I can
put together in english and make any real sense.  But the answer would be
very similar to the answer to why I ride a fixed gear bike.  Perhaps I
just like to be in as direct control of my machines as possible.

J.
-- 
   -----------------
     Jeme A Brelin
    jeme at brelin.net
   -----------------
 [cc] counter-copyright
 http://www.openlaw.org