On Tue, Oct 18, 2011 at 11:49 AM, Bruce Dubbs <[email protected]> wrote:
>
> There are multiple ways of solving a problem.  What new problem does
> your suggestion solve?
>
>   -- Bruce

Well most of it is just a personal intent of building as many binaries
as possible with the new (and fully tested) gcc.  Since file isn't
absolutely necessary so early, it could be postponed until just after
gcc quite easily.

Also, I've always thought of the linux-headers, glibc, zlib, binutils,
and gcc as the essential part of the toolchain that once created
allows the rest of the system to be developed.  Of course this more
recently includes gmp, mpfr and mpc due to gcc dependance on them.
But none of them have a strict dependance on file (by strict I mean it
doesn't appear to link to libmagic).  file is required in the same
fashion that grep, gawk, cat, or pwd is required; It must be present
as it provides runtime support via the binary itself.

The problem stems more from an unfortunate shortcoming of the build
configuration within the three newer libraries that refuses to accept
that file could be located outside of the usual locations.  This is
why I'm reluctant to see file compiled along side the previously
mentioned packages - it is used as a system binary in this context and
is not necessary as a library component for the resulting compilation
system in practice.

I suppose my arguments are a matter of opinion though.  I'm in the
process of a test build with this suggested rearrangement due to
curiosity.  I have my doubts that it will have a major impact on the
build of file, but I'll know in a little bit once I can do a diff.

On a final note, after some more thorough examination of the sources
this appears to be required by libtool, which is present in gmp, mpfr,
and mpc.  So it doesn't seem to be needed by any of them directly, but
as part of libtool's ability to detect what the host system's linker
is capable of.


Jonathan
-- 
http://linuxfromscratch.org/mailman/listinfo/lfs-dev
FAQ: http://www.linuxfromscratch.org/faq/
Unsubscribe: See the above information page

Reply via email to