[OT] gcc manpage and reality...

[Alain Perry <sensei@quarkup.org>]

This mail is probably out of topic, but since I don't really know who to
turn to, and there are many developpers on this list, please read if you
have time...

I'm trying to build a custom firmware image for the Linksys WRT54G. It
runs linux, and the sources and tools to build the image have been
provided. The tools are gnu tools, provided as binarirs and source.
Since the binaries are for the i386 platform, I just thought I would
recompile them. However, the build fails on compilation of ld, because
of an undefined yy_current_buffer variable. I tried to define it from
another source file which used it to, but then ld would segfault when
later used. I then decided to use the version of binutils provided in
debian unstable (which I use) and to compile it instead. I actually did
that for the binutils and gcc. I kept the glibc and uClibc provided. The
compilation went fine, and I obtained the toolchain I needed.

But during the firmware compilation, I get an error saying that "as"
does not understand the -mcpu=r4600 switch. The manpage confirmed that.
So I changed the GCCFLAGS variable in the according Makefile to
-march=r4600 which both gcc and as are supposed to understand. I then
get an error stating that cc1 (gcc as I understand it) does not know the
arch=4600 option.

I've now run out of ideas to build this firmware, but I need it to be
able to tweak it for my project. If anyone of you has an idea about
that, I'd be very grateful. As you may have noticed, I am not very
familiar with cross-compilation, and as a matter of fact not familiar
with compilation at all (I usually only care about ./configure and
missing dependencies...).

Thanks for reading this.

-- 
Alain Perry