On 3/7/06, Dominic Ringuet <[EMAIL PROTECTED]> wrote:
> Dan Nicholson wrote:
> > On 3/7/06, Dominic Ringuet <[EMAIL PROTECTED]> wrote:
> >>
> >>SPECFILE=`dirname $(gcc -print-libgcc-file-name)`/specs &&
> >>gcc -dumpspecs > $SPECFILE &&
> >>unset SPECFILE
> >
> > What is the value of SPECFILE? Is this the whole command? Where's the sed?
>
> That is to intently put a sane specs file on the host.
OK, I wasn't clear here. What's the value of $SPECFILE? Is it in
/tools or in /usr?
> chapter 5.11 gcc-pass2
> dummy.c: readelf -l|grep ld-linux.so.2: /tools/lib/ld-linux.so.2
> chapter 5.12 binutils-pass2
> dummy.c: readelf -l|grep ld-linux.so.2: /lib/ld-linux.so.2
This is crazy. I just can't think of any possible way this could
happen. Could you try this sanity check?
echo 'main(){}' > dummy.c
cc -v -Wl,--verbose dummy.c &> dummy.log
grep cc1 dummy.log
grep collect2 dummy.log
grep include dummy.log
grep SEARCH_DIR dummy.log
> > Either you're writing to /usr/lib/gcc/.../specs, or the host system
> > has installed a specs file, which is unlikely. What is your host
> > system?
> >
>
> Even if unlikely, you're right; that is the whole point as I can see.
> The host is november 2005 SVN. I had problems with the linker and that's
> probably when the specs file got there.
This might have something to do with it. Ch. 5 is supposed to be able
to bootstrap from any host, but there are limitations. If your host
toolchain is hosed, then maybe it's causing these issues.
> At the end, this is probably my host who should'nt had this specs file.
> May be some distros use it. May be not and I'm an isolated case.
Typically, you would change the default gcc specs to what you want
them to be at compile time. I'm pretty sure none of the major distros
rely on a specs file. But the utility is always there.
--
Dan
--
http://linuxfromscratch.org/mailman/listinfo/lfs-dev
FAQ: http://www.linuxfromscratch.org/faq/
Unsubscribe: See the above information page
