Re: [arch/x86/vdso/vdso32-syms.lds] Error 1

[Markus Trippelsdorf <markus@trippelsdorf.de>]

On Wed, Nov 12, 2008 at 10:34:59PM +0100, Markus Trippelsdorf wrote:
> On Wed, Nov 12, 2008 at 10:56:17AM -0800, Roland McGrath wrote:
> > Sorry for the delay, I've been offline.
> > 
> > > All of a sudden I get the following error when I build the kernel:
> > 
> > Suddenly when?  What changed to make this start happening?
> > Is it recent kernel sources?  If so, can you git-bisect to
> > which change did it?  If not, did something else change in
> > your build environment that made this problem appear?
> > Does it come or go with some .config change?
> > 
> > > --- -   2008-11-07 18:51:36.392997346 +0100
> > > +++ arch/x86/vdso/vdso32-int80-syms.lds 2008-11-07 18:51:36.283820343 +0100
> > > @@ -3,5 +3,3 @@
> > >  VDSO32_sigreturn = 0x0400;
> > >  VDSO32_vsyscall = 0x0420;
> > >  VDSO32_vsyscall_eh_frame_size = 0x040;
> > > -VDSO32_vsyscall = 0x0420;
> > > -VDSO32_vsyscall_eh_frame_size = 0x040;
> > > make[1]: *** [arch/x86/vdso/vdso32-syms.lds] Error 1
> > > make: *** [arch/x86/vdso] Error 2
> > 
> > That sure looks odd.  It looks like sort -u failed to uniquify those lines.
> > 
> > Can you try something like this:
> > 
> > > -	else rm -f $(@D)/.tmp_$(@F); exit 1; \
> > > +	else cat >&2 $(@D)/.tmp_$(@F); exit 1; \
> > 
> > and run make V=1 to see the full output on the failure?
> 
> Something odd is going on with my system:

> I will try to figure out what causes this behaviour, but it is not
> kernel related I guess. So I apologize for the noise.

It turned out that my version of glibc, which was compiled with a
gcc-4.4.0 snapshot, was the cause of the problem.

However simply compiling glibc again with gcc-4.3.2 did not work,
because glibc uses "LC_ALL=C sort -u" in many Makefiles. So I had to find
an old binary version of glibc, install that version in a chroot
environment and finally compile glibc again with a sane compiler in that
chroot. 

-- 
Markus