__gmon_start__ issues between 2.1.x and 2.2.x on sparc(?)

[Franz Sirl]

On Wednesday 21 February 2001 22:00, Franz Sirl wrote:
> On Wednesday 21 February 2001 18:20, Andreas Jaeger wrote:
> > Franz Sirl <Franz.Sirl-kernel@lauterbach.com> writes:
> > > At 16:30 21.02.2001, Philip Blundell wrote:
> > > >In message < ho8zn013d8.fsf@gee.suse.de >, Andreas Jaeger writes:
> > > > >Is this correct?  Anything else I can do to check that binutils are
> > > > >correct now?
> > > >
> > > >If you can invent a self-contained testcase, it would be worth adding
> > > > to the ld testsuite.
> > >
> > > See http://sources.redhat.com/ml/libc-alpha/2000-10/msg00056.html
> > >
> > > Unfortunately the html version lost the file names of the attachments,
> > > but you can d/l the mbox formatted archive to get it.
> >
> > Are you sure the testcase is valid?  It fails on sparc - but also on
> > ia32.
> >
> > Can you rerun it with "fixed" binutils?
>
> Oops, I looked at the email again and noticed that I didn't specify the
> compile line for the executable correctly. Add -fpic or -fPIC to it and you
> should see difference:
>
> 	 gcc -fpic tA-22.c -L213 -ltB -o tA
>
> I verified here on PPC that current binutils avoids the segfault.

Actually using -fPIC doesn't work on PPC and that leads to a possible reason 
why this fix won't work for sparc, glibc on sparc uses -fPIC (why?) to 
compile glibc, whereas PPC uses -fpic.

Andreas, can you try to add

pic-ccflag = -fpic

to sysdeps/sparc/Makefile and see if that fixes it? That should also be quite 
a performance improvement on sparc.

Franz.

PS. this reminds me again, there should finally be a way to tell the 
autotools to prefer -fpic over -fPIC (or maybe to prefer -fpic?)