Re: alpha/sigsuspend.S failure due to cfi mismatches

[Mike Frysinger <vapier at gentoo dot org>]

On Thursday 28 June 2007, Balint Cristian wrote:
> On Thursday 28 June 2007, Mike Frysinger wrote:
> > On Thursday 28 June 2007, Mike Frysinger wrote:
> > > On Thursday 31 May 2007, Richard Henderson wrote:
> > > > On Sat, May 19, 2007 at 05:07:26AM -0400, Mike Frysinger wrote:
> > > > > ../sysdeps/unix/sysv/linux/alpha/sigsuspend.S:43: Error:
> > > > > .cfi_endproc without corresponding .cfi_startproc
> > > >
> > > > I havn't been able to reproduce this.
> > >
> > > the trick seems to be old binutils ... using 2.17 triggers the failure,
> > > but using say 2.17.50.0.16 works
> > >
> > > i guess this means it's a bug in binutils, not glibc, that has been
> > > fixed in newer versions of binutils ?
> >
> > to answer my own question, glibc-2.6/alpha requires a binutils which
> > supports cfi in subsections ... this means 2.17.50.0.7 or newer
> > -mike
>
> hmm wierd. once i got that error, but than went away.
> i used than Jakubs a bit modified one patch (hes one didnt worked at that
> time):
>
> --- sysdeps/unix/sysv/linux/alpha/sigsuspend.S.orig     2006-10-22
> 22:17:01.000000000 +0200 +++ sysdeps/unix/sysv/linux/alpha/sigsuspend.S 
> 2006-10-22 22:18:05.000000000 +0200 @@ -31,3 +31,21 @@
>  libc_hidden_def (__sigsuspend)
>  weak_alias (__sigsuspend, sigsuspend)
>  strong_alias (__sigsuspend, __libc_sigsuspend)
> +
> +#ifndef NO_CANCELLATION
> +       .globl __sigsuspend_nocancel;

this patch is for a different issue ... alpha wasnt providing the 
__sigsuspend_nocancel symbol
-mike

Attachment: signature.asc
Description: This is a digitally signed message part.