> I run regression tests on arm-linux-gnueabi with your patch.  There
> are some fails on armv4t arm and thumb mode.  It is an existing bug
> introduced by my patch :( and your patch just exposed it.  I'll fix it
> separately.

Thanks, Yao.

> >    /* If name of symbol doesn't start with '__stack_chk_guard', this
> >       instruction sequence is not for stack protector.  If symbol is
> >       removed, we conservatively think this sequence is for stack protector.  */
> 
> We need to update the comment to sync with the code below.

Indeed!

> > -  if (stack_chk_guard.minsym
> > -      && strncmp (MSYMBOL_LINKAGE_NAME (stack_chk_guard.minsym),
> > +  if (stack_chk_guard.minsym == NULL
> > +      || strncmp (MSYMBOL_LINKAGE_NAME (stack_chk_guard.minsym),
> >  		  "__stack_chk_guard",
> >  		  strlen ("__stack_chk_guard")) != 0)
> 
> Otherwise, it looks good to me.

Thanks! Attached is a revised patch.  I will push it after you
fix the latent bug you noticed - just let me know when?

-- 
Joel