Re: r5717 - in glibc-package/branches/eglibc-2.18/debian: . patches patches/any

[Samuel Thibault]

Thomas Schwinge, le Thu 10 Oct 2013 08:06:52 +0200, a écrit :
> On Wed,  9 Oct 2013 23:37:56 +0000, Samuel Thibault 
> <sthibault@alioth.debian.org> wrote:
> > Author: sthibault
> > Date: 2013-10-09 23:37:56 +0000 (Wed, 09 Oct 2013)
> > New Revision: 5717
> 
> >   * patches/any/multi_arch.diff: New patch, fixes undefined reference 
> > failure
> >     with non-multiarch builds.
> 
> > +--- eglibc-2.18.orig/sysdeps/generic/symbol-hacks.h        2013-06-15 
> > 17:37:04.000000000 +0000
> > ++++ eglibc-2.18/sysdeps/generic/symbol-hacks.h     2013-10-09 
> > 23:31:32.000000000 +0000
> > +@@ -1,6 +1,9 @@
> > + /* Some compiler optimizations may transform loops into memset/memmove
> > +    calls and without proper declaration it may generate PLT calls.  */
> > + #if !defined __ASSEMBLER__ && !defined NOT_IN_libc && defined SHARED
> > ++#include <config.h>
> > ++#  if defined USE_MULTIARCH
> > + asm ("memmove = __GI_memmove");
> > + asm ("memset = __GI_memset");
> > ++#  endif
> > + #endif
> 
> This doesn't look correct to me.

Ah, __GI_ is not related to MULTIARCH, ok. (libc is so full of
exceptions...)

> I think the issue rather is our
> definition of NO_HIDDEN in sysdeps/mach/hurd/configure.

You mean we shouldn't set NO_HIDDEN to 1? Or rather use #if !defined
NO_HIDDEN?

> The problem can
> be reproduced on GNU/Linux in a --disable-hidden-plt configuration.  See
> <https://sourceware.org/bugzilla/show_bug.cgi?id=15605>.

Samuel