Re: PATCH: PR ld/4701: binutils generates invalid klibc-based binary on Linux x86_64

["H. J. Lu" <hjl at lucon dot org>]

On Thu, Jun 28, 2007 at 03:05:15PM +0930, Alan Modra wrote:
> On Wed, Jun 27, 2007 at 09:20:51PM -0700, H. J. Lu wrote:
> > On Thu, Jun 28, 2007 at 12:30:27PM +0930, Alan Modra wrote:
> > > On Wed, Jun 27, 2007 at 07:40:27PM -0700, H. J. Lu wrote:
> > > > Linux ld.so isn't OK with your change. We haven't seen the problem
> > > > because we won't get a PT_LOAD segment with .bss sections only
> > > > using the default linker script. I uploaded a testcase to:
> > > > 
> > > > http://sourceware.org/bugzilla/show_bug.cgi?id=4701
> > > 
> > > Arggh!  I was sure this worked!  So we should go back to using my voff
> > > code probably, or otherwise hack p_offset.
> > 
> > Whatever we do, we should follow gABI.
> > > 
> > > Meanwhile, please let me know exactly why klibc was unhappy, as I'm
> > > concerned that whatever is causing the segv there might also be
> > > affected by overlapping p_offset for bss segments.
> > 
> > My testcase is derived from klibc. Executables have a one PT_LOAD
> > segment with only a .bss section. The only difference is the failed
> > ones in klibc are executables, which are loaded by kernel, and mine
> > is a shared library, which is loaded by ld.so.
> 
> This should comply with alignment requirement but not waste file
> space.  Can you try this against klibc executables?
> 
> 	* elf.c (assign_file_positions_for_load_sections): Ensure bss
> 	segments meet gABI alignment requirements.
> 

It works on klibc. However, I got

FAIL: overlay size

on Linux/x86-64. You should be able to see it with a cross binutils.



H.J.