arm-eabi regressions from Re: mapping of stripped section syms

[Alan Modra]

On Fri, Mar 23, 2012 at 07:09:43PM +0100, Hans-Peter Nilsson wrote:
> > From: Alan Modra <amodra@gmail.com>
> > Date: Fri, 23 Mar 2012 10:34:25 +0100
> 
> > On Fri, Mar 23, 2012 at 07:56:27PM +1030, Alan Modra wrote:
> > > I'll also be committing another patch that disables stripping of
> > > output sections when -q is in effect
> > 
> > bfd/
> > 	* elf.c (assign_file_positions_for_non_load_sections): Don't warn
> > 	on empty alloc sections.
> > ld/
> > 	* ldemul.c (before_allocation_default): When emitrelocations,
> > 	don't strip sections..
> > 	* ldlang.c (lang_add_section): ..and don't set up map_head, map_tail.
> 
> This or another patch in this series of bfd and ld patches
> caused test-suite regressions for arm-eabi:

I see it too in my overnight tests.  It was this patch.  arm creates
both .rel.dyn and .rela.dyn.  One or the other needs to be stripped or
you run foul of reloc size tests in bfd_section_from_shdr.

I'll fix it some time this weekend.  If that's not quick enough for
you, feel free to revert the ldemul.c and ldlang.c changes.

-- 
Alan Modra
Australia Development Lab, IBM