Re: [committed, PATCH] Always create dynsym section with dynamic sections

[Alan Modra <amodra at gmail dot com>]

On Mon, Apr 25, 2016 at 09:03:06AM -0700, H.J. Lu wrote:
> On Mon, Apr 25, 2016 at 6:41 AM, Alan Modra <amodra@gmail.com> wrote:
> > On Mon, Apr 25, 2016 at 05:47:15AM -0700, H.J. Lu wrote:
> >> On Sun, Apr 24, 2016 at 7:18 PM, Faraz Shahbazker
> >> <faraz.shahbazker@imgtec.com> wrote:
> >> > On 04/23/16 18:42, H.J. Lu wrote:
> >> >> Here is a patch.  Does it work for you?
> >> > Yes, this works.
> >> >
> >>
> >> Any other comments, feedbacks for my patch?
> >
> > I think you'd better look at all places that test dynsymcount.  For
> > instance, bfd_elf_size_dynsym_hash_dynstr.
> 
> I checked.  They look OK.
> 
> > This should really have been done when you committed
> > https://sourceware.org/ml/binutils/2016-02/msg00326.html and that
> > patch should have explained why the change was made.  "We should
> > always create the dynsym section, even if it is empty, with dynamic
> > sections" doesn't explain anything about why the change was made.  In
> > fact, the explanation for the patch was no better than "We should do
> > what this patch does".
> 
> The issue is some dynamic tags (sections) are mandatory in .dynamic
> section even if they are empty.  MIPS backend use dynsymcount, which
> is intended for the mandatory DT_SYMTAB (.dynsym section), to build
> static executable, which doesn't have dynamic section.  Here is the
> updated patch.  Is it OK for master?

The dynsymcount != 0 tests in bfd_elf_size_dynsym_hash_dynstr will
never trigger with your patch.  Please remove them.  OK with that
change.

-- 
Alan Modra
Australia Development Lab, IBM