Re: [RFC PATCH] Synthetize some symbols for objdump -d

[Jakub Jelinek <jakub at redhat dot com>]

On Wed, Apr 21, 2004 at 01:15:51PM +0930, Alan Modra wrote:
> On Tue, Apr 20, 2004 at 01:07:56PM +0200, Jakub Jelinek wrote:
> > if they use .got.plt they just need to define elf_backend_plt_entry_size
> > and elf_backend_reserved_plt_size macros, otherwise they don't need
> > anything in most cases.  IA-64 will need to define elf_backend_relplt_name,
> > sparc64 provide its own implementation (as part of PLT is normal PLT like,
> > the second is .got.plt like).
> 
> I think it would be a good idea to define a backend function to process
> each plt reloc.  Given a pointer to the plt section, and a pointer to
> the reloc, it could return the value of the sym or -1 if the reloc
> should be ignored.  That should be general enough to work even with
> something like the sparc plt, and you won't need plt_entry_size or
> reserved_plt_size.

I'll rework the patch so that the backend function is called in the loop
for each PLT, not to handle everything.
I still think plt_entry_size/reserved_plt_size is needed, adding the same
routine on 15 targets which have fixed size PLT entries seems unnecessary
to me.  Wouldn't a comment stating that they should be set only if the
target has fixed size PLT entries and left out (== 0) otherwise be enough?

	Jakub