Re: [rfc, v2] Fix ELF synthetic symbol size bug (affects PPC64)

[Alan Modra <amodra@bigpond.net.au>]

On Tue, Nov 06, 2007 at 04:44:02PM -0500, Daniel Jacobowitz wrote:
> > Is the bfd part OK?
> 
> The elf.c part of this does not make sense to me.  Why should we have
> the same size as that symbol?  The @plt entry is only a couple of
> instructions.  To figure out its size you need to ask the backend
> plt_sym_val routine.

Agreed.  The ppc64 part is OK.  I'm committing the following to fix
a bug I noticed in the elf.c get_synthetic_symtab function.  I also
set udata.p to NULL as this may be all that's needed with a modified
version of Ulrich's patch.

	* elf.c (_bfd_elf_get_synthetic_symtab): Only bump the symbol
        pointer when we have a valid symbol.  Init udata.p to NULL.

Index: bfd/elf.c
===================================================================
RCS file: /cvs/src/src/bfd/elf.c,v
retrieving revision 1.422
diff -u -p -r1.422 elf.c
--- bfd/elf.c	25 Oct 2007 00:40:34 -0000	1.422
+++ bfd/elf.c	12 Nov 2007 03:21:58 -0000
@@ -8678,7 +8678,7 @@ _bfd_elf_get_synthetic_symtab (bfd *abfd
   names = (char *) (s + count);
   p = relplt->relocation;
   n = 0;
-  for (i = 0; i < count; i++, s++, p++)
+  for (i = 0; i < count; i++, p++)
     {
       size_t len;
       bfd_vma addr;
@@ -8695,12 +8695,13 @@ _bfd_elf_get_synthetic_symtab (bfd *abfd
       s->section = plt;
       s->value = addr - plt->vma;
       s->name = names;
+      s->udata.p = NULL;
       len = strlen ((*p->sym_ptr_ptr)->name);
       memcpy (names, (*p->sym_ptr_ptr)->name, len);
       names += len;
       memcpy (names, "@plt", sizeof ("@plt"));
       names += sizeof ("@plt");
-      ++n;
+      ++s, ++n;
     }
 
   return n;

-- 
Alan Modra
Australia Development Lab, IBM