Re: Specifying LMA's after unallocated sections?

[Jonathan Larmour <jlarmour at redhat dot co dot uk>]

Ian Lance Taylor wrote:
> 
>    Date: Thu, 06 Apr 2000 23:31:43 +0100
>    From: Jonathan Larmour <jlarmour@redhat.co.uk>
> 
>    I was thinking of some way to avoid outputting the section regardless.
>    Suppose for example you want to put .bar after .rel.* relocation
>    information. I know (because I accidentally stumbled across it) that
>    outputting an empty .rel section makes BFD very unhappy. And to be honest,
>    I think BFD is right: you shouldn't be emitting "unused" empty output
>    sections in general.
> 
> The only way that comes to mind would be to use memory regions with
> the new AT> construct.  Then you could avoid using LOADADDR and
> SIZEOF.  I can't think of any other way to implicitly define the load
> address, which is basically what you want to do.

Ah, now that is new and useful. I think I can abuse that in exactly the way
I want. Thanks!

> By the way, an empty .rel section should work fine.  If it fails, I
> think that is a bug.

In that case, perhaps the attached trivial patch is required? 

Without it, if you put a ". = ." in e.g. a .rela.text section of a
mips-tx39-elf target linker script defining no actual relocation info, then
you get a BFD assertion failure in bfd_section_from_elf_index(), called
from bfd_section_from_shdr(). This is triggered because of an index of 0
being passed to bfd_section_from_elf_index(). But there's nothing wrong
with an index of 0 AFAIK, so the assert is wrong.

However, even with that fixed, objdump reports "File format not
recognized". I don't know for sure whether the correct behaviour should be
just to ignore empty .rel/.rela sections completely if they are empty, but
it seems reasonable so that's what I suggest.

Jifl

2000-04-07  Jonathan Larmour  <jlarmour@redhat.co.uk>

	* elf.c (bfd_section_from_shdr): Ignore empty SHT_REL/SHT_RELA
	sections
	(bfd_section_from_elf_index): Don't assert on an index of 0


-- 
Red Hat, 35 Cambridge Place, Cambridge, UK. CB2 1NS  Tel: +44 (1223) 728762
"Plan to be spontaneous tomorrow."  ||  These opinions are all my own fault
Index: elf.c
===================================================================
RCS file: /cvs/src/src/bfd/elf.c,v
retrieving revision 1.27
diff -u -5 -p -r1.27 elf.c
--- elf.c	2000/03/27 08:39:12	1.27
+++ elf.c	2000/04/06 23:52:36
@@ -1277,10 +1277,14 @@ bfd_section_from_shdr (abfd, shindex)
       /* *These* do a lot of work -- but build no sections!  */
       {
 	asection *target_sect;
 	Elf_Internal_Shdr *hdr2;
 
+	/* Check for empty section - if so, throw it away */
+	if (hdr->sh_size == 0)
+	    return true;
+
 	/* Check for a bogus link to avoid crashing.  */
 	if (hdr->sh_link >= ehdr->e_shnum)
 	  {
 	    ((*_bfd_error_handler)
 	     (_("%s: invalid link %lu for reloc section %s (index %u)"),
@@ -1400,11 +1404,11 @@ bfd_section_from_shdr (abfd, shindex)
 asection *
 bfd_section_from_elf_index (abfd, index)
      bfd *abfd;
      unsigned int index;
 {
-  BFD_ASSERT (index > 0 && index < SHN_LORESERVE);
+  BFD_ASSERT (index < SHN_LORESERVE);
   if (index >= elf_elfheader (abfd)->e_shnum)
     return NULL;
   return elf_elfsections (abfd)[index]->bfd_section;
 }