size_input_section calcs output_offset incorrectly for merged sections

size_input_section calcs output_offset incorrectly for merged sections

[Michael Eager]

I'm seeing a problem in size_input_section() when it
processes a merged section.  If the merged section is to
be included in the output, the output_offset value is
computed as:

    i->output_offset = dot - output_section_statement->bfd_section->vma;

If the section has been eliminated and is marked as SEC_EXCLUDE,
the section's output_offset value is computed as:

   i->output_offset = i->vma - output_section_statement->bfd_section->vma;

i->vma is zero, which results in output_offset being the inverse
of the output vma.  Later, during relaxation, output_offset is
added to vma resulting in an incorrect symbol address of zero.

It seems to me that output_offset should be the same in both
cases.  I've run a couple tests which work with this change.

Does this seem correct?

-- 
Michael Eager	 eager@eagercon.com
1960 Park Blvd., Palo Alto, CA 94306  650-325-8077

Re: size_input_section calcs output_offset incorrectly for merged sections

[Alan Modra]

On Sat, Jan 05, 2008 at 08:59:04AM -0800, Michael Eager wrote:
> It seems to me that output_offset should be the same in both
> cases.  I've run a couple tests which work with this change.
> 
> Does this seem correct?

No.  The code is correct for excluded sections in general.  You
shouldn't be using symbols from merged sections directly.  See
_bfd_merged_section_offset, _bfd_elf_rel{,a}_local_sym.

-- 
Alan Modra
Australia Development Lab, IBM

Re: size_input_section calcs output_offset incorrectly for merged sections

[Michael Eager]

Alan Modra wrote:
> On Sat, Jan 05, 2008 at 08:59:04AM -0800, Michael Eager wrote:
>> It seems to me that output_offset should be the same in both
>> cases.  I've run a couple tests which work with this change.
>>
>> Does this seem correct?
> 
> No.  The code is correct for excluded sections in general.  You
> shouldn't be using symbols from merged sections directly.  See
> _bfd_merged_section_offset, _bfd_elf_rel{,a}_local_sym.

Thanks.  Calling _bfd_elf_rela_local_sym worked.

-- 
Michael Eager	 eager@eagercon.com
1960 Park Blvd., Palo Alto, CA 94306  650-325-8077