[Patch] Objdump disassembly with dynamic relocation info

[Dirk Thierbach]

Hello all,

this is a patch against binutils 2.12.90.0.1 (from the woody debian
source package) to make objdump disassembly work better with dynamic
relocation information. It does three things:

1) With -dR, in prints dynamic reloction info interspersed with 
   disassembly, like in -dr with normal relocation info.

   The problem here was that dynamic relocations use absolute addresses,
   while normal relocations use addresses relative to the section.

   A possible solution is to modify one of them "in place", so they
   can be treated together. But this is a hack, and would possibly
   interfere with the DISASSEMBLER_NEEDS_RELOCS #ifdef. 

   So instead, both information is processed seperately and in
   parallel, with common code moved into subroutines. This has the
   side effect of making the dissassembly routines somewhat shorter
   and IMHO more readable.

2) When there is initialization code at the beginning of a section
   without symbols, this code is labelled by the following symbol with
   negative offset. To improve this, objdump now makes an attempt
   to re-use debugging symbols that point at the beginning of the
   section as section labels. To achieve this, the symbols are
   modified "in place".

3) For elf-i386, the .got section contains information which can be
   used to label the .plt section. To make this work, we need
   symbols that refer to the .plt section. Here, they are created
   by modifying the address of the symbols that point to the .got
   section, which are not interesting for disassembly anyway.

The last two items are also only activated with the -dR option.
They are a bit questionable, because they modify data "in place",
which may interfere somewhat with later outputs (here: only the
debugging information). Also, (3) is architecture dependent,
though this is marked clearly in the source and the architecture
is checked before using it.

But it works for me, and it makes the disassembly of dynamic code
a lot more readable.
  
Change (1) should be acceptable in any case, IMHO.

- Dirk

P.S: If you answer this mail, please send a copy to my address, since
I am not on any binutils mailing lists.

diff-20020801
Description: Text document