trying to understand some BFD code

trying to understand some BFD code

[shrimpx]

Hi all,

I've asked a similar question before, but I'm still trying to
understand some stuff in how BFD works, for a research project I'm
working on. Help would be much appreciated!

In include/aout/aout64.h there is this type declaration:

struct reloc_std_external
{
  bfd_byte r_address[BYTES_IN_WORD];    /* Offset of of data to relocate.  */
  bfd_byte r_index[3] ;                 /* Symbol table index of symbol.  */
  bfd_byte r_type[1];                /* Relocation type.  */
};

It seems that most uses of this struct in the code are guarded by a

  bfd_header_big_endian(...)

check. E.g. line 420 in bfd/aout-arm.c. If bfd_header_big_endian() is
true, it's assumed that the three bytes in r_index are laid out
right-to-left. When it's false, they are left-to-right. Also,
depending on the value of bfd_header_big_endian(), the *bits* in
r_type[0] are laid out differently -- right to left vs left to right.

Could someone explain why this is the case?

Is it fundamentally tied to the endianness of binary formats? If so,
why are the *bits* inside r_type[0] reversed? To my knowledge,
bfd_header_big_endian() indicates the byte-endianness of the format.

Or is it a high-level programming pattern that helps GDB programmers
work with this code in a way that is harder otherwise?

Thanks very much!

-m

Re: trying to understand some BFD code

[Daniel Jacobowitz]

On Tue, Oct 02, 2007 at 03:31:10PM -0700, shrimpx@gmail.com wrote:
> Hi all,
> 
> I've asked a similar question before, but I'm still trying to
> understand some stuff in how BFD works, for a research project I'm
> working on. Help would be much appreciated!

BFD comes from binutils, not from GDB.  Also, roughly no one works
with a.out any more, so you're unlikely to find much help.  Sorry.

> Could someone explain why this is the case?
> 
> Is it fundamentally tied to the endianness of binary formats? If so,
> why are the *bits* inside r_type[0] reversed? To my knowledge,
> bfd_header_big_endian() indicates the byte-endianness of the format.

I would assume that it is a peculiarity of the a.out format, not
anything more general.

-- 
Daniel Jacobowitz
CodeSourcery