Ulrich Weigand schrieb:
> Markus Deuling wrote:
> 
>> this patch removes BITS_BIG_ENDIAN from defs.h by replacing it with
>> its expression. The way to recognize endianess of a target is
>> gdbarch_byte_order (current_gdbarch) == BFD_ENDIAN_BIG which is
>> widely used by most files.  So this macro is unnecessary.
> 
> As Dan and Eli pointed out, there *is* a difference between the two:
> - gdbarch_byte_order says whether in a multi-byte value, the least
>   or most significant *byte* comes first
> - BITS_BIG_ENDIAN is intended to say whether in a bitfield, the
>   least of most significant *bit* comes first
> 
> Looking at GCC, there are targets with big endian byte order but
> little endian bit order (or vice versa), even though those are 
> not supported in GDB today (or maybe they are but bitfields simply
> do not work correctly).
> 
> Thus I think we should introduce a new gdbarch property to allow
> to set the bit order.
> 


Hi,

thank you very much for all of your valuable feedback. I rewrote the patch
and introduce a new gdbarch property bits_big_endian now to replace BITS_BIG_ENDIAN.

I tested this patch on both Little and Big Endian machine (x86/ppc) without regression.
If this patch is ok I'll post another one for the documentation. Ok to commit?

ChangeLog:

	* gdbarch.sh (function_list): Add new property bits_big_endian to
	gdbarch structure.
	* gdbarch.{c,h}: Regenerate.

	* value.c (struct value): Replace BITS_BIG_ENDIAN by
	gdbarch_bits_big_endian (comment).
	(unpack_field_as_long, modify_field): Likewise.
	* value.h: Likewise (comment).
	* valops.c (value_slice): Likewise.
	* valarith.c (value_subscript, value_bit_index): Likewise.
	* gdbtypes.h (field): Likewise (comment).
	* eval.c (evaluate_subexp_standard): Likewise.
	* dwarf2read.c (dwarf2_add_field): Likewise.
	* ada-lang.c (decode_packed_array, ada_value_primitive_packed_val)
	(move_bits, ada_value_assign, value_assign_to_component): Likewise.
	
	* defs.h (BITS_BIG_ENDIAN): Remove.




-- 
  Markus Deuling
  GNU Toolchain for Linux on Cell BE
  deuling@de.ibm.com