This patch adds the ability for the user to change the floating-point 
model that gdb understands for the session.  Previously we could only 
control this via the OSABI setting, which was not sufficiently flexible.

Now that we have this level of control it seems reasonable to make the 
default setting match what gcc normally generates for arm-elf (ie softfpa, 
the software floating-point with FPA-style mixed-endian doubles on a 
little-endian ARM).

Because we now default to software floating point, this effectively fixes 
8 regression failures in the testsuite.

R.

2003-03-22  Richard Earnshaw  <rearnsha@arm.com>

	* arm-tdep.h (arm_float_model): Add AUTO and LAST values.
	(arm_get_fp_model): Declare.
	* arm-tdep.c (fp_model_strings): New string array.
	(arm_fp_model, current_fp_model): New variables.
	(arm_get_fp_model): New function.
	(arm_set_fp): New function.
	(set_fp_model_sfunc): New function.
	(show_fp_model): New function.
	(_initialize_arm_tdep): Add new command to set/show the FPU.
	(arm_extract_return_value): Use arm_get_fp_model.
	(arm_store_return_value): Likewise.
	(arm_gdbarch_init): Default fpa model is softfpa.  Call arm_set_fp
	to initialize the floating-point data types.
	* arm-linux-tdep.c (arm_linux_init_abi): The default floating point
	model is FPA.