Re: Misc; Was: [RFA]: x86_64 target files

[Andrew Cagney <ac131313 at cygnus dot com>]

> Andrew Cagney <ac131313@cygnus.com> writes:
> 
> 
>>>
>>> I can try to take care of this in binutils.  You suggest something
>>> like get_arm_regnames in opcodes/arm-dis.c I guess?
> 
>>
>> Yes, but having looked at the interface, it isn't the most elegant.
> 
> 
> Do you have a better idea for an interface?

No.

(Just to be clear.  The ``fix'', what ever it is, requires an interface 
change to OPCODES.  I don't think it is reasonable for me to make the 
x86-64 tdep file approval dependant on such an external interface 
change.  Hence me suggestion to Jiri that a PR recording the problem be 
filed and hence the code remain as is.)

I've only got a gut feeling.   The x86-64 makes for a second target (and 
if we add in the x86m at least a third) that wants a mechanism for 
controlling the opcodes assembler.  GDB can certainly keep adding hooks 
to OPCODES, however, I think it might be time to consider a more generic 
interface - that way GDB developers only need to do this once.

Looking at include/dis-asm.h, `struct disassemble_info' already contains 
dissassembler_options.  A simplistic(1) aproach might be to add the command:

	(gdb) set disassembler <blah>

where <blah> was mindlessly assigned to dissassembler_options.  There 
are plenty of variations on this and probably other better ideas.

Enjoy,
	Andrew

(1) And my simplistic suggestion is certainly not as simple as it looks 
:-/ : getting GDB to detect / report a invalid option at the time it is 
set instead of used;  getting GDB to handle this in a multi-arch 
environment (an open problem - avoided so far by people using ``set 
<arch> <option> ...'');  getting GDB to report a list of valid options 
in response to:

	(gdb) set disassembler
or even	(gdb) set disassembler <tab>

hmm ... :-)