Re: [RFC] MI notification on register changes

[Yao Qi <yao at codesourcery dot com>]

On 11/03/2012 12:42 AM, Tom Tromey wrote:
> First, Yao sees that lval_memory on x86, but I don't see it on x86-64.
> I wonder whether it is a bug elsewhere.

Tom,
I also see lval_memory on x86-64, and I am a little surprised that you 
see lval_register.

(gdb) b middle
Breakpoint 1 at 0x40056d
(gdb) run
Starting program: gdb/testsuite/gdb.base/nodebug

Breakpoint 1, 0x000000000040056d in middle ()
(gdb) up
#1  0x000000000040059d in top ()
(gdb) p/x $rbp
$1 = 0x7fffffffe5a0
(gdb) set $rbp=0x7fffffffe5a0

Breakpoint 1, value_assign (toval=0xdaf7a0, fromval=0xe36170) at 
../../gdb/gdb/valops.c:1246
1246	  switch (VALUE_LVAL (toval))
top>p toval->lval
$1 = lval_memory

and I think it is reasonable to see 'lval_memory' instead of 
'lval_register' here.  When executing this 'set' command, the callchain 
looks like,

evaluate_subexp_standard
   |
   ' case OP_REGISTER: value_of_register

In value_of_register,

  frame_register (frame, regnum, &optim, &unavail,
		  &lval, &addr, &realnum, raw_buffer);
....
  VALUE_LVAL (reg_val) = lval;


The 'lval' can be set to 'lval_register' or 'lval_memory' in 
frame_register.  So I don't see anything wrong here.

> Second, Pedro pointed out this PR:
>
>      http://sourceware.org/bugzilla/show_bug.cgi?id=7574
>
> ... which seems apropos.  It seems that the best thing is a generic
> "target changed" notification for the reasons mentioned there.

'target changed' observer and 'register changed' observer serves 
differently.  'target changed' observer is mostly for keeping gdb 
internal states, such as regcache and frames, consistent, however, 
'register changed' observer is for external notifications.

AFAICS, the reason 'target changed' observer is not replaced by 'memory 
changed' observer and 'register change' observer is that GDB doesn't 
know the side effect of setting a register or a piece of memory, so GDB 
conservatively uses 'target changed' observer.

--
Yao