Re: [01/15] Introduce get_current_arch () function

[Pedro Alves <pedro@codesourcery.com>]

On Wednesday 10 June 2009 16:36:14, Ulrich Weigand wrote:
> Tom Tromey wrote:
> > >>>>> "Ulrich" == Ulrich Weigand <uweigand@de.ibm.com> writes:
> > 
> > Ulrich>  printf_command (char *arg, int from_tty)
> > Ulrich>  {
> > Ulrich> +  struct gdbarch *gdbarch = get_current_arch ();
> > 
> > I think that perhaps the printf command should use the arch from the
> > values it is trying to print.
> > 
> > Ulrich>  	      struct type *wctype = lookup_typename (current_language,
> > Ulrich> -						     current_gdbarch,
> > Ulrich> +						     gdbarch,
> > Ulrich>  						     "wchar_t", NULL, 0);
> > 
> > E.g., here it could use the architecture from val_args[i].
> > 
> > The benefit here would be that you could print out two wide strings,
> > each from a different architecture, using a single printf.  This seems
> > like a reasonable thing to want to do.
> 
> There wouldn't really be much of a difference today, as the architecture
> of val_args[i] is the architecture of the expression from which it was
> evaluated, which is likewise get_current_arch ().
> 
> However, you're right that it would be more straightforward to use the
> value arch here, in case we might later decide to allow evaluation of
> an expression to return values in a different arch.

I haven't actually studied the patch set, so sorry if this sounds
way off, but Tom's remark made me wonder how does this interact with
re-printing values from the history ($1, $2, $nnn) or convenience
variables?  As in, what would be the arch used to print $nnn?  Should it
be the current arch, or the arch that was used when the value was produced?  A
simple example where it can be a different arch is if you debug
inferior/target foo, print something, kill/disconnect, debug
inferior/target 2, reprint value from history.

I see from the hunk above that the language used is always the
current language, so it may be right to always use the current arch...

-- 
Pedro Alves