A Saturday 03 May 2008 21:27:44, Daniel Jacobowitz wrote:
> On Sat, May 03, 2008 at 09:19:27PM +0100, Pedro Alves wrote:
> > 2008-05-03  Pedro Alves  <pedro@codesourcery.com>
> >
> > 	gdb/
> > 	* parse.c (parse_exp_in_context): Don't override
> > 	expression_context_pc if get_selected_block returned a valid block.
> >
> > 	gdb/testsuite/
> > 	* gdb.base/macscp.exp, gdb.base/macscp1.c: Add test for printing
> > 	expressions with macros.
>
> This patch is OK since it's clearly progress.  

Thanks, I installed it -- better keep this regression fix separate
from follow up improvements.

> But doesn't this mean 
> that we should be passing a PC to parse_exp_1, not a block?  Otherwise
> macros will not be correct in e.g. breakpoint conditions:
>
>                 loc->cond =
>                   parse_exp_1 (&arg, block_for_pc (loc->address), 0);

Yes, looks like it.  At some point, it may even be better to get rid
of expression_context_block in the language parsers, in favor of 
expression_context_pc for variables declared midblock, I guess:

 {
    a = 1;
    f ();
    B b;
    g ();
 }

I'll submit a patch to the insight side if the attached is OK.  Tested
on x86_64-unknown-linux-gnu.

-- 
Pedro Alves