Re: [RFA] New dwarf2_attribute_true_p function in dwarf2read.c

[Daniel Jacobowitz <drow@false.org>]

On Thu, Apr 15, 2004 at 01:23:18PM -0700, Joel Brobecker wrote:
> Hello,
> 
> Following a suggestion made by JimB in:
> 
> http://sources.redhat.com/ml/gdb-patches/2004-04/msg00297.html
> 
> I made the following change:
> 
> 2004-04-15  Joel Brobecker  <brobecker@gnat.com>
> 
>         * dwarf2read.c (dwarf2_attribute_true_p): New function.
>         (die_is_declaration): Use the function above.
> 
> Tested on x86-linux, no regression.
> 
> OK to apply?

Sorry, this patch is incorrect.

> -  return (dwarf2_attr (die, DW_AT_declaration, cu)
> -	  && ! dwarf2_attr (die, DW_AT_specification, cu));
> +  /* A DIE is a declaration if it has a DW_AT_declaration attribute
> +     which value is non-zero.  However, we have to be careful of DIEs
> +     with a DW_AT_specification attribute, because dwarf2_attribute_p()
> +     follows this attribute, and therefore might cause us to find a
> +     DW_AT_declaration attribute, but that belongs to a different DIE.  */
> +  return (dwarf2_attribute_true_p (die, DW_AT_declaration, cu)
> +	  && ! dwarf2_attribute_true_p (die, DW_AT_specification, cu));
>  }
>  
>  /* Return the die giving the specification for DIE, if there is

First of all, there's no dwarf2_attribute_p, you mean dwarf2_attr and
dwarf2_attr_true_p.

Secondly, the predicate function should be used for flag attributes
like DW_AT_declaration, but checking DW_UNSND for DW_AT_specification
is incorrect - it's a DIE reference.  dwarf2_attr != NULL is the right
check.  Some day I hope to add checking macros to verify we're using
the right member of each union...

-- 
Daniel Jacobowitz
MontaVista Software                         Debian GNU/Linux Developer