This is the mail archive of the gdb-patches@sources.redhat.com mailing list for the GDB project.


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]
Other format: [Raw text]

Re: [RFA] [4/5] Use DWARF-2 DW_AT_artificial information


On Thu, May 16, 2002 at 07:51:38PM -0400, Elena Zannoni wrote:
> Jim Blandy writes:
>  > 
>  > Possibly clueless suggestion:
>  > 
> 
> I think this is where Daniel was headed, he's the one that added that
> 'artificial' member. Except he ran into an HP merge roadblock.

You're thinking too highly of me, but now that you mention it it's a
really good idea... I will investigate the practicality of this.  I
believe it would be a memory increase from where we are now, since we'd
allocate a struct field for every argument, but that's not necessarily
a bad thing.  The simplicity is more important.




> 
> Elena
> 
> 
>  > Couldn't we represent a method's arguments the same way we represent a
>  > function's arguments?  That is, nfields would carry the number of
>  > arguments, and fields[i] would describe the n'th field.  We could use
>  > the `artificial' member of `union field_location' in `struct field' to
>  > hold the information conveyed by DW_AT_artificial.
>  > 
>  > The fact that we've distinguished these has caused problems in the
>  > past.  Check out the following code in hand_function_call:
>  > 
>  >   for (i = nargs - 1; i >= 0; i--)
>  >     {
>  >       /* Assume that methods are always prototyped, unless they are off the
>  > 	 end (which we should only be allowing if there is a ``...'').  
>  >          FIXME.  */
>  >       if (TYPE_CODE (ftype) == TYPE_CODE_METHOD)
>  > 	{
>  > 	  if (i < n_method_args)
>  > 	    args[i] = value_arg_coerce (args[i], TYPE_ARG_TYPES (ftype)[i], 1);
>  > 	  else
>  > 	    args[i] = value_arg_coerce (args[i], NULL, 0);
>  > 	}
>  > 
>  >       /* If we're off the end of the known arguments, do the standard
>  >          promotions.  FIXME: if we had a prototype, this should only
>  >          be allowed if ... were present.  */
>  >       if (i >= TYPE_NFIELDS (ftype))
>  > 	args[i] = value_arg_coerce (args[i], NULL, 0);
>  > 
>  >       else
>  > 	{
>  > 	  param_type = TYPE_FIELD_TYPE (ftype, i);
>  > 	  args[i] = value_arg_coerce (args[i], param_type, TYPE_PROTOTYPED (ftype));
>  > 	}
>  > 
>  > Those two loops could be re-collapsed into one.
> 

-- 
Daniel Jacobowitz                           Carnegie Mellon University
MontaVista Software                         Debian GNU/Linux Developer


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]