Re: Fix frame-args.exp failure

["Ulrich Weigand" <uweigand@de.ibm.com>]

Andreas Schwab wrote:

> Daniel Jacobowitz <drow@false.org> writes:
> > Wouldn't this make any optimized-out locals be treated like frame
> > arguments?  And also see print_frame_arg_vars.  I thought there was an
> > "is an argument" flag on variables nowadays but it doesn't look like
> > it.
> 
> How about this instead?  Tested on ppc-linux.

Did you get a chance to test with non-DWARF-2 debug info as well
(at least stabs would be nice to test)?

> 2008-05-25  Andreas Schwab  <schwab@suse.de>
> 
> 	* symtab.h (enum address_class): Remove LOC_REGPARM and
> 	LOC_COMPUTED_ARG.
> 	(struct symbol): Add is_argument.
> 	(SYMBOL_IS_ARGUMENT): Define.
> 
> 	* ada-lang.c (ada_add_block_symbols): Use SYMBOL_IS_ARGUMENT.
> 	* buildsym.c (finish_block): Likewise.
> 	* stack.c (print_frame_args, print_block_frame_locals)
> 	(print_frame_arg_vars): Likewise.
> 	* symtab.c (lookup_block_symbol): Likewise.
> 	* tracepoint.c (add_local_symbols): Likewise.
> 	* mi/mi-cmd-stack.c (list_args_or_locals): Likewise.
> 
> 	* coffread.c (process_coff_symbol): Set SYMBOL_IS_ARGUMENT.
> 	* dwarf2read.c (new_symbol): Likewise.
> 	* mdebugread.c (parse_symbol): Likewise.
> 	* stabsread.c (define_symbol): Likewise.
> 
> 	* ada-exp.y (select_possible_type_sym): Don't handle LOC_REGPARM
> 	and LOC_COMPUTED_ARG.
> 	* ada-lang.c (resolve_subexp, symtab_for_sym): Likewise.
> 	* ax-gdb.c (gen_var_ref): Likewise.
> 	* eval.c (evaluate_subexp_for_address): Likewise.
> 	* findvar.c (symbol_read_needs_frame, read_var_value): Likewise.
> 	* m2-exp.y (yylex): Likewise.
> 	* printcmd.c (address_info): Likewise.
> 	* symmisc.c (print_symbol, print_partial_symbols): Likewise.
> 	* tracepoint.c (collect_symbol, scope_info): Likewise.
> 
> testsuite/:
> 	* gdb.base/frame-args.exp: Handle arguments that are optimized
> 	out.

Otherwise, this looks good to me, except ...


> Index: gdb/objc-exp.y
> ===================================================================
> RCS file: /cvs/src/src/gdb/objc-exp.y,v
> retrieving revision 1.30
> diff -u -a -p -r1.30 objc-exp.y
> --- gdb/objc-exp.y	19 May 2008 15:50:09 -0000	1.30
> +++ gdb/objc-exp.y	25 May 2008 20:53:31 -0000
> @@ -338,7 +338,7 @@ exp	: 	'[' TYPENAME
>  			    error ("%s is not an ObjC Class", 
>  				   copy_name ($2.stoken));
>  			  write_exp_elt_opcode (OP_LONG);
> -			  write_exp_elt_type (builtin_type_int);
> +			  write_exp_elt_type (builtin_type_void_data_ptr);
>  			  write_exp_elt_longcst ((LONGEST) class);
>  			  write_exp_elt_opcode (OP_LONG);
>  			  start_msglist();
> @@ -353,7 +353,7 @@ exp	: 	'[' TYPENAME
>  exp	:	'[' CLASSNAME
>  			{
>  			  write_exp_elt_opcode (OP_LONG);
> -			  write_exp_elt_type (builtin_type_int);
> +			  write_exp_elt_type (builtin_type_void_data_ptr);
>  			  write_exp_elt_longcst ((LONGEST) $2.class);
>  			  write_exp_elt_opcode (OP_LONG);
>  			  start_msglist();
> Index: gdb/parser-defs.h
> ===================================================================
> RCS file: /cvs/src/src/gdb/parser-defs.h,v
> retrieving revision 1.26
> diff -u -a -p -r1.26 parser-defs.h
> --- gdb/parser-defs.h	1 Jan 2008 22:53:12 -0000	1.26
> +++ gdb/parser-defs.h	25 May 2008 20:53:31 -0000
> @@ -83,7 +83,7 @@ struct objc_class_str
>    {
>      struct stoken stoken;
>      struct type *type;
> -    int class;
> +    CORE_ADDR class;
>    };
>  
>  

... these, which appear to be unrelated changes?


Thanks,
Ulrich

-- 
  Dr. Ulrich Weigand
  GNU Toolchain for Linux on System z and Cell BE
  Ulrich.Weigand@de.ibm.com