Re: [Patch][C++/debug] SDB_DEBUG (-gcoff) vs record_builtin_type vsGAS - refresh

[Danny Smith <dannysmith at clear dot net dot nz>]

Zack Weinberg wrote:
> Danny Smith  writes:
>
>> This is a slightly revised version of patch submitted here:
>> http://gcc.gnu.org/ml/gcc-patches/2004-06/msg01394.html
> ...
>
>> +       /* Don't output intrinsic types.  GAS chokes on SDB .def
>> + statements that contain identifiers with embedded spaces
>> + (eg "unsigned long").  */
>> +       if (DECL_IS_BUILTIN (decl))
>> + return;
>
> I'm not totally sure that DECL_IS_BUILTIN is the right bit to be
> checking.  Can you please check that there's no other kind of
> TYPE_DECL node that gets this bit set?
>

DECL_IS_BUILTIN (decl)  expands to

#ifdef USE_MAPPED_LOCATION
#define DECL_IS_BUILTIN(DECL) \
  (DECL_SOURCE_LOCATION (DECL) <= BUILTINS_LOCATION)
#else
#define DECL_IS_BUILTIN(DECL) (DECL_SOURCE_LINE(DECL) == 0)
#endif

A grep for DECL_IS_BUILTIN finds the macro is always used as its name
indicates.  The only things that I could find that might suggest
a problem with its usage in sdbout.c are these in

dwarf2out.c (dwarf2out_decl):

      /* Don't bother trying to generate any DIEs to represent any of
the
  normal built-in types for the language we are compiling.  */
      if (DECL_IS_BUILTIN (decl))
 {
   /* OK, we need to generate one for `bool' so GDB knows what type
      comparisons have.  */
   if ((get_AT_unsigned (comp_unit_die, DW_AT_language)
        == DW_LANG_C_plus_plus)
       && TREE_CODE (TREE_TYPE (decl)) == BOOLEAN_TYPE
       && ! DECL_IGNORED_P (decl))
     modified_type_die (TREE_TYPE (decl), 0, 0, NULL);

   return;
 }

Should we make a special exception for 'bool' in sdbout.c too?

and this comment in function.c (init_function_start)
  /* ... Also tell final how to output a linenum before the
     function prologue.  Note linenums could be missing, e.g. when
     compiling a Java .class file.  */
  if (! DECL_IS_BUILTIN (subr))
    emit_line_note (DECL_SOURCE_LOCATION (subr));

Danny

> zw