m2-typeprint.c:m2_range's use of TYPE_DOMAIN_TYPE: Eh?

m2-typeprint.c:m2_range's use of TYPE_DOMAIN_TYPE: Eh?

[Doug Evans]

Hi.

Does anyone know why m2_range uses TYPE_DOMAIN_TYPE here?

AFAICT, use of TYPE_DOMAIN_TYPE is wrong here.
It is only used for TYPE_CODE_{MEMBERPTR,METHODPTR,METHOD}.

ref: gdbtypes.h:

  /* * For types with virtual functions (TYPE_CODE_STRUCT),
     VPTR_BASETYPE is the base class which defined the virtual
     function table pointer.

     For types that are pointer to member types (TYPE_CODE_METHODPTR,
     TYPE_CODE_MEMBERPTR), VPTR_BASETYPE is the type that this pointer
     is a member of.

     For method types (TYPE_CODE_METHOD), VPTR_BASETYPE is the aggregate
     type that contains the method.

     Unused otherwise.  */

  struct type *vptr_basetype;

I ask because one symbol table improvement I want to make
may be best implemented if I add a new field to struct main_type.
However, this struct is space-critical, so I'm trying to find
some cleanup that can be done.
Question: Why doesn't vptr_fieldno, vptr_basetype live in
type_specific.cplus_specific?

Alas, vptr_basetype is overloaded and is used also for TYPE_DOMAIN_TYPE:

#define TYPE_DOMAIN_TYPE(thistype) TYPE_MAIN_TYPE(thistype)->vptr_basetype

hence wondering why in the world m2_range is using it. :-)

For TYPE_CODE_{MEMBERPTR,METHODPTR,METHOD},
any reason why their "domain type" (I'll want to rename that)
cannot live in type_specific?

---

void
m2_range (struct type *type, struct ui_file *stream, int show,
	  int level, const struct type_print_options *flags)
{
  if (TYPE_HIGH_BOUND (type) == TYPE_LOW_BOUND (type))
    m2_print_type (TYPE_DOMAIN_TYPE (type), "", stream, show, level,
		   flags);
  else
    {
      struct type *target = TYPE_TARGET_TYPE (type);

      fprintf_filtered (stream, "[");
      print_type_scalar (target, TYPE_LOW_BOUND (type), stream);
      fprintf_filtered (stream, "..");
      print_type_scalar (target, TYPE_HIGH_BOUND (type), stream);
      fprintf_filtered (stream, "]");
    }
}

Re: m2-typeprint.c:m2_range's use of TYPE_DOMAIN_TYPE: Eh?

[Gaius Mulley]

Doug Evans <xdje42@gmail.com> writes:

> Hi.
>
> Does anyone know why m2_range uses TYPE_DOMAIN_TYPE here?
>
> AFAICT, use of TYPE_DOMAIN_TYPE is wrong here.
> It is only used for TYPE_CODE_{MEMBERPTR,METHODPTR,METHOD}.
>
> ref: gdbtypes.h:
>
>   /* * For types with virtual functions (TYPE_CODE_STRUCT),
>      VPTR_BASETYPE is the base class which defined the virtual
>      function table pointer.
>
>      For types that are pointer to member types (TYPE_CODE_METHODPTR,
>      TYPE_CODE_MEMBERPTR), VPTR_BASETYPE is the type that this pointer
>      is a member of.
>
>      For method types (TYPE_CODE_METHOD), VPTR_BASETYPE is the aggregate
>      type that contains the method.
>
>      Unused otherwise.  */
>
>   struct type *vptr_basetype;
>
> I ask because one symbol table improvement I want to make
> may be best implemented if I add a new field to struct main_type.
> However, this struct is space-critical, so I'm trying to find
> some cleanup that can be done.
> Question: Why doesn't vptr_fieldno, vptr_basetype live in
> type_specific.cplus_specific?
>
> Alas, vptr_basetype is overloaded and is used also for TYPE_DOMAIN_TYPE:
>
> #define TYPE_DOMAIN_TYPE(thistype) TYPE_MAIN_TYPE(thistype)->vptr_basetype
>
> hence wondering why in the world m2_range is using it. :-)

Hi Doug,

feel free to clean up the code and make it comply with other languages.
I've a feeling that this code will need an overhaul against gm2-1.1.1
(grafted on gcc-4.7.4).  The gcc/gm2 interface has pretty much all
been rewritten since gm2-1.0 (which was grafted onto gcc-4.1.2).

> For TYPE_CODE_{MEMBERPTR,METHODPTR,METHOD},
> any reason why their "domain type" (I'll want to rename that)
> cannot live in type_specific?
>
> ---
>
> void
> m2_range (struct type *type, struct ui_file *stream, int show,
> 	  int level, const struct type_print_options *flags)
> {
>   if (TYPE_HIGH_BOUND (type) == TYPE_LOW_BOUND (type))
>     m2_print_type (TYPE_DOMAIN_TYPE (type), "", stream, show, level,
> 		   flags);
>   else
>     {
>       struct type *target = TYPE_TARGET_TYPE (type);
>
>       fprintf_filtered (stream, "[");
>       print_type_scalar (target, TYPE_LOW_BOUND (type), stream);
>       fprintf_filtered (stream, "..");
>       print_type_scalar (target, TYPE_HIGH_BOUND (type), stream);
>       fprintf_filtered (stream, "]");
>     }
> }

regards,
Gaius

Re: m2-typeprint.c:m2_range's use of TYPE_DOMAIN_TYPE: Eh?

[Doug Evans]

On Tue, Jan 20, 2015 at 2:25 PM, Gaius Mulley
<gaius.mulley@southwales.ac.uk> wrote:
> Doug Evans <xdje42@gmail.com> writes:
>
>> Hi.
>>
>> Does anyone know why m2_range uses TYPE_DOMAIN_TYPE here?
>>
>> [...]
>> Alas, vptr_basetype is overloaded and is used also for TYPE_DOMAIN_TYPE:
>>
>> #define TYPE_DOMAIN_TYPE(thistype) TYPE_MAIN_TYPE(thistype)->vptr_basetype
>>
>> hence wondering why in the world m2_range is using it. :-)
>
> Hi Doug,
>
> feel free to clean up the code and make it comply with other languages.
> I've a feeling that this code will need an overhaul against gm2-1.1.1
> (grafted on gcc-4.7.4).  The gcc/gm2 interface has pretty much all
> been rewritten since gm2-1.0 (which was grafted onto gcc-4.1.2).

Thanks.
Alas cleaning up this code is non-obvious.
What do I replace TYPE_DOMAIN_TYPE with?
I'm tempted to just go with TYPE_TARGET_TYPE
and let y'all fix it later if that's wrong.

>
>> For TYPE_CODE_{MEMBERPTR,METHODPTR,METHOD},
>> any reason why their "domain type" (I'll want to rename that)
>> cannot live in type_specific?
>>
>> ---
>>
>> void
>> m2_range (struct type *type, struct ui_file *stream, int show,
>>         int level, const struct type_print_options *flags)
>> {
>>   if (TYPE_HIGH_BOUND (type) == TYPE_LOW_BOUND (type))
>>     m2_print_type (TYPE_DOMAIN_TYPE (type), "", stream, show, level,
>>                  flags);
>>   else
>>     {
>>       struct type *target = TYPE_TARGET_TYPE (type);
>>
>>       fprintf_filtered (stream, "[");
>>       print_type_scalar (target, TYPE_LOW_BOUND (type), stream);
>>       fprintf_filtered (stream, "..");
>>       print_type_scalar (target, TYPE_HIGH_BOUND (type), stream);
>>       fprintf_filtered (stream, "]");
>>     }
>> }
>
> regards,
> Gaius

Re: m2-typeprint.c:m2_range's use of TYPE_DOMAIN_TYPE: Eh?

[Gaius Mulley]

Doug Evans <xdje42@gmail.com> writes:

>
> Thanks.
> Alas cleaning up this code is non-obvious.
> What do I replace TYPE_DOMAIN_TYPE with?
> I'm tempted to just go with TYPE_TARGET_TYPE
> and let y'all fix it later if that's wrong.

Hi Doug,

sure this sounds pragmatic - I'll look at the code next week
in the light of gm2-1.1.1

regards,
Gaius