Re: [Patch, Fortran] PR 41580 Add compile-time simplification for Same_Type_As and Extends_Type_Of

[Janus Weil <janus at gcc dot gnu dot org>]

Hi Tobias,

> This patch adds support for Same_Type_As and Extends_Type_Of; the patch is
> relatively straight forward, except that after gfc_resolve_* the simplify
> functions are called again - thus one needs to exit early otherwise the
> BT_DERIVED (as the vtab is internally represented) is simplified to .FALSE.
>
> Build and regtested on x86-64-linux.
> OK for the trunk? -- If so, which one: 4.6 or 4.7?

I think in principle your patch is ok, and I would like to see it in 4.6.

Some comments:

1) I don't particularly like the following part:

@@ -240,6 +240,7 @@ gfc_build_class_symbol (gfc_typespec *ts,
symbol_attribute *attr,
 	}
       c->attr.access = ACCESS_PRIVATE;
       c->attr.pointer = 1;
+      c->attr.vtab = 1;
     }


Up to now we are using the "vtab" attribute to indicate that a symbol
is a vtable entry for a particular type. Now here you use it to mark a
"_vptr" component of a class container, which is a different thing (it
usually *points* to a vtab), and I'm not sure if this mix-up could
create problems somehow.

If you want to check for a _vptr component, you could just check for
the name (I think I've also been doing this in a recent patch).


2) A very minor thing: I think "DONT_EXISTS" is not a very
grammatically correct name for a subroutine. Not that subroutine names
are required to be grammatically correct in any Fortran standard I
know of, but I just stumbled over it ;)


3) Could you expand a bit on why a BT_DERIVED would be simplified to
.false.? I don't quite get that part ...


Thanks for working on this!

Cheers,
Janus