https://gcc.gnu.org/bugzilla/show_bug.cgi?id=91606
--- Comment #7 from Jakub Jelinek <jakub at gcc dot gnu.org> ---
(In reply to Richard Biener from comment #5)
> for example _31->__delta gets alias-set 4 and the store alias-set 6. It
> looks
> like the C++ FE creates separate pfn structure types without any TBAA measure
> here? In this case the method type we create them are different.
>
> void MyStruct::<T4316> (struct MyStruct *, int)
>
> and
>
> int MyStruct::<T431a> (struct MyStruct *, char, short int, bool)
If you are looking at these two, you are clearing looking at different
pointer-to-member.
But if I compare say the:
# .MEM_11 = VDEF <.MEM_10(D)>
D.10827._M_elems[0].D.10649.f1.__pfn = fun1;
store with
# VUSE <.MEM_8>
_1 = _31->__pfn;
it seems they actually use the same pointer-to-member struct (well, the first
one is
struct
{
void MyStruct::<T1bbe> (struct MyStruct *, int) * __pfn;
long int __delta;
}
and the second is:
const struct
{
void MyStruct::<T1bbe> (struct MyStruct *, int) * __pfn;
long int __delta;
}
but TYPE_MAIN_VARIANT of the latter is the former type.
The first of these type and a type with the different method type are in a
union (f1 is this one, f2 the other one).
Yet, get_alias_set on D.10827._M_elems[0].D.10649.f1.__pfn returns 3 and on
_31->__pfn returns 5, even when TYPE_MAIN_VARIANT (TREE_TYPE
(D.10827._M_elems[0].D.10649.f1)) == TYPE_MAIN_VARIANT (TREE_TYPE (*_31)).
Alias set for the former comes from TYPE_ALIAS_SET of the anonymous union
containing f1 and f2 fields. When that alias set has been created, it calls
record_component_aliases, but both f1 and f2 fields have these
TYPE_PTRMEMFUNC_P
types for which cxx_get_alias_set returns 0:
/* Punt on PMFs until we canonicalize functions properly. */
if (TYPE_PTRMEMFUNC_P (t)
|| (INDIRECT_TYPE_P (t)
&& TYPE_PTRMEMFUNC_P (TREE_TYPE (t))))
return 0;
The alias machinery then records that the anonymous union has zero child.
