This is the mail archive of the
gcc-patches@gcc.gnu.org
mailing list for the GCC project.
Re: Adjust empty class parameter passing ABI (PR c++/60336)
On Mon, Nov 6, 2017 at 10:37 AM, Marek Polacek <polacek@redhat.com> wrote:
> On Fri, Nov 03, 2017 at 12:19:05PM -0400, Jason Merrill wrote:
>> On Fri, Nov 3, 2017 at 9:55 AM, Marek Polacek <polacek@redhat.com> wrote:
>> > + TYPE_EMPTY_P (t) = targetm.calls.empty_record_p (t);
>>
>> I think we want to set this in finalize_type_size; since the point of
>> all this is becoming compliant with the psABI (and compatible with the
>> C front end), I wouldn't think it should be specific to the C++ front
>> end.
>
> Sure, that works. I'd wanted to do the TYPE_EMPTY_P setting in layout_type,
> but that wasn't called for the classes in my testcases. I've moved the
> setting of TYPE_EMPTY_P to finalize_type_size now.
>
>> > + TYPE_WARN_EMPTY_P (t) = warn_abi && abi_version_crosses (12);
>>
>> Can this flag go on the TRANSLATION_UNIT_DECL rather than the type?
>
> Yeah, that works, too. To avoid needing a lang hook, I'm setting the
> flag in cxx_init_decl_processing, hope that's ok.
Sure.
>> > + if (TREE_CODE (field) == FIELD_DECL
>> > + && (DECL_NAME (field)
>> > + || RECORD_OR_UNION_TYPE_P (TREE_TYPE (field)))
>> > + && !default_is_empty_type (TREE_TYPE (field)))
>> > + return false;
>> > + return true;
>>
>> Hmm, this assumes that any unnamed field can be ignored; I'm concerned
>> that some front end might clear DECL_NAME for a reason that doesn't
>> imply that the field is just padding.
>
> In that case I guess we need a new lang hook, right? Because in
> default_is_empty_type we can't check FE-specific flags such as
> DECL_C_BIT_FIELD. For C++, should that lang hook be just
> is_really_empty_class? Or is there anything else I can do?
Hmm, maybe leave it as it is and just document this assumption about
FIELD_DECL with null DECL_NAME, both here and in tree.def.
Jason