From mboxrd@z Thu Jan  1 00:00:00 1970
Return-Path: <gcc-bugzilla@gcc.gnu.org>
Received: by sourceware.org (Postfix, from userid 48)
	id 2A3F838582A5; Sun, 18 Sep 2022 12:28:03 +0000 (GMT)
DKIM-Filter: OpenDKIM Filter v2.11.0 sourceware.org 2A3F838582A5
DKIM-Signature: v=1; a=rsa-sha256; c=relaxed/relaxed; d=gcc.gnu.org;
	s=default; t=1663504083;
	bh=i40/juoWXtwiXFJObur/Y3iHJeQVTa1Ei98x7NY8wH4=;
	h=From:To:Subject:Date:In-Reply-To:References:From;
	b=qF0AYLyOUQ2OQVb08gNy18voQsI8DmCfNkUPyvWGn3CMtpb/4WXuyGxkxTBS1t6/e
	 IQ2JUlmzA4nBFoDliWDEVLQrnwriys2N0T6V897pIj/9dk4zPJnQdiaAvpAg/Zpnnn
	 mbXds+oz68s0nDwJnjprukBFhDgHyGbkS8tu9k1w=
From: "andysem at mail dot ru" <gcc-bugzilla@gcc.gnu.org>
To: gcc-bugs@gcc.gnu.org
Subject: [Bug libstdc++/98978] Consider packing _M_Engaged in the tail
 padding of T in optional<>
Date: Sun, 18 Sep 2022 12:27:03 +0000
X-Bugzilla-Reason: CC
X-Bugzilla-Type: changed
X-Bugzilla-Watch-Reason: None
X-Bugzilla-Product: gcc
X-Bugzilla-Component: libstdc++
X-Bugzilla-Version: 11.0
X-Bugzilla-Keywords: ABI, missed-optimization
X-Bugzilla-Severity: enhancement
X-Bugzilla-Who: andysem at mail dot ru
X-Bugzilla-Status: NEW
X-Bugzilla-Resolution: 
X-Bugzilla-Priority: P3
X-Bugzilla-Assigned-To: unassigned at gcc dot gnu.org
X-Bugzilla-Target-Milestone: ---
X-Bugzilla-Flags: 
X-Bugzilla-Changed-Fields: 
Message-ID: <bug-98978-4-0vNulqZCP3@http.gcc.gnu.org/bugzilla/>
In-Reply-To: <bug-98978-4@http.gcc.gnu.org/bugzilla/>
References: <bug-98978-4@http.gcc.gnu.org/bugzilla/>
Content-Type: text/plain; charset="UTF-8"
Content-Transfer-Encoding: quoted-printable
X-Bugzilla-URL: http://gcc.gnu.org/bugzilla/
Auto-Submitted: auto-generated
MIME-Version: 1.0
List-Id: <gcc-bugs.sourceware.org>

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=3D98978

--- Comment #6 from andysem at mail dot ru ---
(In reply to Jonathan Wakely from comment #5)
> (In reply to andysem from comment #3)
> > Is there no way to improve standard components implementation? I'd imag=
ine
> > you could provide the new implementation in the new version inline name=
space
> > and still support the old ABI for backward compatibility.
>=20
> To give a more complete answer: Inline namespaces don't help, that's a my=
th.
>=20
> struct X { std::optional<bool> b; };
>=20
> Now one translation unit has X using the old ABI and one has X using the =
new
> ABI, but they're the same X. The fact that the two versions of optional a=
re
> in different namespaces is no help at all. You still have an ABI break.
>=20
> We did it for std::string and it was a multi-year effort that caused
> disruption across the industry. It's not worth doing that again to save a
> few bytes in std::optional.

You could include a hash of all members' namespaces in the X's mangled name=
. Or
maybe add an attribute that would propagate a tag to the outer class' mangl=
ed
name. Yes, that's also an ABI change in itself, but a useful one that would
allow inline namespaces to become useful.

I think a way to improve standard library components is essential. If not
inline namespaces then something else. Otherwise there is no evolution path=
 of
the standard library components.=