public inbox for gcc-bugs@sourceware.org
help / color / mirror / Atom feed
From: "420 at zerberste dot es" <gcc-bugzilla@gcc.gnu.org>
To: gcc-bugs@gcc.gnu.org
Subject: [Bug c++/114709] New: Incorrect handling of inactive union member access via pointer to member in constant evaluated context
Date: Sat, 13 Apr 2024 01:59:22 +0000	[thread overview]
Message-ID: <bug-114709-4@http.gcc.gnu.org/bugzilla/> (raw)

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=114709

            Bug ID: 114709
           Summary: Incorrect handling of inactive union member access via
                    pointer to member in constant evaluated context
           Product: gcc
           Version: unknown
            Status: UNCONFIRMED
          Severity: normal
          Priority: P3
         Component: c++
          Assignee: unassigned at gcc dot gnu.org
          Reporter: 420 at zerberste dot es
  Target Milestone: ---

According to https://standards.pydong.org/c++/class.union#general-note-1 an
inactive union member may be used to inspect the initial sequence common with
the active union member, given both are struct types. Furthermore
https://standards.pydong.org/c++/class.mem#general-26 tells us the behavior of
reading a non-static member m of such a common initial sequence is _as if_ the
corresponding member were nominated. Since it is not expression-equivalent, we
are still accessing the wrong member as far as the abstract machine is
concerned. Hence https://standards.pydong.org/c++/expr.const#5.10 still
applies.

GCC implements this correctly for the simple case. To reuse the example from
[class.mem]/26 consider this code fragment:
```cpp
struct T1 { int a, b; };
struct T2 { int c; double d; };
union U { T1 t1; T2 t2; };

consteval int f() {
  U u = { { 1, 2 } }; // active member is t1
  return u.t2.c;      // access through t2
}

static constexpr auto foo = f();
```
https://godbolt.org/z/x14n47T45

GCC correctly errors with the diagnostic 
> accessing 'U::t2' member instead of initialized 'U::t1' member in constant expression


However, consider the following code fragment:
```cpp
struct T1 { int a, b; };
struct T2 { int c; double d; };
union U { T1 t1; T2 t2; };

consteval int f() {
  U u = { { 1, 2 } };  // active member is t1
  return u.t2.*&T2::c; // access through t2
}

static constexpr auto foo = f();
```
https://godbolt.org/z/dK95jEPjq

GCC 12 up until trunk accept this code. GCC 11 and Clang don't.

             reply	other threads:[~2024-04-13  1:59 UTC|newest]

Thread overview: 9+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2024-04-13  1:59 420 at zerberste dot es [this message]
2024-04-13  8:51 ` [Bug c++/114709] " pinskia at gcc dot gnu.org
2024-04-13  8:52 ` [Bug c++/114709] [12/13/14 Regression] " pinskia at gcc dot gnu.org
2024-04-17 13:33 ` law at gcc dot gnu.org
2024-04-19  7:19 ` [Bug c++/114709] [12/13/14 Regression] Incorrect handling of inactive union member access via pointer to member in constant evaluated context since r12-6425 jakub at gcc dot gnu.org
2024-04-19 17:51 ` ppalka at gcc dot gnu.org
2024-04-24 21:50 ` cvs-commit at gcc dot gnu.org
2024-04-24 21:52 ` [Bug c++/114709] [12/13 " ppalka at gcc dot gnu.org
2024-05-13 16:43 ` cvs-commit at gcc dot gnu.org

Reply instructions:

You may reply publicly to this message via plain-text email
using any one of the following methods:

* Save the following mbox file, import it into your mail client,
  and reply-to-all from there: mbox

  Avoid top-posting and favor interleaved quoting:
  https://en.wikipedia.org/wiki/Posting_style#Interleaved_style

* Reply using the --to, --cc, and --in-reply-to
  switches of git-send-email(1):

  git send-email \
    --in-reply-to=bug-114709-4@http.gcc.gnu.org/bugzilla/ \
    --to=gcc-bugzilla@gcc.gnu.org \
    --cc=gcc-bugs@gcc.gnu.org \
    /path/to/YOUR_REPLY

  https://kernel.org/pub/software/scm/git/docs/git-send-email.html

* If your mail client supports setting the In-Reply-To header
  via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for read-only IMAP folder(s) and NNTP newsgroup(s).