public inbox for gcc-bugs@sourceware.org
help / color / mirror / Atom feed
From: "redi at gcc dot gnu.org" <gcc-bugzilla@gcc.gnu.org>
To: gcc-bugs@gcc.gnu.org
Subject: [Bug libstdc++/103240] New: std::type_info::before gives wrong answer for ARM EABI
Date: Mon, 15 Nov 2021 00:20:33 +0000	[thread overview]
Message-ID: <bug-103240-4@http.gcc.gnu.org/bugzilla/> (raw)

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=103240

            Bug ID: 103240
           Summary: std::type_info::before gives wrong answer for ARM EABI
           Product: gcc
           Version: 12.0
            Status: UNCONFIRMED
          Keywords: ABI
          Severity: normal
          Priority: P3
         Component: libstdc++
          Assignee: unassigned at gcc dot gnu.org
          Reporter: redi at gcc dot gnu.org
  Target Milestone: ---
            Target: arm*-*-*gnueabi

This passes on x86_64 but fails on armv7hl:

// f1.C
#include <typeinfo>

namespace
{
  struct S { };
}

std::type_info const& t1 = typeid(S);

// f2.C
#include <typeinfo>
#include <cassert>

extern std::type_info const& t1;

namespace
{
  struct S { };
}

std::type_info const& t2 = typeid(S);

int main()
{
  if (t1 == t2)
  {
    assert( !t1.before(t2) );
    assert( !t2.before(t1) );
  }
  else
  {
    assert( t1.before(t2) || t2.before(t1) );
  }
}

$ g++ f1.C f2.C
$ ./a.out
a.out: f2.C:22: int main(): Assertion `t1.before(t2) || t2.before(t1)' failed.
qemu: uncaught target signal 6 (Aborted) - core dumped
Aborted (core dumped)

The problem is in libsupc++/tinfo2.cc

bool
type_info::before (const type_info &arg) const _GLIBCXX_NOEXCEPT
{
#if __GXX_MERGED_TYPEINFO_NAMES
  return name () < arg.name ();
#else
  return (name ()[0] == '*') ? name () < arg.name ()
    :  __builtin_strcmp (name (), arg.name ()) < 0;
#endif
}

This should have been fixed like operator== was in r179236 so that it checks
__name[0] not name()[0], otherwise the '*' is skipped and so is never found
there. That means we do a strcmp comparison of the mangled names, even when the
'*' says the types must be compared by pointer address. In the testcase the two
structs have the same mangled name, even though they are distinct types (due to
the unnamed namespaces). The string comparison therefore returns 0, and so
neither type is "before" the other, despite being distinct types.

             reply	other threads:[~2021-11-15  0:20 UTC|newest]

Thread overview: 12+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2021-11-15  0:20 redi at gcc dot gnu.org [this message]
2021-11-15  0:21 ` [Bug libstdc++/103240] " redi at gcc dot gnu.org
2021-11-15  0:36 ` redi at gcc dot gnu.org
2021-11-17 17:22 ` cvs-commit at gcc dot gnu.org
2021-11-17 17:42 ` redi at gcc dot gnu.org
2021-11-23 21:17 ` cvs-commit at gcc dot gnu.org
2023-04-29 16:15 ` frankhb1989 at gmail dot com
2023-04-29 18:01 ` frankhb1989 at gmail dot com
2023-05-02 16:56 ` redi at gcc dot gnu.org
2023-05-04 10:31 ` frankhb1989 at gmail dot com
2023-05-04 11:04 ` redi at gcc dot gnu.org
2023-05-05 16:28 ` frankhb1989 at gmail dot com

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-103240-4@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).