[Bug libstdc++/54025] New: atomic<chrono::duration> won't compile: chrono::duration::duration() is not C++11 compliant

[Bug libstdc++/54025] New: atomic<chrono::duration> won't compile: chrono::duration::duration() is not C++11 compliant

[chip at pobox dot com]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=54025

             Bug #: 54025
           Summary: atomic<chrono::duration> won't compile:
                    chrono::duration::duration() is not C++11 compliant
    Classification: Unclassified
           Product: gcc
           Version: 4.7.1
            Status: UNCONFIRMED
          Severity: normal
          Priority: P3
         Component: libstdc++
        AssignedTo: unassigned@gcc.gnu.org
        ReportedBy: chip@pobox.com


Attempting to compile atomic<duration> fails, because the duration default
constructor is not " = default" as required by the standard, but instead
explicitly initializes its representation.  Here is what libstdc++ says:

    constexpr duration() : __r() { }

but here is what the standard says should be there, and if I make the change,
compilation succeeds:

    constexpr duration() = default;

Test source:

#include <atomic>
#include <chrono>
using namespace std;
using namespace chrono;
int main() {
    atomic<duration<long, micro>> dur;
}

Error before patch:

/usr/include/c++/4.7/atomic: In instantiation of ‘struct
std::atomic<std::chrono::duration<long int, std::ratio<1ll, 1000000ll> > >’:
atdur.cc:6:35:   required from here
/usr/include/c++/4.7/atomic:160:7: error: function ‘std::atomic<_Tp>::atomic()
[with _Tp = std::chrono::duration<long int, std::ratio<1ll, 1000000ll> >]’
defaulted on its first declaration with an exception-specification that differs
from the implicit declaration ‘constexpr std::atomic<std::chrono::duration<long
int, std::ratio<1ll, 1000000ll> > >::atomic()’

[Bug libstdc++/54025] atomic<chrono::duration> won't compile: chrono::duration::duration() is not C++11 compliant

[chip at pobox dot com]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=54025

--- Comment #1 from Chip Salzenberg <chip at pobox dot com> 2012-07-19 02:56:57 UTC ---
Created attachment 27829
  --> http://gcc.gnu.org/bugzilla/attachment.cgi?id=27829
patch to duration default ctor

[Bug libstdc++/54025] atomic<chrono::duration> won't compile: chrono::duration::duration() is not C++11 compliant

[redi at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=54025

Jonathan Wakely <redi at gcc dot gnu.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
           Keywords|                            |rejects-valid
             Status|UNCONFIRMED                 |NEW
   Last reconfirmed|                            |2012-07-19
     Ever Confirmed|0                           |1

--- Comment #2 from Jonathan Wakely <redi at gcc dot gnu.org> 2012-07-19 08:34:03 UTC ---
confirmed

[Bug libstdc++/54025] atomic<chrono::duration> won't compile: chrono::duration::duration() is not C++11 compliant

[paolo.carlini at oracle dot com]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=54025

Paolo Carlini <paolo.carlini at oracle dot com> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
                 CC|                            |redi at gcc dot gnu.org

--- Comment #3 from Paolo Carlini <paolo.carlini at oracle dot com> 2012-07-19 09:40:07 UTC ---
Jon, shall I just apply patch and testcase?

[Bug libstdc++/54025] atomic<chrono::duration> won't compile: chrono::duration::duration() is not C++11 compliant

[redi at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=54025

--- Comment #4 from Jonathan Wakely <redi at gcc dot gnu.org> 2012-07-19 10:20:07 UTC ---
Yes, I think so.  I initially wondered if there was some interaction with PR
53901 but I think that just makes the bug visible, duration's constructor
should be explicitly-defaulted anyway.

[Bug libstdc++/54025] atomic<chrono::duration> won't compile: chrono::duration::duration() is not C++11 compliant

[paolo.carlini at oracle dot com]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=54025

--- Comment #5 from Paolo Carlini <paolo.carlini at oracle dot com> 2012-07-19 11:57:04 UTC ---
If we just do the change and nothing else, 20_util/duration/cons/constexpr.cc
doesn't compile anymore with this error:

/scratch/Gcc/svn-dirs/trunk/libstdc++-v3/testsuite/20_util/duration/cons/constexpr.cc:27:42:
  required from here
/scratch/Gcc/svn-dirs/trunk/libstdc++-v3/testsuite/util/testsuite_common_types.h:698:18:
error: uninitialized const '__obj' [-fpermissive]
  { constexpr _Tp __obj; }
                  ^
In file included from
/scratch/Gcc/svn-dirs/trunk/libstdc++-v3/testsuite/20_util/duration/cons/constexpr.cc:21:0:
/home/paolo/Gcc/svn-dirs/trunk-build/x86_64-unknown-linux-gnu/libstdc++-v3/include/chrono:220:14:
note: 'const struct std::chrono::duration<long int>' has no user-provided
default constructor
       struct duration
              ^
/home/paolo/Gcc/svn-dirs/trunk-build/x86_64-unknown-linux-gnu/libstdc++-v3/include/chrono:231:12:
note: constructor is not user-provided because it is explicitly defaulted in
the class body
  constexpr duration() = default;
            ^
/home/paolo/Gcc/svn-dirs/trunk-build/x86_64-unknown-linux-gnu/libstdc++-v3/include/chrono:349:6:
note: and the implicitly-defined constructor does not initialize
'std::chrono::duration<long int>::rep std::chrono::duration<long int>::__r'
  rep __r;
      ^

[Bug libstdc++/54025] atomic<chrono::duration> won't compile: chrono::duration::duration() is not C++11 compliant

[redi at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=54025

--- Comment #6 from Jonathan Wakely <redi at gcc dot gnu.org> 2012-07-19 12:06:32 UTC ---
I think that test is wrong, a duration is only constexpr_default_constructible
if the rep type has a default-constructor, but std::chrono::seconds uses a
scalar rep.

I think the test should use something like this to check duration is
constexpr-default-constructible:

struct Seconds {
    constexpr Seconds() = default;
    std::chrono::seconds s{};
};
test1.operator()<std::chrono::duration<Seconds>>();

[Bug libstdc++/54025] atomic<chrono::duration> won't compile: chrono::duration::duration() is not C++11 compliant

[redi at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=54025

--- Comment #7 from Jonathan Wakely <redi at gcc dot gnu.org> 2012-07-19 12:08:29 UTC ---
(In reply to comment #6)
> I think that test is wrong, a duration is only constexpr_default_constructible
> if the rep type has a default-constructor

... that initializes all its members.

[Bug libstdc++/54025] atomic<chrono::duration> won't compile: chrono::duration::duration() is not C++11 compliant

[paolo.carlini at oracle dot com]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=54025

--- Comment #8 from Paolo Carlini <paolo.carlini at oracle dot com> 2012-07-19 12:17:13 UTC ---
Understood, thanks. A couple of times in the past I already tweeked a bit some
of those tests. Let me see...

[Bug libstdc++/54025] atomic<chrono::duration> won't compile: chrono::duration::duration() is not C++11 compliant

[redi at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=54025

--- Comment #9 from Jonathan Wakely <redi at gcc dot gnu.org> 2012-07-19 12:19:05 UTC ---
Another option would be to give duration::__rep a NSDMI, which leaves the
testcase valid (as a QoI feature, I believe)

[Bug libstdc++/54025] atomic<chrono::duration> won't compile: chrono::duration::duration() is not C++11 compliant

[paolo.carlini at oracle dot com]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=54025

--- Comment #10 from Paolo Carlini <paolo.carlini at oracle dot com> 2012-07-19 12:23:40 UTC ---
I see. For now I would rather just minimally tweak to testcase. If you want to
play a bit more with this and experiment with the various options, just let me
know.

[Bug libstdc++/54025] atomic<chrono::duration> won't compile: chrono::duration::duration() is not C++11 compliant

[redi at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=54025

--- Comment #11 from Jonathan Wakely <redi at gcc dot gnu.org> 2012-07-19 12:34:19 UTC ---
Oh, I think our current code might be intentional, we should ask Benjamin:
http://www.open-std.org/jtc1/sc22/wg21/docs/papers/2011/n3303.html

See c++std-lib-32464 for rational for the std semantics i.e.
duration<trivial_type> is a trivial type, and leaves the rep uninitialized.

[Bug libstdc++/54025] atomic<chrono::duration> won't compile: chrono::duration::duration() is not C++11 compliant

[paolo.carlini at oracle dot com]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=54025

Paolo Carlini <paolo.carlini at oracle dot com> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
                 CC|                            |bkoz at gcc dot gnu.org,
                   |                            |daniel.kruegler at
                   |                            |googlemail dot com

--- Comment #12 from Paolo Carlini <paolo.carlini at oracle dot com> 2012-07-19 12:48:19 UTC ---
Ah! Let's add in CC both Benjamin and Daniel then.

[Bug libstdc++/54025] atomic<chrono::duration> won't compile: chrono::duration::duration() is not C++11 compliant

[daniel.kruegler at googlemail dot com]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=54025

--- Comment #13 from Daniel Krügler <daniel.kruegler at googlemail dot com> 2012-07-19 15:18:16 UTC ---
(In reply to comment #12)
> Ah! Let's add in CC both Benjamin and Daniel then.

I more and more tend to change my mind: I recommend to ignore the
recommendation of N3303 and to make duration conforming with the current
library spec. The described use-case here makes very much sense to me and
demonstrates to me that N3303 did overreact here. But I would be happy to hear
Benjamin's position as well.

[Bug libstdc++/54025] atomic<chrono::duration> won't compile: chrono::duration::duration() is not C++11 compliant

[paolo.carlini at oracle dot com]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=54025

Paolo Carlini <paolo.carlini at oracle dot com> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|NEW                         |ASSIGNED
         AssignedTo|unassigned at gcc dot       |paolo.carlini at oracle dot
                   |gnu.org                     |com
   Target Milestone|---                         |4.8.0

--- Comment #14 from Paolo Carlini <paolo.carlini at oracle dot com> 2012-07-19 15:25:39 UTC ---
Thanks Daniel. Therefore let's wait a bit in case Benjamin disagrees, otherwise
I'll take care of applying patchlet + testcase + tweak to the existing
constexpr testcase,

[Bug libstdc++/54025] atomic<chrono::duration> won't compile: chrono::duration::duration() is not C++11 compliant

[paolo at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=54025

--- Comment #15 from paolo at gcc dot gnu.org <paolo at gcc dot gnu.org> 2012-07-20 09:49:02 UTC ---
Author: paolo
Date: Fri Jul 20 09:48:57 2012
New Revision: 189711

URL: http://gcc.gnu.org/viewcvs?root=gcc&view=rev&rev=189711
Log:
2012-07-20  Chip Salzenberg  <chip@pobox.com>
        Jonathan Wakely  <jwakely.gcc@gmail.com>

    PR libstdc++/54025
    * include/std/chrono (duration<>::duration): Fix per C++11.
    * testsuite/20_util/duration/cons/54025.cc: New.
    * testsuite/20_util/duration/cons/constexpr.cc: Adjust.

Added:
    trunk/libstdc++-v3/testsuite/20_util/duration/cons/54025.cc
Modified:
    trunk/libstdc++-v3/ChangeLog
    trunk/libstdc++-v3/include/std/chrono
    trunk/libstdc++-v3/testsuite/20_util/duration/cons/constexpr.cc

[Bug libstdc++/54025] atomic<chrono::duration> won't compile: chrono::duration::duration() is not C++11 compliant

[paolo.carlini at oracle dot com]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=54025

Paolo Carlini <paolo.carlini at oracle dot com> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|ASSIGNED                    |RESOLVED
            Version|4.7.1                       |4.8.0
         Resolution|                            |FIXED
         AssignedTo|paolo.carlini at oracle dot |unassigned at gcc dot
                   |com                         |gnu.org

--- Comment #16 from Paolo Carlini <paolo.carlini at oracle dot com> 2012-07-20 09:51:39 UTC ---
Done.