[Bug libstdc++/40297] New: [C++0x] debug mode vs atomics

[jwakely dot gcc at gmail dot com]

The atomic types in <cstdatomic> have debug-mode assertions that will always
fail.

#define _GLIBCXX_DEBUG
#include <cstdatomic>

int main()
{
    std::atomic_address ptr;
    return ptr.load() ? 0 : 1;
}

Fails with:

/dev/shm/wakelyjo/insroot/gcc/lib/gcc/x86_64-unknown-linux-gnu/4.5.0/../../../../include/c++/4.5.0/bits/atomic_2.h:114:
void* std::__atomic2::atomic_address::load(std::memory_order) const volatile:
Assertion '__m == memory_order_release' failed.

Due to:
    void*
    load(memory_order __m = memory_order_seq_cst) const volatile
    {
      __glibcxx_assert(__m == memory_order_release);
      __glibcxx_assert(__m == memory_order_acq_rel);

      __sync_synchronize();
      void* __ret = _M_i;
      __sync_synchronize();
      return __ret;
    }

Since memory_order is an enumeration and each enumerator has a distinct value,
it's not possible for both assertions to pass.

I'm not sure what they were intended to check, unless it's to prevent atomics
being used in debug mode!


-- 
           Summary: [C++0x] debug mode vs atomics
           Product: gcc
           Version: 4.5.0
            Status: UNCONFIRMED
          Severity: normal
          Priority: P3
         Component: libstdc++
        AssignedTo: unassigned at gcc dot gnu dot org
        ReportedBy: jwakely dot gcc at gmail dot com


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=40297