[Moved to DR at the November, 2016 meeting as paper P0507R0.]

Proposed resolution (June, 2016):

The resolution is given in paper P0507R0.

Additional note (October, 2015):

For additional examples, consider:

   int i = i++;
   int j = j;

The current sequencing rules do not cover these initializations.

Also, in

  const int& f(const int& x) { return x; }
  int y = f(5);

it doesn't appear that 6.9.1 [intro.execution] paragraph 10 requires the temporary for 5 to persist until the initialization of y is complete.

Additional note (February, 2014):

Aggregate initialization could also involve more than one full-expression, so the limitation above to “initialization of a non-class object” is not correct.

Additional note, April, 2013:

There is implementation variance in the treatment of the following example:

  struct A {
    A() { puts("ctor"); }
    A(const A&) { puts("copy"); }
    const A&get() const { return *this; }
    ~A() { puts("dtor"); }
  };
  struct B { B(A, A) {} };

  typedef A A2[2];
  A2 a = { A().get(), A().get() };
  B b = { A().get(), A().get() };
  int c = (A2{ A().get(), A().get() }, 0);
  int d = (B{ A().get(), A().get() }, 0);

  int main() {}

The current wording does not indicate that initialization of a non-class object is a full-expression, but presumably should do so.