[Bug middle-end/59470] [4.8 Regression] libstdc++ miscompilation after r205709

[jakub at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=59470

--- Comment #10 from Jakub Jelinek <jakub at gcc dot gnu.org> ---
Re: the #c7 second testcase, rth thinks it is undefined behavior since there is
no sequence point.  So here is a better testcase that should have defined
behavior, still before r205709 we miscompile it at -O2:

int a, b, d[1024];
int
main ()
{
  int c = a;
  asm ("movl $6, (%2); movl $1, %0" : "=r" (d[c])
       : "rm" (b), "r" (&a) : "memory");
  if (d[0] != 1 || d[6] != 0)
    __builtin_abort ();
  return 0;
}