[Bug middle-end/103006] [9/10/11/12 Regression] wrong code at -O1 or -O2 on x86_64-linux-gnu

[pinskia at gcc dot gnu.org via Gcc-bugs]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=103006

Andrew Pinski <pinskia at gcc dot gnu.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
            Summary|[9/10/11/12 Regression]     |[9/10/11/12 Regression]
                   |wrong code at -O2 (only) on |wrong code at -O1 or -O2 on
                   |x86_64-linux-gnu            |x86_64-linux-gnu
          Component|tree-optimization           |middle-end
      Known to fail|                            |8.1.0
   Target Milestone|---                         |9.5
           Keywords|                            |wrong-code

--- Comment #2 from Andrew Pinski <pinskia at gcc dot gnu.org> ---
Since noipa does not work with GCC 7, the following fails all the back to GCC
7:

__attribute__((noipa,noinline,noclone)) void ff(void){asm("":::"memory");}
int a, *b, c, e, f;
__attribute__((always_inline))
static inline void g() {
  int *d[7];
  d[6] = b = (int *)d;
  ff();
}
__attribute__((noinline))
int i() {
  for (c = 0; c < 2; c++) {
    long h[6][2];
    for (e = 0; e < 6; e++)
      for (f = 0; f < 2; f++)
        h[e][f] = 1;
    if (c) {
      g();
      return h[3][0];
    }
  }
  return 0;
}
int main() {
  if (i() != 1)
    __builtin_abort ();
  return 0;
}