[Bug tree-optimization/78895] [6 Regression] wrong code with -O1 when setting union twice since 6.3

[jakub at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=78895

Jakub Jelinek <jakub at gcc dot gnu.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
           Priority|P3                          |P1
             Status|UNCONFIRMED                 |NEW
   Last reconfirmed|                            |2016-12-22
                 CC|                            |jakub at gcc dot gnu.org
   Target Milestone|---                         |6.4
            Summary|[6 regression] wrong code   |[6 Regression] wrong code
                   |with -O1 when setting union |with -O1 when setting union
                   |twice since 6.3             |twice since 6.3
     Ever confirmed|0                           |1

--- Comment #1 from Jakub Jelinek <jakub at gcc dot gnu.org> ---
Reduced testcase:

union U { int i; _Bool b; };

void __attribute__((noinline, noclone))
foo (union U u)
{
  if (u.b)
    __builtin_abort ();
}

int
main ()
{
  union U u;
  u.i = 10;
  u.b = 0;
  foo (u);
  return 0;
}

It is fre1 that removes the u.b = 0; store.  Doesn't fail on the trunk, testing
now if it ever failed on trunk this fall.