On Wed, Oct 29, 2014 at 11:24 AM, Marek Polacek <pola...@redhat.com> wrote: > PR52769 reports a bug that has been fixed in 4.7, but the test case > was never added. So I'd like to put this test in and close PR52769. > > Ok?
Ok everywhere. Thanks, Richard. > 2014-10-29 Marek Polacek <pola...@redhat.com> > > PR c/52769 > * gcc.dg/pr52769.c: New test. > > diff --git gcc/testsuite/gcc.dg/pr52769.c gcc/testsuite/gcc.dg/pr52769.c > index e69de29..138cecb 100644 > --- gcc/testsuite/gcc.dg/pr52769.c > +++ gcc/testsuite/gcc.dg/pr52769.c > @@ -0,0 +1,24 @@ > +/* PR c/52769 */ > +/* { dg-do run } */ > +/* { dg-options "-O3" } */ > + > +typedef struct > +{ > + int should_be_zero; > + char s[6]; > + int x; > +} foo_t; > + > +int > +main (void) > +{ > + volatile foo_t foo = { > + .s = "123456", > + .x = 2 > + }; > + > + if (foo.should_be_zero != 0) > + __builtin_abort (); > + > + return 0; > +} > > Marek