Re: [PATCH] Fix PR rtl-optimization/67736 in combine.c

[Steve Ellcey <sellcey at imgtec dot com>]

On Thu, 2015-10-22 at 13:46 -0500, Segher Boessenkool wrote:

> One nit and maybe a problem:
> 
> > --- a/gcc/testsuite/gcc.dg/combine-subregs.c
> > +++ b/gcc/testsuite/gcc.dg/combine-subregs.c
> > @@ -0,0 +1,36 @@
> > +/* { dg-do run } */
> > +/* { dg-options "-O2 -fexpensive-optimizations" } */
> 
> -fexpensive-optimizations is default at -O2.

You are right, the original report I got mentioned this flag and then I
was comparing -fexpensive-optimization and -fno-expensive-optimization
so I didn't notice that it was on by default.  I will remove it from the
test.

> > +
> > +#include <inttypes.h>
> 
> Does every target have that header?  Shouldn't it be <stdint.h>?
> 
> > --- a/gcc/testsuite/gcc.dg/torture/pr67736.c
> > +++ b/gcc/testsuite/gcc.dg/torture/pr67736.c
> > @@ -0,0 +1,32 @@
> > +/* { dg-do run } */
> > +
> > +#include <stdlib.h>
> > +
> 
> And here you don't need inttypes at all?  Confused.
> 
> 
> Segher

Not including the include in pr67736.c was an oversight, I meant to
include it.  I didn't notice that Andrew had put in his own definition
of uint64_t into this test so it didn't need any includes, that is why
it still worked.  But you are right I should be using stdint.h instead
of inttypes.h.  It looks like most tests have:

/* { dg-do run { target { stdint_types } } } */
#include <stdint.h>

So I will use that in both tests and remove the local definition of
uint64_t from pr67736.c.

Steve Ellcey
sellcey@imgtec.com