On 2014.11.19 at 11:20 -0500, David Edelsohn wrote:
> On Wed, Nov 19, 2014 at 10:24 AM, Markus Trippelsdorf
> wrote:
> > bootstrap-ubsan on gcc112 shows a couple of signed integer overflows:
>
> > The fix was tested on powerpc64-unknown-linux-gnu.
> > OK for trunk?
> >
> > Thank you.
> >
> > 2014
On Wed, Nov 19, 2014 at 10:24 AM, Markus Trippelsdorf
wrote:
> bootstrap-ubsan on gcc112 shows a couple of signed integer overflows:
> The fix was tested on powerpc64-unknown-linux-gnu.
> OK for trunk?
>
> Thank you.
>
> 2014-11-19 Markus Trippelsdorf
>
> * config/rs6000/constraints.md
On Wed, Nov 19, 2014 at 04:41:49PM +0100, Marek Polacek wrote:
> > - HOST_WIDE_INT c, lsb;
> > + unsigned HOST_WIDE_INT c, lsb;
> >
> >c = INTVAL (op);
>
> Shouldn't you use UINTVAL then?
That doesn't really matter here. It looks a bit cleaner, of course.
I wouldn't spend too much time
On Wed, Nov 19, 2014 at 04:24:25PM +0100, Markus Trippelsdorf wrote:
> --- a/gcc/config/rs6000/constraints.md
> +++ b/gcc/config/rs6000/constraints.md
> @@ -140,7 +140,7 @@
> (define_constraint "I"
>"A signed 16-bit constant"
>(and (match_code "const_int")
> - (match_test "(unsigned
On Wed, Nov 19, 2014 at 04:24:25PM +0100, Markus Trippelsdorf wrote:
> ;; Return 1 if op is a constant integer valid for addition
> @@ -827,7 +827,7 @@
> (define_predicate "mask_operand"
>(match_code "const_int")
> {
> - HOST_WIDE_INT c, lsb;
> + unsigned HOST_WIDE_INT c, lsb;
>
>c =
