Re: [Qemu-devel] [PATCH 2/2] target-i386: Don't left shift negative constant

[Richard Henderson]

On 10/23/2015 05:07 AM, Eduardo Habkost wrote:
> On Thu, Oct 01, 2015 at 02:06:36PM -0300, Eduardo Habkost wrote:
> > On Thu, Oct 01, 2015 at 11:35:52AM +1000, Richard Henderson wrote:
> > > On 09/30/2015 06:34 AM, Eduardo Habkost wrote:
> > > > Left shift of negative values is undefined behavior. Detected by clang:
> > > >    qemu/target-i386/translate.c:2423:26: runtime error:
> > > >      left shift of negative value -8
> > > >
> > > > This changes the code to reverse the sign after the left shift.
> > > >
> > > > Signed-off-by: Eduardo Habkost <address@hidden>
> > > > ---
> > > >   target-i386/translate.c | 2 +-
> > > >   1 file changed, 1 insertion(+), 1 deletion(-)
> > > >
> > > > diff --git a/target-i386/translate.c b/target-i386/translate.c
> > > > index 8b35de1..cc59b7e 100644
> > > > --- a/target-i386/translate.c
> > > > +++ b/target-i386/translate.c
> > > > @@ -2420,7 +2420,7 @@ static void gen_pusha(DisasContext *s)
> > > >   {
> > > >       int i;
> > > >       gen_op_movl_A0_reg(R_ESP);
> > > > -    gen_op_addl_A0_im(-8 << s->dflag);
> > > > +    gen_op_addl_A0_im(-(8 << s->dflag));
> > >
> > > Better as -8U << s->d_flag?
> >
> > That's even more confusing to me. I wouldn't want to require other
> > people to read the C specification to find out how many type conversions
> > are happening in that statement. (Because I will have to do that, to
> > find out what's the type of "-8U").
> >
> > I would prefer an expression that doesn't involve any type conversion.
> > But you are more familiar with that code, so it's up to you.
>
> Ping? I would really like to fix this warning to be able to enable
> check-mode by default in QEMU 2.5. Are there objections to get this
> patch included?

None.


r~