On Fri, Oct 25, 2013 at 12:50 PM, Jens Maurer <Jens.Maurer@gmx.net> wrote:Nope: [expr.shift]p1 says, "The type of the result is that of the
> On 10/25/2013 09:36 PM, John Regehr wrote:
>>> What reason do you have to believe that crypto is using any signed
>>> arithmetic? I would not.
>> Here's an example that's at least slightly interesting, from the latest
>> version of LibTomCrypt:
>> kappa[i] =
>> (key[pos ] << 24) ^
>> (key[pos + 1] << 16) ^
>> (key[pos + 2] << 8) ^
>> (key[pos + 3] );
>> key is a pointer to unsigned char. Of course, the array element becomes
>> signed after promotion. The shift by 24 then executes an undefined
>> behavior whenever the shifted value is >127.
>> So the interesting thing is that the developer is basically doing things
>> right and getting hosed by the arithmetic conversions.
> If I'm reading 5p10 correctly, this should help (and is consistently
> expressing intent):
> kappa[i] =
> (key[pos ] << 24u) ^
> (key[pos + 1] << 16u) ^
> (key[pos + 2] << 8u) ^
> (key[pos + 3] );
promoted left operand."
However, I think p2 saves our intrepid developer in C++14: "Otherwise,
if E1 has a signed type and non-negative value, and E1 × 2^E2 is
representable in the corresponding unsigned type of the result type,
then that value, converted to the result type, is the