On Fri, Oct 25, 2013 at 1:10 PM, Jeffrey Yasskin <jyasskin@google.com> wrote:
On Fri, Oct 25, 2013 at 12:50 PM, Jens Maurer <Jens.Maurer@gmx.net> wrote:
> 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]      );
>
> Jens

Nope: [expr.shift]p1 says, "The type of the result is that of the
promoted left operand."

I believe John is correct. 5.8/1 says we perform integral promotions on the operands, not the usual arithmetic conversions. Clang, g++, and EDG agree -- the type of (unsigned char)0 << 0u is int.
 
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
resulting value;"

We gave this defined behavior as a DR, so I view this code has having de facto defined behavior in C++11 and C++98 too. But it's UB in C.

In other words, we've already fixed this one (for some value of "fixed").