Is integer overflow defined (for instance, as being mod 2^size)? Can I reliably say that -long.min == 0L?

On 09/01/2011 06:20 PM, Sean Eskapp wrote:
> Is integer overflow defined (for instance, as being mod 2^size)?

I am quite sure that all operations are defined as operations on two's complement integers, which would mean overflow is defined, but I cannot find the respective part of the specification...

> Can I reliably say that -long.min == 0L?

Well no, -long.min == long.min .

according to TDPL p53., that fact is defined. (unary minus: -x == ~x+1)

== Quote from Timon Gehr (timon.gehr@gmx.ch)'s article
> On 09/01/2011 06:20 PM, Sean Eskapp wrote:
> > Is integer overflow defined (for instance, as being mod 2^size)?
> I am quite sure that all operations are defined as operations on two's
> complement integers, which would mean overflow is defined, but I cannot
> find the respective part of the specification...
>  > Can I reliably say that -long.min == 0L?
> Well no, -long.min == long.min .
> according to TDPL p53., that fact is defined. (unary minus: -x == ~x+1)

Ah, good points, thanks!

Timon Gehr:

> according to TDPL p53., that fact is defined. (unary minus: -x == ~x+1)

Uhmmmm.

Bye,
bearophile

On 09/02/2011 02:45 AM, bearophile wrote:
> Timon Gehr:
>
>> according to TDPL p53., that fact is defined. (unary minus: -x == ~x+1)
>
> Uhmmmm.
>
> Bye,
> bearophile

For unsigned integers it is clearly defined:

1000... => 0111... => 1000...

And according to TDPL p53: "This manipulation does not raise particular questions about the signedness of val".

So, it is guaranteed that it will work for long.min .