[m-rev.] for discussion: undefined behaviours in int.m

[Peter Wang]

On Fri, 21 Oct 2016 17:53:29 +1100, Peter Wang <novalazy at gmail.com> wrote:
> 
> We'd need to use unsigned integer operations to get _defined_ overflow
> behaviour, or rely on compiler switches like gcc -fwrapv or gcc
> -fno-strict-overflow.  Otherwise, the C compiler can assume that signed
> integer overflow does not happen, and generate code accordingly.

Actually -fno-strict-overflow doesn't _define_ signed integer overflow;
-fwrapv does.

Note that UndefinedBehaviorSanitizer won't report that an operation
overflowed if the overflow is defined (see the name).  That's a shame if
you consider an overflow to be a likely indicator of a bug.

Peter