This fix looks good to me. As you pointed out, I didn't see that theRnd::operator() gives a value in the range[0,n) instead of[0,n] as I expected.

The same issue actually is also present for floating point, although in a different form:

const FloatVal rndVal = (static_cast<FloatVal>(_random(INT_MAX)) / INT_MAX)*(range.second - range.first) + range.first;

This will never reach the upper bound, as the division will never reach one becauseINT_MAX is excluded. To fix this, we might have to give up some of the granularity in the division and reduce the right-hand side of the division:

const FloatVal rndVal = (static_cast<FloatVal>(_random(INT_MAX)) / (INT_MAX-1))*(range.second - range.first) + range.first;

This seems fine, and I think any other fix would require an inclusive version of the random number generator, which would be more work.

Updated for uniform_on_researt for float. Thank@Dekker1