Why isn't the cout overflow in this situation?

I have the following code patch running on my ubuntu20.04, compiled by gcc-9.

#include <iostream>

int main()
{
    uint16_t a = 0xFFFF;
    uint16_t b = 1;

    uint16_t ret = a + b;
    std::cout << a + b << std::endl;
    std::cout << ret << std::endl;
}

I expect both of the results are overflowed, but the running results show that only the second one overflows and the first one can get correct result.

What makes them behave differently. The right value of the + operation should also be uint16_t, so why the first cout can get the right result?

P.S., when I change the type from uint16_t to uint32_t, both of them are overflowed.