On Mon, 25 Jan 2016 22:20:53 -0500, Michael McConville wrote:
> The third case seems questionable. However, (IIUC) integer conversion
> rules dictate that the int is implicitly cast to an unsigned int[1] and
> that two's complement arithmetic is preserved in this cast.[2]
>
> C integer type rules
I've discussed the below with otto@, who thought it was sound but wanted
confirmation before I commit.
While debugging rarpd(8) for unrelated reasons, I noticed a loop of the
following form:
> for (i = 1; i; i <<= 1)
where i is an int.
This relies on shifting into and then out of the sign bit,
