> I think perhaps Soda-san misinterpreted my remarks. :-)

Oops, perhaps I misunderstood your intention. :-)

> What is the correct prototype for foo()?  (It may depend on whether
> xyz_t is an alias for an unsigned type, and in that case, it can
> also depend on the relationship between the maximum values of an
> "xyz_t" and an int.)
> 
> The problem is most easily solved by always using prototypes and
> never using K&R definitions, but this precludes K&R-compatibility.

Yes, this is real problem, and breaking K&R compatibility is easiest
solution from library authors's POV.

But since it will introduces incompatibility with 3rd party applications,
I wish that library authors be more careful to keep K&R ABI compatibility.
--
soda