Question

In WinSock2.h, the invalid socket and socket error are defined as these? Is there any significance to this?

#define INVALID_SOCKET  (SOCKET)(~0)
#define SOCKET_ERROR            (-1)

Answer 1

On a two’s complement system (and Windows is always two’s complement), ~0 is equal to -1, so there’s no significance to the compiler.

There may be a significance to the reader: ~0 emphasizes that it’s a value with all bits set, whereas -1 emphasizes that it’s a value 1 less than 0.

Aside:

On a system which is not two’s complement, and assuming that SOCKET is an unsigned type, it is generally wrong to write (SOCKET)(~0). The reason is that on such systems, ~0 does not represent the value -1, it’s one of INT_MIN, negative zero, or a trap representation. Hence it will not necessarily convert to type SOCKET as the value with all bits zero, rather it will convert as INT_MAX+2, 0, or goodness-knows-what (perhaps the value with all bits set).

So generally you should initialize unsigned types with -1 to get the value with all bits set. You could use UINT_MAX, or ~0UL, or similar, if you know which unsigned type you’re dealing with. But it’s not worth it, because -1 works for all unsigned types.