Alan Barrett wrote:
> Perhaps we should define something like
> 
>     #define __MAX_DIGITS_FOR_UNSIGNED_TYPE(t) \
> 	((sizeof(t) * CHAR_BIT + 2) / 3)
>     #define __MAX_DIGITS_FOR_SIGNED_TYPE(t) \
> 	((sizeof(t) * CHAR_BIT + 2) / 3 + 1)

I don't think the user has to worry about signed or unsigned himself:

	#define __MAX_STRLEN_FOR_INT_TYPE(t) \
		((t) -1 < 0 ? __MAX_DIGITS_FOR_SIGNED_TYPE(t) : __MAX_DIGITS_FOR_UNSIGNED_TYPE(t))

Of course, this doesn't allow

	int x;
	char buf[__MAX_STRLEN_FOR_INT_TYPE(x)];

but you can always use the explicit variant in such cases.

-- 
Christian