Reply to thread

Message

[QUOTE="Eric Sosman, post: 2416305"]
Agreed.  My point is that if even the inventor of the interface
can make such a simple oversight, the interface may in and of itself
be error-prone.  It's at least worth giving thought to what sort of
interface might be less likely to be, er, oversought.

This particular error isn't as bad as it used to be in the pre-
Standard days, because if there's a prototype for vast() in scope
the compiler will catch the omission.  The same can't be said for
using a sentinel; the compiler will *not* catch a failure to append
`(char*)NULL' after the "real" arguments.  This leaves you with a
balancing act: The existing interface may provoke more coding
errors that the compiler will catch, while a sentinel method
might provoke fewer errors but nastier.

Note that the compiler will not detect the problem in

tmp = vast(4, "Welcome to ", HOST, ", you are logged in as ",
	           name, ".");

or (worse) in

#define HOST "kremvax"
	...
	tmp = vast(5, "Welcome to " HOST ", you are logged in as ",
	           name, ".");

That is, omitting the count is not the only error mode you need to
think about.
[/QUOTE]

Verification