That's mostly true enough (especially about the satanic nature of [macro]s), but point 2 is mushy because the [C++] [type system] is designed to be broken anyway, and will remain so until [C-style casts] are removed from the language -- in other words, [as long as earth and sky abide].

(That used to say "[deprecated]" instead of "removed from the language", but [Furious George] was kind enough to point out that they already are [deprecated], and what I meant was something different -- and he was right, too. Ack!)

Secondly, [sprintf]() is indispensable because there's no other equally powerful way to format things. And how many people seriously use the [iostream] library? Not me. It's annoying and weak and it requires too much support in my own [code] for too little benefit. I'll give up my [FILE *] when you pry my cold dead fingers from it. Then again, I learned [C] first (not necessarily best) and that colors my view of things.

On the other hand, ever since I learned to use the [varargs.h] stuff, I've been looking for uses for it in my own [code], and I've yet to find a compelling one¹. Whenever it seems convenient, I aways run up against the problem of determining how many [argument]s are actually on the damn [stack]. The [format string] in the [printf]() family does the job nicely (assuming the [programmer] doesn't goof, which is a hell of a dangerous assumption), but without something like that, you're left with adding a dummy [NULL] or else making the first [argument] a count of the others. Both of the latter two solutions are dismal [kludge]s. If anybody knows any others, I'd love to hear 'em.

Hm. Come to think of it, since you've got the [address] of the [argument] that was [push]ed last, I'm a bit fuzzy about why you can't just compare the [address] of any given [argument] to the [stack pointer] and know where you are without having to be told. You must be able to get the [stack pointer] somehow; I mean, it's [C], right? So there's obviously something about that which eludes my grasp.

"[there are always other, better ways]"?

Them sounds like [famous last words] to me, ha ha.

¹ The sole exception would be a [sprintf]()ish member of a string class, which would [realloc]() as needed, so it wouldn't be subject to [buffer overrun]s. [MFC]'s [string] [class], [CString], does that. The [class] is a dog in general, but boy that one [feature] is [nifty as all heck]. [CString] also has an odd and ill-advised [kludge] in it which addresses [Gorgonzola]'s second point, and which is discussed in gruesome and drooling detail in the [CString] node.

never use variable arguments in C++