What is &&& operation in C

Question

#include <stdio.h>

volatile int i;

int main()
{
    int c;

    for (i = 0; i < 3; i++) 
    {
         c = i &&& i;
         printf("%d\n", c);
    }

    return 0;
}

The output of the above program compiled using gcc is

0
1
1

With the -Wall or -Waddress option, gcc issues a warning:

warning: the address of ‘i’ will always evaluate as ‘true’ [-Waddress]

How is c being evaluated in the above program?

I believe it's `i && (&i)`? Interesting that I can't find a duplicate post on SO. — irrelephant, Dec 19 '12 at 06:49
Left-to-right (except where otherwise specified, which would only be `=`), so `&&` then `&i` — Voo, Dec 19 '12 at 06:50
Why do you answer in comments rather than answering the question itself? — Spook, Dec 19 '12 at 06:51
@KennyTM: I didn't quite understand `while (i &&& i <-- j) {}` — manav m-n, Dec 19 '12 at 09:51
possible duplicate of [Can you have a triple minus signs in C programming? What does it mean?](http://stackoverflow.com/questions/13874179/can-you-have-a-triple-minus-signs-in-c-programming-what-does-it-mean) — Anirudh Ramanathan, Dec 19 '12 at 10:14
Not a duplicate, but s a similar question and that's a good link — Nathan Cooper, Dec 19 '12 at 10:43
@Manav http://stackoverflow.com/questions/1642028/what-is-the-name-of-this-operator?rq=1 — Izkata, Dec 19 '12 at 11:33
@LuchianGrigore Maximum munch, it is the same answer, for `&&&`, `+++`, `---`, etc, is it not? — Anirudh Ramanathan, Dec 19 '12 at 12:04
@Cthulhu well I'm sure there are better fits than `---`, no? — Luchian Grigore, Dec 19 '12 at 12:06
its the same classic maximal munch rule, all over again as @Cthulhu already mentioned. Bear in mind that this question is again a dulpicate, I have also found this once already, but wasn't marked as duplicate. — aiao, Dec 19 '12 at 13:08
You tagged the question "gcc-warning", but you didn't show the gcc warning (I've added it). You also tagged it "linux", which is irrelevant (I've removed the tag). — Keith Thompson, Jan 01 '13 at 21:24
0 && (&&j). It uses a GCC extension to take the address of label j, which should always evaluate to a non-zero value. — nneonneo, Jun 05 '15 at 16:37

score 275 · Accepted Answer · answered Dec 19 '12 at 06:51

275

It's c = i && (&i);, with the second part being redundant, since &i will never evaluate to false.

For a user-defined type, where you can actually overload unary operator &, it might be different, but it's still a very bad idea.

If you turn on warnings, you'll get something like:

warning: the address of ‘i’ will always evaluate as ‘true’

answered Dec 19 '12 at 06:51

Luchian Grigore

253,575
64
457
625

1

Why not : `c = i & (&&i);`; Where `i` is some label? – anishsane Dec 20 '12 at 09:21
4

@anishsane `i` is defined as `int` and there's no labels in the question. Also, maximal munch... – Luchian Grigore Dec 20 '12 at 09:35
5

@anishsane: And the `&&` operator to take the address of a label is non-standard gcc extension anyway. But even if it were standard, the maximal munch rule would prevent it from being parsed that way (unless you insert a space). – Keith Thompson Dec 27 '12 at 20:36
Actually, `&i` can evaluate to `false`. Its used sometimes for defaults. Example: `void fn(type& x = *reinterpret_cast(NULL));` What is the value of `&x` in `fn` if `fn` is called without parameters? It's 0 a.k.a. false. However, using it the way described, it'll always be true unless `i == 0`, and if one was using it as I described, it would be `c = &i && i`. – Adrian May 24 '13 at 22:09
@Adrian you can't dereference a null pointer, which is what the `*` before `reinterpret_cast(NULL)` does. – Luchian Grigore May 24 '13 at 22:14
@LuchianGrigore: Although the type is dereferenced, it's not an error till the value is accessed. That's my understanding. Is it defined otherwise in the standard? – Adrian May 24 '13 at 22:17
@Adrian there's no value to speak of. Dereferencing a NULL pointer is undefined behavior. – Luchian Grigore May 24 '13 at 22:18
let us [continue this discussion in chat](http://chat.stackoverflow.com/rooms/30599/discussion-between-adrian-and-luchian-grigore) – Adrian May 24 '13 at 22:18
I stand corrected. void fn(type& x = *reinterpret_cast(NULL)); is invalid. – Adrian May 24 '13 at 22:51
There is a misconception: UB != invalid. UB means "the language spec. don't define it", but compiler implementer can. And there are compilers that allow `int& a = *(int*)0;` and exhibit exactly that behavior (`&a == nullptr`). Of course code like that is not portable (is in fact compiler specific), but the assertion "&i is always true" is not correct: being UB, it *may* be false. No-one grants it, but it can be. – Emilio Garavaglia Jul 27 '14 at 19:26
@EmilioGaravaglia any statements about the behaviour of a code snippet should be assumed to have "(unless there is already UB in which case anything could happen)" implied. – M.M Jun 03 '15 at 05:38

Keith Thompson · Answer 2 · 2013-08-13T20:52:37.127

There is no &&& operator or token in C. But the && (logical "and") and & (unary address-of or bitwise "and") operators do exist.

By the maximal munch rule, this:

c = i &&& i;

is equivalent to this:

c = i && & i;

It sets c to 1 if both i and &i are true, and to 0 if either of them is false.

For an int, any non-zero value is true. For a pointer, any non-null value is true (and the address of an object is always non-null). So:

It sets c to 1 if i is non-zero, or to 0 if i is equal to zero.

Which implies that the &&& is being used here just for deliberate obfuscation. The assignment might as well be any of the following:

c = i && 1;
c = !!i;
c = (bool)i;          // C++ or C with <stdbool.h>
c = i ? 1 : 0;        /* C */
c = i ? true : false; // C++

What is &&& operation in C

2 Answers2

Linked