Incrementing reference variable doesn't work

Question

The following code is supposed to output 6, but instead it outputs 5. I can't figure out why. What's going on?

#include <iostream>

template <typename T>
void foo(T& y)
{
   y++;
}

int main()
{
   int x = 5;

   // Why won't this line work???/
   foo(x);

   std::cout << x;
}

Live demo

[SPOILER WARNING!](http://coliru.stacked-crooked.com/view?id=965ca9098aaef2bd5099a7c94ae60595-f674c1a6d04c632b71a62362c0ccfc51) — Mark Garcia, Jul 03 '13 at 08:16
@Bathsheba: Yes, but I didn't write that in my code so I don't see the relevance. — Lightness Races in Orbit, Jul 03 '13 at 08:19
This blew my mind. Can you imagine receiving this sort of problem in a job interview or something?! Probability of looking stupid = 100%. — Ed King, Jul 03 '13 at 08:23
@LightnessRacesinOrbit If you [already knew the answer](http://stackoverflow.com/revisions/10f4c328-8bff-4b46-84e1-69ec7004762f/view-source), why didn't you answer your own question? — Bernhard Barker, Jul 03 '13 at 08:24
This is basically a glorified bug report for the pretty printer syntax highlighter...? (Not sure whether to +1 or -1.) — Kerrek SB, Jul 03 '13 at 08:25
wow, I hitted submit answer ans saw other ten at the same time, people likes trigraphs! — DRC, Jul 03 '13 at 08:25
@LightnessRacesinOrbit: Maybe this should be called a "smug report". — Kerrek SB, Jul 03 '13 at 08:26
@EdKing: If any interviewer ever asks me that question, I'll walk right out. — Xeo, Jul 03 '13 at 08:26
@DRC: I don't get any awards for it, unless you count the warm fuzzy feeling I receive when my good friends on Stack Overflow learn something new about C++ and have a little smile in the morning. — Lightness Races in Orbit, Jul 03 '13 at 08:32
This question appears to be off-topic because it is just a joke. See http://stackoverflow.com/revisions/10f4c328-8bff-4b46-84e1-69ec7004762f/view-source — johnchen902, Jul 03 '13 at 08:35
You need another `153` upvotes so you can reach [this one](http://stackoverflow.com/questions/15736282/what-is-this-smiley-with-beard-expression) which I consider equivalent. — Jesse Good, Jul 03 '13 at 08:37
@johnchen902: It's not a joke if someone learns from it. In fact, it's no more a joke than a self-answered Q&A... that I _didn't_ post an A doesn't change that. — Lightness Races in Orbit, Jul 03 '13 at 08:39
The fact that this has been posted multiple times illustrates that there is little one would *learn* from it. — devnull, Jul 03 '13 at 08:58
@devnull: On the contrary - the fact that this has been posted multiple times illustrates that there are plenty of people who have plenty to learn from it. — Lightness Races in Orbit, Jul 03 '13 at 09:21
@LightnessRacesinOrbit True, but yet this particular question doesn't contribute much to a long-time learning effect, disguising the actual problem with a completely unrelated question. — Christian Rau, Jul 03 '13 at 09:42
@LightnessRacesinOrbit: No, the fact that this has been posted multiple times illustrates that it should be closed as duplicate. Unfortunately my duplicate vote was lumped in with the off-topic votes which made it practically invisible, and now the question has been reopened. — interjay, Jul 03 '13 at 09:43
That happens when people carry their awful behaviour of multiple question marks over from their SO questions into their code!!! — Christian Rau, Jul 03 '13 at 09:46
@LightnessRacesinOrbit Compilers having been ignoring digraphs/trigraphs for a while now, at least gcc does. I'm not sure what is there to learn here. — devnull, Jul 03 '13 at 10:03
@devnull: I'm sorry you feel that way. 18 people saw fit to disagree with you enough to upvote. — Lightness Races in Orbit, Jul 03 '13 at 10:05
@LightnessRacesinOrbit People upvote pretty much anything. Maybe if you ask why does `<:]{%>;` compile you'll gather enough upvotes. — devnull, Jul 03 '13 at 10:08
@LightnessRacesinOrbit I heard somebody say `Sometimes, stackoverflow is the most depressing site on the internet.` a while back. You seem to prove that. — devnull, Jul 03 '13 at 10:13
@LightnessRacesinOrbit I give up. I'm reminded of what I was once told: `Don't stoop down to anybody's level. What if they start digging!` — devnull, Jul 03 '13 at 10:23
Terrible duplicate. It's not even about the same trigraph or operator. It bears essentially no resemblance. — Lightness Races in Orbit, Jul 04 '13 at 10:23

score 21 · Accepted Answer · edited May 23 '17 at 12:09

Trigraphs

You're using the good ol' trick of trigraphs.

// Why won't this line work???/
                            | |
                            \ /
                             |
                         ~trigraph~

The ??/ trigraph is in turn converted to \ which basically concatenates the current line with the next line and thus your code becomes more or less like this:

// Why won't this line work? foo(x);

A fine trick indeed.

Quoting from the C++11 standard:

§2.2.2:

Each instance of a backslash character (\) immediately followed by a new-line character is deleted, splicing physical source lines to form logical source lines. ...

§2.4.1:

Table 1 - Trigraph sequences
...
==========================
| Trigraph | Replacement |
==========================
|          ...           |
==========================
|   ??/    |      \      |
==========================

Fortunately, GCC seems to detect this kind of trickery, emitting a warning (just set -Wall):

main.cpp:13:32: warning: trigraph ??/ converted to \ [-Wtrigraphs]
    // Why won't this line work???/
 ^

main.cpp:13:4: warning: multi-line comment [-Wcomment]
    // Why won't this line work???/
    ^

_{Related references:}

meaning of `???-` in C++ code

What is this smiley-with-beard expression: "<:]{%>"?

What does the C ??!??! operator do?

And all other similar questions out there. ??)

_{PS: That's a smiley.}

@LightnessRacesinOrbit Interesting? Yes! Neat? No! Useful? Absolutely Not! When do they finally drop rubbish like that from the standard (together with the literal versions of the logical operators). Show me a single person that needs those nowadays for more than just "party knowledge" (or maybe entries to a hypothetical *IOC++CC*). — Christian Rau, Jul 03 '13 at 09:39
@LightnessRacesinOrbit you make out like you didn't know very convincingly. I'd like to see the programmer that writes a comment `// Why won't this line work???/` before a line that they haven't tested yet. That must be a very pessimistic person... — Sam Holder, Jul 03 '13 at 16:14

score 8 · Answer 2 · answered Jul 03 '13 at 08:19

8

??/ is a Trigraph sequence which is replaced with \.
For the compiler \ means that immediate line that follows is a part of the current line. In this case the current line is a comment. The effective result is:

// Why won't this line work foo(x);

answered Jul 03 '13 at 08:19

Alok Save

202,538
53
430
533

Oh! You're right! (BTW, 99.9k.. nice) – Lightness Races in Orbit Jul 03 '13 at 08:21
3

@LightnessRacesinOrbit: We both know that you knew it already & you are just trolling ;-) Ah, I missed the moment though shows 100k already, not that I am complaining. – Alok Save Jul 03 '13 at 08:34
@AlokSave Congrats for your 100k RP – Grijesh Chauhan Jul 03 '13 at 10:44

johnchen902 · Answer 3 · 2013-07-03T08:28:59.823

It's Trigraphs. In c++11 specification

2.4 Trigraph sequences [lex.trigraph]

¹ Before any other processing takes place, each occurrence of one of the following sequences of three characters (“trigraph sequences”) is replaced by the single character indicated in Table 1.

                     Table 1 — Trigraph sequences
    Trigraph Replacement │ Trigraph Replacement │ Trigraph Replacement
    ─────────────────────┼──────────────────────┼─────────────────────
       ??=      #        │    ??(      [        │    ??<      {
       ??/      \        │    ??)      ]        │    ??>      }
       ??’      ˆ        │    ??!      |        │    ??-      ~

Now replace ??/ with \ and you'll find out.

Yu Hao · Answer 4 · 2013-07-03T08:33:55.367

2

// Why won't this line work???/
   foo(x);

The trigraph ??/turns to \, so the code will be translated to:

// Why won't this line work?\
   foo(x);

And string concatenation works.

When I tested it in g++, trigraph is turned off by default(generates a warning), it will output 6. If compiled using g++ t.cpp -trigraphs, will output 5.

edited Jul 03 '13 at 08:33

answered Jul 03 '13 at 08:19

Yu Hao

119,891
44
235
294

score 2 · Answer 5 · edited May 23 '17 at 12:28

2

probably your comment is interpreted as a trigraph (not treegraph!) that "deletes" the function call.

http://ideone.com/sU4YGc works for me deleting that ??/ in the comment.

// Why won't this line work?
foo(x);

see also Purpose of Trigraph sequences in C++?

edited May 23 '17 at 12:28

Community

1
1

answered Jul 03 '13 at 08:24

DRC

4,898
2
21
35

2

Well. Is a tree a graph. Mmm. Some graphs can be a forest.... Mmm. I'm lost – sehe Jul 03 '13 at 08:35
ehehe, just mispelled tri for tree. – DRC Jul 03 '13 at 08:42

Stefano Falasca · Answer 6 · 2013-07-03T08:40:09.543

1

The compiler is your friend. This

   // Why won't this line work???/
   foo(x);

is a multiline comment, foo(x) is not executed!

The trigraph "??/" is converted to "\" which indicates a linebreak into the comment.

edited Jul 03 '13 at 08:40

answered Jul 03 '13 at 08:22

Stefano Falasca

8,837
2
18
24

What does the `/` do? – Lightness Races in Orbit Jul 03 '13 at 08:25
it converts to \ when it follows ??. Trigraphs used in c and c++ can be found at: http://en.wikipedia.org/wiki/Digraphs_and_trigraphs – Stefano Falasca Jul 03 '13 at 08:27
@StefanoFalasca you have a strange and roundabout way of arriving at the point. And your explanation is ... slightly misleading. '/' is not converted to '\\' when it follows anything. '??/' is _seen_ by the compiler as '\\' iff trigraphs are enabled. (Urffff. F*cken comment markdown) – sehe Jul 03 '13 at 08:33
3

actually, the trigraph ??/ doesn't even get to the compiler, the substitution is made by the preprocessor – Stefano Falasca Jul 03 '13 at 08:41
1

@StefanoFalasca precisely what I said. The compiler _sees_ a '/' – sehe Jul 03 '13 at 08:45
@sehe: No, it sees a '\'. – Lightness Races in Orbit Jul 14 '13 at 17:44
@LightnessRacesinOrbit not if run it through `cpp` first. [Note to self: can't win a pedantry war] – sehe Jul 14 '13 at 19:15

Incrementing reference variable doesn't work

Live demo

6 Answers6

Trigraphs

2.4 Trigraph sequences [lex.trigraph]