Too many actual parameters for macro?

Question

Code:

#include <iostream>

using namespace std;

#define ADD(x,y)  ((x)+(y))

int main( int argc, char** argv )
{
    cout << ADD(1,2,) << endl;
    return 0;
}

Compiler output:

1>Compiling...
1>main.cpp
1>c:\\warn_test\\main.cpp(9) : warning C4002: too many actual parameters for macro 'ADD'

Why isn't this an error?

g++ (GCC) 4.2.1 20070719 [FreeBSD] gives more reasonable (in my mind) output:

main.cpp:9:18: error: macro "ADD" passed 3 arguments, but takes just 2
main.cpp: In function 'int main(int, char**)':
main.cpp:9: error: 'ADD' was not declared in this scope

Though I'm not entirely sure what either compiler thinks the third argument is.

EDIT: Added complete gcc output and version info.

Answer 1

I'm going to throw out a complete guess, inspired by Steve Jessop's comment that it's related to variadic macro support.

Possibly it was easier to make it a warning when the visual studio team implemented variadic macros? I've noticed varying levels of tolerance when implementing code like:

#define MACRO(...) my_func(true, __VA_ARGS__);

MACRO(1,,2); // Missing argument
MACRO(1,); // missing tail
MACRO(); // no arguments

Some compilers error, warn or ignore the various situations. I don't know what the standards says tho.

Answer 2

You use ADD(1,2,) , note the second , . Remove that and it will compile just fine!

@schnaader: You are right, I read too fast. Sorry.

[edit] Please provide more details about the compiler in question. I use: g++ (Ubuntu/Linaro 4.4.4-14ubuntu5) 4.4.5, and this is the result I get:

test.cpp:9: error: macro "ADD" passed 3 arguments, but takes just 2
test.cpp: In function ‘int main(int, char**)’:
test.cpp:9: error: ‘ADD’ was not declared in this scope

[edit2] Sorry, again a bit too fast :-). I see you tagged it with visual studio. VS is more tolerant than g++. I suppose that -- because it is easy to resolve in this case -- it automatically corrects it.

Answer 3

I guess this is somewhat compiler's choice. If there was a third parameter, it would perhaps be more problematic, but as there isn't, you can argue about just ignoring the comma or throwing an error. Microsoft seems to be more error tolerant often (like in IE HTML parsing).