is extern int i - a declaration or defination?

[shraddha18]

is the following statement a declaration or defination ?
extern int i

[Banfa]

It's nothing it has not ending semi-colon however

extern int i;

Is a declaration, it declares that somewhere in the program a variable called i with type int exists at global scope. It does not define the variable though, that is it does not allocate memory to contain the variable.

[weaknessforcats]

Depends.

extern int i;

is an assertion that there is an int named i defined in another file .c with external linkage. External linkage means accessible from outside the file such as a global variable. So in this case this is a declaration. You will need to define the int in some other .c file that is part of your build.

On the other hand:

extern int i = 0;

Is a definition since the compiler needs to create an int to contain that zero.

[donbock]

It depends on the C implementation.
In general, compilers use one of these four models for top-level variables.

The Initializer Model...

• Exactly one of the declarations has an initializer. That is the definition and the others are references.
• The extern keyword is assumed if it is not present.

The Omitted Storage Class Model...

• Exactly one of the declarations must omit the extern keyword. That is the definition and all of the others are references.
• Only the definition is allowed to have an [optional] initializer.

The Common Model...

• Same as The Initializer Model except the initializer is optional.
• If there is no initializer, then the linker chooses one instance to be the definition. The rest are references.

The Mixed Common Model...

• Same as The Common Model except the optional initializer can only be present with a declaration that does not have the extern keyword.

I don't know if any of this applies to C++.

[weaknessforcats]

Just when I thought I knew how this worked.

[donbock]

Yet another reason why global variables are a bad idea.

[donbock]

I tend not to use global variables any more; but that wasn't always the case. I recall developing the following idiom some years ago to insure that the definitions and declarations are consistent.

Create a header file with all the global variables:

#ifdef DEFINITIONS
#define GLOBAL(var,val) var = val
#else
#define GLOBAL(var,val) extern var
#endif

GLOBAL(v1,0);
GLOBAL(v2,10);
...
GLOBAL(v99,0);

Then #include the global-variable header in every source file that needs them. However, in one of those source files I would #define DEFINITIONS before including the header. Notice that the macro requires an initial value for every global variable.

What's good about this: the code (after macro expansion) is compatible with all four models. (But you could do the same manually.)

What's very bad about this: if I don't define DEFINITIONS in any source file and if my compiler uses The Common Model or the Mixed Common Model then I get no warnings but none of my global variables are initialized!