Thread: Forwards declaration question?

Hello,

I have recently looked into forwards declarations and what they do. I understand how to code forward declarations but I don't understand why we would need them in this particular example.

Please view the small code sample and I will pose my question at the end:

Code:

=================mainc. (does nothing)
#include <stdio.h>
#include "PC.h"
#include "REC.h"
 
int main()
{
int i; 
}

==================PC.h
#ifndef PC_H
#define PC_H

#include "REC.h"
typedef struct tag_pc 			
{unsigned short	gso_flash_address;              
} pc;                                                 		

#endif // PC_H //

==================PC.c
#include <stdio.h>
#include "PC.h"
// Methods here...

==================REC.h
#ifndef REC_H
#define REC_H

//struct tag_pc;	// Forwards declaration commented on purpose

typedef struct tag_rec_data
{unsigned short LKdct__F1;          
}rec_data;

typedef struct tag_rec                  				
{unsigned short f__pc_xchoice; 				     
rec_data *records;
}rec;                           

int REC_match_link_id(pc  *pObj);           
                    // struct tag_pc     *pObj);           

#endif // REC_H //

===================REC.c
#include <stdio.h>
#include "PC.h"
#include "REC.h"

int REC_match_link_id(pc     *pObj)           
                      //struct tag_pc      *pObj) 
{
// ... other code
}

========================

Okay, the above code gives an error at the "int REC_match_link_id(....)" function declaration and implementation levels. My question is:

If REC.c includes REC.h and REC.h include PC.h (which is where pc is declared), why is it still not sufficient for the compiler to recognize the "pc *pObj" parameter in the int REC_match_link_id(....) function parameter list?

I mean isn't the whole PC.h header included in the REC.c compiler's translation unit???? Then, why is it that I have use a forwards declaration for the code above to compile correctly??? I know that if I uncomment the forwards declaration in REC.h and replace the "pc *pObj" parameter to "struct tag_pc" in the function's parameter list everything will compile... but I don't see why it wouldn't compile if I leave the code as above!

Basically, when we include the PC.h in the REC.h, then, if pc is used anywhere in REC.h/.c it should have visible scope... no?

All feedback sincerely appreciated!

Originally Posted by see the big C

If REC.c includes REC.h and REC.h include PC.h (which is where pc is declared), why is it still not sufficient for the compiler to recognize the "pc *pObj" parameter in the int REC_match_link_id(....) function parameter list?

You need to look more closely at your own code. REC.h does not #include PC.h. PC.h declares the pc type after #include'ing REC.h.

Therefore, when reading REC.h, the compiler has no information about what the pc type is (typedefs that come later as you - or the compiler - reads the file don't count).

That means REC.h never has any information about what the type pc is. Since REC.h declares a function that accepts a pointer to pc, but the type pc has not been defined at any point when REC.h is included, a forward declaration is required.