Headers circular referencing problem

[Malefaust]

Hi,
The problem is:

Header A-> include B
Header B-> include C
Header B-> include A
Header C-> include A

All the header have a pointer to struct that reside in include header.
How can i resolve this problem with preprocessor directivies?

Thanx

[Ancient Dragon]

1. redesign them. put everything into one header file.

2. use preprocessor directives to insure the header is included only once.

#if defined(_HEADER_A_H)
#define _HEADER_A_H
// forward declarations here
struct xyz; // this is in another header file
// rest of header file here
// blabla

#endif

[dwks]

Ancient Dragon, close, but not quite:

#if !defined(_HEADER_A_H)
#define _HEADER_A_H
// forward declarations here
struct xyz; // this is in another header file
// rest of header file here
// blabla

#endif

Remember, you want to compile the following code if it has not been compiled.

Look at the standard header files for examples. I usually use ifndef:

#ifndef FILENAME_H
#define FILENAME_H 1

/* stuff in header file that can only be included once */

#endif

[Narf]

I usually use ifndef:

Me too. For me it's a nice little mnemonic because ifdef and define have the same number of characters. In my editor if the two lines for inclusion guards don't line up, that sets off a warning flag in my mind and I usually need to change ifdef to ifndef to fix it.

Just on a side note since no one has mentioned it yet, identifiers leading underscores followed by an upper case letter are always reserved by the implementation. So it's possible that mysterious breakage could happen with Ancient Dragon's example.

[Ancient Dragon]

if you use !defined(...) you can have more than one expresssion

#if defined(_MSVC) || defined(_BORLAND) || defined(_GCC)
// blabla
#elif defined(_UNIX) && !defined(_X11)
// blabla
#elif defined(_UNIX) && !defined(_MAC)
// blabla
#endif

You can't do something like above with #ifdef statements


> identifiers leading underscores followed by an upper case letter are always reserved by the implementation

I often ignore rules like that. That's just a guideline, not a rule that is enforced by any compiler.

[Narf]

if you use !defined(...) you can have more than one expresssion

Which is irrelevant with inclusion guards because you only need one unique identifier per header. In general I agree that defined() is more flexible, but for inclusion guards I don't see the need.

You can't do something like above with #ifdef statements

But that wasn't a multiple inclusion guard, which is what I was talking about since that's what the topic of the thread is.

That's just a guideline, not a rule that is enforced by any compiler.

It doesn't have to be enforced. It'll just break silently and with strange behavior. You'll be searching for days to find a bug like that. Ignore it if you'd like, but I've been bitten by that particular bug before, and it's not pretty.