C programming question ....

[Roger Oberholtzer]

On Wed, 2006-11-08 at 14:13 -0500, Matthew Carpenter wrote:
> On Wednesday 08 November 2006 11:16, Ben Duncan wrote:
> >
> > And it COMPILES and works. SO, it seems that
> >
> > struct rlib;
> > typedef struct rlib rlib;
> >
> > Is a way to prototype a Structure ...
> >
> > Ok, my questions to the REAL C guru's out there is:
> > Why would you want to do such a thing when simply moving
> > the ACTUAL structure ahead of things would solve having to
> > create extra lines of code protyping the structure ?
> 
> It's about scope.  Allowing the struct to be defined at the place it most 
> makes sense is a good way to manage large projects (like, say, a kernel ;)
> 
> The first "struct rlib;" is actually implicitly "extern struct rlib;" and the 
> compiler is sated.  The struct could then be defined in a header file or some 
> other code file.

Not so sure about that. In C, data structure definitions (struct and
typedef struct) do not have a concept of extern. They are simply defined
or undefined. I personally would declare it as:

typedef struct {

	int whatever;

} rlib;

After all, when all those disconnected declarations are done, that is
what rlib is. So why mess about? If it needs to refer to itself, try
this:

typedef struct rlib_s {

	int		whatever;

	struct rlib_s	*next, *previous;

} rlib;


This compiles on every C compiler I have encountered (GNU, MS, TI, SDCC,
OS/9, HPUX, SUN come immediately to mind). Don't let all those c++ types
tell you this is wrong. We are talking C. It is fully allowed.


--
Roger Oberholtzer