Path: utzoo!utgpu!jarvis.csri.toronto.edu!mailrus!ames!haven!adm!smoke!gwyn
From: gwyn@smoke.BRL.MIL (Doug Gwyn)
Newsgroups: comp.lang.c
Subject: Re: forward references in typedefs
Message-ID: <10567@smoke.BRL.MIL>
Date: 20 Jul 89 22:56:21 GMT
References: <55480@tut.cis.ohio-state.edu>
Reply-To: gwyn@brl.arpa (Doug Gwyn)
Organization: Ballistic Research Lab (BRL), APG, MD.
Lines: 40

In article <55480@tut.cis.ohio-state.edu> George M. Jones <george@cis.ohio-state.edu> writes:
>The following code compiles (Sun, GCC, Green Hills) just fine as expected:
>  struct one
>    {
>      struct two_t *foo;
>    } one_t;
>  struct two
>    {
>      struct one_t *bar;
>    } two_t;

If you were to change the member types to point to "struct two" and
"struct one" types, it would be correct.  The lack of a diagnostic at
the end of the translation unit ("struct two_t and struct one_t missing
declarations") can be considered a compiler deficiency.

C deliberately allows forward struct references in such contexts.
It may not have been completely clear in K&R 1st Edition, but it's
officially specified in the proposed Standard.

>However all the compilers I tried appear to be choaking on the forward 
>reference to two_t in the following chunk of code
>  typedef struct one
>    {
>      two_t *foo;
>    } one_t;

Sure; "two_t" is neither a C keyword nor a defined type at the point
that it is first used.  The compiler cannot know that it is eventually
going to be a structure type, unless it were to scan ahead (which is
contrary to the intent of C).

You might consider writing something like:
	typedef struct two two_t;	/* incomplete type */
	typedef struct one {
		two_t *foo;
	) one_t;
	struct two {			/* completes the type */
		one_t *bar;
	};