On 23 July 2013 03:43, Andreas Färber <[email protected]> wrote:
> v2 improves internal vs. "public" header separation for GIC.
> As before, no feedback was received to address PMM's QOM concerns,
> so this is what we have as design patterns for the moment.
I had a thought about this. Suppose we have our class header
files do something like this:
#ifdef MYCLASS_IMPLEMENTATION
#define PRIVATE
#else
#ifdef __GNUC__
#define PRIVATE __attribute__((deprecated("this is a private field")))
#else
#define PRIVATE
#endif
typedef struct MyObject {
int publicfield;
int privatefield PRIVATE;
} MyObject;
Then we can allow both users of the class and the implementation
to share the same header file (obviously only the implementation
defines MYCLASS_IMPLEMENTATION before using it). The users can
embed the struct MyObject with no problems, but if they try
to directly access a private field this happens:
/tmp/zz9.c:22:5: warning: ‘privatefield’ is deprecated (declared at
/tmp/zz9.c:12): this is a private field [-Wdeprecated-declarations]
(Since this is only a safety-guard against accidental uses,
it's OK that it's only present when building with gcc.)
thanks
-- PMM
