Re: Problem with Objective-C++ Header Files

On Fri, Jan 8, 2010 at 2:41 PM, Christopher Stern <[email protected]>wrote:

>
> On Jan 8, 2010, at 1:18 PM, Stephen Furlani wrote:
>
>
>
> On Fri, Jan 8, 2010 at 1:18 PM, Christopher Stern <[email protected]>wrote:
>
> Thanks for the reply.
>
> The C++ library uses glxew, and the code claims to support AGL -
> http://www.equalizergraphics.com/, so I'm not intentionally using
> XWindows.  I suspect that XWindows is not being properly excluded for
> compiling on a Mac.  I'll have to dig deep for that.
>
> I'm still a bit uncertain about using #define (and I'm wearing my C++
> ignorance here on my sleeve) to re-define something in a header file that
> isn't redefined in the library.  If I do this:
>
> *#define id eqID*
> *#include "eq/eq.h"*
> *#undef id*
>
> and eq.h has
>
> *class somethingA {*
> *private:*
> *void id ();*
> *}*
>
> #define'd to:
>
> *class somethingA {*
> *private:*
> *void eqID ();*
> *}*
>
> then it doesn't mess up eq.dylib which has
>
> *void somethingA::id() {}*
>
> and not
>
> *void somethingA::eqID() {}*
>
>
> You are correct that this could confuse the linker.
>
> There's also no way I can get XCode to just ignore Obj-C keywords in
> "eq/eq.h"?
>
>
> No, the source file, with all '#include's '#include'ed is one stream of
> text to the compiler.
> The good news is that I don't think there are any true ObjC keywords that
> don't start with '@'.
> There are some things defined in standard headers like BOOL that may be
> coming into conflict
> but that's a problem that exists in pure 'C' space the Obj-ness or ++-ness
> doesn't come into it.
>
> Note 'id' in ObjC is a typedef, not a keyword:
> typedef struct objc_object {
>     Class isa;
> } *id;
> the example you give above, with a struct or class having a member called
> 'id' shouldn't produce any conflict
> shouldn't need any defines or anything, in fact, there are standard MacOSX
> headers that have this.
>
>
>
I have the source for the library, so I could technically refactor (or
hackfactor) it to make it work, I'd just really rather not.

Thanks for the help.  I'll hunt down the typedef issues and see if I can't
be more intelligent with compiling under OBJC.

For the sake of clarity, (and in case there's another larger unseen issue)
I'll include two typical errors that show up when compiling as a .mm, that
aren't there when compiled as a .cpp.

The following code included in the headers:

185    /** Write a base::UUID. */
186    template<>
187    inline DataOStream& DataOStream::operator << ( const base::UUID& id )
188    {
189        base::UUID out( id );
190        out.convertToNetwork();
191        write( &out, sizeof( out ));
192        return *this;
193    }

gives the following error:

/Users/pebble/Documents/equalizer/build/XCode/include/eq/net/dataOStream.h:191:
error: invalid conversion from 'eq::base::UUID (*)(objc_object*)' to 'const
void*'

including a couple other places where 'id' is used as a member or passed
value.

    struct SessionGetIDMasterReplyPacket : public SessionPacket
    {
        SessionGetIDMasterReplyPacket( const SessionGetIDMasterPacket*
request )
            {
                command   = CMD_SESSION_GET_ID_MASTER_REPLY;
                size      = sizeof( SessionGetIDMasterReplyPacket );
                requestID = request->requestID;
                *id*        = request->*id*; // error shows up here,
"unqualified-id before '=' token"
            }

        NodeID   masterID;
        uint32_t requestID;
        uint32_t *id;*
    };

Again, thanks for the help.

-- 
-Stephen Furlani
Biomedical Software Engineer
e: [email protected]
c: 240-491-7288