linux- windows cross c++ application - c++

Question

I'm developing an application which has to run on Linux and Windows.
I have an object called obj which I want to use in the code and it has different behavior on Linux and Windows. so I inherit aaa and called WindowsObj for Windows object and LinuxObj for Linux object.
My question is: How to use this object in the code? what do I have to write that it will run both for Linux and Windows?
For swiching types I use typedef like:
typedef uint32_t DWORD;
but what do I have to use for objects?
I want to write this code:
tr1::shared_ptr<WindowsObj> windowsobj (new WindowsObj(parameter));
tr1::shared_ptr<LinuxObj> linuxobj (new LinuxObj(parameter));
Any idea?

Answer 1

The same thing :)
class _object
{
};
class WindowsObject : public _object
{
};
class LinuxObject public _object
{
};
#if defined(WIN32)
typedef WindowsObject Object;
#else
typedef LinuxObject Object;
#endif
Object myObject;
EDIT: Naturally, the interface that WindowsObject and LinuxObject expose must be the same. In this example, _object would be an abstract base-class that defined the interface, and LinuxObject and WindowsObject would then implement this interface, hiding away the platform-specific stuff in their implementation files.
Sample
_object.h
class _object
{
public:
virtual void doSomething() = 0;
}; // eo class _object
WindowsObject.h
#include "_object.h"
class WindowsObject : public _object
{
public:
virtual void doSomething();
}; // eo class WindowsObject
WindowsObject.cpp
#if defined(WIN32)
#include <windows.h>
void WindowsObject::doSomething()
{
// do something totally reliant on windows here
}; // eo doSomething
#endif
Then you would do the same for LinuxObject.h and LinuxObject.cpp, the latter having completely different preprocessor instructions. e.g, #if defined(UNIX) or some such flavor. Note the WIN32 guards around the implementation. Then you'd have some core header file you'd use:
#if defined(WIN32)
#include "WindowsObject.h"
typedef WindowsObject Object;
#else
#include "LinuxObject.h"
typedef LinuxObject Object;
#endif
Now, in your program
Object a;
a.doSomething();
It's worth noting that, if it's just the odd line of code that differs in your complex object (like a init call at initialisation, destruction) you might be better off with a single platform-agnostic Object and put guards in the implementation. This solution makes more sense when there are huge differences.

Answer 2

I tend to use conditional compilation for this, e.g.
#ifdef WIN32
// windows-specific code
#else
// non-windows
#endif

Answer 3

I suggest to use a cross platform framework, like GTK+ or wxWidgets, so that you hadn't to reinvent the wheel...
Then, you should have the least possible platform depended code, and possibly well deep-inside your classes. There you can use #ifdef to conditionally include code for Windows or for Unix.

Answer 4

You may use the same header file with two different implementations in two different .cpp files; the PIMPL idiom makes this possible even if the two implementations require completely different data members. You simply compile and link one source when compiling for Windows and a different one when compiling for Linux.
One area that I've found this useful is for email interfaces. The two OS have completely different methods for sending emails.

Answer 5

Hi if you can use c++, use boost and size_t type its better way.