C / C++ / MFC

13-Jun-10 20:34

Hi All,

I have a scenario where I need to implement something like what setjmp does. I am well aware of using the setjmp API but the problem statement requires this to be implemented in a different manner.

I have a variable that would store the address of the next function to be called. This variable could be passed as a parameter to another function and to another one from that function and so on...

For this I will need the address of the function that is currently executing and the address to which I want to jump to. Any ideas on how can I read the addresses or implement this?

You need to google first, if you have "It's urgent please" mentioned in your question.
Wink | ;-)

_AnShUmAn_

«_Superman_»13-Jun-10 21:00

13-Jun-10 21:00

I do not understand why you want the address of a function at runtime.
You could simply store the addresses of all available functions in a pointer array and use that array in code.

«_Superman_»

I love work. It gives me something to do between weekends.

Microsoft MVP (Visual C++)

Polymorphism in C

_AnsHUMAN_ 13-Jun-10 22:25

13-Jun-10 22:25

How do I store the addresses of functions?
These functions may also be defined in different files across the applications code.

You need to google first, if you have "It's urgent please" mentioned in your question.
Wink | ;-)

_AnShUmAn_

«_Superman_»13-Jun-10 22:27

13-Jun-10 22:27

If you have a function -

void fun() {}

you can take its address as -

void* funptr = fun;

«_Superman_»

I love work. It gives me something to do between weekends.

Microsoft MVP (Visual C++)

Polymorphism in C

_AnsHUMAN_ 13-Jun-10 22:50

13-Jun-10 22:50

Thanks for that Smile | :)

.
But I have a different issue over here. With all the functions loaded into the memory and millions of LOC I doubt this to be a feasible solution in my case.

You need to google first, if you have "It's urgent please" mentioned in your question.
Wink | ;-)

_AnShUmAn_

Aescleal13-Jun-10 21:27

13-Jun-10 21:27

Why do you need the address of the function that's currently executing? When a function returns you end up back at where you were executing before. So if I have a function pointer (declared as void (*function_pointer() ) and call it:

function_ptr();

when whatever function the pointer points to returns you'll be back executing the next statement in the calling function.

Incidentally a lot of finite state machines use the technique of storing their states as a function pointer, at least when implemented in C. It might pay to do some research on them - they may hold the key to what you want.

Cheers,

Ash

PS: If you're using C++ you don't want a function pointer, look at using an interface pointer instead.

_AnsHUMAN_ 13-Jun-10 22:23

13-Jun-10 22:23

I will explain the situation.
File A.cpp:

jmp_buf buf;
void func1()
{
cout<<"abc";
longjmp(1,buf);
}
void func2()
{
cout<<"def";
}
int main()
{
if (!setjmp(buf))
func1();
else
cout<<"Do nothing";
}
In this case I am able to do the navigation from func1 and back to main. Just like we do for a LABEL and GOTO case but this happens across function and file calls as well.

In my case I have another file say DEF.cpp wherein resides a function called CallMe
I am able to make the call to that function as well from main. but if I want that CallMe should call another function then I should also return back to the main function, for which I would need the address in the call stack to which I want to return to.

Let me know in case I have misunderstood something

You need to google first, if you have "It's urgent please" mentioned in your question.
Wink | ;-)

_AnShUmAn_

Aescleal13-Jun-10 22:37

13-Jun-10 22:37

Ah, I see what you want to do and now I see it I don't think it's a good idea in C++. setjmp and longjmp have so many problems in C++ that if you think you understand the complexity of using them you're usually wrong. You'd have all the same problems (what destructors get called, what happens if an exception is thrown which are undefined by the standard) with any hand rolled system you come up with.

Cheers,

Ash

_AnsHUMAN_ 13-Jun-10 22:47

13-Jun-10 22:47

Thanks for your inputs. I have already done some investigation regarding the use of setjmp and longjmp and the other overheads underlying their usage, but still just wanted to get a second opinion before I throw it away...
Thanks again.

You need to google first, if you have "It's urgent please" mentioned in your question.
Wink | ;-)

_AnShUmAn_

_AnsHUMAN_ 14-Jun-10 19:03

14-Jun-10 19:03

HI,
Is there a mechanism in C++ apart from GNU where we can get the address of a label. I noticed in GNU compiler there is an operator '&&' (ITS NOT AN AND OPERATOR) ex:void *ptr;
...
ptr = &&foo; //(where foo stands to be the name of a label)
prefixed to the name of the label to get the address.
Can we do something like this on the visual studio compiler.

You need to google first, if you have "It's urgent please" mentioned in your question.
Wink | ;-)

_AnShUmAn_

Aescleal15-Jun-10 0:21

15-Jun-10 0:21

&& refers to an R-value reference, it's not the sort of thing you're looking for.

You really ought to get yourself a good book on C++. Yes, there is an operator for getting the address of an object in a C++ program, it's the unary & operator. So:

int *p = &n;

assigns the address of the variable n to the variable p.

Cheers,

Ash

_AnsHUMAN_ 15-Jun-10 1:51

15-Jun-10 1:51

Ah! I specifically mentioned that this operator was on C++ for the GNU compiler. Probably you missed it, it does acquire the address of the label using "&&" but it is not there in C++ with the VS compiler.
This link[^] states the implementation and yes I have read a few books but none of them explains this concept for VS Roll eyes | :rolleyes:

I am a HUMAN. I have that keyword (??? too much) in my name........
Wink | ;-)

_AnsHUMAN_b>

Aescleal15-Jun-10 8:35

15-Jun-10 8:35

&& is an operator supported by g++ 4.1 and later and VC++ 2010. It might have been something vendor specific in earlier versions - I can't comment as I try and avoid anything outside of the C++ standard as it's not portable.

You can read more about r-value references here[^].

Cheers,

Ash

norish14-Jun-10 9:17

norish

14-Jun-10 9:17

Why don't you use c++ exception semantics?
It might be as below.

struct TException;

void func1()
{
    cout<<"abc";
    throw new TException("func1");
}
void func2()
{
    cout<<"def";
}
void main()
{
    try {
        func1();
    }
    catch (TException* e) {
        e->printSomthing();
        delete e;
    }
    catch(...) {
        cerr<<"general exception which I dont know"<<endl;
    }
}

Aescleal14-Jun-10 9:34

14-Jun-10 9:34

That's not what the author wanted - he wanted to be able to implement something like co-routines or resumable exceptions with his code (which is why set/longjmp attracted him, he could branch from an arbitrary lump of code back to another arbitrary lump). Exceptions are just good for a one way trip up the call stack.

Oh, and as an aside - if you use C++ exceptions follow the rule "throw by value, catch by reference." The last thing you want to be doing is resource management when you're trying to handle exceptions - it's complete madness. And if you think I'm talking complete rubbish consider what happens if printSomething throws - you'll leak a pointer to a TException object.

And I've just noticed you're using that cardinal sin void main(). Guess what? That doesn't compile on any standard conforming C++ compiler. Use one of the two standard forms instead:

int main()

or:

int main( int argv, char *argv[] )

Ash

norish14-Jun-10 10:07

norish

14-Jun-10 10:07

Thanks your pointing out.
That was a Java style coding. In c++, ordinarily;

throw TException("func1");

catch(TException& e) {
}

And I know void main is not compatible in c++.
That was only concept codes, but may not be so good example.

Aescleal14-Jun-10 11:08

SendMessage and PostMessage

14-Jun-10 11:08

The Java style of exceptions was actually around a couple of years before Java first came out. Microsoft had an exception handling mechanism in MFC 1 and 2 that threw by pointer and caught by pointer requiring the caller to delete the exceptions. It also used setjmp and longjmp to transfer control despite all the problems with that. It's still there in MFC 19 years later (but uses real C++ exception handling these days).

The only reason I brought the point up was I didn't want another neophyte C++ programmer getting the idea that the correct way to throw and catch was some knackered old view of the world imposed by a second rate C++ compiler from 20 years ago. The poor souls have enough to cope with!

Cheers,

Ash

Pryabu13-Jun-10 20:23

Pryabu

13-Jun-10 20:23

In which scenario,Sendmessage and Postmessage should be used?

Thanks,

«_Superman_»13-Jun-10 20:27

13-Jun-10 20:27

You can send a standard Windows message or a user defined message using both SendMessage and PostMessage.
SendMessage will wait for the message handler to complete the action, whereas PostMessage will simply queue the message to the target window's message queue.
These APIs can be use to send messages to any window whether in the same process or not.

«_Superman_»

I love work. It gives me something to do between weekends.

Microsoft MVP (Visual C++)

Polymorphism in C

Pryabu13-Jun-10 20:53

Pryabu

13-Jun-10 20:53

thanks. I want to know in which case we should use sendmessage and postmessage functions?

«_Superman_»13-Jun-10 20:57

13-Jun-10 20:57

There can be many uses.
One such use is for simple interprocess communication like sending a WM_CLOSE message to an external application.
Another frequently used scenario is when you declare your own custom messages like WM_USER + 1 or WM_APP + 1.

«_Superman_»

I love work. It gives me something to do between weekends.

Microsoft MVP (Visual C++)

Polymorphism in C

Pryabu13-Jun-10 21:02

Pryabu

13-Jun-10 21:02

thanks