«

#include "project.h"


class ThatDoesProcessing;
typedef int (ThatDoesProcessing::*PFUNC)(int);


class ThatDoesProcessing {

public:

int ProcessThis( PFUNC pFunc ) {

int iRet = (this->*pFunc)( 2 );

return iRet;

}

};


class ThatNeedsToBeProcessed : public ThatDoesProcessing {

public:

void Test( void ) {

PFUNC pMyFunc;

pMyFunc = (PFUNC)CompareFunc;

ProcessThis( pMyFunc );

}

int CompareFunc( int i ) {

return i;

}

};


int _tmain( int /* argc */, TCHAR * /* argv[] */ ) {

ThatNeedsToBeProcessed b;

b.Test();

return 0;

}

---------------------------------------------------
The above compiles without any warnings or errors at all even
with warning level set to 4 with the MS compiler.

Compliling it with GCC version 3.3.1 (cygming special)
gives the following output:

$ make
g++ -O2 -fno-strength-reduce -I/cygdrive/m/include -I/usr/X11R6/include
-D__i386__ -DWIN32_LEAN_AND_MEAN -DX_LOCALE -D_X86_ -D__CYGWIN__
-D_XOPEN_SOURCE -D_POSIX_C_SOURCE=199309L -D_BSD_SOURCE
-D_SVID_SOURCE -D_GNU_SOURCE -c -o main.o main.cpp
main.cpp: In member function `void ThatNeedsToBeProcessed::Test()':
main.cpp:33: error: argument of type `int (ThatNeedsToBeProcessed:(int)' does
not match `int (ThatDoesProcessing::*)(int)'
make: *** [main.o] Error 1

---------------------------------------------------

The error is on the line which reads

pMyFunc = (PFUNC)CompareFunc;

in ThatNeedsToBeProcessed::Test

I thought the whole idea of a cast was to coerce the compiler into seeing
one type as another, so why does it refuse to do so here? How do I get
around it?

Thanks in advance.

"John Doe" <> wrote in message news n...

>
> I thought the whole idea of a cast was to coerce the compiler into seeing
> one type as another, so why does it refuse to do so here? How do I get
> around it?

Even if it compiled (which it shouldn't), your code exhibits undefined behavior. There's no
guarantee that it is going to work. You are not allowed to use the cast-bashed pointer
value until you cast it back to what it was before. The cast here is a reinterpret cast
and all you are guaranteed to be able to do is park the pointer value in the other pointer
and then cast it back to what it was before before you use it (you can also check it
against null).

It shouldn't even compile. You don't form pointer to members like you wrote it.
pMyFunc = reinterpret_cast<PFUNC>(&ThatNeedsToBeProcessed::C ompareFunc);

The fact that you even need the cast is telling. If this were to work nicely the implicit
conversion would suffice. The problem is that not every member of the derived class
is valid in base (and specifically this one isn't).

How about a functor or something.

"John Doe" <> threw a soggy newspaper against the
wall, and here's what stuck:

>
> The error is on the line which reads
>
> pMyFunc = (PFUNC)CompareFunc;
>
> in ThatNeedsToBeProcessed::Test
>
> I thought the whole idea of a cast was to coerce the compiler into
> seeing one type as another, so why does it refuse to do so here? How
> do I get around it?


This seemed to do the trick for me:


START CODE --

// #include "project.h"
#include <iostream>

class ThatDoesProcessing;
typedef int (ThatDoesProcessing::*PFUNC)(int);

class ThatDoesProcessing
{
public:
int ProcessThis( PFUNC pFunc ) {
int iRet = (this->*pFunc)( 2 );
return iRet;
}

virtual int CompareFunc( int i ) {
std::cout << "Oops!" << std::endl;
return 0;
}

};


class ThatNeedsToBeProcessed : public ThatDoesProcessing
{
public:
void Test( void ) {
PFUNC pMyFunc;
pMyFunc = &ThatDoesProcessing::CompareFunc;
ProcessThis( pMyFunc );
}

virtual int CompareFunc( int i ) {
std::cout << "Made it!" << std::endl;
return i;
}

};


// I eliminated the _tmain and TCHAR here so I could
// build it.
int main( int /* argc */, char * /* argv[] */ )
{

ThatNeedsToBeProcessed b;
b.Test();
return 0;
}

-- END CODE

PMP

John Doe escribió:

> #include "project.h"
>
> class ThatDoesProcessing;
> typedef int (ThatDoesProcessing::*PFUNC)(int);
>
> class ThatDoesProcessing {
>
> public:
>
> int ProcessThis( PFUNC pFunc ) {
>
> int iRet = (this->*pFunc)( 2 );
>
> return iRet;
>
> }
>
> };
>
> class ThatNeedsToBeProcessed : public ThatDoesProcessing {
>
> public:
>
> void Test( void ) {
>
> PFUNC pMyFunc;
>
> pMyFunc = (PFUNC)CompareFunc;

pMyFunc= (PFUNC) & ThatNeedsToBeProcessed::CompareFunc;

Regards.

"Paul M. Parks" <> wrote in message news:Xns941DA52E1CC1BDarthPaulOfDuluth@216.168.3.4 4...

> This seemed to do the trick for me:
>

Much better...well formed and no undefined behavior.

Ron Natalie wrote:

> "Paul M. Parks" <> wrote in message news:Xns941DA52E1CC1BDarthPaulOfDuluth@216.168.3.4 4...
>
>
>>This seemed to do the trick for me:
>>
>
>
> Much better...well formed and no undefined behavior.
>

The msdev compiler accepts a whole load of nonesense which
it ought to complain about. An example is:

int ff() {
return 0;
}

int main()
{
if (ff) { // this ought to have been 'if (ff()) {'
return 1;
}
return 0;
}

jeffc wrote:

> "lilburne" <> wrote in message
> news:bn9fh5$v9uis$...
>
>>The msdev compiler accepts a whole load of nonesense which
>>it ought to complain about. An example is:
>>
>>int ff() {
>> return 0;
>>}
>>
>>int main()
>>{
>> if (ff) { // this ought to have been 'if (ff()) {'
>> return 1;
>> }
>> return 0;
>>}
>
>
> My current compiler allows that too.
>

Fortunately the sun solaris and irix compilers generate
warnings. Try turning up the warning level on your compiler
and force all warnings to be errors.

"lilburne" <> wrote in message
news:bn9fh5$v9uis$...
>
> The msdev compiler accepts a whole load of nonesense which
> it ought to complain about. An example is:
>
> int ff() {
> return 0;
> }
>
> int main()
> {
> if (ff) { // this ought to have been 'if (ff()) {'
> return 1;
> }
> return 0;
> }

My current compiler allows that too.

"lilburne" <> wrote in message news:bn9fh5$v9uis$...

> The msdev compiler accepts a whole load of nonesense which
> it ought to complain about. An example is:
>
> int ff() {
> return 0;
> }
>
> int main()
> {
> if (ff) { // this ought to have been 'if (ff()) {'

Perfectly valid C++. ff is the address of the function. The statement is always true.

Ron Natalie wrote:

> "lilburne" <> wrote in message news:bn9fh5$v9uis$...
>
>
>>The msdev compiler accepts a whole load of nonesense which
>>it ought to complain about. An example is:
>>
>>int ff() {
>> return 0;
>>}
>>
>>int main()
>>{
>> if (ff) { // this ought to have been 'if (ff()) {'
>
>
> Perfectly valid C++. ff is the address of the function. The statement is always true.
>

It might be valid but it is never what was intended. Quite a
few compilers generate a warning.