bug in release-1.0 with respect to method pointers?

bug in release-1.0 with respect to method pointers?

[Dale Martin]

Hello.  I'm trying to compile a large project I'm working on with
egcs-1.0.  I'm immediatley having compilation problems.  Here is some code
that precisely illustrates the problem:

[ begin included source ]

#include <iostream.h>

class abstractBase {
public:
  abstractBase(){}
  ~abstractBase(){}
  
  virtual int returnInt(){
    return -1;
  }
};

class commonAbstract : abstractBase {
public:
  commonAbstract(){}
  ~commonAbstract(){}
  
  virtual int returnInt(){
    return -1;
  }
};

class derived1 : public commonAbstract {
public:
  derived1(){}
  ~derived1(){}

  int returnInt(){
    return 1;
  }
};

class derived2 : public commonAbstract {
public:
  derived2(){}
  ~derived2(){}
  
};

class unrelated {
public:
  unrelated(){}
  ~unrelated(){}
  
  void printResult(commonAbstract *callOn, int(commonAbstract::*memberFunction)() ){
    cout << (callOn->*memberFunction)() << endl;
  }
  
};

int 
main( int argc, char *argv[] ){
  commonAbstract *ca1, *ca2;
  unrelated *u;

  u = new unrelated();
  ca1 = new derived1();
  ca2 = new derived2();

  cout << "Testing on a derived1" << endl;
  u->printResult( ca1, commonAbstract::returnInt );

  cout << "Testing on a derived2" << endl;
  u->printResult( ca2, commonAbstract::returnInt );
}

[ end included source ]

This code compiles and behaves as expected with g++-2.7.2.3 and SunPro
4.2.

(Here is the output:)
Testing on a derived1
1
Testing on a derived2
-1

With egcs-1.0 I get:
bug.cc: In function `int main(int, char **)':
bug.cc:61: no matching function for call to `unrelated::printResult (commonAbstract *&, int (commonAbstract::)())'
bug.cc:45: candidates are: unrelated::printResult(commonAbstract *, int (commonAbstract::*)())
bug.cc:64: no matching function for call to `unrelated::printResult (commonAbstract *&, int (commonAbstract::)())'
bug.cc:45: candidates are: unrelated::printResult(commonAbstract *, int (commonAbstract::*)())

I'm compiling on a linux/libc6 system.  I enable haifa extensions when I
configured egcs.

I realize this could be a problem in my code - it's kind of a weird thing
to do :-)

Thanks,
	Dale

-- 
+--------------------  finger for pgp public key  ---------------------+
| Dale E. Martin | University of Cincinnati Savant Research Laboratory |
| dmartin@ececs.uc.edu    |     http://www.ececs.uc.edu/~dmartin       |
+----------------------------------------------------------------------+

Re: bug in release-1.0 with respect to method pointers?

[Olivier Galibert]

On Thu, Dec 04, 1997 at 11:18:02AM -0500, Dale Martin wrote:
> Hello.  I'm trying to compile a large project I'm working on with
> egcs-1.0.  I'm immediatley having compilation problems.  Here is some code
> that precisely illustrates the problem:

Your code is invalid. Use (note the &) :
  u->printResult( ca1, &commonAbstract::returnInt );

and then it works.

  OG.

Re: bug in release-1.0 with respect to method pointers?

[H.J. Lu]

> 
> 
> Hello.  I'm trying to compile a large project I'm working on with
> egcs-1.0.  I'm immediatley having compilation problems.  Here is some code
> that precisely illustrates the problem:
> 
> [ begin included source ]
> 
> #include <iostream.h>
> 
> class abstractBase {
> public:
>   abstractBase(){}
>   ~abstractBase(){}
>   
>   virtual int returnInt(){
>     return -1;
>   }
> };
> 
> class commonAbstract : abstractBase {
> public:
>   commonAbstract(){}
>   ~commonAbstract(){}
>   
>   virtual int returnInt(){
>     return -1;
>   }
> };
> 
> class derived1 : public commonAbstract {
> public:
>   derived1(){}
>   ~derived1(){}
> 
>   int returnInt(){
>     return 1;
>   }
> };
> 
> class derived2 : public commonAbstract {
> public:
>   derived2(){}
>   ~derived2(){}
>   
> };
> 
> class unrelated {
> public:
>   unrelated(){}
>   ~unrelated(){}
>   
>   void printResult(commonAbstract *callOn, int(commonAbstract::*memberFunction)() ){
>     cout << (callOn->*memberFunction)() << endl;
>   }
>   
> };
> 
> int 
> main( int argc, char *argv[] ){
>   commonAbstract *ca1, *ca2;
>   unrelated *u;
> 
>   u = new unrelated();
>   ca1 = new derived1();
>   ca2 = new derived2();
> 
>   cout << "Testing on a derived1" << endl;
>   u->printResult( ca1, commonAbstract::returnInt );
> 
>   cout << "Testing on a derived2" << endl;
>   u->printResult( ca2, commonAbstract::returnInt );
> }
> 

I think there is a difference in g++ between the address of a
member function and a pointer to a member function. In g++,
"commonAbstract::returnInt" gives the memory address of
commonAbstract::returnInt and "&commonAbstract::returnInt"
gives a pointer to commonAbstract::returnInt.

Not all C++ compilers do that. VC++ 5.0 treats them the same.
Personally I like the g++ way. I have some code depending on it.
I know it is very compiler specific. But it seems to be a simple
way to pass a non-virtual member function and THIS to a thread
creatation function. Without it, I have to write more code
for it.


H.J.