pointer-to-member function question

[everid]

#include <iostream>

class test
{
public:
test(int i){ m_i=i;}
test(){};
void hello()
{
  std::cout<<"hello()"<<std::endl;
}
private:
int m_i;
};

int main()
{
 test *p=new test();
 p->hello(); 
 p=0; 
 p->hello();
} 

the output is

hello()
hello()

using g++ 4.2.4 and VC2005.
why even with NULL pointer, we can still call the function correctly?

[PanGalactic]

You don't declare it as such, but hello() is a static function. It doesn't matter what value p holds and the compiler doesn't care. It's going to call hello(). This is a silent optimization.

You can verify that this is the case by replacing hello() with this:

void hello()
{
  std::cout<<"hello()"<< m_i << std::endl;
}

The result? Segmentation fault.

[everid]

sure, what you said about dereference p to get data member is correct.
but why the value p holds the compiler doesn't care?
as I tried in vc2005,
the first call f(), "p" and "this" has address as 0x003b6ca0.
the second call f(), "p" and "this" has address as 0x00000000.

so the "this" pointer is a temporary??? and can be generated as an arbitrary value???

[Disch]

why even with NULL pointer, we can still call the function correctly?

Because 'this' isn't used in the function in this case.

What you're doing is similar to this:

void func(test* foo)
{
  cout << "func()";
}

int main()
{
  test* t = 0;
  func( t );  // works just fine, even though 't' is null because it's never used
}

That's why it's working for you.

However it is not guaranteed to always work (if, for example, 'hello' was a virtual function, your program would explode) so it's ill advised to do such a thing.