std::list question

**void_main** · 11-10-2003

hi,

i want to use a stl list of pointers in my program, so i made a little test application to find out exactly how much "maintenance" i need to perform on them (after removing them from the list). however, the behavior of the program has got me a little confused. the code below gives no errors (as i thought it would):

Code:

#include <list>
#include <iostream>
using namespace std;

class dummy
{
public:
	dummy() {};
	~dummy() { cout << "destructor" << endl; }

	void blah() { cout << "blah" << endl; }
};

int main()
{
	list<dummy*> theList;
	list<dummy*>::iterator it;

	dummy *a = new dummy();
	dummy *b = new dummy();
	theList.push_back(a);
	theList.push_back(b);

	cout << theList.size() << endl;

	for (it = theList.begin(); it != theList.end(); it++)
	{
		(*it)->blah();
	}

	theList.remove(a);

	delete a;
	a = NULL;
	a->blah();

	return 0;
}

it seems to me that the last a->blah() call should throw an access violation, yet it works fine (i even checked, and the destructor is being called like it should be on the delete a; line). can anyone tell me what's going on here? and what clean up i will need to worry about when handling a list of pointers? thanks!

**grib** · 11-10-2003

a->blah() is undefined behaviour, that includes actually working. Since dummy has no actual data members there is nothing to acess and therefore violate. The normal expected (but not defined) behavior of methods on a pointer to a deleted object is to go ahead and use all the memory it was using before, regardless of wether or not it makes any sence to do so. NULL just happens to be a memory location that it's easy for the hardware to inform the operating system that your program no longer knows what the hell it's doing, but c++ no longer cares, it's undefined.

As far as containers of pointers to dynamic memory, it's genrally not worth it. Make your copy ctor as cheap as posible or if it's a real problem use boost::shared_ptr.

**Fordy** · 11-10-2003

Using a pointer after a call to delete is effectively undefined....it may crash...it may not, but if you want reliable code, you wont take the risk.

As to using pointers as part of a container....I dislike doing this as it's cumbersome and often error prone......my preference is to either allow the object to be created naturally (use a container of objects...not pointers) or if I really need pointers, I'll try to use a smart pointer (have a look at the boost library)

[edit]..read grib's post as he is quicker at typing than me

**void_main** · 11-10-2003

thanks very much for the info.

**grib** · 11-10-2003

Originally posted by Fordy

[edit]..read grib's post as he is quicker at typing than me

Only because I use words like wether rite good I should have learned in collage, but the beer was too cheap.

**Cat** · 11-10-2003

In fact, THIS gives no errors on most compilers:

Code:

#include <iostream>
class C{
public:
	void dummy(){ std::cout << "Who's a dummy?" << std::endl;}
};

int main(){
	C* cptr = NULL;
	cptr->dummy();
}

The reason, as grib said, is that your function call does not attempt to use any member variables. If it doesn't need to access variables inside the class, it never truly dereferences the pointer -- that is, it never attempts to access memory using the pointer. The function is also not virtual, so it does not need to use a vtable in the object (which would again be dereferencing a pointer).

Of course, it's very bad to do something like this, and you can't expect it to work on all compilers, but it typically will, because there is no need to dereference the pointer when you don't need to access data.

Also, a function that was not virtual and that did not access any class data members doesn't need to be part of a class anyhow, so this result happens because it is a special case, and not typical of any real-world object.

Oh, and as Fordy said, I also highly recommend boost for low-maintenence, container-safe smart pointers.

Thread: std::list question

Thread Tools

Search Thread

Display

std::list question

Similar Threads

Alice....

Debugging question

Question about pointers #2

Question...

Question, question!