deque memory leak

Hi,
I can't figure out why the following is leaking memory - when I run the application the memory usage just keeps going up. Anyone see something wrong with this code:

I'm adding some initial values to the 2d deque like so:

Code:

deque<deque<MyType*> > dq;
for(int i = 0; i < 10; i++) {
	dq.push_back(deque<MyType*>());
	for(int j = 0; j < 10; j++) {
		dq[i].push_back(new MyType(1,2));
	}
}

And then during the running of the application, the user periodically performs actions that invoke this:

Code:

for(int j = 0; j < 10; j++) {
    delete dq.back()[j];
}
dq.pop_back();
			
dq.push_front(deque<MyType*>());
for(int j = 9; j >= 0; j--) {
	dq.front().push_front(new MyType(1,2));
}

or this:

Code:

for(int i = 0; i < 10; i++) {
    dq[i].push_front(new MyType(1,2));
    delete dq[i].back();
    dq[i].pop_back();
}

Much thanks for any help!

Do you really need a deque of deque of pointers to MyType rather than a deque of deque of MyType objects? If so, have you considered using a deque<deque<tr1::shared_ptr<MyType> > > instead?

Well, MyType contains a lot of data, so I was reluctant to pass copies. I guess I can reconsider that a bit. I hadn't considered using shared_ptr as I don't feel I have a firm grasp on what that is. How would that relate to this problem? Given your suggestion, should I take you to mean that the deque of deque of pointers to MyType is a bad way to do things?
Thanks

Originally Posted by pollypocket4eva

I hadn't considered using shared_ptr as I don't feel I have a firm grasp on what that is. How would that relate to this problem? Given your suggestion, should I take you to mean that the deque of deque of pointers to MyType is a bad way to do things?

Well, you say that you have a memory leak problem, then I see that you are doing manual memory management. The idea behind what I am saying is simply: do not do manual memory management. Storing the objects by value accomplishes that, as does using a smart pointer like std::tr1::shared_ptr. If you have access to Boost, another possibility is to use a boost::ptr_deque.

Hey,
Thanks for your help. I've just found my problem. Turns out the deque was actually fine all along. I thought I'd ruled out all other possible causes of the leak, but there was something I'd missed.
Thanks!