Skip to main content
GameDev.net gamedev.net
🔒 Locked

vector subscript out of range

Started by lride Dec 8, 2012 at 5:48 PM 7 replies 2.2k views
Original Post
lride
lride
I have a vector of pointer to entity. I also have a GameWorld class that contains this vector.
[source lang="cpp"]std::vector entities;

GameWorld::GameWorld()
{
entities.push_back(new Character("YOURS", CharacterType::BLACK));
}
GameWorld::~GameWorld()
{
for(auto e: entities)
delete e;
}[/source]

When I close the game I get assertion failure "vector subscript out of range". What did I do wrong?

Edit:
I still get the vector assertion failure even when I changed the std::vector to std::list
I also found out that I can use the std::vector even I don't #include

OK I solved my problem. The GameWorld object is inside Game class which is a singleton. So the GameWorld is statically allocated.
But when I allocated Game class on stack, it worked fine. I have no idea why..
An invisible text.
MarkS_
MarkS_
Why not use "delete entities;"? The vector destructor takes care of calling the Entity destructor for each object in the list..
Radikalizm
Radikalizm

Why not use "delete entities;"? The vector destructor takes care of calling the Entity destructor for each object in the list..


The vector will destroy the pointers, but not the objects the pointers are pointing to. Big difference!
Also 'delete entities' wouldn't even work as the entities vector wasn't allocated on the heap. It wouldn't even compile either, as entities is not a pointer type.
I gets all your texture budgets!
MarkS_
MarkS_
I see. He is creating a vector of Entity pointers, not a pointer to a vector of Entity objects.
lride
lride
Anyways getting rid of evil singleton design solved the problem
An invisible text.
Radikalizm
Radikalizm

Still, doesn't the vector destructor call the object's destructor upon deletion?


A vector does call a destructor for an object it holds when that object is erased, but this vector is storing pointer objects, not Entity objects.
So basically the vector is calling a destructor for the pointer object (which is perfectly legal), but that doesn't mean the destructor for the object the pointer is pointing to gets called or that the memory consumed by the object the pointer is pointing to gets released.
I gets all your texture budgets!
MarkS_
MarkS_

[quote name='MarkS' timestamp='1354997809' post='5008591']
Still, doesn't the vector destructor call the object's destructor upon deletion?


A vector does call a destructor for an object it holds when that object is erased, but this vector is storing pointer objects, not Entity objects.
So basically the vector is calling a destructor for the pointer object (which is perfectly legal), but that doesn't mean the destructor for the object the pointer is pointing to gets called or that the memory consumed by the object the pointer is pointing to gets released.
[/quote]

That is very interesting! I don't create vectors of pointers, but I'll need to keep that in mind.

Since we are on the subject, why create a vector of pointers in the first place?
iMalc
iMalc

When I close the game I get assertion failure "vector subscript out of range". What did I do wrong?
You either corrupted your vector somewhere, or you tried to access outside of its bounds.

Edit:
I still get the vector assertion failure even when I changed the std::vector to std::list
I also found out that I can use the std::vector even I don't #include [/quote]Both of those statements are inaccurate.
If you change it to a list then you wont get assert about the vector because it isn't a vector any more. A std::list does not use code from within the std::vector. You're probably just getting the same error from a different piece of the std library. Getting the same behviour probably means that you are going past the end of the container rather than corrupting it.
If code with vector still compiles, then something else you are including is pulling in the vector header for you. You should probably not be relying on that.

If you need your vector to hold pointers for reasons of polymorphism, then you might be better off with a boost ptr_container rather than a vector. Otherwise you should problably just store instances directly rather than pointers.
lride
lride

Since we are on the subject, why create a vector of pointers in the first place?

Because I want to store different kinds of object in the same vector. This can only be done by storing the pointers or references to the object and using polymorphism.
Also, I store pointers to objects in a vector when the object is too expensive to copy. (When you use std::vector::push_back(), the copy of the object gets stored.)
However, this extra copy can be avoided by using c++11 'move', .



Both of those statements are inaccurate.
If you change it to a list then you wont get assert about the vector because it isn't a vector any more. A std::list does not use code from within the std::vector. You're probably just getting the same error from a different piece of the std library. Getting the same behviour probably means that you are going past the end of the container rather than corrupting it.
If code with vector still compiles, then something else you are including is pulling in the vector header for you. You should probably not be relying on that.

If you need your vector to hold pointers for reasons of polymorphism, then you might be better off with a boost ptr_container rather than a vector. Otherwise you should problably just store instances directly rather than pointers.


This is the only place I'm using vector in my entire project. and I've searched #include in my entire solution and there was none.
But now that I already solved the problem, what is the advantage of using a boost ptr vector instead of using a normal vector
An invisible text.

Topic Locked

This topic has been locked by a moderator. New replies are not allowed.

Sign in to reply to this topic.