Traversal & Deletion in an STL List

Started by
5 comments, last by Brother Bob 11 years, 5 months ago
I've been having problems with some code I used to traverse the dynamic objects and update them. Strangely enough the code works fine on my Ubuntu Linux computer while crashing on a Windows 7 PC after some items are juggled around. Here it is:
//The horde is defined as so:
std::list<zombie*> horde;
typedef std::list<zombie*>::iterator hordeIt;
//Update the zombies
for (hordeIt zIt=horde.begin(); zIt !=horde.end(); ++zIt)
{
//The zombie will know if it should die. If a zombie returns DELETE, remove the list item
if ((*zIt)->onUpdate()==DELETE) //Note that DELETE is just #define DELETE 1, and that under Windows I had to rename it to DELETE_D due to some error (which didn't help anything)
{
delete (*zIt);
horde.erase(zIt);
//Make sure we don't try to access a nonexistent location
--zIt;
continue;
}
(*zIt)->onRender();
//Resolve health events with collisions (More zombie stuff here) ...
}
//New zombies are created like so:
horde.push_back(new zombie(/*various args*/));

When I lint the code with CppCheck I get "Dereferenced iterator zIt has been deleted", but it compiles fine under GCC and MinGW for Windows. No crashes occur under Linux but the Windows build crashes very frequently.
Here are the questions I have regarding this code:

  1. Do I need to use a separate iterator for the deletion of zIt?
  2. Is it alright if I continue to use the STL heavily? (I actually made my own doubly linked list and replaced the STL list in the same project as a test but it had phantom segfaults and thread sync errors [despite the project being single-threaded])
  3. Why do you think this code worked on Linux but not on Windows?

Want to get to know my work and I better? See my website: Au 79 Games

I wrote General Tips on the Process of Solo Game Development

Advertisement
Erase returns the a new iterator value; you should be using it instead of the dodgey "--" trick.
for (hordeIt zIt=horde.begin(); zIt !=horde.end(); /*N.B. nothing here*/)
{
if (...)
{
zIt = horde.erase(zIt);
}
else
{
++zIt;
}
}

2) Replacing STL data-structures with custom implementations that do the same thing isn't that helpful (especially when you replace them with buggy versions). Avoiding STL structures is more useful if you require another kind of structure altogether (e.g. something other than a doubly linked list, e.g. an enumerable pool of zombies, etc...).
3) You were lucky in the game of russian-roulette that is undefined-behaviour. In your code, --zIt is making use of an invalidated iterator. Depending on the STL implementation, this could do anything (including 'work as intended').

Erase returns the a new iterator value; you should be using it instead of the dodgey "--" trick.
for (hordeIt zIt=horde.begin(); zIt !=horde.end(); /*N.B. nothing here*/)
{
if (...)
{
zIt = horde.erase(zIt);
}
else
{
++zIt;
}
}

I did come across using the returned value of erase() while doing some research on this problem, but I implemented it incorrectly when I did adjust the code. Thanks for showing me how to do it right!

2) Replacing STL data-structures with custom implementations that do the same thing isn't that helpful (especially when you replace them with buggy versions). Avoiding STL structures is more useful if you require another kind of structure altogether (e.g. something other than a doubly linked list, e.g. an enumerable pool of zombies, etc...).

That makes a lot of sense. I should probably not try to make one tool do a different one's job.

3) You were lucky in the game of russian-roulette that is undefined-behaviour. In your code, --zIt is making use of an invalidated iterator. Depending on the STL implementation, this could do anything (including 'work as intended').

Ah, I see.
Again, thanks for the answer, I'll rework the code & see if I can get that Windows build going smile.png.


UPDATE: After modifying the list code it built perfectly!

Want to get to know my work and I better? See my website: Au 79 Games

I wrote General Tips on the Process of Solo Game Development

It's worth noting that some std containers don't implement a return of an iterator form erase. For example although std::map on Microsoft's implementation does, std::map doesn't return an iterator in the standard implementation.

A way around this problem in those circumstances is store the iterator in another variable, and then increment the iterator prior to erasing the stored iterator.

[source lang="cpp"]hordeIt toEraseIt = zIt;
++zIt;
horde.erase( toEraseIt );[/source]

It's worth noting that some std containers don't implement a return of an iterator form erase. For example although std::map on Microsoft's implementation does, std::map doesn't return an iterator in the standard implementation.

This was changed in C++11. Now associative containers return an iterator to the next element on the iterator based erase() calls.

It's worth noting that some std containers don't implement a return of an iterator form erase. For example although std::map on Microsoft's implementation does, std::map doesn't return an iterator in the standard implementation.

A way around this problem in those circumstances is store the iterator in another variable, and then increment the iterator prior to erasing the stored iterator.

[source lang="cpp"]hordeIt toEraseIt = zIt;
++zIt;
horde.erase( toEraseIt );[/source]


You can also postincrement it when calling erase.

[source lang="cpp"]if (wantToDelete)
horde.erase(zIt++);
else
++zIt;[/source]
A thread like this is typically not complete without a standard library-centric answer.

horde.remove_if(
[](zombie *z) {return z->onUpdate() == DELETE;}
);

std::for_each(
std::begin(horde),
std::end(horde),
[](zombie *z) {z->onRender();}
);

Or if you want to to the update and the rendering in the same place.

horde.remove_if(
[](zombie *z)->bool {
if(z->onUpdate() == DELETE) {
return true;
}
z->onRender();
return false;
}
);

This one is less obvious and self-explaning though as you're mixing rendering code with the update and remove logic.

This topic is closed to new replies.

Advertisement