Newton crash on switching level

[AggrorJorn]

From time to time my game crashes when a collision hook is triggered.

It is difficult to see where it goes exactly wrong. I can send you the project if you like.

 

I always get the following call stack error:

[Josh]

I can debug it if you send the project.

[AggrorJorn]

I send you a pm with the source code.

[AggrorJorn]

I noticed that the stack call has changed slightly. It still happens at the same time:

[Josh]

Not a good idea. I have hundreds of messages. Please post it here if you can.

[AggrorJorn]

I really would like to see this solved, so I will just put the source code here although I would have prefer it if we could have some sort of secure environment for this.

http://www.leadwerks.com/werkspace/index.php?app=core&module=attach&section=attach&attach_id=4856

[Josh]

testing now...

[Josh]

Is this using a callback or script to change the map inside the collision hook? This would definitely produce a crash. For stuff like this, it's best to set a global variable and then handle that in your main loop.

[Rick]

Check out some of the posts below this to get more detail Aggror: http://www.leadwerks.com/werkspace/files/file/430-next-level/

 

I don't agree with the global idea but the idea Josh is trying to say is that you can't change levels when inside the engines World::Update() method, which is the case inside a callback since World::Update() is actually calling that callback. You can set some flags and such inside this, and then change it at the start of your game loop. Again, I would avoid globals personally as there is for sure a better way to handle map switching but to each their own.

[AggrorJorn]

The hook only sets a global bool to true. The level loading itself is done within a class. I am not using any lua to load levels.

 

After replacing the collision hook (which triggered a next level loading outside the hook) with a simple distance checker, I still get this error. So right now I don't have any collision hooks.

[Josh]

I found the cause of this. Old bodies that were scheduled for deletion were hanging around causing errors. This was a mistake in my threading implementation.