Crashing in ActionManager on Scene transition

cocos2d for iPhone » Programming » cocos2d support (graphics engine)

Crashing in ActionManager on Scene transition

(24 posts) (10 voices)

Started 2 years ago by malphigian
Latest reply from Gamer

Tags:

malphigian
Member

I'm having an intermittent problem with my game where it crashes in the middle of a scene transition. Each scene has a single layer. The crash happens right around when the layer of the first scene is being dealloc'd and the second scene and layer inits have run.

The message in the debugger is:
Program received signal: “EXC_BAD_ACCESS”.

The crash appears to occur at Line 316 of ActionManager, inside tick:,
currentTarget = (tHashElement*) bin->elt;

This is called by Scheduler tick which in turn called by Director main loop.

Obviously ActionManager is trying to access an object that is already gone, but I'm unclear how to figure out what object and why. Probably something in the Layer that just got dealloc'd? I'm not using the Scheduler at all myself.

Any tips on how to track this down?

Thanks in advance!

Posted 2 years ago #
tehsuck
Member

We're having a similar problem, but it doesn't seem to be related to anything in particular (randomly occurs usually quickly after app start, or about 30-60 minutes in)... then :KA-BLAMM: we get EXC_BAD_ACCESS and I can't really figure out what's causing it. Is this a bug becuase I've seen other posts on the forum / bug-list that seem to concern Scheduler / ActionManager?

Thanks for any help!

---------------------
Date/Time: 2009-08-12 07:51:43.893 -0700 OS Version: iPhone OS 3.0 (7A341) Report Version: 104
Exception Type: EXC_BAD_ACCESS (SIGSEGV) Exception Codes: KERN_INVALID_ADDRESS at 0x3d2aaaab Crashed Thread: 0
Thread 0 Crashed: 0 xxOurAppxx 0x00030daa -[ActionManager tick:] (ActionManager.m:316) 1 xxOurAppxx 0x00048d56 -[Timer fire:] (Scheduler.m:83) 2 xxOurAppxx 0x000493c0 -[Scheduler tick:] (Scheduler.m:211) 3 xxOurAppxx 0x0003848a -[Director mainLoop] (Director.m:182) 4 xxOurAppxx 0x000392f4 -[FastDirector preMainLoop] (Director.m:887) 5 CoreFoundation 0x3020ee3c __invoking___ + 60 6 CoreFoundation 0x30262ac0 -[NSInvocation invoke] + 108 7 CoreFoundation 0x30262a42 -[NSInvocation invokeWithTarget:] + 30 8 CoreFoundation 0x3025af5a -[NSObject performSelector:withObject:] + 18 9 Foundation 0x305025b6 __NSThreadPerformPerform + 230 10 CoreFoundation 0x302548ec CFRunLoopRunSpecific + 1908 11 CoreFoundation 0x30254164 CFRunLoopRunInMode + 44 12 GraphicsServices 0x3204529c GSEventRunModal + 188 13 UIKit 0x308f0374 -[UIApplication _run] + 552 14 UIKit 0x308eea8c UIApplicationMain + 960 15 xxOurAppxx 0x0000223c main (main.m:15) 16 xxOurAppxx 0x000021d8 start + 44

Posted 2 years ago #
SmallMike
Member

Similar problem here. Unfortunately it is really hard pinpoint the problem since it only happens intermittently. We do have a scheduled timer in the Layer that is being disposed, I might try manually un-scheduling that before the transition.

Crash log:

0 Sir Revs-a-Lot 0x0001e8c0 -[ActionManager tick:] (ActionManager.m:320)
1 Sir Revs-a-Lot 0x00042594 -[Timer fire:] (Scheduler.m:83)
2 Sir Revs-a-Lot 0x00042f7c -[Scheduler tick:] (Scheduler.m:210)
3 Sir Revs-a-Lot 0x00028bfc -[Director mainLoop] (Director.m:182)
4 Foundation 0x306942a0 __NSFireTimer + 132
5 CoreFoundation 0x30269d88 CFRunLoopRunSpecific + 2642
6 CoreFoundation 0x30269320 CFRunLoopRunInMode + 44
7 GraphicsServices 0x31567e58 GSEventRunModal + 268
8 UIKit 0x30a4fa6c -[UIApplication _run] + 520
9 UIKit 0x30a591d0 UIApplicationMain + 1132
10 Sir Revs-a-Lot 0x000037e8 main (main.m:6)
11 Sir Revs-a-Lot 0x00003768 start + 44

Posted 2 years ago #
Irish
Member

I've had the same problem since switching to cocos 0.8.

The ActionManager crashes at line 316:
currentTarget = (tHashElement*) bin->elt;

I've got quite a lot of AtlasSprites (around 60) in my layer, which are all animating using actions. When they need to change to a different animation, I call stopAction on the running animation, and then start the new one.

Having a lot of these calls to stopAction seems to affect this bug, as if I take it out and just run the new animations on top of the running one, it doesn't seem to crash. (May be unrelated but i will do some more investigation).

Posted 2 years ago #
malphigian
Member

I did some more testing on this, setting up my game so it could "play itself", which would reliably cause a crash sooner or later.

It always crashes around the same area of ActionManager (in tick:), and *always* on the same reference.

Specifically when it tries to reference the element of the ccHashSetBin bin (bin = targets->table[i]) -- this is a struct.

I tried to enable NSZombie, but I think it's no help because we're dealing with a struct here not an object that is retained and released.

I'm out of my depth here, but if I had to guess deleteHashElement: is being called on an element, which in turn calls free() on that element, so it's gone. Yet the ccHashSetBin still exists inside targets->table -- thus the bad access.

I know absolutely nothing about multithreaded programming, but is it possible the element is getting freed *while* the tick: is in the middle of trying to use it?

Probably not, I'm just sort of at a loss here and don't really know what, if anything, in my code might be causing ActionManager to crash out like this.

Posted 2 years ago #
riq
Key Master

there is an open bug:
http://code.google.com/p/cocos2d-iphone/issues/detail?id=490

but I would like to point out that cocos2d (the whole API) is not thread safe.
You should NOT call any cocos2d API from a different thread other than the main-thread.

Since I can't reproduce the crash (mono-thread), in order to fix it I need a an example to reproduce it.

I would appreciate if anyone of you could attach the code that reproduces the crash.
Please attach it in the open issue.

thanks.

Posted 2 years ago #
tehsuck
Member

Well right now we can't really tell what code is causing the crash, but my co-worker and I have found a level on our game that seems to reproduce the crash quicker... we did notice (and I'm kind of a Obj-C/C n00b) that it's bin->elt is always crashing on the same memory address
-----
0x3d2aaaab
-----
And that is coming from two builds on separate phones (one is a 3G, the other a 3GS).

Posted 2 years ago #
riq
Key Master

@tehsuck: thanks. what I need is an example that reproduces the issue since I can't reproduce it. Could you attach your example in the issue ? thanks.

Posted 2 years ago #
malphigian
Member

I just linked a very stripped down version of our game in the open bug thread ( http://code.google.com/p/cocos2d-iphone/issues/detail?id=490 ).

Let me know if that version gets the bug for you or not, it happens every time for me. I set up the game to flip scene all by itself so you can just let it run after hitting play. It usually takes a few minutes to crash.

Posted 2 years ago #
riq
Key Master

thanks. I'm testing it right now.

Posted 2 years ago #
riq
Key Master

ok. I found the bug.
Anyone want to try the fix ?
r1283

thanks.

Posted 2 years ago #
tehsuck
Member

It seems we fixed our issue for now by removing some calls to [CocosNode stopAllActions] - seems a lot more stable (as Irish said above - thanks).

Posted 2 years ago #
riq
Key Master

tehsuck: could you try with r1283 ? thanks.
it was an interesting bug.

Posted 2 years ago #
tehsuck
Member

Riq - it appears that solved the bug (at least in our code) many, many thanks for being so active in the forum and helping us out!

Posted 2 years ago #
malphigian
Member

Thanks for addressing this bug so quickly.

Unfortunately, I have not been able to test the fix, because r1283 will not compile for me. I get this:
---
"_inflateEnd", referenced from:
_inflateMemory_ in ZipUtils.o
_inflateMemory_ in ZipUtils.o
_inflateMemory_ in ZipUtils.o
"_inflate", referenced from:
_inflateMemory_ in ZipUtils.o
"_inflateInit2_", referenced from:
_inflateMemory_ in ZipUtils.o
ld: symbol(s) not found
collect2: ld returned 1 exit status
---

Any idea why?

Posted 2 years ago #
riq
Key Master

@malphigian:
http://www.cocos2d-iphone.org/wiki/doku.php/release_notes:0_8_1_beta#added_removed_files

Posted 2 years ago #
malphigian
Member

Great, thanks.

So far no crashes, the new code seems to be working great.

Posted 2 years ago #
crmagicxxx
Member

I had similar errors (can't recall exact ones) as malphigian but it wasn't due to missing files, it was about missing libz.dylib library. Do I need it or I made some mistake with cocos2d files?

Posted 2 years ago #
riq
Key Master

@crmagicxxx: yes, you need to link against libz if you use the v0.8.1 files.
If you don't want to update to v0.8.1 yet, you can just replace the ActionManager.m file without adding the new files.

Posted 2 years ago #
Tone
Member

I have the necessary files as malphigian but I am still getting the error. I cleaned it with no success. What am I missing?

Posted 2 years ago #
Tone
Member

When I try and use 8.1 I get the same error.

---
"_inflateEnd", referenced from:
_inflateMemory_ in ZipUtils.o
_inflateMemory_ in ZipUtils.o
_inflateMemory_ in ZipUtils.o
"_inflate", referenced from:
_inflateMemory_ in ZipUtils.o
"_inflateInit2_", referenced from:
_inflateMemory_ in ZipUtils.o
ld: symbol(s) not found
collect2: ld returned 1 exit status
---

Even though ZipUtils is present in my Exisiting files.

Posted 2 years ago #
Clain
Member

In your linked frameworks, did you add libz.1.2.3.dylib ?

Posted 2 years ago #
lbalves
Member

Tricky... I just started on Cocos2D and I'm using the tutorials from bit101. The _inflate crash on his project template made me think twice, but now it's fine after including the libz. Thanks guys! Hopefully, I will be able to help others soon.

Posted 2 years ago #
Gamer
Member

Hi, You all getting this error:

Undefined symbols for architecture armv6:
"_inflateEnd", referenced from:
_inflateMemoryWithHint in ZipUtils.o

only due to missing of -lz in other linking flags,

Just go to build setting and add -lz in other linking flags.

Posted 4 weeks ago #

RSS feed for this topic

Reply

You must log in to post.