I'm getting a crash periodically with a gpus_ReturnGuiltyForHardwareRestart.
The crash is sporadic, and occurs in a complex multi-threaded app. Sample stack trace below.
It seems to occur (which is rare) when the UI and related handlers are being stressed (think hyperactive user making lots of gestures as quickly as the app and system will allow), with lots of resulting calls to behind-the-scenes and up-front rendering.
Researching gpus_ReturnGuiltyForHardwareRestart suggests this may be due to a buffer issue, e.g. a buffer overrun due to an incorect binding or failure to unbind.
gpus_ReturnGuiltyForHardwareRestart crash
(not relevant, but I did look: gpus_ReturnGuiltyForHardwareRestart)
My understanding is that a buffer gets corrupted in some way, and the crash happens sometime later when the corrupted buffer is accessed.
I've been through the code and made sure every bound buffer and texture is subsequently unbound to prevent unwanted/unintentional changes by later code; still getting the crash.
I did just come across these, suggesting the possibility of a bug in the OS:
iOS 9 Beta 4 crash: gpus_ReturnGuiltyForHardwareRestart (original poster indicated was fixed in 9 beta 5)
Re: OpenGLES driver crash in iOS9 Beta2/3
However, I'm currently testing on 9.3.4, so it seems this would be fixed.I've tried making sure all buffers are properly unbound after use, and tried periodic use of glFlush(), both without success.
Does anybody have experience with this, any knowledge on potential sources, means of tracking down causes, or fixes?
Stack trace:
Date/Time: 2016-08-11T20:20:40Z
Launch Time: 2016-08-11T20:15:22Z
OS Version: iPhone OS 9.3.4 (13G35)
Report Version: 104
Exception Type: SIGSEGV
Exception Codes: SEGV_ACCERR at 0x1
Crashed Thread: 0
Thread 0 Crashed:
0 libGPUSupportMercury.dylib 0x0000000191bd9f28 gpus_ReturnGuiltyForHardwareRestart + 12
1 libGPUSupportMercury.dylib 0x0000000191bdaec4 gpusSubmitDataBuffers + 168
2 GLEngine 0x0000000195e9f1e4 gliPresentViewES_Exec + 172
3 GLEngine 0x0000000195e9f0fc gliPresentViewES + 80
4 OpenGLES 0x000000018576bc44 -[EAGLContext presentRenderbuffer:] + 68
5 NWFPApp 0x00000001001ef8fc -[NWFPIOSGLView renderNormalBuffers] (NWFPIOSGLView.mm:543)
6 NWFPApp 0x00000001001ef814 -[NWFPIOSGLView renderAll] (NWFPIOSGLView.mm:516)
7 NWFPApp 0x00000001001ee780 -[NWFPIOSGLView doInContext:] (NWFPIOSGLView.mm:135)
8 NWFPApp 0x00000001001ef770 -[NWFPIOSGLView drawView] (NWFPIOSGLView.mm:494)
9 NWFPApp 0x0000000100203cc8 -[NWFPGLChoreographer displayLinkEvent:] (NWFPGLChoreographer.m:128)
10 QuartzCore 0x000000018614022c CA::Display::DisplayLinkItem::dispatch() + 36
11 QuartzCore 0x00000001861400e0 CA::Display::DisplayLink::dispatch_items(unsigned long long, unsigned long long, unsigned long long) + 420
12 IOKit 0x0000000183885e54 IODispatchCalloutFromCFMessage + 368
13 CoreFoundation 0x00000001835ad030 __CFMachPortPerform + 176
14 CoreFoundation 0x00000001835c57d4 __CFRUNLOOP_IS_CALLING_OUT_TO_A_SOURCE1_PERFORM_FUNCTION__ + 52
15 CoreFoundation 0x00000001835c4f0c __CFRunLoopDoSource1 + 432
16 CoreFoundation 0x00000001835c2c64 __CFRunLoopRun + 1796
17 CoreFoundation 0x00000001834ecc50 CFRunLoopRunSpecific + 380
18 GraphicsServices 0x0000000184dd4088 GSEventRunModal + 176
19 UIKit 0x00000001887ce088 UIApplicationMain + 200
20 NWFPApp 0x00000001002af1a0 main (main.m:30)
21 ??? 0x000000018308a8b8 0x0 + 0
Related
I've been dealing with a pernicious crash bug in my iPhone app, which is consistently reproducible in the version downloaded from the App Store, but not at all in the debug version. I have produced a workaround, which works by excluding the critical drawing code from running during a UIViewController animation.
The crash log indicates that there is an assert failure in CGPathAddLineToPoint.
The app uses the drawing functions CGContextMoveToPoint, CGContextAddLineToPoint, and CGContextStrokePath. The functions CGPathCreateMutable, CGPathAddLines, and CGContextAddPath are also used, but the app still crashes even in configurations that (should) avoid invoking them.
The crash happens consistently when lines are drawn on a view while its UIViewController is being presented (animated). Otherwise, the drawing works without issue.
I have confirmed in the debug version that every point of every line drawn is real and within the boundaries of the view.
The app crashes immediately on iOS7 when the drawing is attempted. On iOS5 and iOS6, the app stops responding for 30 seconds before crashing, leaving behind an "Unknown" crash log with no information.
The only cases I could find on the web similar to this problem were CGFloatIsValid assert failures in CGPathAddLineToPoint in OS X apps (including iPhoto crash logs posted to Apple's support forums).
I can't say much more about it because the app never crashes or even gives a warning in the debug version. Has anyone else encountered this bug? Is there a better way to resolve this issue?
You can view the workaround in action in the release version on the App Store. You can see that the line drawing on the graph is delayed until after the animation completes when you push the [Graph] button, when the calculator is in 2D mode.
The important part of the crash log (from an iPhone 5S running iOS 7.0.3, app armv7):
Exception Type: EXC_CRASH (SIGABRT)
Exception Codes: 0x0000000000000000, 0x0000000000000000
Triggered by Thread: 0
Thread 0 Crashed:
0 libsystem_kernel.dylib 0x3a5e11fc __pthread_kill + 8
1 libsystem_pthread.dylib 0x3a648a4f pthread_kill + 55
2 libsystem_c.dylib 0x3a592029 abort + 73
3 libsystem_c.dylib 0x3a571c67 __assert_rtn + 179
4 CoreGraphics 0x2fd566f5 CGPathAddLineToPoint + 149
5 (My App) 0x001464cd 0xd7000 + 455885
6 UIKit 0x324d4dd5 -[UIView(CALayerDelegate) drawLayer:inContext:] + 369
7 QuartzCore 0x3210aa75 -[CALayer drawInContext:] + 97
8 QuartzCore 0x320f4241 CABackingStoreUpdate_ + 1857
9 QuartzCore 0x321ce2e9 ___ZN2CA5Layer8display_Ev_block_invoke + 49
10 QuartzCore 0x320f3af1 x_blame_allocations + 81
11 QuartzCore 0x320f37a3 CA::Layer::display_() + 1115
12 QuartzCore 0x320d7365 CA::Layer::display_if_needed(CA::Transaction*) + 205
13 QuartzCore 0x320d6ffd CA::Layer::layout_and_display_if_needed(CA::Transaction*) + 21
14 QuartzCore 0x320d6a09 CA::Context::commit_transaction(CA::Transaction*) + 225
15 QuartzCore 0x320d681b CA::Transaction::commit() + 311
16 QuartzCore 0x320d0549 CA::Transaction::observer_callback(__CFRunLoopObserver*, unsigned long, void*) + 53
17 CoreFoundation 0x2fc9bf67 __CFRUNLOOP_IS_CALLING_OUT_TO_AN_OBSERVER_CALLBACK_FUNCTION__ + 19
18 CoreFoundation 0x2fc998f3 __CFRunLoopDoObservers + 283
19 CoreFoundation 0x2fc99c3f __CFRunLoopRun + 735
20 CoreFoundation 0x2fc0446d CFRunLoopRunSpecific + 521
21 CoreFoundation 0x2fc0424f CFRunLoopRunInMode + 103
22 GraphicsServices 0x349382e7 GSEventRunModal + 135
23 UIKit 0x324b9841 UIApplicationMain + 1133
24 (My App) 0x000de723 0xd7000 + 30499
25 (My App) 0x000dde0c 0xd7000 + 28172
Several things:
Try opening the crash log on the machine from which you submitted the version in the app store. That version should have the archive containing the symbol file that you need to symbolicate the crash log.
If you only crash when you add lines to a path while your view controller is being animated onto the screen, why not add code to the completion block of your call to presentViewController:animated:completion: that invokes a method enabling drawing. Set up your view controller so drawing is disabled when it's first displayed, and gets enabled by the method call in the completion block.
To clarify (for people of the future), there is a workaround if your app is crashing during an animation, only in the App Store version, and the crash logs indicate an assert failure in CGPathAddLineToPoint. Just exclude all line-drawing code from running during the animation, and it won't crash.
I have been getting rather strange crash reports from my live app with stack traces like the following:
Thread 0: Crashed: com.apple.main-thread
0 libobjc.A.dylib 0x38af7942 realizeClass(objc_class*) + 117
1 libsystem_malloc.dylib 0x390dbef5 szone_malloc_should_clear + 1376
2 libobjc.A.dylib 0x38af976f lookUpImpOrForward + 74
3 libobjc.A.dylib 0x38af1feb _class_lookupMethodAndLoadCache3 + 34
4 libobjc.A.dylib 0x38af1db9 _objc_msgSend_uncached + 24
5 UIKit 0x30e571bf __57-[_UIDelayedPresentationContext beginDelayedPresentation]_block_invoke + 26
6 libdispatch.dylib 0x38fd9d07 _dispatch_client_callout + 22
7 libdispatch.dylib 0x38fe2803 _dispatch_source_invoke$VARIANT$mp + 262
8 libdispatch.dylib 0x38fe073d _dispatch_main_queue_callback_4CF$VARIANT$mp + 188
9 CoreFoundation 0x2e3ef819 __CFRUNLOOP_IS_SERVICING_THE_MAIN_DISPATCH_QUEUE__ + 8
10 CoreFoundation 0x2e3ee0ed __CFRunLoopRun + 1300
11 CoreFoundation 0x2e358c27 CFRunLoopRunSpecific + 522
12 CoreFoundation 0x2e358a0b CFRunLoopRunInMode + 106
13 GraphicsServices 0x3302c283 GSEventRunModal + 138
14 UIKit 0x30bfc049 UIApplicationMain + 1136
This is rather mysterious because neither the main thread nor any of the other live threads in the reports seem to imply that this is caused by my code, though of course I am skeptical of this.
This seems to be a rather common crash according to the number of reports I receive from Crashlytics, yet I have not been able to reproduce it on my own devices. I suspect this is probably related to some memory management issues because the various crashes always end up being some bad pointers being sent messages.
This always happens on this thread and following the -[_UIDelayedPresentationContext beginDelayedPresentation]_block_invoke call. This is obviously a private class being referenced from within some Apple framework, however I am at a loss to figure out exactly which one could be calling this.
The app is an educational game and I suspect this could be related to the GameKit API (particularly the Game Center authentication dialogs).
All of these crashes have been happening exclusively on iOS 7 and on iPad only. The app is universal so it is interesting to see that iPhone users seem to be unaffected.
Does anybody have any previous experience with these private classes that could give me some hints?
In one of our iPhone apps, we have gotten a few Crashlytics crashes that we have, so far, been unable to reproduce. However, on closer inspection, I noticed a bit of a trend...
Crash 1: -[UINavigationTransitionView _notifyDelegateTransitionDidStopWithContext:]
Free RAM: 10%
Crash 2: -[UINavigationTransitionView _notifyDelegateTransitionDidStopWithContext:]
Free RAM: 8%
Crash 3: HostBase::lock()
Free RAM: 8%
Crash 4:-[UINavigationTransitionView _notifyDelegateTransitionDidStopWithContext:]
Free RAM: 22%
Does the amount of free RAM mean anything? Can we be sure that these are actual crashes and not our app getting killed by the OS due to low memory? Does Crashlytics send crash reports if the app crashes while in the background?
Added a couple stack traces...
Exception Type: SIGSEGV, Nav Crashes
0 libobjc.A.dylib
objc_msgSend + 15
1 UIKit ✭
-[UINavigationTransitionView _notifyDelegateTransitionDidStopWithContext:] + 274
2 UIKit
-[UIViewAnimationState sendDelegateAnimationDidStop:finished:] + 158
3 UIKit
-[UIViewAnimationState animationDidStop:finished:] + 50
4 QuartzCore
CA::Layer::run_animation_callbacks(void*) + 208
5 libdispatch.dylib
_dispatch_client_callout + 22
6 libdispatch.dylib
_dispatch_main_queue_callback_4CF$VARIANT$mp + 224
7 CoreFoundation
__CFRunLoopRun + 1290
8 CoreFoundation
CFRunLoopRunSpecific + 356
9 CoreFoundation
CFRunLoopRunInMode + 104
10 GraphicsServices
GSEventRunModal + 74
11 UIKit
UIApplicationMain + 1120
Exception Type: SIGSEGV, HostLock crash
0 libobjc.A.dylib
objc_msgSend + 15
1 CFNetwork ✭
HostBase::lock() + 14
2 CFNetwork
DispatchHost::performInvocation(void const*) + 12
3 CFNetwork
__setupTCPConnection_block_invoke_2 + 290
4
...
libsystem_network.dylib
__tcp_connection_notify_complete_block_invoke_1 + 18
7 libsystem_c.dylib
_pthread_wqthread + 294
8% memory can be a lot, I highly doubt that this is an issue. I don't know if they send crashes while running in the background, you have to ask them directly since their SDK is not open source.
Both exception stack traces you added hint at a memory problem.
The first one looks like it tries to call a delegate at the end of an animation of an object, which does not exist any more.
The second one looks like a similar problem, where the completion block is accessing an object that is gone. So you might check your blocks for such patterns.
My application is crashing when one of the views is purged from memory because of low-memory condition. At least this is what I understand from the crashlog. It happens on numerous screens but only when opening a Facebook dialog (using the Facebook SDK). Basically, looks like the system sometimes runs out of memory when we need to present a Facebook dialog (e.g. to let user post something on the Facebook timeline).
Date/Time: 2012-03-14 19:47:33.819 +0000
OS Version: iPhone OS 5.1 (9B176)
Report Version: 104
Exception Type: EXC_BAD_ACCESS (SIGSEGV)
Exception Codes: KERN_INVALID_ADDRESS at 0x30000008
Crashed Thread: 0
Thread 0 name: Dispatch queue: com.apple.main-thread
Thread 0 Crashed:
0 libobjc.A.dylib 0x30f2bf78 objc_msgSend + 16
1 MyApp 0x00003c0e -LTBaseViewController viewDidUnload (LTBaseViewController.m:145)
2 MyApp 0x00004ea2 -LTBaseTableViewController viewDidUnload (LTBaseTableViewController.m:90)
3 UIKit 0x33766bd8 -[UIViewController unloadViewForced:] + 244
4 UIKit 0x338ae492 -[UIViewController purgeMemoryForReason:] + 58
5 Foundation 0x3071a4f8 __57-NSNotificationCenter addObserver:selector:name:object:_block_invoke_0 + 12
6 CoreFoundation 0x30e95540 ___CFXNotificationPost_block_invoke_0 + 64
7 CoreFoundation 0x30e21090 _CFXNotificationPost + 1400
8 Foundation 0x3068e3e4 -NSNotificationCenter postNotificationName:object:userInfo: + 60
9 Foundation 0x3068fc14 -NSNotificationCenter postNotificationName:object: + 24
10 UIKit 0x3387926a -UIApplication _performMemoryWarning + 74
11 UIKit 0x33879364 -UIApplication _receivedMemoryNotification + 168
12 libdispatch.dylib 0x36a12252 _dispatch_source_invoke + 510
13 libdispatch.dylib 0x36a0fb1e _dispatch_queue_invoke$VARIANT$up + 42
14 libdispatch.dylib 0x36a0fe64 _dispatch_main_queue_callback_4CF$VARIANT$up + 152
15 CoreFoundation 0x30e9c2a6 __CFRunLoopRun + 1262
16 CoreFoundation 0x30e1f49e CFRunLoopRunSpecific + 294
17 CoreFoundation 0x30e1f366 CFRunLoopRunInMode + 98
18 GraphicsServices 0x33fb6432 GSEventRunModal + 130
19 UIKit 0x336f5e76 UIApplicationMain + 1074
20 MyApp 0x00004818 main (main.m:16)
21 MyApp 0x000023b4 0x1000 + 5044
I checked and there are almost no memory leaks, e.g. when testing the app for an hour the total memory leaked was around 2-3Kb caused by some string-copying libraries. So I don't believe this is caused by the application. I guess that when the phone is not restarted for some time there are applications running in the background and when using Facebook SDK the memory becomes a problem and the system tries to recover the memory from random applications, including my application.
My question is, how can I prevent this crash from happening? How should I handle unloadViewForced on a view controller to make the app more robust in low-memory conditions? And lastly, am I right that this crashlog suggests the crash occurred because the system tried to free memory and my application didn't handle it properly?
Any help greatly appreciated.
What is most likely happening is that one of the objects being referred to and probably being released by LTBaseViewController viewDidUnload method is being doubly released. In fact, since the backtrace indicates that the crash is happing on line 145 of LTBaseViewController.m you should be able to quickly see which object is the culprit.
An iPhone app is crashing on the device but not on the simulator. So I'm trying to learn how to interpret the crash log. I read lots of forum posts saying that the symbolicated crash log shows a back trace that gives the method and line number of the calls leading to the crash but I don't see anything useful. Maybe I'm not looking at the symbolicated crash log. Here is the beginning of what I see:
Incident Identifier: 432A8974-1661-409F-B5A6-970148550A46
CrashReporter Key: db93147c0a70a5f4c60dc92f826e72d5a74477c8
Hardware Model: iPhone3,3
Process: Darken [57959]
Path: /var/mobile/Applications/CB27C10F-CD3B-4148-8321-2C251888B27B/Darken.app/Darken
Identifier: Darken
Version: ??? (???)
Code Type: ARM (Native)
Parent Process: launchd [1]
Date/Time: 2012-02-25 10:43:47.753 -0500
OS Version: iPhone OS 4.2.10 (8E600)
Report Version: 104
Exception Type: EXC_BAD_ACCESS (SIGBUS)
Exception Codes: KERN_PROTECTION_FAILURE at 0x00000008
Crashed Thread: 0
Thread 0 Crashed:
0 libobjc.A.dylib 0x32716464 objc_msgSend + 16
1 UIKit 0x3245e6fe -[UIScrollView(UIScrollViewInternal) _scrollViewAnimationEnded] + 90
2 CoreFoundation 0x32071bb8 -[NSObject(NSObject) performSelector:withObject:] + 16
3 UIKit 0x3245e5b8 -[UIAnimator stopAnimation:] + 276
4 UIKit 0x323efbf2 -[UIAnimator(Static) _advance:] + 214
5 UIKit 0x323efb0e LCDHeartbeatCallback + 10
6 GraphicsServices 0x35474362 HeartbeatVBLCallback + 86
7 IOMobileFramebuffer 0x34739bf4 IOMobileFramebufferVsyncNotifyFunc + 68
8 IOKit 0x348e5e64 IODispatchCalloutFromCFMessage + 192
9 CoreFoundation 0x32070be0 __CFMachPortPerform + 204
10 CoreFoundation 0x320686f8 __CFRUNLOOP_IS_CALLING_OUT_TO_A_SOURCE1_PERFORM_FUNCTION__ + 20
11 CoreFoundation 0x320686bc __CFRunLoopDoSource1 + 160
12 CoreFoundation 0x3205af76 __CFRunLoopRun + 514
13 CoreFoundation 0x3205ac80 CFRunLoopRunSpecific + 224
14 CoreFoundation 0x3205ab88 CFRunLoopRunInMode + 52
15 GraphicsServices 0x354724a4 GSEventRunModal + 108
16 GraphicsServices 0x35472550 GSEventRun + 56
17 UIKit 0x323c7d1a -[UIApplication _run] + 406
18 UIKit 0x323c5884 UIApplicationMain + 664
19 Darken 0x000029d6 0x1000 + 6614
20 Darken 0x00002998 0x1000 + 6552
... Threads other than 0 listed here
Is anything here useful for finding out which line of my code led to the crash? Darken is the name of the application -- I already knew that. The only method name I recognize is UIApplicationMain but the crash didn't happen when the app was first launched -- I was running it about a minute and doing dozens of functions before the crash.
You may want to try and set NSZombieEnabled to YES in your project and let it crash with your device running in debug. It should stop at the line of code causing your crash. Your error looks like it is an EXC_BAD_ACCESS which usually means you were trying to access some deallocated memory.
You won't get a line number from a crash dump (unless you compiled your app with -g and run in GDB, but I doubt it since you don't seem to know what these are at all).
You ARE looking at the symbolicated crash dump: you DO have the names of the functions in the call stack. The crash occurs in the last called (topmost) function, which is objc_msgSend. That means you're not properly balancing your alloc/retain/copy methods with autorelease/release, and the messenger function tries to access already freed/corrupted/nonexistent memory, hence the crash (EXC_BAD_ACCESS is similar to a segfault, in fact you'll get one of these when you make such a mistake).
So my advice is, triple check your code for method calls modifying the reference count.