Unity crashes (signal 11) - how to diagnose or fix

[liortal]

We are using an automated crash reporting service, and we’re seeing a very high number of crashes which do not occur inside our code.

Here are a few examples:

0java.lang.Error: signal 11 (SIGSEGV), code 1 (SEGV_MAPERR), fault addr 00000024
1Build fingerprint: ‘samsung/m3xx/m3:4.4.4/KTU84P/I9305XXUFPB1:user/release-keys’
2Revision: ‘2’
3pid: 8452, tid: 11239, name: UnityPreload >>> com.moonactive.coinmaster <<<
4 r0 00000000 r1 657e0048 r2 000000cc r3 00000000
5 r4 653af7e0 r5 000000cc r6 00000000 r7 657e0054
6 r8 00000000 r9 653af7e0 sl 00000002 fp 64fb1c40
7 ip 00000000 sp 64fb1bd8 lr 6339413c pc 63394488 cpsr 42504655
8
9 at libunity.0040c488(Native Method)
10 at libunity.0040c138(Native Method)
11 at libunity.004075a0(Native Method)
12 at libunity.003e3ac4(Native Method)
13 at libunity.003e2140(Native Method)
14 at libunity.003e2090(Native Method)
15 at libunity.0043bf5c(Native Method)
16 at libc.__thread_entry(__thread_entry:72)
17 at libc.pthread_create(pthread_create:240)

Here’s the device breakdown:
samsung GT-N7100 27%
samsung GT-I9060I 24%
samsung SM-T116NU 11%
samsung SM-G7102 9%
samsung GT-I9301I 9%
samsung SM-G360H 9%
samsung SGH-M919 8%
samsung SM-T560 8%

And OS breakdown:
4.4.2 58%
4.4.4 42%
4.4.3 1%
4.4 1%
5.0.1 1%
4.4.2_Eyal.H_Cell 1%

We have other reports of similar issues (they are classified as different crashes because the stack trace is slightly different), but overall, the top of the stack is similar for all (Unity library):

9 at libunity.0040c488(Native Method)
10 at libunity.0040c138(Native Method)
11 at libunity.004075a0(Native Method)
12 at libunity.003e3ac4(Native Method)
13 at libunity.003e2140(Native Method)
14 at libunity.003e2090(Native Method)
15 at libunity.0043bf5c(Native Method)

The game was released with Unity 5.3.6, our next release will use 5.3.6p8.

1. Is this a known Unity bug? we have no repro scenario so we cannot send a project, etc.
2. This occurs on the scale of hundreds of thousands of times (around 150k crashes in the past month).
3. Is there any way we can convert the addresses from the stack trace into actual useful information ? (function names) ?

Any help will be appreciated !! CC: @Yury-Habets

[liortal]

BUMP

[JC_SummitTech]

Not to be a jerk and tell you to RTFM, but I have found that page from the manual. It didn’t really help me, objdump didn’t give me anything that makes sense, but it might help you.

https://docs.unity3d.com/Manual/MobileCrashes.html

Best of luck, there is nothing I hate more than debugging on mobile

[liortal]

I already got a nicer stack trace by using tools from the NDK:

PersistentManager::PostLoadStreamNameSpace(StreamNameSpace&, int)
PersistentManager::GetStreamNameSpaceInternal(int)
PersistentManager::GetLoadErrorMessageFromPath(std::string const&)
LoadSceneOperation::Perform()
PreloadManager::Run()
PreloadManager::Run(void*)
Thread::RunThreadWrapper(void*)

This is a known bug that is said to be worked on (not too comforting as its open for months now). Looks likeit is related to resource or scene loafing.

Here’s a link to the currently open bug: Unity Issue Tracker - [Android] Unity crashes loading a scene

[Yury-Habets]

@liortal yeah we’ve seen your bug report. Something crashes in serialization/caching system, device specific. Out Best People (c) are on it.

[liortal]

Yes you already mentioned that @superpig is on top of it. waiting for more info… this is really killing us (over 100k crashes so far).

[InspiredSquare]

1% user affected due to this crash after updating Unity to 5.5.