[Android] SIGSEGV (out of memory) crashes dont get grouped up properly

[tayl0r]

Description

I am looking at our Bugsnag dashboard and there are thousands of issues, all with 1 user 1 event, for SIGSEGV errors. They should be grouped together but they aren't. It breaks the way you are supposed to use Bugsnag with this behavior.

Issue

SIGSEGV errors should be grouped together as 1 issue.

Environment

• bugsnag-unity version: 4.2.0
• Unity version: 2017.4.14f1
• Operating system name and version: building with OSX
• Target platform names and versions: Android 8.0 on HUAWEI ANE-LX2 (P20 Lite)
• Initializing bugsnag via the Unity UI or in code? UI

Example code snippet

Reproduce a memory crash like in this ticket: https://github.com/bugsnag/bugsnag-unity/issues/134 Each time you do, it creates a new issue rather than grouping them up. I am assuming it is because the "callstack" is actually just a single memory address which is always different.

Screenshots:

[tayl0r]

I figured out that you can use Group by error context in Projects -> Error Processing and then it will start to group these up.

I still think it should try to group these up automatically but this is an acceptable solution as well.

[abigailbramble]

Hi @tayl0r - glad you managed to work around this! We are actually looking to discard these particular errors.

We have seen similar reports from other customers of ungroupable SIGSEGV errors with a stacktrace of a single unresolvable address (like yours 0xd89797c0 above). It appears to be caused by some bug in Unity which is only triggered when the app is force closed while certain kinds of Asset Bundles are loaded into memory. So users would force close the game and then you'd get this SIGSEGV error.

We have managed to reproduce this by loading an invalid asset then force-closing the app, which triggers that bug in Unity which causes a SIGSEGV, and the previous handler does not terminate the process.

We are currently working on a change which will discard these errors and therefore they will not be shown in your dashboard. Note this is only for this specific case of SIGSEGV.

[technicalDeath]

@phillipsam thanks for providing more insight into this issue we are also seeing. Would it be possible to share your reproduction of the issue? I am curious what this "invalid" asset looks like. Ideally we could address this at the root.

[abigailbramble]

Hi @technicalDeath - I have managed to reproduce this by creating a script which takes a GameObject and then never assigning the GameObject in the editor. I receive the NullReferenceException in Bugsnag and then when I close the application I also receive the SIGSEGV with a single unresolvable address.