Incorrect line number in Sentry event when running app on release mode.

[rodolfoBee]

Integration

sentry-android

Build System

Other

AGP Version

NA

Proguard

Enabled

Version

6.23.0

Steps to Reproduce

1. Create a simple Android app
2. Run the following code in debug mode to trigger a ArrayIndexOutOfBoundsException
   
3. Note the line number in Sentry is correct (525)
4. Run the same code again but now in release mode

Expected Result

The line number in the crash report from release mode is the same is in the debug mode. This is the screenshot from the debug mode:

Actual Result

The line number in the crash report is incorrect (line 3)

Proguard mapping is uploaded to the project and applied to the affected event:
Event:

"uuid": "123e8712-e89b-12d3-a456-426614174910",
"type": "proguard"

The result using retrace also shows line 3 as the affected line:

java -jar retrace.jar mapping.txt stacktrace.txt
java.lang.ArrayIndexOutOfBoundsException: length=3; index=5
    at com.sp.[redacted].views.fragment.ConsentFragment$onclickListener$1.invoke(ConsentFragment.java:3)
    at com.sp.[redacted].views.fragment.ConsentFragment$onclickListener$1.invoke(ConsentFragment.java:1)
    at i.o$a.invoke(o.java:30)
    at com.sp.[redacted].utilities.SPUtilsKt$sam$androidx_lifecycle_Observer$0.onClick(SPUtilsKt.java:5)
    at android.view.View.performClick(View.java:7441)
    at android.view.View.performClickInternal(View.java:7418)
    at android.view.View.access$3700(View.java:835)
    at android.view.View$PerformClick.run(View.java:28676)
    at android.os.Handler.handleCallback(Handler.java:938)
    at android.os.Handler.dispatchMessage(Handler.java:99)
    at android.os.Looper.loopOnce(Looper.java:201)
    at android.os.Looper.loop(Looper.java:288)
    at android.app.ActivityThread.main(ActivityThread.java:7842)
    at java.lang.reflect.Method.invoke(Method.java)
    at com.android.internal.os.RuntimeInit$MethodAndArgsCaller.run(RuntimeInit.java:548)
    at com.android.internal.os.[redacted].main([redacted].java:1003)

Links to events and project are available in the internal ticket.

[rodolfoBee]

Internal ticket with information: https://getsentry.atlassian.net/browse/FEEDBACK-2393

[romtsn]

@rodolfoBee if retrace yields the same result I doubt we can do much more about it from our side. This is how lambdas are optimized by r8, and if they themselves cannot recover the original stacktrace I'm not sure we can.

[rodolfoBee]

hi @romtsn thank you for the info.

I got confirmation it also happens on v7.20.0. The raw stack trace also points to the same location (it does not get resolved by the proguard mapping):

java.lang.ArrayIndexOutOfBoundsException: length=3; index=5
    at com.sp.[redacted].views.fragment.ConsentFragment$onclickListener$1.invoke(ConsentFragment.kt:3)
    at com.sp.[redacted].views.fragment.ConsentFragment$onclickListener$1.invoke(ConsentFragment.kt:1)

do you believe is caused by how the code is obfuscated?

[romtsn]

@rodolfoBee you can disregard the SDK version, the problem is usually either in our proguard processing or mapping upload logic, but in this particular case this is caused by how the code is optimized, yes. If you look at the mapping file, it's not possible to trace back the com.sp.shuftipro_sdk.views.fragment.ConsentFragment$onclickListener$1.invoke stackframe to the original line - there's no such mapping exists.

[romtsn]

@rodolfoBee perhaps the customer could report this to https://issuetracker.google.com/issues and provide a reproducible sample of their code, so google could shed some light on this.