Status Update
Comments
su...@google.com <su...@google.com> #2
Can you include a sample project that crashes in release mode?
ru...@gmail.com <ru...@gmail.com> #3
For us this is only happening in places where LifecycleEventEffect
is used. Example stacktrace:
Fatal Exception: java.lang.IllegalStateException
CompositionLocal LocalLifecycleOwner not present
androidx.lifecycle.compose.LocalLifecycleOwnerKt$LocalLifecycleOwner$1$1.invoke (LocalLifecycleOwner.android.kt:63)
androidx.lifecycle.compose.LocalLifecycleOwnerKt$LocalLifecycleOwner$1$1.invoke (LocalLifecycleOwner.android.kt:62)
kotlin.SynchronizedLazyImpl.getValue (LazyJVM.kt:74)
androidx.compose.runtime.LazyValueHolder.getCurrent (LazyValueHolder.java:29)
androidx.compose.runtime.LazyValueHolder.getValue (LazyValueHolder.java:31)
androidx.compose.runtime.CompositionLocalMapKt.read (CompositionLocalMap.kt:90)
androidx.compose.runtime.ComposerImpl.consume (Composer.kt:2135)
androidx.lifecycle.compose.LifecycleEffectKt.LifecycleEventEffect (LifecycleEffect.kt:748)
com.freeletics.feature.profile.ProfileUiKt$ProfileUi$1.invoke (ProfileUi.kt:62)
This is also on Lifecycle 2.8.2 with Compose 1.6.
an...@google.com <an...@google.com> #4
Here is a sample project
- Checkout repository
- Build release apk with ./gradlew :app:assembleRelease -PenableReleaseSigning=true
- App starts without crash
- Comment or remove line 14 in
https://github.com/nilsjr/Koncept/blob/develop/app/proguard-rules.pro - Build release apk with ./gradlew :app:assembleRelease -PenableReleaseSigning=true
- App crashing on start
FATAL EXCEPTION: main (Ask Gemini)
Process: de.nilsdruyen.koncept, PID: 5880
java.lang.IllegalStateException: CompositionLocal LocalLifecycleOwner not present
...
ru...@gmail.com <ru...@gmail.com> #5
Thank you for sharing the sample project. I can confirm that I was able to reproduce the issue, and that
ra...@google.com <ra...@google.com> #6
It appears that the custom ProGuard rule was not working as intended across all projects.
However, using public static *** getLocalLifecycleOwner();
(with the wildcard type ***
) seems to work consistently in the sample project and our internal experiments:
-if public class androidx.compose.ui.platform.AndroidCompositionLocals_androidKt {
public static *** getLocalLifecycleOwner();
}
-keep public class androidx.compose.ui.platform.AndroidCompositionLocals_androidKt {
public static *** getLocalLifecycleOwner();
}
We are investigating further, and we will be working on a fix for the issue.
ru...@gmail.com <ru...@gmail.com> #7
Branch: androidx-main
commit 79f5644cb937d950318c3c5ef2aca70ab1413119
Author: Marcello Galhardo <mgalhardo@google.com>
Date: Fri Jun 14 16:34:11 2024
Fix Lifecycle 2.8 custom ProGuard rule
* The custom ProGuard rule was not working as intended across all projects.
* Replacing by `public static *** getLocalLifecycleOwner();` (with the wildcard type `***`) seems to work consistently in all projects.
Fixes:
Test: manual
Change-Id: I4cfdecc0bbfc0be02d66efcee2c63bb5b025dca2
M lifecycle/lifecycle-runtime-compose/proguard-rules.pro
Description
Version used: 2.1.0
Devices/Android versions reproduced on: 6.0
We're getting this crash on multiple devices:
Fatal Exception: java.lang.NullPointerException: Attempt to invoke virtual method 'boolean android.util.ArrayMap.containsKey(java.lang.Object)' on a null object reference
at android.os.BaseBundle.containsKey + 277(BaseBundle.java:277)
at androidx.work.impl.background.systemjob.SystemJobScheduler.getPendingJobIds + 331(SystemJobScheduler.java:331)
at androidx.work.impl.background.systemjob.SystemJobScheduler.schedule + 138(SystemJobScheduler.java:138)
at androidx.work.impl.Schedulers.schedule + 93(Schedulers.java:93)
at androidx.work.impl.WorkManagerImpl.rescheduleEligibleWork + 587(WorkManagerImpl.java:587)
at androidx.work.impl.utils.ForceStopRunnable.run + 82(ForceStopRunnable.java:82)
at androidx.work.impl.utils.SerialExecutor$Task.run + 75(SerialExecutor.java:75)
at java.util.concurrent.ThreadPoolExecutor.runWorker + 1113(ThreadPoolExecutor.java:1113)
at java.util.concurrent.ThreadPoolExecutor$Worker.run + 588(ThreadPoolExecutor.java:588)
at java.lang.Thread.run + 833(Thread.java:833)
For some reason, the mMap in BaseBundle is null and getPendingJobIds doesn't catch that:
for (JobInfo jobInfo : jobs) {
PersistableBundle extras = jobInfo.getExtras();
// The map inside extras can be null, so extras.containsKey can cause a NPE
if (extras != null && extras.containsKey(EXTRA_WORK_SPEC_ID)) {
if (workSpecId.equals(extras.getString(EXTRA_WORK_SPEC_ID))) {
jobIds.add(jobInfo.getId());
}
}
}