StackOverflowError in WorkManager on KitKat [129091233]

Fixed

Bug

Status Update

No update yet.

Description

ka...@gmail.com

created issue #1

Mar 22, 2019 02:16PM

Component used: android.arch.work:work-runtime
Version used: 1.0.0-rc01
Devices/Android versions reproduced on: We've seen it on 4.4.4 and 4.4.2

We haven't been able to reproduce this ourselves, but we're seeing it show up in our tracking. There's a StackOverflowError caused somewhere inside WorkManager. Here's the stack trace we get:

Fatal Exception: java.lang.StackOverflowError
at java.util.HashMap.secondaryHash(HashMap.java:350)
at java.util.LinkedHashMap.get(LinkedHashMap.java:251)
at android.util.LruCache.get(LruCache.java:118)
at android.database.sqlite.SQLiteConnection.acquirePreparedStatement(SQLiteConnection.java:1100)
at android.database.sqlite.SQLiteConnection.prepare(SQLiteConnection.java:689)
at android.database.sqlite.SQLiteSession.prepare(SQLiteSession.java:588)
at android.database.sqlite.SQLiteProgram.(SQLiteProgram.java:58)
at android.database.sqlite.SQLiteQuery.(SQLiteQuery.java:37)
at android.database.sqlite.SQLiteDirectCursorDriver.query(SQLiteDirectCursorDriver.java:44)
at android.database.sqlite.SQLiteDatabase.rawQueryWithFactory(SQLiteDatabase.java:1433)
at android.database.sqlite.SQLiteDatabase.rawQueryWithFactory(SQLiteDatabase.java:1408)
at android.arch.persistence.db.framework.FrameworkSQLiteDatabase.query(myapp:161)
at android.arch.persistence.room.RoomDatabase.query(myapp:233)
at androidx.work.impl.model.DependencyDao_Impl.getDependentWorkIds(myapp:117)
at androidx.work.impl.WorkerWrapper.recursivelyFailWorkAndDependents(myapp:503)
<snip ~50 recursive frames>
at androidx.work.impl.WorkerWrapper.recursivelyFailWorkAndDependents(myapp:503)
at androidx.work.impl.WorkerWrapper.setFailedAndResolve(myapp:488)
at androidx.work.impl.WorkerWrapper.runWorker(myapp:242)
at androidx.work.impl.WorkerWrapper.run(myapp:127)
at java.util.concurrent.ThreadPoolExecutor.runWorker(ThreadPoolExecutor.java:1112)
at java.util.concurrent.ThreadPoolExecutor$Worker.run(ThreadPoolExecutor.java:587)
at java.lang.Thread.run(Thread.java:841)

#0. Crashed: WorkManager-WorkManagerTaskExecutor-thread-0
at java.util.HashMap.secondaryHash(HashMap.java:350)
at java.util.LinkedHashMap.get(LinkedHashMap.java:251)
at android.util.LruCache.get(LruCache.java:118)
at android.database.sqlite.SQLiteConnection.acquirePreparedStatement(SQLiteConnection.java:1100)
at android.database.sqlite.SQLiteConnection.prepare(SQLiteConnection.java:689)
at android.database.sqlite.SQLiteSession.prepare(SQLiteSession.java:588)
at android.database.sqlite.SQLiteProgram.(SQLiteProgram.java:58)
at android.database.sqlite.SQLiteQuery.(SQLiteQuery.java:37)
at android.database.sqlite.SQLiteDirectCursorDriver.query(SQLiteDirectCursorDriver.java:44)
at android.database.sqlite.SQLiteDatabase.rawQueryWithFactory(SQLiteDatabase.java:1433)
at android.database.sqlite.SQLiteDatabase.rawQueryWithFactory(SQLiteDatabase.java:1408)
at android.arch.persistence.db.framework.FrameworkSQLiteDatabase.query(myapp:161)
at android.arch.persistence.room.RoomDatabase.query(myapp:233)

Comments

su...@google.com <su...@google.com> #2Mar 22, 2019 06:32PM

Seems like it happens on API 21-22 too.

ka...@gmail.com <ka...@gmail.com> #3Mar 22, 2019 07:31PM

This is expected. You are calling: enqueueUniquePeriodicWork() with an ExistingPeriodicWorkPolicy.REPLACE. This will cancel running work, and kick off new instances of all Workers.

https://github.com/rubensousa/WorkManagerKitkatBug/blob/master/app/src/main/java/com/github/rubensousa/workkitkatbug/MainActivity.kt#L24

su...@google.com <su...@google.com> #4Mar 22, 2019 08:25PM

I just ran the sample again and now the behaviour is different for some weird reason. I forgot to mention that I only used REPLACE to test this properly. Because it also happens with KEEP.

ka...@gmail.com <ka...@gmail.com> #5Mar 22, 2019 08:40PM

And shouldn't the behaviour be consistent across all Android versions?

[Deleted User] <[Deleted User]> #6Mar 24, 2019 01:23PM

We don't see any issues with KEEP. If you have a way for us to reproduce any problems, please send us the appropriate instructions.

su...@google.com <su...@google.com> #7Mar 28, 2019 08:09PM

Marked as fixed, reassigned to su...@google.com.

Ok, I found something odd. Seems like setRequiredNetworkType(NetworkType.UNMETERED) causes this for some reason. I'm going to update the sample so you can review this again. Please, reopen this since it's not intended behavior.

ka...@gmail.com <ka...@gmail.com> #8Mar 28, 2019 08:38PM

This happens on a fresh install with KEEP instead of REPLACE if the constraints have setRequiredNetworkType(NetworkType.UNMETERED). I've updated the sample

su...@google.com <su...@google.com> #9Mar 28, 2019 09:11PM

"And shouldn't the behaviour be consistent across all Android versions?"
The behavior is actually consistent. Once your worker starts running, if we get a call to enqueueUniquePeriodicWork a new work request and with ExistingPeriodicWorkPolicy.REPLACE we will stop existing work.
If your Worker was not running, then there is nothing to cancel. This is just a matter of timing.

ka...@gmail.com <ka...@gmail.com> #10Mar 29, 2019 11:44AM

I'm not arguing against the expected behavior of ExistingPeriodicWorkPolicy.REPLACE, since that's working as intended. But this is happening only when setRequiredNetworkType(NetworkType.UNMETERED) is used while having ExistingPeriodicWorkPolicy.KEEP.