Tracking: Reported issues Jan 2025

[nikclayton]

The Pachli Current releases in Jan 2025 are crashing for some but not all users. This is to track the reported issues, collect feedback.

Current state

• Releasing Released 2.9.1+ab958916. This contains fixes to PruneCacheWorker queries that may prevent the database from locking. Feedback / bug reports on earlier versions is not necessary.

Versions / bug reports

Versions, newest first:

• 4234, 2.9.1+353f123b - (release in progress) Fix for "NPE handling SetActiveAccountError.Unexpected" (displays error instead of crashing)
• 4233, 2.9.1+ab958916 - ? (PruneCacheWorker query fixes)
• 4232, 2.9.1+b1ed66fe (coroutine cancellation fixes, logs after updating marker) - known crashes
• 4228, 2.9.1+11480d80 - known crashes
• 4226, 2.9.1+05c68f6d - known crashes

Reported issues

App crashing at startup

Fixed with 4233, 2.9.1+ab958916

Seen in:

• 4232, 2.9.1+b1ed66fe
• 4228, 2.9.1+11480d80
• 4226, 2.9.1+05c68f6d

Crash reports from:

• kallekn
• radgryd
• keefmarshall

Updating the user's notification marker (the ID of the most recent fetched notification) crashes, and looks like this:

2025-01-24T18:02:08.203Z, priority=3, tag=NotificationFetcher, message=updating notification marker for @[redacted](mailto:redacted) to: 653334, t=null
2025-01-24T18:02:08.296Z, priority=6, tag=NotificationFetcher, message=Error while fetching notifications
android.database.sqlite.SQLiteException: cannot start a transaction within a transaction (code 1 SQLITE_ERROR[1])
at android.database.sqlite.SQLiteConnection.nativeExecute(Native Method)
at android.database.sqlite.SQLiteConnection.execute([SQLiteConnection.java:1034](http://sqliteconnection.java:1034/))
at android.database.sqlite.SQLiteSession.beginTransactionUnchecked([SQLiteSession.java:321](http://sqlitesession.java:321/))
at android.database.sqlite.SQLiteSession.beginTransaction([SQLiteSession.java:300](http://sqlitesession.java:300/))
at android.database.sqlite.SQLiteDatabase.beginTransaction([SQLiteDatabase.java:972](http://sqlitedatabase.java:972/))
at android.database.sqlite.SQLiteDatabase.beginTransactionNonExclusive([SQLiteDatabase.java:906](http://sqlitedatabase.java:906/))
at androidx.sqlite.db.framework.FrameworkSQLiteDatabase.M(SourceFile:3)
at [androidx.room.RoomDatabase.c(SourceFile:26)](http://androidx.room.RoomDatabase.c(SourceFile:26))
at app.pachli.core.database.dao.AccountDao_Impl$[40.call(SourceFile:23)](http://40.call(SourceFile:23))
at [androidx.room.CoroutinesRoom$Companion$execute$2.t(SourceFile:8)](http://androidx.room.CoroutinesRoom$Companion$execute$2.t(SourceFile:8))
at kotlin.coroutines.jvm.internal.BaseContinuationImpl.g(SourceFile:6)
at [kotlinx.coroutines.DispatchedTask.run(SourceFile:113)](http://kotlinx.coroutines.DispatchedTask.run(SourceFile:113))
at [a.f.run(SourceFile:531)](http://a.f.run(SourceFile:531))
at java.util.concurrent.ThreadPoolExecutor.runWorker([ThreadPoolExecutor.java:1145](http://threadpoolexecutor.java:1145/))
at java.util.concurrent.ThreadPoolExecutor$[Worker.run(ThreadPoolExecutor.java:644)](http://Worker.run(ThreadPoolExecutor.java:644))
at [java.lang.Thread.run(Thread.java:1012)](http://java.lang.Thread.run(Thread.java:1012)

I know it's the marker update that crashes because there should be a log message immediately after the marker is updated and it's not present, so the crash is there.

Suspect this is a symptom, not the cause -- some other query is crashing and left an open transaction, then all future transactions crash.

Working theory: The queries executed by PruneCacheWorker could trigger FK constraints and fail

NPE handling SetActiveAccountError.Unexpected

**Note**: This is now handled by https://github.com//pull/1235. The app is probably still in an unusable state if this happens, but it should show the error.

java.lang.NullPointerException: Attempt to invoke virtual method 'java.lang.String app.pachli.core.database.model.AccountEntity.c()' on a null object reference
at app.pachli.MainActivity.bindUiResult(MainActivity.kt:665)
at app.pachli.MainActivity.access$bindUiResult(MainActivity.kt:189)
at app.pachli.MainActivity$onCreate$9$1$1.emit(MainActivity.kt:458)
at app.pachli.MainActivity$onCreate$9$1$1.emit(MainActivity.kt:458)
at kotlinx.coroutines.flow.FlowKt__ChannelsKt.emitAllImpl$FlowKt__ChannelsKt(Channels.kt:33)
at kotlinx.coroutines.flow.FlowKt__ChannelsKt.access$emitAllImpl$FlowKt__ChannelsKt(Channels.kt:1)
at kotlinx.coroutines.flow.FlowKt__ChannelsKt$emitAllImpl$1.invokeSuspend(Channels.kt)
at kotlin.coroutines.jvm.internal.BaseContinuationImpl.resumeWith(ContinuationImpl.kt:33)
at kotlinx.coroutines.internal.DispatchedContinuation.resumeUndispatchedWith$kotlinx_coroutines_core(DispatchedContinuation.kt:237)
at kotlinx.coroutines.DispatchedTaskKt.resume(DispatchedTask.kt:164)
at kotlinx.coroutines.CancellableContinuationImpl.cancelCompletedResult$kotlinx_coroutines_core(CancellableContinuationImpl.kt:177)
at kotlinx.coroutines.DispatchedTaskKt.resumeUnconfined(DispatchedTask.kt:177)
at kotlinx.coroutines.DispatchedTaskKt.dispatch(DispatchedTask.kt:149)
at kotlinx.coroutines.CancellableContinuationImpl.dispatchResume(CancellableContinuationImpl.kt:470)
at kotlinx.coroutines.CancellableContinuationImpl.completeResume(CancellableContinuationImpl.kt:591)
at kotlinx.coroutines.channels.BufferedChannelKt.tryResume0(BufferedChannel.kt:2957)
at kotlinx.coroutines.channels.BufferedChannelKt.access$tryResume0(BufferedChannel.kt:1)
at kotlinx.coroutines.channels.BufferedChannel$BufferedChannelIterator.tryResumeHasNext(BufferedChannel.kt:1719)
at kotlinx.coroutines.channels.BufferedChannel.tryResumeReceiver(BufferedChannel.kt:662)
at kotlinx.coroutines.channels.BufferedChannel.updateCellSend(BufferedChannel.kt:478)
at kotlinx.coroutines.channels.BufferedChannel.access$updateCellSend(BufferedChannel.kt:33)
at kotlinx.coroutines.channels.BufferedChannel.sendImpl$default(BufferedChannel.kt:301)
at kotlinx.coroutines.channels.BufferedChannel.send$suspendImpl(BufferedChannel.kt:110)
at kotlinx.coroutines.channels.BufferedChannel.send(BufferedChannel.kt)
at app.pachli.MainViewModel.onUiAction(MainViewModel.kt:213)
at app.pachli.MainViewModel.access$onUiAction(MainViewModel.kt:144)
at app.pachli.MainViewModel$onUiAction$1.invokeSuspend(MainViewModel.kt)
at kotlin.coroutines.jvm.internal.BaseContinuationImpl.resumeWith(ContinuationImpl.kt:33)
at kotlinx.coroutines.DispatchedTask.run(DispatchedTask.kt:99)
at android.os.Handler.handleCallback(Handler.java:991)
at android.os.Handler.dispatchMessage(Handler.java:102)
at android.os.Looper.loopOnce(Looper.java:232)
at android.os.Looper.loop(Looper.java:317)
at android.app.ActivityThread.main(ActivityThread.java:8787)
at java.lang.reflect.Method.invoke(Native Method)
at com.android.internal.os.RuntimeInit$MethodAndArgsCaller.run(RuntimeInit.java:591)
at com.android.internal.os.ZygoteInit.main(ZygoteInit.java:871)

Code is:

                            builder.setTitle(uiError.cause.wantedAccount.fullName)

wantedAccount is null here even though it shouldn't be. Fix is to add an error case for SQLiteException and display an error instead of crashing.

TImeline stops loading posts

Seen in:

• 4232, 2.9.1+b1ed66fe

My timeline stopped at 40 minutes today (apart from my own posts), but I noticed when I clicked into the last post and then went back, it loaded some more posts below that. So whatever triggers the automatic "load more" when scrolling seems to be slightly broken but not entirely broken.

Timeline jumps when boosting

Seen in:

• ?

So it seems when I hit "load newest posts" it's erasing anything older than 1 hour from my timeline (apart from posts that are also in my notifications).. at least that's the symptom. If I leave it without hitting that menu it retains the timeline back to the last point it had, but doesn't load anything older.

Also, boosting a post makes my timeline jump back by quite a few minutes.. possibly by the number of new posts that happened since I last interacted?

Other notes

---

[nikclayton]

Keith, I need some extra information about the "Timeline stops loading" issue you reported.

Here's what's supposed to happen:

On your device the database is treated as a local cache that can be discarded at any time. When that happens fresh data is fetched from your server and used to re-populate the cache.

When you "refresh" your home timeline (either the "refresh" menu, or by pulling down on the list of posts):

1. The cached statuses in that timeline are deleted from your device (statuses referenced by notifications remain)
2. Network requests are made to your server to fetch the posts that were on-screen when you refreshed, as well as posts immediately before and after them (so you can start scrolling without immediately triggering another network fetch)
3. When the first set of posts are loaded from your server your reading position is adjusted so the post you most recently saw is on-screen
4. If any newer posts are also loaded then the list scrolls down about 30 pixels to "peek" as a hint that new content was loaded.

When you "Load from newest" from the menu much the same thing happens, except instead of trying to load the posts around where you were reading, your server is asked for the newest posts. When they load you're scrolled to the top of the newest posts, ignoring whatever your previous reading position was.

In both cases, when you scroll, and you get close to the end of the posts that have been loaded so far, your server is contacted to get newer (or older, if you're scrolling down) posts.

If the server responds with new posts they're added to the list. But if the server responds with no new posts then Pachli will stop trying to fetch new data until you explicitly refresh (either from the menu, or by swiping down on the list).

[keefmarshall]

So I can't see any specific questions to answer above, but i can think of a scenario that explains two of the issues I've seen. I haven't looked at the code so this is speculation based on behaviour, and what you describe above.

Assume I have just opened the app, and clicked "load newest posts" to jump to the top. Everything is cleared except the latest posts.

I scroll downwards, going back in time.

Two symptoms can occur (these definitely do not happen in the mainline Pachli, only Current):

1. I reach a point at which there are no more posts to scroll down, except a few stray ones I posted myself, or were in my notifications (DMs etc)
2. If I boost a post, the timeline suddenly jumps up to a newer point above.

Both of these could be explained by the same thing: if Pachli has somehow lost track of where I am and thinks I'm higher up - e.g. 20-30 posts higher up in the list.

• it doesn't load any more posts because it hasn't realised I've got to the bottom of the viewable list.
• it jumps back up to where it thinks I should be after having boosted a post, rather than where I actually was.

Speculating even more wildly - could this be because more posts have since come in at the top of the stream, but not been accounted for in the position that I'm in? e.g. let's say I scrolled down 100 posts, but in the meantime another 30 posts came in. Pachli thinks I'm on post 100, but it's forgotten to add the 30 new ones, so actually it should think I'm at post 130..? Dunno - that feels like a stretch. I don't have time to dive into the code to help track this down further I'm afraid.

[keefmarshall]

Version 2.9.1+353f123b hasn't crashed yet but is still only loading the last hour or so of my timeline, after which it's just posts that reference my username directly.

[nikclayton]

Version 2.9.1+353f123b hasn't crashed yet but is still only loading the last hour or so of my timeline, after which it's just posts that reference my username directly.

@keefmarshall and if you check your timeline on the website and scroll down past that point, posts show up that don't appear in Pachli?

If so, please can you:

1. Restart Pachli
2. Scroll to the top of your home time line as normal
3. Use the "Load newest posts" menu option. This will clear the cache and insert just the most recent items in your timeline
4. Scroll down until you start to see the problem
5. Use "send error report" from the top right menu
6. In the error report note that it's not a crash, it's this loading issue

That should let me see the network requests and response codes (not the actual content) so I can figure out what's happening.

[keefmarshall]

Yes, not only if I check on the website, but also in the mainline version of Pachli, there are many more posts in my timeline. This behaviour is specific to Current.

I'll try to send a report as asked.