Update notification text when import is paused due to gameplay

[peppy]

Addresses #30388.

osu.2024-10-23.at.08.50.25.mp4

[tovc]

more of a point regarding design, but the spinning circle doesn't change, which might, at a glance, imply that the import isn't paused. something like a yellow pause icon might be more appropriate?

actually, for that matter, there is already the yellow spinning circle design, might as well just reuse that

[smoogipoo]

It's a little bit scary that this is the first entrypoint to the async context (there's no await prior to this line). For example if Import is ever called from the update/bdl thread, then this is going to potentially deadlock the game.

I suggest either await Task.Yield() before the pause, or wrapping the entire thing in a Task.Run().

Or.... if it can, use await Task.Delay() instead of Thread.Sleep(). Doesn't look like it can though.

[peppy]

It's a fair point. I started by duplicating the method and making an async version, but then ended up refactoring things a bit to not require this.

Of note, the Parallel.ForEachAsync change was required because without it, the previous Select loop would seem to cause all imports to get past the pause check and run up to the Import part which would mean they never actually pause.

The whole import async logic is a bit of a mess but as far as I can tell it's working as expected.

[smoogipoo]

Test failures need resolution here.

[smoogipoo]

Are you sure about this? Won't this be logged (to the user also) as an unobserved exception? I mean... see the try-catch below which even swallows the OperationCanceledException

[smoogipoo]

The pauseIfNecessaryAsync method also throws and will likely do the same thing too outside of the ForEachAsync here (maybe a reproduction for that is enter gameplay, start import from stable, then cancel the notification?).

[peppy]

It definitely doesn't come up as an unobserved else I would have noticed.

You can test using:

diff --git a/osu.Game/Database/RealmArchiveModelImporter.cs b/osu.Game/Database/RealmArchiveModelImporter.cs
index c4eb93b754..addba3f338 100644
--- a/osu.Game/Database/RealmArchiveModelImporter.cs
+++ b/osu.Game/Database/RealmArchiveModelImporter.cs
@@ -119,6 +119,9 @@ public async Task<IEnumerable<Live<TModel>>> Import(ProgressNotification notific
 
             await Parallel.ForEachAsync(tasks, notification.CancellationToken, async (task, cancellation) =>
             {
+                while (!notification.CancellationToken.IsCancellationRequested)
+                    Thread.Sleep(1000);
+
                 cancellation.ThrowIfCancellationRequested();
 
                 try
@@ -139,6 +142,7 @@ await Parallel.ForEachAsync(tasks, notification.CancellationToken, async (task,
                 }
                 catch (OperationCanceledException)
                 {
+                    throw;
                 }
                 catch (Exception e)
                 {
@@ -531,7 +535,8 @@ protected virtual void PostImport(TModel model, Realm realm, ImportParameters pa
         /// <param name="model">The new model proposed for import.</param>
         /// <param name="realm">The current realm context.</param>
         /// <returns>An existing model which matches the criteria to skip importing, else null.</returns>
-        protected TModel? CheckForExisting(TModel model, Realm realm) => string.IsNullOrEmpty(model.Hash) ? null : realm.All<TModel>().OrderBy(b => b.DeletePending).FirstOrDefault(b => b.Hash == model.Hash);
+        protected TModel? CheckForExisting(TModel model, Realm realm) =>
+            string.IsNullOrEmpty(model.Hash) ? null : realm.All<TModel>().OrderBy(b => b.DeletePending).FirstOrDefault(b => b.Hash == model.Hash);
 
         /// <summary>
         /// Whether import can be skipped after finding an existing import early in the process.

Debugging through I see a TaskCanceledException but it gracefully handles somewhere (debugger doesn't let me see). But maybe there's a scenario where that's not the case? I couldn't find one.

Using ThrowIfCancellationRequested is recommended way to handle this, so I'd be inclined to add a catch somewhere if it's required, rather than change the calls.