Roll Flutter master to 8b22f67c85114

[stuartmorgan-g]

Rolls the Flutter master pin to 8b22f67

flutter/flutter@d8baa77...8b22f67

Makes the following changes:

• Updates the Pigeon example app's iOS channel registration for the breaking change caused by UISceneDelegate automigration in the flutter tool
• Updates the video_player_avfoundation XCTests to use mock registrars instead of real registrars created by the real registry. This was tech debt, since unit tests shouldn't be coupled to real app state, and it broke because of the registry no longer being available when the unit tests run (due to the same migration).
• Updates video_player_avfoundation to take a view provider, so that tests can inject a fake provider, decoupling unit test behavior from the actual app's root view controller, fixing a potential race after the UISceneDelegate migration.
• Skips quick_actions_ios XCUITests until the plugin can be fixed. Tracked in [iOS] Quick Actions Not Triggered When App is in Background flutter#169928

[stuartmorgan-g]

This is a subset of #9402 to narrow down the range where the various failures there started.

[stuartmorgan-g]

It looks like this narrow range has the iOS quick actions failure, as expected, and also a Pigeon integration test failure that's also in the larger-range roll PR:

00:21 +0 -1: loading /Volumes/Work/s/w/ir/x/w/packages/packages/pigeon/example/app/integration_test/example_app_test.dart [E] Failed to load "/Volumes/Work/s/w/ir/x/w/packages/packages/pigeon/example/app/integration_test/example_app_test.dart": WebSocketChannelException: SocketException: Connection refused (OS Error: Connection refused, errno = 61), address = 127.0.0.1, port = 55185

I'm not sure what would be causing that.

[stuartmorgan-g]

From local testing, flutter/flutter#169365 broke the Pigeon example, although the error message is slightly different if I check out that hash (maybe the follow-up to do a runtime warning changed the behavior slightly). That suggests that the automigration is not always safe. /cc @gaaclarke

[stuartmorgan-g]

Oh, I see. The example is doing

letcontroller= window?.rootViewController as!FlutterViewController

so it's crashing. This is flagged by the runtime warning if I flutter run.

[stuartmorgan-g]

This is going to be a problem. Fixing this within the context of our CI, which runs against both stable and master, is non-trivial, because the symbols I need (FlutterPluginRegistrant and friends) don't existing on stable. I'm trying to figure out if there's a solution that doesn't involve re-adding the pre-compilation file rewriting system we had briefly several years ago (IIRC) for a somewhat similar situation on the Android side.

[gaaclarke]

Fixing this within the context of our CI, which runs against both stable and master

You can mimic the behavior of the FlutterPluginRegistrant for now by subclassing the view controller and putting your setup code in the awakeFromNib method.

[stuartmorgan-g]

Thanks, I was trying to figure out what the least disruptive way to get a call at the right time was. A custom view controller makes sense!

[stuartmorgan-g]

Now investigating the video_player_avfoundation failures, which I can repro locally.

[stuartmorgan-g]

// TODO(stuartmorgan): Convert to using mock registrars instead.NSObject<FlutterPluginRegistry> *GetPluginRegistry(void) {
#if TARGET_OS_IOS
return (NSObject<FlutterPluginRegistry> *)[[UIApplication sharedApplication] delegate];
#elsereturn (FlutterViewController *)NSApplication.sharedApplication.windows[0].contentViewController;
#endif
}

🤦🏻

[stuartmorgan-g]

@bparrishMines I believe this will be landable now (modulo flake) as another incremental to continue working toward getting the roller caught up and healthy again.

[bparrishMines]

LGTM

[auto-submit]

autosubmit label was removed for flutter/packages/9407, because - The status or check suite Linux_web web_platform_tests_shard_1 master has failed. Please fix the issues identified (or deflake) before re-applying this label.

[stuartmorgan-g]

Overrides: the changes to the example app aren't relevant to clients.

[stuartmorgan-g]

(Force pushed due to flutter/flutter#170300)

[stuartmorgan-g]

This testBlankVideoBugWithEncryptedVideoStreamAndInvertedAspectRatioBugForSomeVideoStream failure may be real? It's at least a frequent flake since it showed up in the last run too. It didn't repro locally for me though.

Test Case '-[VideoPlayerTests testBlankVideoBugWithEncryptedVideoStreamAndInvertedAspectRatioBugForSomeVideoStream]' started.
2025-06-09 15:55:24.753451-0700 video_player_avfoundation_example[53023:1014858] [aqme] LoudnessManager.mm:413 PlatformUtilities::CopyHardwareModelFullName() returns unknown value: Macmini9,1, defaulting hw platform key
/Volumes/Work/s/w/ir/x/w/packages/packages/video_player/video_player_avfoundation/darwin/RunnerTests/VideoPlayerTests.m:216: error: -[VideoPlayerTests testBlankVideoBugWithEncryptedVideoStreamAndInvertedAspectRatioBugForSomeVideoStream] : ((player.playerLayer.superlayer) != nil) failed - AVPlayerLayer should be added on screen.
Test Case '-[VideoPlayerTests testBlankVideoBugWithEncryptedVideoStreamAndInvertedAspectRatioBugForSomeVideoStream]' failed (0.061 seconds).

[stuartmorgan-g]

Maybe a race condition. We are getting what should end up being the superlayer via UIApplication.sharedApplication.keyWindow.rootViewController, and with the delegate migration maybe that's not reliably available when the test runs.

[stuartmorgan-g]

video_player_avfoundation should get a re-review; I changed the production code to allow an injectable view provider rather than iOS hard-coding getting it from the app's root view controller.