Zachary Anderson

Sure, if this is mostly just noise from SkiaPerf, we can just close. We'll get a clearer signal when these benchmarks are running on dedicated machines in the device lab.

Looking at the logs, I'm not able to discern why the tests failed. Routing to the Android team.

Routing to @elliette who added the test, though it may be a dwds issue. Also reducing the priority to P3 since this is only just above the flake threshold.

It does not get disabled automatically. You'd have to open a PR to add the `skip: true` parameter to the test. (And tag this issue in a comment.)

2933 was a real crash in Impeller 2916 was an Observatory discovery timeout 2895 was not a flake. Dropping the priority to P3 and repurposing to track the Impeller crash.

malloc debug info is here. The logs for 2933 contain backtraces for all threads. ``` [backdrop_filter_perf_impeller_ios__timeline_summary] [STDOUT] stdout: [+1944 ms] Runner(437,0x16f593000) malloc: Incorrect checksum for freed object 0x11230e800: probably modified...

3197 is the Observatory discovery timeout 3177 was a Dart VM crash while reporting a timeline event ``` [backdrop_filter_perf_impeller_ios__timeline_summary] [STDOUT] stdout: [ ] * thread #10, name = 'io.flutter.1.raster', stop...

3437 is another crash in Dart timeline event initialization: ``` [2023-03-24 10:27:34.873475] [STDOUT] stdout: [ ] * thread #9, name = 'io.flutter.1.ui', stop reason = signal SIGABRT [2023-03-24 10:27:34.873519] [STDOUT]...

3616 and 3559 are observatory discovery timeouts 3593 and 3565 are crashes in the Dart VM timeline I'm going to file a separate issue to track the Dart VM timeline...

All three of these are timeouts on Observatory discovery. lldb is reporting the stacks for all engine threads indicating that is waiting for input. The log links don't include a...