tree: 352e6c06a57f48e913deb9c06f9059b3126b20ab [path history] [tgz]
  1. bin/
  2. lib/
  3. test/
  4. manifest.yaml
  5. pubspec.yaml
  6. README.md
dev/devicelab/README.md

Flutter devicelab

“Devicelab” (a.k.a. Cocoon) is a physical lab that tests Flutter on real Android and iOS devices.

This package contains the code for test framework and the tests. More generally the tests are referred to as “tasks” in the API, but since we primarily use it for testing, this document refers to them as “tests”.

Current statuses for the devicelab are available at https://flutter-dashboard.appspot.com.

Dashboards

Build dashboard

The build page is accessible at https://flutter-dashboard.appspot.com/#/build. This page reports the build statuses of commits to the flutter/flutter repo.

Tasks

Task statuses are color-coded in the following manner:

New task (blue): the task is waiting for an agent to pick it up and start the build.

Task is running (blue with clock): an agent is currently building the task.

Task succeeded (green): an agent reported a successful completion of the task.

Task is flaky (yellow): the task was attempted multiple time, but only the latest attempt succeeded (we currently only try twice).

Task failed (red): the task failed all of the attempts.

Task is rerunning (orange): the task is being rerun.

Task was skipped (transparent): the task is not scheduled for a build. This usually happens when a task is removed from manifest.yaml file.

In addition to color-coding, a task may display a question mark. This means that the task was marked as flaky manually. The status of such task is ignored when considering whether the build is broken or not. For example, if a flaky task fails, GitHub will not prevent PR submissions. However, if the latest status of a non-flaky task is red, all pending PRs will contain a warning about the broken build and recommend caution when submitting.

Clicking a cell will pop up an overlay with information about that task. It includes information such as the task name, number of attempts, run time, queue time, whether it is manually marked flaky, and the agent it was run on. It has actions to download the log, rerun the task, and view the agent on the agent dashboard.

Why is a task stuck on “new task” status?

The dashboard aggregates build results from multiple build environments, including Cirrus, Chrome Infra, and devicelab. While devicelab tests every commit that goes into the master branch, other environments may skip some commits. For example, Cirrus will only test the last commit of a PR that's merged into the master branch. Chrome Infra may skip commits when they come in too fast.

Agent dashboard

Agent statuses are available at https://flutter-dashboard.appspot.com/#/agents.

A green agent is considered healthy and ready to receive new tasks to build. A red agent is broken and does not receive new tasks.

Performance dashboard

Flutter benchmarks are available at https://flutter-dashboard.appspot.com/benchmarks.html.

How the devicelab runs tasks

The devicelab agents have a small script installed on them that continuously asks the CI server for tasks to run. When the server finds a suitable task for an agent it reserves that task for the agent. If the task succeeds, the agent reports the success to the server and the dashboard shows that task in green. If the task fails, the agent reports the failure to the server, the server increments the counter counting the number of attempts it took to run the task and puts the task back in the pool of available tasks. If a task does not succeed after a certain number of attempts (as of this writing the limit is 2), the task is marked as failed and is displayed using red color on the dashboard.

Running tests locally

Do make sure your tests pass locally before deploying to the CI environment. Below is a handful of commands that run tests in a similar way to how the CI environment runs them. These commands are also useful when you need to reproduce a CI test failure locally.

Prerequisites

You must set the ANDROID_HOME or ANDROID_SDK_ROOT environment variable to run tests on Android. If you have a local build of the Flutter engine, then you have a copy of the Android SDK at .../engine/src/third_party/android_tools/sdk.

You can find where your Android SDK is using flutter doctor.

Warnings

Running devicelab will do things to your environment.

Notably, it will start and stop gradle, for instance.

Running all tests

To run all tests defined in manifest.yaml, use option -a (--all):

../../bin/cache/dart-sdk/bin/dart bin/run.dart -a

This defaults to only running tests supported by your host device's platform (--match-host-platform) and exiting after the first failure (--exit).

Running specific tests

To run a test, use option -t (--task):

# from the .../flutter/dev/devicelab directory
../../bin/cache/dart-sdk/bin/dart bin/run.dart -t {NAME_OR_PATH_OF_TEST}

Where NAME_OR_PATH_OF_TEST can be either of:

  • the name of a task, which you can find in the manifest.yaml file in this directory. Example: complex_layout__start_up.
  • the path to a Dart file corresponding to a task, which resides in bin/tasks. Tip: most shells support path auto-completion using the Tab key. Example: bin/tasks/complex_layout__start_up.dart.

To run multiple tests, repeat option -t (--task) multiple times:

../../bin/cache/dart-sdk/bin/dart bin/run.dart -t test1 -t test2 -t test3

To run tests from a specific stage, use option -s (--stage). Currently there are only three stages defined, devicelab, devicelab_ios and devicelab_win.

../../bin/cache/dart-sdk/bin/dart bin/run.dart -s {NAME_OF_STAGE}

Running tests against a local engine build

To run device lab tests against a local engine build, pass the appropriate flags to bin/run.dart:

../../bin/cache/dart-sdk/bin/dart bin/run.dart --task=[some_task] \
  --local-engine-src-path=[path_to_local]/engine/src \
  --local-engine=[local_engine_architecture]

An example of a local engine architecture is android_debug_unopt_x86.

Reproducing broken builds locally

To reproduce the breakage locally git checkout the corresponding Flutter revision. Note the name of the test that failed. In the example above the failing test is flutter_gallery__transition_perf. This name can be passed to the run.dart command. For example:

../../bin/cache/dart-sdk/bin/dart bin/run.dart -t flutter_gallery__transition_perf

Writing tests

A test is a simple Dart program that lives under bin/tasks and uses package:flutter_devicelab/framework/framework.dart to define and run a task.

Example:

import 'dart:async';

import 'package:flutter_devicelab/framework/framework.dart';

Future<void> main() async {
  await task(() async {
    ... do something interesting ...

    // Aggregate results into a JSONable Map structure.
    Map<String, dynamic> testResults = ...;

    // Report success.
    return new TaskResult.success(testResults);

    // Or you can also report a failure.
    return new TaskResult.failure('Something went wrong!');
  });
}

Only one task is permitted per program. However, that task can run any number of tests internally. A task has a name. It succeeds and fails independently of other tasks, and is reported to the dashboard independently of other tasks.

A task runs in its own standalone Dart VM and reports results via Dart VM service protocol. This ensures that tasks do not interfere with each other and lets the CI system time out and clean up tasks that get stuck.

Adding tests to the CI environment

The manifest.yaml file describes a subset of tests we run in the CI. To add your test edit manifest.yaml and add the following in the “tasks” dictionary:

  {NAME_OF_TEST}:
    description: {DESCRIPTION}
    stage: {STAGE}
    required_agent_capabilities: {CAPABILITIES}

Where:

  • {NAME_OF_TEST} is the name of your test that also matches the name of the file in bin/tasks without the .dart extension.
  • {DESCRIPTION} is the plain English description of your test that helps others understand what this test is testing.
  • {STAGE} is devicelab if you want to run on Android, or devicelab_ios if you want to run on iOS.
  • {CAPABILITIES} is an array that lists the capabilities required of the test agent (the computer that runs the test) to run your test. As of writing, the available capabilities are: linux, linux/android, linux-vm, mac, mac/ios, mac/iphonexs, mac/ios32, mac-catalina/ios, mac-catalina/android, ios/gl-render-image, windows, windows/android.

If your test needs to run on multiple operating systems, create a separate test for each operating system.