Skip to content

Latest commit

 

History

History
179 lines (117 loc) · 5.33 KB

DETECT.md

File metadata and controls

179 lines (117 loc) · 5.33 KB
Raw

Coming soon! See flutter/devtools#3951.

The text below is under construction.

Detect Memory Leaks

The documentation below is valid for Flutter SDKs >= 3.18.0.

This page describes how to auto-detect certain types of memory leaks. Read more about leak tracking in overview.

Quick start to track leaks for Flutter

Test cover with testWidgets

The Flutter test method testWidgets can be configured to track and detect leaks from all instrumented classes. To enable leak tracking for your entire test suite and to configure leak tracking settings:

  1. Add dev_dependency on leak_tracker_flutter_testing. Put any instead of version, because the version is defined by your Flutter SDK.
dev_dependencies:
  ...
  leak_tracker_flutter_testing: any
  1. Add or modify your test/flutter_test_config.dart file:
import 'dart:async';

import 'package:leak_tracker_flutter_testing/leak_tracker_flutter_testing.dart';

...

Future<void> testExecutable(FutureOr<void> Function() testMain) async {
  LeakTesting.enable();
  LeakTesting.settings = LeakTesting.settings
    .withIgnored(createdByTestHelpers: true);

  ...

  await testMain();
}

You can also adjust leak tracking settings for individual tests:

testWidgets('Images are rendered as expected.',
// TODO(polina-c): make sure images are disposed, https://github.com/polina-c/my_repo/issues/141
experimentalLeakTesting: LeakTesting.settings.withIgnored(classes: ['Image']),
(WidgetTester tester) async {
  ...

See documentation for testWidgets for more information.

Instrument more disposables

To instrument a disposable class for leak tracking, you need to dispatch object creation and disposal events to leak_tracker. Use the example as a guide.

See leaks in a running Flutter application

TODO(polina-c): implement and test this scenario #172

  1. Add leak_tracker to dependencies in pubspec.yaml.

  2. Before runApp invocation, enable leak tracking, and connect the Flutter memory allocation events:

import 'package:flutter/foundation.dart';
import 'package:leak_tracker/leak_tracker.dart';

...

enableLeakTracking();
FlutterMemoryAllocations.instance
      .addListener((ObjectEvent event) => dispatchObjectEvent(event.toMap()));
runApp(...);
  1. Run the application in debug mode and watch for a leak related warnings.

TODO(polina-c): add example of the warning #172

Limitations

By environment

At this time, leak tracking is supported only for unit tests within Flutter packages.

With the latest version of leak_tracker, leak tracking is not supported for:

  1. Web platform
  2. Running applications
  3. Pure Dart packages

By leak types

Leak tracking for not-GCed leaks is experimental and is off by default. At this time, it is recommended to track only not-disposed leaks.

By tracked classes

The leak_tracker will catch leaks only for instrumented objects (see concepts for details).

However, the good news is:

  1. Disposables in the Flutter Framework are instrumented. If your Flutter app manages widgets in a way that results in leaks, leak_tracker will catch them.

  2. If a leak involves at least one instrumented object, the leak will be caught and all other objects, even non-instrumented, will stop leaking as well.

See the instrumentation guidance.

By build mode

The leak_tracker availability differs by build modes. See Dart build modes or Flutter build modes.

Dart development and Flutter debug

Leak tracking is fully available.

Flutter profile

Leak tracking is available, but FlutterMemoryAllocations that listens to Flutter instrumented objects, should be turned on if you want to track Flutter Framework objects.

Dart production and Flutter release

Leak tracking is disabled.

NOTE: If you are interested in enabling leak tracking for release mode, please, comment here.

Performance impact

Memory

The leak_tracker stores a small additional record for each tracked alive object and for each detected leak, that increases the memory footprint.

For the Gallery application in profile mode on macos the leak tracking increased memory footprint of the home page by ~400 KB that is ~0.5% of the total.

CPU

Leak tracking impacts CPU in two areas:

  1. Per object tracking. Added ~0.05 of millisecond (~2.7%) to the total load time of Gallery home page in profile mode on macos.

  2. Regular asynchronous analysis of the tracked objects. Took ~2.5 milliseconds for Gallery home page in profile mode on macos.