Debugging Flutter apps

There’s a wide variety of tools and features to help debug Flutter applications. Here are some of the available tools:

  • DevTools, a suite of performance and profiling tools that run in a browser.
  • Android Studio/IntelliJ, and VS Code (enabled with the Flutter and Dart plugins) support a built-in source-level debugger with the ability to set breakpoints, step through code, and examine values.
  • Flutter inspector, a widget inspector available in DevTools, and also directly from Android Studio and IntelliJ (enabled with the Flutter plugin). The inspector allows you to examine a visual representation of the widget tree, inspect individual widgets and their property values, enable the performance overlay, and more.

DevTools

For debugging and profiling apps, DevTools might be the first tool you reach for. DevTools runs in a browser and supports a variety of features:

  • source-level debugger
  • widget inspector that displays a visual widget tree, and “widget select” mode where you select a widget in the app and it drills down to that widget in the tree
  • memory profiler
  • timeline view that supports tracing, and importing and exporting trace information
  • logging view

If you run your application in debug mode or profile mode, while it’s running you can open DevTools in the browser to connect to your app. DevTools doesn’t work well with an app compiled to release mode, as the debugging and profiling information has been stripped away.

If you use DevTools for profiling, make sure to run your application in profile mode. Otherwise, the main output that appears on your profile are the debug asserts verifying the framework’s various invariants (see Debug mode assertions).

GIF showing DevTools features

For more information, see the DevTools documentation.

Setting breakpoints

You can set breakpoints directly in your IDE/editor (such as Android Studio/IntelliJ and VS Code), in the DevTools debugger, or programmatically.

The Dart analyzer

If you’re using a Flutter enabled IDE/editor, the Dart analyzer is already checking your code and looking for possible mistakes.

If you run from the command line, test your code with flutter analyze.

The Dart analyzer makes heavy use of type annotations that you put in your code to help track problems down. You are encouraged to use them everywhere (avoiding var, untyped arguments, untyped list literals, and so on) as this is the quickest and least painful way of tracking down problems.

For more information, see Using the Dart analyzer.

Logging

Another useful debugging tool is logging. You set logging up programmatically then view the output in the DevTools logging view, or in the console.

Debugging application layers

Flutter was designed with a layered architecture that includes widget, rendering, and painting layers. For links to more information and videos, see The Framework architecture on the GitHub wiki, and the community article, The Layer Cake.

The Flutter widget inspector provides a visual representation of the widget tree, but if you want a greater level of detail, or you want a verbose text-based dump of the widget, layer, or render trees, see Debug flags: application layers in the Debugging Flutter apps programmatically page.

Debug mode assertions

During development, you are highly encouraged to use Flutter’s debug mode. This is the default if you use bug icon in Android Studio, or flutter run at the command line. Some tools support assert statements through the command-line flag --enable-asserts.

In this mode, Dart assert statements are enabled, and the Flutter framework evaluates the argument to each assert statement encountered during execution, throwing an exception if the result is false. This allows developers to enable or disable invariant checking, such that the associated performance cost is only paid during debugging sessions.

When an invariant is violated, it’s reported to the console, with some context information to help track down the source of the problem.

For more information, see Assert in the Dart language tour.

Debugging animations

The easiest way to debug animations is to slow them down. The Flutter inspector provides a Slow Animations button, or you can slow the animations programmatically.

For more information on debugging janky (non-smooth) applications, see Flutter performance profiling.

Measuring app startup time

To gather detailed information about the time it takes for your Flutter app to start, you can run the flutter run command with the trace-startup and profile options.

$ flutter run --trace-startup --profile

The trace output is saved as a JSON file called start_up_info.json under the build directory of your Flutter project. The output lists the elapsed time from app startup to these trace events (captured in microseconds):

  • Time to enter the Flutter engine code.
  • Time to render the first frame of the app.
  • Time to initialize the Flutter framework.
  • Time to complete the Flutter framework initialization.

For example:

{
  "engineEnterTimestampMicros": 96025565262,
  "timeToFirstFrameMicros": 2171978,
  "timeToFrameworkInitMicros": 514585,
  "timeAfterFrameworkInitMicros": 1657393
}

Tracing Dart code

To perform a performance trace, you can use the DevTools Timeline view. The Timeline view also supports importing and exporting trace files. For more information, see the Timeline view docs.

You can also perform traces programmatically, though these traces can’t be imported into DevTool’s Timeline view.

Be sure to use run your app in profile mode before tracing to ensure that the runtime performance characteristics closely matches that of your final product.

Performance overlay

To get a graphical view of the performance of your application, turn on the performance overlay. You can do this in the by clicking the Performance Overlay button in the Flutter inspector.

You can also turn on the overlay programmatically.

For information on how to interpret the graphs in the overlay, see The performance overlay in the Flutter performance profiling guide.

Debug flags

In most cases, you won’t need to use the debug flags directly, as you’ll find the most useful debugging functionality in the DevTools suite. But if you prefer to use the debug flags directly, see Debug flags: performance in the Debugging Flutter apps programmatically page.

Common problems

The following is a problem that some have encountered on MacOS.

“Too many open files” exception (MacOS)

The default limit for Mac OS on how many files it can have open at a time is rather low. If you run into this limit, increase the number of available file handlers using the ulimit command:

ulimit -S -n 2048

If you use Travis or Cirrus for testing, increase the number of available file handlers that they can open by adding the same line to flutter/.travis.yml, or flutter/.cirrus.yml, respectively.

Widgets marked const that should be equal to each other, aren’t

In debug mode, you might find that two const widgets that should to all appearances be equal (because of Dart’s constant deduplication) are not.

For example, this code should print 1:

print(<Widget>{
  // this is the syntax for a Set<Widget> literal
  const SizedBox(),
  const SizedBox(),
}.length);

It should print 1 (rather than 2) because the two constants are the same and sets coalesce duplicate values (and indeed the analyzer complains that “Two elements in a set literal shouldn’t be equal”). As expected, in release builds, it does print 1. However, in debug builds it prints 2. This is because the flutter tool injects the source location of Widget constructors into the code at compile time, so the code is effectively:

print(<Widget>{
  const SizedBox(/* location: Location(file: 'foo.dart', line: 12) */),
  const SizedBox(/* location: Location(file: 'foo.dart', line: 13) */),
}.length);

This results in the instances being different, and so they are not deduplicated by the set. We use this injected information to make the error messages clearer when a widget is involved in an exception, by reporting where the relevant widget was created. Unfortunately, it has the visible side-effect of making otherwise-identical constants be different at compile time.

To disable this behavior, pass --no-track-widget-creation to the flutter run command. With that flag set, the code above prints “1” in debug and release builds, and error messages include a message saying that they cannot provide all the information that they would otherwise be able to provide if widget creation tracking was enabled.

See also:

Other resources

You might find the following docs useful: