This pull request fixes#143803 by taking advantage of Dart's null-aware operators.
And unlike `switch` expressions ([9 PRs](https://github.com/flutter/flutter/pull/143634) and counting), the Flutter codebase is already fantastic when it comes to null-aware coding. After refactoring the entire repo, all the changes involving `?.` and `??` can fit into a single pull request.
This PR implements the functionality described above and hides it behind
the `--experimental-faster-testing` flag of `flutter test`.
### The following are some performance measurements from test runs
conducted on GitHub Actions
run 1 logs:
https://github.com/derekxu16/flutter_test_ci/actions/runs/8008029772/attempts/1
run 2 logs:
https://github.com/derekxu16/flutter_test_ci/actions/runs/8008029772/attempts/2
run 3 logs:
https://github.com/derekxu16/flutter_test_ci/actions/runs/8008029772/attempts/3
**length of `flutter test --reporter=expanded test/animation
test/foundation` step**
run 1: 54s
run 2: 52s
run 3: 56s
average: 54s
**length of `flutter test --experimental-faster-testing
--reporter=expanded test/animation test/foundation` step**
run 1: 27s
run 2: 27s
run 3: 29s
average: 27.67s (~48.77% shorter than 54s)
**length of `flutter test --reporter=expanded test/animation
test/foundation test/gestures test/painting test/physics test/rendering
test/scheduler test/semantics test/services` step**
run 1: 260s
run 2: 270s
run 3: 305s
average: 278.33s
**length of `flutter test --experimental-faster-testing
--reporter=expanded test/animation test/foundation test/gestures
test/painting test/physics test/rendering test/scheduler test/semantics
test/services` step**
from a clean build (right after deleting the build folder):
run 1: 215s
run 2: 227s
run 3: 245s
average: 229s (~17.72% shorter than 278.33s)
Note that in reality, `test/material` was not passed to `flutter test`
in the trials below. All of the test files under `test/material` except
for `test/material/icons_test.dart` were listed out individually
**length of `flutter test --reporter=expanded test/material` step**
run 1: 408s
run 2: 421s
run 3: 451s
average: 426.67s
**length of `flutter test --experimental-faster-testing
--reporter=expanded test/material` step**
run 1: 382s
run 2: 373s
run 3: 400s
average: 385s (~9.77% shorter than 426.67s)
---------
Co-authored-by: Dan Field <dnfield@google.com>
Fixes https://github.com/flutter/flutter/issues/143405.
It was counter-intuitive that a SemanticsFinder without specifying a FlutterView would only search the nodes in the default view. This change makes it so that when no view is specified the semantics trees of all known FlutterViews are searched.
Reverts flutter/flutter#143386
Initiated by: Piinks
Reason for reverting: This broke a customer test.
Original PR Author: Piinks
Reviewed By: {goderbauer}
This change reverts the following previous change:
Original Description:
Fixes https://github.com/flutter/flutter/issues/143173
The `start` and `end` parameters of `SemanticsController.simulatedAccessibilityTraversal` were deprecated in https://github.com/flutter/flutter/issues/112413, but no tests were added that verified the new API. 😳
This change
- fixes a typo in an error message
- fixes the new `startNode` and `endNode` not being accounted for in setting the traversal range
- adds dart fixes for the deprecations
- adds tests for the new API that is meant to replace the deprecated one.
- Filed https://github.com/flutter/flutter/issues/143405 to follow up on the new API not working in multiple views.
Reverts flutter/flutter#143334
Initiated by: hangyujin
Reason for reverting: broke g3 tests
Original PR Author: hangyujin
Reviewed By: {LongCatIsLooong}
This change reverts the following previous change:
Original Description:
Add a semantics flag to text field to fix https://github.com/flutter/flutter/issues/143337 (in IOS the disabled text field is not read `dimmed`)
internal: b/322345393
Fixes https://github.com/flutter/flutter/issues/143173
The `start` and `end` parameters of `SemanticsController.simulatedAccessibilityTraversal` were deprecated in https://github.com/flutter/flutter/issues/112413, but no tests were added that verified the new API. ð³
This change
- fixes a typo in an error message
- fixes the new `startNode` and `endNode` not being accounted for in setting the traversal range
- adds dart fixes for the deprecations
- adds tests for the new API that is meant to replace the deprecated one.
- Filed https://github.com/flutter/flutter/issues/143405 to follow up on the new API not working in multiple views.
During golden test image comparison 2 lists of a different type are compared with the method "identical", so this will never be true. The test image is a _Uint8ArrayView while the master image is an Uint8List. So that results in always a heavy computation to get the difference between the test and the master image.
When you run this test snippet I go from 51 seconds to 14 seconds:
```dart
import 'package:flutter/material.dart';
import 'package:flutter_test/flutter_test.dart';
void main() {
for (int i = 0; i < 100; i++) {
testWidgets('Small test', (WidgetTester tester) async {
await tester.pumpWidget(Directionality(textDirection: TextDirection.ltr, child: Text('jo')));
await expectLater(find.byType(Text), matchesGoldenFile('main.png'));
});
}
}
```
Update the `matchesGoldenFile()` / `LocalComparisonOutput` code to generate failure images for golden tests that fail when the image sizes do not match. This can make it far quicker to identify what is wrong with the test image.
Fixes https://github.com/flutter/flutter/issues/141488
- [ x I read and followed the [Flutter Style Guide], including [Features we expect every widget to implement].
Reverts flutter/flutter#142339
In the original change one of the tests included the same view twice which resulted in a different failure than the expected one. The second commit contains the fix for this. I don't understand how this wasn't caught presubmit on CI.
Reverts flutter/flutter#141484
Initiated by: eliasyishak
This change reverts the following previous change:
Original Description:
The existing `runApp` bootstraps the widget tree and renders the provided widget into the default view (which is currently the implicit View from `PlatformDispatcher.implicitView` and - in the future - may be a default-created window). Apps, that want more control over the View they are rendered in, need a new way to bootstrap the widget tree: `runWidget`. It does not make any assumptions about the View the provided widget is rendered into. Instead, it is up to the caller to include a View widget in the provided widget tree that specifies where content should be rendered. In the future, this may enable developers to create a custom window for their app instead of relying on the default-created one.
The existing `runApp` bootstraps the widget tree and renders the provided widget into the default view (which is currently the implicit View from `PlatformDispatcher.implicitView` and - in the future - may be a default-created window). Apps, that want more control over the View they are rendered in, need a new way to bootstrap the widget tree: `runWidget`. It does not make any assumptions about the View the provided widget is rendered into. Instead, it is up to the caller to include a View widget in the provided widget tree that specifies where content should be rendered. In the future, this may enable developers to create a custom window for their app instead of relying on the default-created one.
When a `TextField` is rendered before a `Navigator`, it breaks in semantics mode. This is because the framework generates the incorrect semantics tree (excludes the TextField) and when that tree gets sent to the engine, we don't get the signal to create the corresponding `<input>` element.
This happens for a few reasons:
* `ModalBarrier` uses `BlockSemantics` to drop the semantics of routes beneath the current route in `Navigator`
* `ModalBarrier` mistakenly recognizes the widget outside of the `Navigator` to be its sibling
* So we end up dropping the semantics node of the `TextField` rendered before it.
The fix is to let `Navigator` generate a semantics node so that `ModalBarrier` doesn't mistakenly think widgets outside of `Navigator` are its siblings.
`Navigator` doesn't currently do this, which causes all the nodes generated from its widget subtree to be directly attached to the parent semantics node above `Navigator` - since this is also the parent of `TextField`, it considers them siblings.
Fixes https://github.com/flutter/flutter/issues/129324
Much nicer calling API and simplifies evolving this API in the future.
I wish we could write a dart fix for this, but that's blocked on https://github.com/dart-lang/sdk/issues/54668.
The old doc says that, AutomatedTestWidgetsFlutterBinding for `flutter test` and LiveTestWidgetsFlutterBinding for `flutter run`. However, suppose we `flutter test integration_test/simple_test.dart` with the following code:
```
void main() {
testWidgets('hi', (WidgetTester tester) async {
print('hi ${TestWidgetsFlutterBinding.instance} ${Platform.operatingSystem}');
});
}
```
We will see: `hi <IntegrationTestWidgetsFlutterBinding> ios`. Therefore, we see `IntegrationTestWidgetsFlutterBinding` is used in a `flutter test` command, which is contrary to the documentation.
This reverts commit
d24c01bd0c.
The original change was reverted because it caused some apps to get
stuck on the splash screen on some phones.
An investigation determined that this was due to a rounding error.
Example: The device reports a physical size of 1008.0 x 2198.0 with a
dpr of 1.912500023841858. Flutter would translate that to a logical size
of 527.0588169589221 x 1149.2810314243163 and use that as the input for
its layout algorithm. Since the constraints here are tight, the layout
algorithm would determine that the resulting logical size of the root
render object must be 527.0588169589221 x 1149.2810314243163.
Translating this back to physical pixels by applying the dpr resulted in
a physical size of 1007.9999999999999 x 2198.0 for the frame. Android
now rejected that frame because it didn't match the expected size of
1008.0 x 2198.0 and since no frame had been rendered would never take
down the splash screen.
Prior to dynamically sized views, this wasn't an issue because we would
hard-code the frame size to whatever the requested size was.
Changes in this PR over the original PR:
* The issue has been fixed now by constraining the calculated physical
size to the input physical constraints which makes sure that we always
end up with a size that is acceptable to the operating system.
* The `ViewConfiguration` was refactored to use the slightly more
convenient `BoxConstraints` over the `ViewConstraints` to represent
constraints. Both essentially represent the same thing, but
`BoxConstraints` are more powerful and we avoid a couple of translations
between the two by translating the` ViewConstraints` from the
`FlutterView` to `BoxConstraints` directly when the `ViewConfiguration`
is created.
All changes over the original PR are contained in the second commit of
this PR.
Fixes b/316813075
Part of https://github.com/flutter/flutter/issues/134501.
Reverts flutter/flutter#139717
Initiated by: LongCatIsLooong
This change reverts the following previous change:
Original Description:
Fixes https://github.com/flutter/flutter/issues/131435, #104594, #43400
Needs https://github.com/flutter/engine/pull/48774 (to fix the web test failure).
Currently the method we use for text span hit testing `TextPainter.getPositionForOffset` always returns the closest `TextPosition`, even when the given offset is far away from the text.
The new TextPaintes method tells you the layout bounds (`width = letterspacing / 2 + x_advance + letterspacing / 2`, `height = font ascent + font descent`) of a character, the PR changes the hit testing implementation such that a TextSpan is only considered hit if the point-down event landed in one of it's character's layout bounds.
Potential issues:
1. In theory since the text is baseline aligned, we should use the max ascent and max descent of each character to calculate the height of the text span's hit-test region, in case some characters in the span have to fall back to a different font, but that will be slower and it typically doesn't make a huge difference.
This is a breaking change. It also introduces a new finder and a new method `WidgetTester.tapOnText`: `await tester.tapOnText('string to match')` for ease of migration.
Fixes https://github.com/flutter/flutter/issues/131435, #104594, #43400
Needs https://github.com/flutter/engine/pull/48774 (to fix the web test failure).
Currently the method we use for text span hit testing `TextPainter.getPositionForOffset` always returns the closest `TextPosition`, even when the given offset is far away from the text.
The new TextPaintes method tells you the layout bounds (`width = letterspacing / 2 + x_advance + letterspacing / 2`, `height = font ascent + font descent`) of a character, the PR changes the hit testing implementation such that a TextSpan is only considered hit if the point-down event landed in one of it's character's layout bounds.
Potential issues:
1. In theory since the text is baseline aligned, we should use the max ascent and max descent of each character to calculate the height of the text span's hit-test region, in case some characters in the span have to fall back to a different font, but that will be slower and it typically doesn't make a huge difference.
This is a breaking change. It also introduces a new finder and a new method `WidgetTester.tapOnText`: `await tester.tapOnText('string to match')` for ease of migration.
1. Move leak_tracker and leak_tracker_testing out of direct dependencies.
2. Move leak_tracker_flutter_testing from dev to prod dependencies for flutter_test
It is prerequisite for https://github.com/flutter/flutter/issues/135856