Before this change, having an Android app depend on a plugin that has no android implementation resulted in a Gradle build failure.
This scenario is likely to become more common if we're enabling federated plugins, as the package implementing just the desktop implementation of a plugin won't have an Android implementation.
This changes the Gradle plugin to not try to build any plugins that doesn't have an android/build.gradle file.
This reverts commit af9f424d5d.
Reverts flutter/flutter#39439
Reason: this passed the local device lab test on my Macbook but it failed in the actual device lab. Will investigate, fix, and reland.
TBR: @dnfield @goderbauer @tvolkert @Hixie
For https://github.com/flutter/flutter/issues/33899
Test added:
- simple_animation_perf_ios
Test modified:
- backdrop_filter_perf_ios__timeline_summary
We'll add the CPU/GPU measurement to more iOS tests
once it's proven to be non-flaky.
`flutter build aar`
This new build command works just like `flutter build apk` or `flutter build appbundle`, but for plugin and module projects.
This PR also refactors how plugins are included in app or module projects. By building the plugins as AARs, the Android Gradle plugin is able to use Jetifier to translate support libraries into AndroidX libraries for all the plugin's native code. Thus, reducing the error rate when using AndroidX in apps.
This change also allows to build modules as AARs, so developers can take these artifacts and distribute them along with the native host app without the need of the Flutter tool. This is a requirement for add to app.
`flutter build aar` generates POM artifacts (XML files) which contain metadata about the native dependencies used by the plugin. This allows Gradle to resolve dependencies at the app level. The result of this new build command is a single build/outputs/repo, the local repository that contains all the generated AARs and POM files.
In a Flutter app project, this local repo is used by the Flutter Gradle plugin to resolve the plugin dependencies. In add to app case, the developer needs to configure the local repo and the dependency manually in `build.gradle`:
repositories {
maven {
url "<path-to-flutter-module>build/host/outputs/repo"
}
}
dependencies {
implementation("<package-name>:flutter_<build-mode>:1.0@aar") {
transitive = true
}
}
`flutter build aar`
This new build command works just like `flutter build apk` or `flutter build appbundle`, but for plugin and module projects.
This PR also refactors how plugins are included in app or module projects. By building the plugins as AARs, the Android Gradle plugin is able to use Jetifier to translate support libraries into AndroidX libraries for all the plugin's native code. Thus, reducing the error rate when using AndroidX in apps.
This change also allows to build modules as AARs, so developers can take these artifacts and distribute them along with the native host app without the need of the Flutter tool. This is a requirement for add to app.
`flutter build aar` generates POM artifacts (XML files) which contain metadata about the native dependencies used by the plugin. This allows Gradle to resolve dependencies at the app level. The result of this new build command is a single build/outputs/repo, the local repository that contains all the generated AARs and POM files.
In a Flutter app project, this local repo is used by the Flutter Gradle plugin to resolve the plugin dependencies. In add to app case, the developer needs to configure the local repo and the dependency manually in `build.gradle`:
repositories {
maven {
url "<path-to-flutter-module>build/host/outputs/repo"
}
}
dependencies {
implementation("<package-name>:flutter_<build-mode>:1.0@aar") {
transitive = true
}
}
This is the first step in getting esoteric devices wired
up to device lab to ensure that we can at least start Flutter
apps in profile mode on such devices.
https://github.com/flutter/flutter/issues/35838
* Introduce image_list performance benchmark that runs on jit(debug) build.
We want to track performance of debug build because ongoing work on reducing the cost of spanwing new isolate will impact jit(debug) more than it will aot(release) configurations.
* Newlines and spaces
* Add test case for Flutter Issue #27677 as a benchmark.
See https://github.com/flutter/flutter/issues/27677
I got the following results running the test on a Moto E2 which
will help us determine how much we can gain by analyzing the
operations and eliminating unnecessary repaints based on
dirty rectangles/regions.
no blurs - avg 216.0 fps over 3 samples
blur the group - avg 22.9 fps over 3 samples
blur each txt - avg 3.4 fps over 3 samples
* Added the new benchmark to be tracked on the dashboard.
* Gradle generates ELF shared libraries instead of AOT snapshots.
* `flutter build apk/appbundle` supports multiple `--target-platform` and defaults to `android-arm` and `android-arm64`.
* `flutter build apk` now has a flag called `--split-per-abi`.
* Gradle generates ELF shared libraries instead of AOT snapshots.
* `flutter build apk/appbundle` supports multiple `--target-platform` and defaults to `android-arm` and `android-arm64`.
* `flutter build apk` now has a flag called `--split-per-abi`.
* Added support for authentication codes for the VM service.
Previously, a valid web socket connection would use the following URI:
`ws://127.0.0.1/ws`
Now, by default, the VM service requires a connection to be made with a
URI similar to the following:
`ws://127.0.0.1:8181/Ug_U0QVsqFs=/ws`
where `Ug_U0QVsqFs` is an authentication code generated and shared by
the
service.
This behavior can be disabled with the `--disable-service-auth-codes`
flag.
Previously, I used the Android emulator for testing and everything
seemed to work fine with stdout (if I remember correctly). But our
devicelab uses real Android devices and the warnings are routed to
stderr. Hence change stdout to stderr in the test.
## Description
Print actionable warnings if `flutter drive` (where most of our performance benchmarks come from) is run in debug mode and it tries to gather benchmarks using `traceAction`.
## Related Issues
https://github.com/flutter/flutter/issues/30625
## Tests
I added the following tests:
* drive_perf_debug_warning devicelab test
* add trailing commas on list/map/parameters
* add trailing commas on Invocation with nb of arg>1
* add commas for widget containing widgets
* add trailing commas if instantiation contains trailing comma
* revert bad change
This patch adds a default shader warm up process which moves shader compilation from the animation time to the startup time. This also provides an extension for `runApp` so developers can customize the warm up process.
This should reduce our worst_frame_rasterizer_time_millis from ~100ms to ~20-30ms for both flutter_gallery and complex_layout benchmarks. Besides, this should also have a significant improvement on 90th and 99th percentile time (50%-100% speedup in some cases, but I haven't tested them thoroughly; I'll let our device lab collect the data afterwards).
The tradeoff the is the startup time (time to first frame). Our `flutter run --profile --trace-startup` seems to be a little noisy and I see about 100ms-200ms increase in that measurement for complex_layout and flutter_gallery. Note that this only happens on the first run after install or data wipe. Later the Skia persistent cache will remove the overhead.
This also adds a cubic_bezier benchmark to test the custom shader warm up process.
This should fix https://github.com/flutter/flutter/issues/813 (either by `defaultShaderWarmUp`, or a `customShaderWarmUp`).
Before this, we had several places where an isReleaseMode was defined, all with the same definition. This just makes it more broadly visible to allow our users to use it, as well as creating debug and profile versions, and adding a device lab test for it.
Since this is a const value, this makes it possible for a developer to easily mark blocks that can be removed at AOT compile time.
These are essentially self-inflicted race conditions. Instead of timeouts we're going to try a more verbose logging mechanism that points out when things are taking a long time.
* Remove many timeouts.
These are essentially self-inflicted race conditions. Instead of timeouts we're going to try a more verbose logging mechanism that points out when things are taking a long time.
* Get the attach tests to pass.
* Apply review comments from Todd
* More review comment fixes
* Put back the extended timeouts here now that I know why we have them...
This reverts commit 93573de216.
Fails in the devicelab:
```
run:stderr: A problem occurred evaluating project ':app'.
run:stderr: > Could not resolve all files for configuration 'classpath'.2018-10-23T10:46:46.335864:
run:stderr: > Could not find aapt2-proto.jar (com.android.tools.build:aapt2-proto:0.3.1).2018-10-23T10:46:46.335960:
run:stderr: Searched in the following locations:2018-10-23T10:46:46.336048:
```
Adds an integration devicelab test that runs an Android app with two
custom named isolates. Tests that the isolate names are present and that
it's possible to attach to just one of the isolates.
Fixesflutter/flutter#22009
We decided that redefining the default for templates was premature. We're going to go back to having "module" in experimental land again, and we'll try again when we have the feature set fully baked.
This keeps the writing of the .metadata files, and writing the template type to them, because that was a good improvement, and there are still a bunch of added tests that improve our coverage.
* Prototype
* Fix paths to Flutter library resources
* Invoke pod install as necessary for materialized modules
* Add devicelab test for module use on iOS
* Remove debug output
* Rebase, reame materialize editable
* Add devicelab test editable iOS host app
* Removed add2app test section
* Add a detach command to detach without terminating (#21376)
* Add a detach command to detach without terminating
Fixes#21154.
* Bump protocol version for app.detach
* Tweak to detach/quit text
* Change logPrefix to named param
* Fix the text that the devicelab attach test looks for
* Track number of package dependencies in Flutter
Relands #20774.
* Use evalFlutter instead of startProcess.
That way we don't need `flutter` on the PATH.
* Revert "Deprecated CupertinoDialog in favor of a new widget called CupertinoP… (#20649)"
This reverts commit 3a4ae280e3.
* Revert "Feature pesto new recipes (#19415)"
This reverts commit e2f3b3d6f2.
* Revert "Use markdown table (#20721)"
This reverts commit 82d43b952a.
* Revert "Documentation regarding tap gesture callbacks (#20647)"
This reverts commit 3acc278521.
* Revert "Add branch to footer information in flutter docs (#20711)"
This reverts commit 9118d3d715.
* Revert "Performance test cleanup (#20652)"
This reverts commit 1993a67381.
* Revert "Track number of package dependencies in Flutter (#20722)"
This reverts commit 03d6f18f4a.
All temporary directory start with `flutter_` and have their random component separated from the name by a period, as in `flutter_test_bundle.YFYQMY`.
I've tried to find some of the places where we didn't cleanly delete temporary directories, too. This greatly reduces, though it does not entirely eliminate, the directories we leave behind when running tests, especially `flutter_tools` tests.
While I was at it I standardized on `tempDir` as the variable name for temporary directories, since it was the most common, removing occurrences of `temp` and `tmp`, among others.
Also I factored out some common code that used to catch exceptions that happen on Windows, and made more places use that pattern.
* Upgrade everything except matcher.
* Roll matcher (and test)
* Adjust tests that depend on flutter:test directly to depend on a shim
* Require use of package:test shim and remove other references to package:test
Somehow I forgot to say "super.tap()" when calling "tap()" on the new
superclass, so it was just recursing infinitely but ended up actually
crashing on the first reuse of the finder.
The error was previously swallowed, I made this print it instead.
Disallow calling stop() or cancel() multiple times. This means that
when you use startProgress you have to more carefully think about what
exactly is going on.
Properly cancel startProgress in non-ANSI situations, so that
back-to-back startProgress calls all render to the console.
fuchsia_tester.dart still assumes Dart 1. Previously, it ran tests directly
from source, flutter_platform.dart automatically runs a kernel compile when
operating in Dart 2 mode, but this assumes a functional Dart SDK is available
in the artifacts directly, and fuchsia_tester.dart mocks out the artifacts
directory with an empty temp dir.
Remaining work is:
1. Get the frontend server building as a dependency on Fuchsia.
2. Patch fuchsia_tester.dart to use a valid Dart SDK and frontend server.
This also reverts migration to Dart 2 typedef syntax.
This reverts commit 6c56bb2. (#18362)
This reverts commit 3daebd0. (#18316)
* It's time to #deleteDart1 (#18293)
Eliminates support for Dart 1 in flutter_tools, and drops our Dart 1
benchmarks. All commands now run in Dart 1 mode only.
Eliminates --preview-dart-2 / --no-preview-dart-2 support.
* Fix indentation, remove no longer necessary .toList()
* Only push udpated kernel if >0 invalidated srcs
Eliminates support for Dart 1 in flutter_tools, and drops our Dart 1
benchmarks. All commands now run in Dart 1 mode only.
Eliminates --preview-dart-2 / --no-preview-dart-2 support.
* With dart being default, move preview_dart_2 tests to dart1.
* Swap out defaults for preview-dart-2 option in tests. Make dart1 explicit
* Move suffix around
* Rename files
* Roll engine to b6df7a637498ca9beda1fa9cd7210e3202ea599f.
Changes since last roll:
```
b6df7a637 Roll dart to 290c576264faa096a0b3206c71b2435309d9f904. (#4771)
a6764dbd5 Add sources for Fuchsia target. (#4763)
2d5900615 [fuchsia] Remove unused header file. (#4769)
9717063b7 Revert "Roll dart to c080951d45e79cd25df98036c4be835b284a269c. (#4767)" (#4768)
9a9814312 Roll dart to c080951d45e79cd25df98036c4be835b284a269c. (#4767)
e74e8b35c [async] Update includes of async headers to new path (#4760)
e2c4b2760 Use Dart 2 camel case constants in the engine Dart libraries (#4766)
9c1e48434 Updates for Fuchsia roll. (#4765)
14c940e27 Switch from fxl::Mutex to std::mutex (#4764)
debf82c0b Roll Garnet (#4759)
5bffdefbb Use weak pointers to the accesibility bridge from objects vended to the UIKit accessibility framework. (#4761)
```
* Reject requests for hot reload if a hot reload is already in progress.
Fixes#14184
* Implement TODO, verifying further hot reloads complete sucessfully.
* Fix year on new file.
* Add missing type annotations to fix lints
* Add run_machine_concurrent_hot_reload to manifest for CI
* Reformat document
... but undo things that cause lints (like single-line ifs)
* Extract std stream transformations
* Make inProgressHotReload private
* Disallow all types of reload while hot reload in progress
* Simplify code handling in-progress hot reloads
This fills out the following matrix:
App: `flutter create` default app, Complex Layout, and Gallery.
Platforms: Linux/Android, Mac/iOS, Windows/Android
Build modes: full release, aot blobs, debug
Dart: normal, 2.0 preview
It also renames all the tests that used "build" to mean compilation
build to use "compile" instead, so that it's much clearer on the
dashboard.
Also, mark some flaky tests that are no longer flaky as not flaky.
Also, remove uses of call() since that's probably going to go away at
some point.
* Clone hot reload benchmark for --preview-dart-2 option.
* Get rid of linux and win preview_dart_2 (only android would be sufficient for now). Refactor code into lib/tasks
* Revert 2016 to 2017
* Mark new test as flaky
* Revert "Revert "fire service protocol events for frames (#11565)" (#11727)"
This reverts commit f25e2f5213.
* move the postEvent() call into a separate method
- Wait for full Hot Reload
- Wait for full Restart
- Fallback if checkHealth throws METHOD_NOT_FOUND.
We try to wait for the service extensions to be registered and retry.
* fire service protocol extension events for frames
* start time in micros
* introduce a profile() function; only send frame events when in profile (or debug) modes
* moved the profile() function to foundation/profile.dart
* refactor to make the change more testable; test the change
* fire service protocol events by listening to onFrameInfo
* remove the frame event stream; add a devicelab test
* remove a todo
* final
* add Android instrumentation test
* add devicelab test
* add to manifest.yaml
* rename _smoke_test.dart to _smoketest.dart to prevent flutter test from picking it up
* volatile fields; style fixes
* use ConditionVariable; fix sh script
The main purpose of this PR is to make it so that when you set the
initial route and it's a hierarchical route (e.g. `/a/b/c`), it
implies multiple pushes, one for each step of the route (so in that
case, `/`, `/a`, `/a/b`, and `/a/b/c`, in that order). If any of those
routes don't exist, it falls back to '/'.
As part of doing that, I:
* Changed the default for MaterialApp.initialRoute to honor the
actual initial route.
* Added a MaterialApp.onUnknownRoute for handling bad routes.
* Added a feature to flutter_driver that allows the host test script
and the device test app to communicate.
* Added a test to make sure `flutter drive --route` works.
(Hopefully that will also prove `flutter run --route` works, though
this isn't testing the `flutter` tool's side of that. My main
concern is over whether the engine side works.)
* Fixed `flutter drive` to output the right target file name.
* Changed how the stocks app represents its data, so that we can
show a page for a stock before we know if it exists.
* Made it possible to show a stock page that doesn't exist. It shows
a progress indicator if we're loading the data, or else shows a
message saying it doesn't exist.
* Changed the pathing structure of routes in stocks to work more
sanely.
* Made search in the stocks app actually work (before it only worked
if we happened to accidentally trigger a rebuild). Added a test.
* Replaced some custom code in the stocks app with a BackButton.
* Added a "color" feature to BackButton to support the stocks use case.
* Spaced out the ErrorWidget text a bit more.
* Added `RouteSettings.copyWith`, which I ended up not using.
* Improved the error messages around routing.
While I was in some files I made a few formatting fixes, fixed some
code health issues, and also removed `flaky: true` from some devicelab
tests that have been stable for a while. Also added some documentation
here and there.
Adds an initial integration test that:
1. captures laid-out widget size
2. focuses a text field to show the keyboard and resize the view
3. unfocuses the text field to dismiss the keyboard and resize the view
4. verify that the final widget size matches the initial size
Turns out MacOS has a really low open files limit and so if you try to
open EVERY FILE AT THE SAME TIME it falls over.
This just opens the files one at a time, the way we used to back in
the old days.
