Reverts flutter/flutter#134031
context: b/301051367
Looked at the error message from the broken TAP target, but seems like the failure might be non trivial to resolve. Would it be okay if we revert this for now while it is being triaged?
This PR modifies the web build slightly to create an `AssetManifest.json`, that is a JSON(base64)-encoded version of the `AssetManifest.bin` file.
_(This should enable all browsers to download the file without any interference, and all servers to serve it with the correct headers.)_
It also modifies Flutter's `AssetManifest` class so it loads and uses said file `if (kIsWeb)`.
### Issues
* Fixes https://github.com/flutter/flutter/issues/124883
### Tests
* Unit tests added.
* Some tests that run on the Web needed to be informed of the new filename, but their behavior didn't have to change (binary contents are the same across all platforms).
* I've deployed a test app, so users affected by the BIN issue may take a look at the PR in action:
* https://dit-tests.web.app
## Description
This removes all of the comments that are of the form "so-and-so must not be null" or "so-and-so must be non-null" from the cases where those values are defines as non-nullable values.
This PR removes them from the library in the repo that don't have anything to do with the framework.
This was done by hand, since it really didn't lend itself to scripting, so it needs to be more than just spot-checked, I think. I was careful to leave any comment that referred to parameters that were nullable, but I may have missed some.
In addition to being no longer relevant after null safety has been made the default, these comments were largely fragile, in that it was easy for them to get out of date, and not be accurate anymore anyhow.
This did create a number of constructor comments which basically say "Creates a [Foo].", but I don't really know how to avoid that in a large scale change, since there's not much you can really say in a lot of cases. I think we might consider some leniency for constructors to the "Comment must be meaningful" style guidance (which we de facto have already, since there are a bunch of these).
## Related PRs
- https://github.com/flutter/flutter/pull/134984
- https://github.com/flutter/flutter/pull/134991
- https://github.com/flutter/flutter/pull/134992
- https://github.com/flutter/flutter/pull/134993
## Tests
- Documentation only change.
Support for FFI calls with `@Native external` functions through Native assets on Linux. This enables bundling native code without any build-system boilerplate code.
For more info see:
* https://github.com/flutter/flutter/issues/129757
### Implementation details for Linux.
Mainly follows the design of https://github.com/flutter/flutter/pull/130494.
Some differences are:
* Linux does not support cross compiling or compiling for multiple architectures, so this has not been implemented.
* Linux has no add2app.
The assets copying is done in the install-phase of the CMake build of a flutter app.
CMake requires the native assets folder to exist, so we create it also when the feature is disabled or there are no assets.
### Tests
This PR adds new tests to cover the various use cases.
* packages/flutter_tools/test/general.shard/linux/native_assets_test.dart
* Unit tests the Linux-specific part of building native assets.
It also extends various existing tests:
* packages/flutter_tools/test/integration.shard/native_assets_test.dart
* Runs (incl hot reload/hot restart), builds, builds frameworks for Linux and flutter-tester.
Speeds up the native assets target in the backend by
1. changing other targets `gen_dart_plugin_registrant` and
`release_unpack_ios` to do async I/O,
2. not reparsing the package config, and
3. not calling `dart pub deps --json` for 0 or 1 packages (fixed
package:native_assets_builder).
* https://github.com/flutter/flutter/issues/134427
```
[ +2 ms] native_assets: Starting due to {}
[ +2 ms] Skipping target: gen_localizations
[ +1 ms] gen_dart_plugin_registrant: Starting due to {InvalidatedReasonKind.inputChanged: The following inputs have updated contents: /Users/dacoharkes/flt/engine/flutter/examples/hello_world/.dart_tool/package_config_subset}
[ +33 ms] gen_dart_plugin_registrant: Complete
[ +107 ms] release_unpack_ios: Complete
[ +60 ms] Writing native_assets.yaml.
[ +7 ms] Writing /Users/dacoharkes/flt/engine/flutter/examples/hello_world/.dart_tool/flutter_build/be2692bbfbc0b9a27fcd2422d52354c6/native_assets.yaml done.
[ ] native_assets: Complete
```
->
```
[ +4 ms] native_assets: Starting due to {}
[ ] Skipping target: gen_localizations
[ +1 ms] gen_dart_plugin_registrant: Starting due to {InvalidatedReasonKind.inputChanged: The following inputs have updated contents: /Users/dacoharkes/flt/engine/flutter/examples/hello_world/.dart_tool/package_config_subset}
[ +31 ms] Writing native_assets.yaml.
[ +8 ms] Writing /Users/dacoharkes/flt/engine/flutter/examples/hello_world/.dart_tool/flutter_build/f9451a65a465bfab70d004e21d6cc1d6/native_assets.yaml done.
[ +1 ms] native_assets: Complete
```
## Pre-launch Checklist
- [x] I read the [Contributor Guide] and followed the process outlined
there for submitting PRs.
- [x] I read the [Tree Hygiene] wiki page, which explains my
responsibilities.
- [x] I read and followed the [Flutter Style Guide], including [Features
we expect every widget to implement].
- [x] I signed the [CLA].
- [x] I listed at least one issue that this PR fixes in the description
above.
- [x] I updated/added relevant documentation (doc comments with `///`).
- [ ] I added new tests to check the change I am making, or this PR is
[test-exempt].
- [x] All existing and new tests are passing.
<!-- Links -->
[Contributor Guide]:
https://github.com/flutter/flutter/wiki/Tree-hygiene#overview
[Tree Hygiene]: https://github.com/flutter/flutter/wiki/Tree-hygiene
[test-exempt]:
https://github.com/flutter/flutter/wiki/Tree-hygiene#tests
[Flutter Style Guide]:
https://github.com/flutter/flutter/wiki/Style-guide-for-Flutter-repo
[Features we expect every widget to implement]:
https://github.com/flutter/flutter/wiki/Style-guide-for-Flutter-repo#features-we-expect-every-widget-to-implement
[CLA]: https://cla.developers.google.com/
[flutter/tests]: https://github.com/flutter/tests
[breaking change policy]:
https://github.com/flutter/flutter/wiki/Tree-hygiene#handling-breaking-changes
[Discord]: https://github.com/flutter/flutter/wiki/Chat
Fixes https://github.com/flutter/flutter/issues/134566.
Prior to this fix, `ShutdownHooks` were run in the private helper
function `_exit()` defined in the `package:flutter_tools/runner.dart`
library. Independent of this, the tool had signal handling logic that
traps SIGINT and SIGTERM. However, these handlers called `exit()` from
`dart:io`, and didn't run these hooks.
This PR moves the `_exit()` private helper to
`package:flutter_tools/src/base/process.dart` and renames it to
`exitWithHooks()`, so that it can be called by the signal handlers in
`package:flutter_tools/src/base/signals.dart`.
Xcode uses the CONFIGURATION_BUILD_DIR build setting to determine the location of the bundle to build and install. When launching an app via Xcode with the Xcode debug workflow (for iOS 17 physical devices), temporarily set the CONFIGURATION_BUILD_DIR to the location of the bundle so Xcode can find it.
Also, added a Xcode Debug version of the `microbenchmarks_ios` integration test since it uses `flutter run --profile` without using `--use-application-binary`.
Fixes https://github.com/flutter/flutter/issues/134186.
Fixes https://github.com/flutter/flutter/issues/133404.
Per the dart2js team on the linked issue, `-O0` is not intended for end users, but more for actual debugging/development of the compiler.
Support for FFI calls with `@Native external` functions through Native assets on MacOS and iOS. This enables bundling native code without any build-system boilerplate code.
For more info see:
* https://github.com/flutter/flutter/issues/129757
### Implementation details for MacOS and iOS.
Dylibs are bundled by (1) making them fat binaries if multiple architectures are targeted, (2) code signing these, and (3) copying them to the frameworks folder. These steps are done manual rather than via CocoaPods. CocoaPods would have done the same steps, but (a) needs the dylibs to be there before the `xcodebuild` invocation (we could trick it, by having a minimal dylib in the place and replace it during the build process, that works), and (b) can't deal with having no dylibs to be bundled (we'd have to bundle a dummy dylib or include some dummy C code in the build file).
The dylibs are build as a new target inside flutter assemble, as that is the moment we know what build-mode and architecture to target.
The mapping from asset id to dylib-path is passed in to every kernel compilation path. The interesting case is hot-restart where the initial kernel file is compiled by the "inner" flutter assemble, while after hot restart the "outer" flutter run compiled kernel file is pushed to the device. Both kernel files need to contain the mapping. The "inner" flutter assemble gets its mapping from the NativeAssets target which builds the native assets. The "outer" flutter run get its mapping from a dry-run invocation. Since this hot restart can be used for multiple target devices (`flutter run -d all`) it contains the mapping for all known targets.
### Example vs template
The PR includes a new template that uses the new native assets in a package and has an app importing that. Separate discussion in: https://github.com/flutter/flutter/issues/131209.
### Tests
This PR adds new tests to cover the various use cases.
* dev/devicelab/bin/tasks/native_assets_ios.dart
* Runs an example app with native assets in all build modes, doing hot reload and hot restart in debug mode.
* dev/devicelab/bin/tasks/native_assets_ios_simulator.dart
* Runs an example app with native assets, doing hot reload and hot restart.
* packages/flutter_tools/test/integration.shard/native_assets_test.dart
* Runs (incl hot reload/hot restart), builds, builds frameworks for iOS, MacOS and flutter-tester.
* packages/flutter_tools/test/general.shard/build_system/targets/native_assets_test.dart
* Unit tests the new Target in the backend.
* packages/flutter_tools/test/general.shard/ios/native_assets_test.dart
* packages/flutter_tools/test/general.shard/macos/native_assets_test.dart
* Unit tests the native assets being packaged on a iOS/MacOS build.
It also extends various existing tests:
* dev/devicelab/bin/tasks/module_test_ios.dart
* Exercises the add2app scenario.
* packages/flutter_tools/test/general.shard/features_test.dart
* Unit test the new feature flag.
Sometimes `ios-deploy` loses connection to the device after installing, starting debugserver, and launching. This is shown with an error message like:
```
Process 579 exited with status = -1 (0xffffffff) lost connection
```
This happens frequently in our CI system: https://github.com/flutter/flutter/issues/120808
Usually in CI, on retry it'll work and pass - so this is an attempt to retry without failing the test first. It's not guaranteed to fix since we're unable to recreate this error locally.
â¦on file
The deeplink validation tool will become an static app so it can't no longer access vm services.
The goal will be then to turn them into flutter analyze command similar to `flutter analyze --android --[options]` that static app can use on.
This pr only removes vm services and turn the api to dump a output file instead of printing everything to stdout.
Since applink validation tool is going to be a static tool, It won't have access to vmservices.
[flutter.dev/go/static-tooling-in-devtools](http://flutter.dev/go/static-tooling-in-devtools)
I remove the vm services and also update the deeplink task to also include path pattern and custom scheme
http://go/android-applink-apis (internal only)
Fixes https://github.com/flutter/flutter/issues/133093
When I introduced the new, more robust version file `//flutter/bin/cache/version.json` in https://github.com/flutter/flutter/pull/124558, I changed `class FlutterVersion` into an abstract interface, implemented by `_FlutterVersionFromGit` (which is essentially the previous behavior) and `_FlutterVersionFromFile`, which merely reads the data it would have computed via git from `//flutter/bin/cache/version.json`.
While doing this, I made `_FlutterVersionFromGit.ensureVersionFile()` to be a no-op, since I assumed this would not be necessary since we already had a version file in the cache. However, this method was what was previously responsible for ensuring `//flutter/version` existed on disk. This means that if, for whatever reason, the user had `//flutter/bin/cache/flutter.version.json` present but NOT `//flutter/version`, the tool would have never created that file, and they would hit the tool crash seen in https://github.com/flutter/flutter/issues/133093.
This fixes the tool by ensuring `//flutter/version` exists regardless of if we're hydrating `FlutterVersion` from `//flutter/bin/cache/flutter.version.json` or not.
This makes two changes to prepare for incoming changes to skwasm in the web engine:
* We will (at least for now) be depending on the `WebAssembly.Function` constructor in `skwasm`, which is hidden behind the `--experimental-wasm-type-reflection` flag. We need to pass that when running skwasm benchmarks.
* We are going to be upgrading the skwasm build to a newer version of emscripten, which exposes the wasm exports via the `wasmExports` property instead of the `asm` property. Make sure to support either, if passed.
1. Remove vm service registration
2. combine print<variant>ApplicationId and print<variant>AppLinkDomain into one task dump<variant>AppLinkSettings, which dump all the data in a json file
The deeplink validation tool will be a static app in devtool instead of regular app. A Static app doesn't require a running app; therefore, we can't call these API through vmservices. I decided to convert these API into flutter analyzer command, which will be done in a separate PR https://github.com/flutter/flutter/pull/131009.
The reason these print tasks are converted into file dumps is to reduce the amount of data encoding and decoding. Instead of passing data through stdout, the devtool can read the files generated by gradle tasks instead.
Starting in Xcode 15, when building macOS, DT_TOOLCHAIN_DIR cannot be used to evaluate LD_RUNPATH_SEARCH_PATHS or LIBRARY_SEARCH_PATHS. `xcodebuild` error message recommend using TOOLCHAIN_DIR instead.
Since Xcode 15 isn't in CI, I tested it in a one-off `led` test:
* [Pre-fix failure](04e485a0b1/+/build.proto)
* [Post-fix success](d454a3e181/+/build.proto)
Fixes https://github.com/flutter/flutter/issues/132755.
Currently, flutter pub get generates localizations if there exists an l10n.yaml file where synthetic-package is not false. However, for any user who needs to turn off synthetic-package, their localizations are not generated. This PR should make the behavior more consistent. (Also it seems good to make it so that running flutter pub get once resolves all the dependencies so that people can get to work without running flutter gen-l10n manually.)
Fixes https://github.com/flutter/flutter/issues/84979.
This PR adjusts the quoting of the project description not not apply twice in the index.html / manifest.json of web builds.
*List which issues are fixed by this PR. You must list at least one issue.*
Fixes https://github.com/flutter/flutter/issues/131834
*If you had to change anything in the [flutter/tests] repo, include a link to the migration guide as per the [breaking change policy].*
Partial work towards https://github.com/flutter/flutter/issues/132245.
The goal here is to "sniff" out any missing pieces that would block engine builds, rolls, benchmarks and so on before requiring humans to provide the parameter. The implementation is based on a [short discussion with @christopherfujino](https://discord.com/channels/608014603317936148/608022056616853515/1141503921546875110):
@matanlurey:
> Not sure whether to post here or â hackers-infra-ð¡ , but is there a way to (and is it advisable to) detect whether the tool is running in a CI environment? I'd like to "soft enforce" --local-engine-host being provided strictly on CI, make sure that lands well, and then "upgrade" it to being non-CI invocations as well (re: https://github.com/flutter/flutter/issues/132245).
>
> Also happy to get talked out of this idea ð
@christopherfujino:
> we have a check, lemme find it
> whether or not it is advisable, idk
> https://github.com/flutter/flutter/blob/flutter-3.14-candidate.0/packages/flutter_tools/lib/src/base/bot_detector.dart#L30
>
> (...)
>
> is your desire to get early signal before enforcing t his for humans to prevent functionality churn of landing and reverting and re-landing?
>
> (yes)
>
> uhh, sure, that's advisable ð
Partial work towards https://github.com/flutter/flutter/issues/132245.
I also couldn't help myself to do a very minor refactor and add some comments to `LocalEngineInfo` because I was getting confused myself implementing it.
Partial work towards https://github.com/flutter/flutter/issues/132245.
I made a minor refactor to test-only code because it was too confusing
to have 2 optional parameters that are technically required together,
but otherwise all other changes *should* be pass throughs. That being
said, I can't say I totally understand the Gradle stuff so I could use a
hand double checking that.
At some point in the near future, we'll start an Impeller on Android preivew - and later make it the default. Unlike Skia, Impeller does not have a fallback software rendering mode. We'd like to stop suggesting this as an option now, and in the future remove the option to force software rendering.
Once impeller is the default, asking for software rendering on Android will result in either an error or falling back to Skia.
Fixes a couple of issues introduced in new iOS 17 physical device tooling: https://github.com/flutter/flutter/pull/131865.
1) Duplicate messages were being filtered out too aggressively.
For example, if on the counter app, you printed "Increment!" on button click, it would only print once no matter how many times you clicked.
Sometimes more than one log source is used at a time and the original intention was to filter duplicates between two log sources, so it wouldn't print the same message from both logs. However, it would also filter when the same message was added more than once via the same log.
The new solution distinguishes a "primary" and a "fallback" log source and prefers to use the primary source unless it's not working, in which it'll use the fallback. If the fallback is faster than the primary, the primary will exclude the logs received by the fallback in a 1-to-1 fashion to prevent too-aggressive filtering. Once a flutter-message has been received by the primary source, fallback messages will be ignored.
Note: iOS < 17 did not regress.
2) There was a race condition between the shutdown hooks and exiting XcodeDebug that was causing a crash when deleting a file that doesn't exist. This only affects CI - for the new integration tests and when testing with iOS 17 physical devices.
Towards https://github.com/dart-lang/test/issues/2065
The flutter test runner uses the copy of `host.dart.js` from the copy of
`package:test` that surfaces in the pub solve for `flutter_tool`. This
copy has been updated to allow either the old pattern of communication,
or this new pattern. The new pattern removes an extra hop and use of the
frame `window.onMessage` messages.
Implement expected functionalities when supplying `--web-launch-url` and/or `--web-hostname` arguments to `flutter drive`.
- `--web-launch-url` now sets the starting url for the (headless) browser
- Which for example means you can start at a certain part of the app at the start of your integration test
- `--web-hostname` now sets the hostname where the target of flutter drive will be hosted
- Which allows you to set something other than localhost (allowing access via a reverse-proxy for example)
Fixes#118028
Fixes: https://github.com/flutter/flutter/issues/124970
Part of https://github.com/flutter/flutter/issues/47161
Before this change, there were two places we overrode the `Artifacts` in a Zone:
1. if/when we parse local-engine CLI options: 1cf3907407/packages/flutter_tools/lib/src/runner/flutter_command_runner.dart (L281)
2. an additional override for fuchsia platform dill (no longer used, deleted in this PR): 1cf3907407/packages/flutter_tools/lib/src/commands/attach.dart (L274)
Note 1 above creates a new instance of `Artifacts.getLocalEngine()`. In this flow, there exist two instances of `Artifacts`:
1. The default fallback instance of `CachedArtifacts` (which gets all artifacts from flutter/bin/cache), instantiated in context_runner.dart: 1cf3907407/packages/flutter_tools/lib/src/context_runner.dart (L137)
2. An instance of `CachedLocalEngineArtifacts` created in the command runner once the CLI options have been parsed: 1cf3907407/packages/flutter_tools/lib/src/runner/flutter_command_runner.dart (L281)
The regression happened when we direct injected the Artifacts 1 from above BEFORE we parsed the local-engine flag, and then used this in the second zone override, and then when creating the `FlutterDevice` there are multiple calls to `globals.artifacts` returned it when it should have returned Artifacts 2: 1cf3907407/packages/flutter_tools/lib/src/resident_runner.dart (L80)
Device.artifactOverrides was originally introduced in https://github.com/flutter/flutter/pull/32071, but is no longer used, so I deleted it.
I also removed direct injection of `Artifacts` to the attach sub-command, because that class now no longer references artifacts.
I believe the ideal true fix for this would be to:
1. Migrate all leaf calls to `globals.artifacts` to use direct injection (in this case, the offending invocations were in [`FlutterDevice.create()`](1cf3907407/packages/flutter_tools/lib/src/resident_runner.dart (L80-L218)), but I'm not sure that something else would not have broken later)
2. Ensure we are always direct injecting the desired instance of `Artifacts`--that is, if the user desires local engine artifacts, that we are passing an instance of `CachedLocalEngineArtifacts`.
a. Alternatively, and probably simpler, teach `CachedArtifacts` to know about the local engine. This would mean parsing the global CLI options BEFORE we ever construct any instance of `Artifacts`.
As an overall recommendation for implementing https://github.com/flutter/flutter/issues/47161, in the overall tree of tool function calls, we should probably migrate the leaves first (that is, migrate the sub-commands last). We should also audit and reconsider any usage of `runZoned()` or `context.run()` for the purpose overriding zoneValues.
Closes https://github.com/flutter/flutter/issues/132162.
I did a tiny bit of minor cleanup but didn't want to go 🍔 beyond the
scope of this change. After it lands I'll update the Wiki.
## Before
```bash
$ fl run \
--local-engine-src-path=$ENGINE \
--local-engine=android_debug_unopt_arm64
```
... would try to use `host_debug_unopt` (i.e. Rosetta).
## After
```bash
$ fl run \
--local-engine-src-path=$ENGINE \
--local-engine=android_debug_unopt_arm64 \
--local-engine-host=host_debug_unopt_arm64
```
... uses `host_debug_unopt_arm64`, as specified.
---
/cc @jonahwilliams @gaaclarke @zanderso
