Fixes https://github.com/flutter/flutter/issues/127090.
https://github.com/flutter/flutter/pull/122505 did a few things to speed up the first asset load that a flutter app performs. One of those things was to not include the main asset in its own list of variants in the asset manifest. The idea was that we know that the main asset always exists, so including it in its list of variants is a waste of storage space and loading time (even if the cost was tiny).
However, the assumption that the main asset always exists is wrong. From [Declaring resolution-aware image assets](https://docs.flutter.dev/ui/assets-and-images#resolution-aware), which predates https://github.com/flutter/flutter/pull/122505:
> Each entry in the asset section of the pubspec.yaml should correspond to a real file, with the exception of the main asset entry. If the main asset entry doesnât correspond to a real file, then the asset with the lowest resolution is used as the fallback for devices with device pixel ratios below that resolution. The entry should still be included in the pubspec.yaml manifest, however.
For example, it's valid to declare `assets/image.png` as an asset even if only `assets/3x/image.png` exists on disk.
This fix restores older behavior of including a main asset as a variant of itself in the manifest if it exists.
This fix also includes a non-user-visible behavior change:
* `"dpr"` is no longer a required field in the asset manifest's underlying structure. For the main asset entry, we do not include `"dpr"`. It makes less sense for the tool to decide what the default target dpr for an image should be. This should be left to the framework.
Always pass the space code point (`0x20`) to the font subsetter when targetting web because the web engine relies on that character to calculate a font's height.
Fixes#127270
Reverts flutter/flutter#126698
There are a bunch of tool crashes on CI that start with this commit. I'm
not sure this PR is the cause because there is no backtrace from the
tool on the crashes. The only error message is `Oops; flutter has exited
unexpectedly: "Null check operator used on a null value`.
Service extensions are unable to handle requests when the isolate they were registered on is paused. The DevTools launcher logic was waiting for some service extension invocations to complete before advertising the already active DevTools instance, but when --start-paused was provided these requests would never complete, preventing users from using DevTools to resume the paused isolate.
Fixes https://github.com/flutter/flutter/issues/126691
Fixes#122081.
When validating an Android Studio installation, add a warning validation message when we are unable to detect the version. This is because we have logic throughout the tool (JDK/JRE-searching) that is at higher risk of failing when we don't know the version.
Fixes#124252, finishing work on the umbrella tracking issue, #126126.
Essentially, after this PR, no (non-test) code should be be referencing/invoking the java home or binary paths.
Replace imports of `src/remote_listener.dart` with `backend.dart` which exports `RemoteListener`.
Remove the unused imports of `src/backend/stack_trace_formatter.dart` (no uses of `StackTraceFormatter`) and
`src/backend/suite_channel_manager.dart` (no uses of `SuiteChannelManager`).
This reverts commit a19b3436eead2e64bb2b014f14afcd8fab8c9d32.
We added this logging to try and determine if the reason for Dart VM errors (https://github.com/flutter/flutter/issues/121231) was caused by some issue with the streams.
A recent test proves that is not the case:
https://ci.chromium.org/ui/p/flutter/builders/prod/Mac_ios%20platform_view_ios__start_up/11046/overview
The test shows the Dart VM url in the device log. However, the test log does NOT show a log for the Dart VM url but does show the stack trace, which all come from the main stream, which means it's not an issue with the secondary streams not receiving the log.
So reverting the debugging we added.
## How we determine the channel name
Historically, we used the current branch's upstream to figure out the current channel name. I have no idea why. I traced it back to https://github.com/flutter/flutter/pull/446/files where @abarth implement this and I reviewed that PR and left no comment on it at the time.
I think this is confusing. You can be on a branch and it tells you that your channel is different. That seems weird.
This PR changes the logic to uses the current branch as the channel name.
## How we display channels
The main reason this PR exists is to add channel descriptions to the `flutter channel` list:
```
ianh@burmese:~/dev/flutter/packages/flutter_tools$ flutter channel
Flutter channels:
master (tip of tree, for contributors)
main (tip of tree, follows master channel)
beta (updated monthly, recommended for experienced users)
stable (updated quarterly, for new users and for production app releases)
* foo_bar
Currently not on an official channel.
ianh@burmese:~/dev/flutter/packages/flutter_tools$
```
## Other changes
I made a few other changes while I was at it:
* If you're not on an official channel, we used to imply `--show-all`, but now we don't, we just show the official channels plus yours. This avoids flooding the screen in the case the user is on a weird channel and just wants to know what channel they're on.
* I made the tool more consistent about how it handles unofficial branches. Now it's always `[user branch]`.
* I slightly adjusted how unknown versions are rendered so it's clearer the version is unknown rather than just having the word "Unknown" floating in the output without context.
* Simplified some of the code.
* Made some of the tests more strict (checking all output rather than just some aspects of it).
* Changed the MockFlutterVersion to implement the FlutterVersion API more strictly.
* I made sure we escape the output to `.metadata` to avoid potential injection bugs (previously we just inlined the version and channel name verbatim with no escaping, which is super sketchy).
* Tweaked the help text for the `downgrade` command to be clearer.
* Removed some misleading text in some error messages.
* Made the `.metadata` generator consistent with the template file.
* Removed some obsolete code to do with the `dev` branch.
## Reviewer notes
I'm worried that there are implications to some of these changes that I am not aware of, so please don't assume I know what I'm doing when reviewing this code. :-)
Log all output of `ios-deploy` to try and determine if the issue of https://github.com/flutter/flutter/issues/121231 is with stream or with `ios-deploy`.
Note: This will cause some duplicate logs like example below but only in verbose mode
```
(lldb) 2023-05-19 13:48:19.107935-0500 Runner[2521:390363] [VERBOSE-2:FlutterDarwinContextMetalImpeller.mm(35)] Using the Impeller rendering backend.
(lldb) 2023-05-19 13:48:19.107935-0500 Runner[2521:390363] [VERBOSE-2:FlutterDarwinContextMetalImpeller.mm(35)] Using the Impeller rendering backend.
2023-05-19 13:48:19.156866-0500 Runner[2521:390612] flutter: The Dart VM service is listening on http://127.0.0.1:63508/IsFnhXJykCM=/
VM Service URL on device: http://127.0.0.1:63508/IsFnhXJykCM=/
```
## Background
The Windows runner has a race at startup:
1. **Platform thread**: creates a hidden window
2. **Platform thread**: launches the Flutter engine
3. **UI/Raster threads**: renders the first frame
4. **Platform thread**: Registers a callback to show the window once the next frame has been rendered.
Steps 3 and 4 happen in parallel and it is possible for step 3 to complete before step 4 starts. In this scenario, the next frame callback is never called and the window is never shown.
As a result the `windows_startup_test`'s test, which [verifies that the "show window" callback is called](1f09a8662d/dev/integration_tests/windows_startup_test/windows/runner/flutter_window.cpp (L60-L64)), can flake if the first frame is rendered before the show window callback has been registered.
## Solution
This change makes the runner schedule a frame after it registers the next frame callback. If step 3 hasn't completed yet, this no-ops as a frame is already scheduled. If step 3 has already completed, a new frame will be rendered, which will call the next frame callback and show the window.
Part of https://github.com/flutter/flutter/issues/119415
See this thread for alternatives that were considered: https://github.com/flutter/engine/pull/42061#issuecomment-1550080722
xcodebuild command generates a xcresult bundle file on each run, however, it doesn't delete the file generated from previous run and will throw an error if the exact file already exists.
In tool, we manually delete the file after each `flutter build` or `flutter run` command. However, there are some internal logic where xcodebuild retries multiple times.
This PR deletes the xcresult bundle file at the start of each retry if it exists.
Fixes https://github.com/flutter/flutter/issues/127119
Check what is available in the device's iOS DeviceSupport folder to check if symbols were properly fetched. Also, add some logging to track what status the debugger is in.
Debugging for https://github.com/flutter/flutter/issues/121231.
Reverts flutter/flutter#126086.
This PR changed the interfaces of some classes, namely `AndroidSdk`, and deleted a global.
These classes had custom overrides in g3 that were not updated with a g3fix, so this PR has broken tests. See https://b.corp.google.com/issues/281945232 (non-public link)
Sync lints with https://github.com/dart-lang/linter/blob/master/example/all.yaml and enable `implicit_reopen` and `type_literal_in_constant_pattern` (which have no violations). Also contains some clean-up work towards enabling `matching_super_parameters`, which is not quite ready yet due to its handling of "private" arguments.
This is the first step in unifying Java-finding logic across the tool. If curious, see #126029 for an example of what all the changes will probably entail.
Moves java-related functionality like `AndroidSdk.findJavaHome` to a new class, `Java`.
See tracking issue https://github.com/flutter/flutter/issues/126126 for more.
This change makes `flutter pub` show the help usage:
```
PS > flutter pub
Commands for managing Flutter packages.
Global options:
-h, --help Print this usage information.
-v, --verbose Noisy logging, including all shell commands executed.
If used with "--help", shows hidden options. If used with "flutter
doctor", shows additional diagnostic information. (Use "-vv" to force
verbose logging in those cases.)
-d, --device-id Target device id or name (prefixes allowed).
--version Reports the version of this tool.
--suppress-analytics Suppress analytics reporting for the current CLI invocation.
--disable-telemetry Disable telemetry reporting when this command runs.
Usage: flutter pub <subcommand> [arguments]
-h, --help Print this usage information.
Available subcommands:
add Add a dependency to pubspec.yaml.
cache Work with the Pub system cache.
deps Print package dependencies.
downgrade Downgrade packages in a Flutter project.
get Get the current package's dependencies.
global Work with Pub global packages.
login Log into pub.dev.
logout Log out of pub.dev.
outdated Analyze dependencies to find which ones can be upgraded.
pub Pass the remaining arguments to Dart's "pub" tool.
publish Publish the current package to pub.dartlang.org.
remove Removes a dependency from the current package.
run Run an executable from a package.
test Run the "test" package.
token Manage authentication tokens for hosted pub repositories.
upgrade Upgrade the current package's dependencies to latest versions.
uploader Manage uploaders for a package on pub.dev.
version Print Pub version.
Run "flutter help" to see global options.
```
Previously it showed an error message:
```
PS > flutter pub
Missing subcommand for "flutter pub".
Run 'flutter -h' (or 'flutter <command> -h') for available flutter commands and options.
```
Addresses https://github.com/flutter/flutter/issues/110025
This should fix https://github.com/flutter/flutter/issues/126178
When we don't pass a `--concurrency` flag to the test package, it uses a default based on the number of cores that are on the machine. However, the web test platform itself serializes all these requests anyway, which can lead to the test package timing out. This is because from the test package's perspective, it has already started the loading process on a number of suites which are simply waiting for other test suites to compile and run. The ones that wait the longest can run up against the test packages 12 minute timeout for loading a given suite, even though they haven't actually started to try to load.
Instead, we should always pass `--concurrency=1` to the test package so that it doesn't attempt to start loads concurrently in the first place.
Fixes:
- https://github.com/flutter/flutter/issues/124411
This PR is cleaning up the `--disable-telemetry` help message to make it clear that opting out will opt out of all telemetry collection for flutter and dart commands. It is also adding the opposite flag `--enable-telemetry` which will enable telemetry collection
*Replace this paragraph with a description of what this PR is changing or adding, and why. Consider including before/after screenshots.*
*List which issues are fixed by this PR. You must list at least one issue.*
*If you had to change anything in the [flutter/tests] repo, include a link to the migration guide as per the [breaking change policy].*
