The libimobiledevice suite of tools do not include version information.
A simple way to verify they meet our version requirements is to run
idevice_id -l, which will fail when older versions are run against iOS
devices with newer versions of iOS installed.
Unfortunately, idevice_id -l will also fail when libimobiledevice is up
to date, but the attached devices have never been paired with the host
machine in Xcode.
This patch updates the error message to help guide the user in such
situations.
On iOS 11.0.x ideviceinfo fails with the following message
ERROR: Could not connect to lockdownd, error code -3
This workaround should also work for #12330
1. Migrate simulator device log tailing to os_log toolchain
2. When the log tag (component) is available (iOS 11/Xcode 9), filter to
the set of log lines with tag 'Flutter'.
As of iOS 11 / Xcode 9, Flutter engine logs are no longer recorded in the
simulator's syslog file, which we previously read using tail -f. Instead
they're now accessible through Apple's new macOS/iOS os_log facility,
via /usr/bin/log, which supports a relatively flexible query language.
When run in non-interactive mode, /usr/bin/log buffers its output in 4k
chunks, which is significantly smaller than what's emitted up to the
point where the observatory/diagnostics port information is logged. As a
workaround we force it to run in interactive mode via the script tool.
* Add --trace-skia parameter to flutter run
Skia tracing is extremely useful for internal debug, but reduces the
amount of space available in the Dart Timeline buffers.
Disable skia tracing by default and expose them via the --trace-skia
flag.
* Roll Engine to 57a1445a45964d386500c39f5e8d06db060abadb
This was introduced to suppress libMobileGestalt noise originating from
libsystem_asl.dylib. Commit 39680ebfbd
suppresses all application log messages not originating from the
app/engine iteself on iOS 10 and above. Since the log message in
question is only emitted on devices running iOS >= 10.3.0, this
blacklist no longer necessary.
On iOS 10 and above, suppress engine log messages from system components
other than Flutter. This eliminates a large amount of keyboard/plugin
related noise during edit-refresh development.
ios-deploy 1.9.2 includes fixes for a common source of Xcode breakage
(flutter/flutter#4326) with Xcode 8.3.3 + iOS 10.3.3, and is required to
to support Xcode 9 (flutter/flutter#11875).
Opening Xcode is no longer sufficient to enable develop mode in Xcode 9.
Update the message to run the command-line tool. Alternatively users can
launch an app in the Xcode debugger to do this.
This patch migrates iOS device listing from using Xcode instruments to
using the libimobiledevice tools idevice_id and ideviceinfo.
ideviceinfo was previously incompatible with iOS 11 physical devices;
this has now been fixed.
In 37bb5f1300 flutter_tools migrated from
libimobiledevice-based device listing on iOS to using Xcode instruments
to work around the lack of support for iOS 11. Using instruments entails
several downsides, including a significantly higher performance hit, and
leaking hung DTServiceHub processes in certain cases when a simulator is
running, necessitating workarounds in which we watched for, and cleaned
up leaked DTServiceHub processes. This patch returns reverts the move to
instruments now that it's no longer necessary.
This patch supports basic filtering of log lines from physical iOS
devices, similar to existing functionality for iOS simulator logging.
This patch also suppresses the following two log messages which are
emitted at app startup on iOS 10.3 devices:
libMobileGestalt MobileGestaltSupport.m:153: pid 123 (Runner) does not have sandbox access for frZQaeyWLUvLjeuEK43hmg and IS NOT appropriately entitled
libMobileGestalt MobileGestalt.c:550: no access to InverseDeviceID (see <rdar://problem/11744455>)
In some cases, we've seen interactions between Instruments and the iOS
simulator that cause hung instruments and DTServiceHub processes. If
enough instances pile up, the host machine eventually becomes
unresponsive.
Until the underlying issue is resolved, manually kill any orphaned
instances (where the parent process has died and PPID is 1) before
launching another instruments run.
Previously, xcodeMajorVersion and xcodeMinorVersion returned null unless
xcodeVersionSatisfactory had been called first. We now compute them on
demand, and cache the resultant values.
This allows us to take advantage of improved command-line tooling (e.g.,
improvements in device listing in Instruments). Now that the engine is
built with Xcode 8 and the framework is tested against Xcode 8, this
reduces the set of configurations we need to support to allow us to
focus on the supported ones: Xcode 8 and Xcode 9.
This reverts commit b2909a245a.
This resubmits the following patches:
1. Use Xcode instruments to list devices (#10801)
Eliminates the dependency on idevice_id from libimobiledevice. Instead,
uses Xcode built-in functionality.
2. Make device discovery asynchronous (#10803)
Migrates DeviceDiscovery.devices and all device-specific lookup to be
asynchronous.
* Revert "Make device discovery asynchronous (#10803)"
This reverts commit 972be9c8b4.
* Revert "Use Xcode instruments to list devices (#10801)"
This reverts commit 37bb5f1300.
This is to resolve a failure that looks related to a bad install of Xcode 8.0
on our build bots and should be reinstated when the infra issue is diagnosed
and resolved.
Instruments worked well when this was originally landed, and on the
following commit, but started failing two commits after this originally
landed. Manual invocation of instruments on the build host currently
results in:
```
dyld: Library not loaded: @rpath/InstrumentsAnalysisCore.framework/Versions/A/InstrumentsAnalysisCore
Referenced from: /Applications/Xcode8.0.app/Contents/Developer/usr/bin/instruments
Reason: image not found
Abort trap: 6
```
It appears the /Applications/Xcode8.0.app/Contents/Applications
directory (which contains Instruments) is missing on the host.
Moves all remaining calls to tools that are part of the libimobiledevice
suite of tools to the IMobileDevice class. This allows for better
tracking of this dependency, and easier mocking in tests.
Use a top-level getter in mac.dart rather than a static instance getter
and a top-level getter in ios_workflow.dart. Makes this code consistent
with how we do context lookups elsewhere.
Extract out IMobileDevice class, move class to idevice_id, ideviceinfo
(and eventually other libimobiledevice tools such as iproxy) behind this
interface.
Add tests for the case where libimobiledevice is not installed, the case
where it returns no devices, and the case where it returns device IDs.
Eliminates the need for the device/daemon code to get at the iOS/Android
tooling indirectly via Doctor. In tests, we now inject the workflow
objects (or mocks) directly.
This code is unused in any test. In upcoming changes that migrate to
Xcode instruments based device listing, we'll mock out the instruments
output separately.
* Before tests
* Add the part to trust the cert on the device
* flip the error checks since some are more specific and are more actionable
* add tests
* review
Eliminates nearly-duplicate install instructions for libimobiledevice,
ideviceinstaller.
Since ideviceinstaller depends on libimobiledevice, it's almost certain
that if libimobiledevice isn't installed, or needs updating, so does
ideviceinstaller.
This message will be emitted both when libimobiledevice requires
updating, or when it has not yet been installed.
It's also not specifically the version of Xcode that it's incompatible
with, it's the lockdownd daemon, which is actually more closely tied to
iTunes.
* first pass
* improvements
* extract terminal.dart
* rebase
* add default terminal to context
* The analyzer wants the ../ imports in front of the ./ imports
* review notes
For some reaosn, when we discovered our URI, we were re-instantiating
the `Completer` instance variable whose future we listen to in `nextUri()`.
This led to a race between a caller calling `nextUri()` and us discovering
the URI. If we happened to discover our URI before a caller called
`nextUri()`, then they would be left waiting on a future from the newly
allocated `Completer` (which would never complete).
Fixes#10064
* blind wrote everything except the user prompt
* works
* Add some logical refinements
* Make certificates unique and add more instructinos
* print more info
* Add test
* use string is empty
* review notes
* some formatting around commands
* add a newline
`adb` can sometimes hang, which will in turn hang the Dart isolate if
we're using `Process.runSync()`. This changes many of the `Device` methods
to return `Future<T>` in order to allow them to use the async process
methods. A future change will add timeouts to the associated calls so
that we can properly alert the user to the hung `adb` process.
This is work towards #7102, #9567
* Fix tests to use Ahem, and helpful changes around that
- Fix fonts that had metric-specific behaviours.
- LiveTestWidgetsFlutterBinding.allowAllFrames has been renamed
to LiveTestWidgetsFlutterBinding.framePolicy.
- LiveTestWidgetsFlutterBinding now defaults to using a frame policy
that pumps slightly more frames, to animate the pointer crosshairs.
- Added "flutter run --use-test-fonts" to enable Ahem on devices.
- Changed how idle() works to be more effective in live mode.
- Display the test name in live mode (unless ahem fonts are enabled).
- Added a toString to TextSelectionPoint.
- Style nit fixes.
* Roll engine to get Ahem changes.
* Update tests for dartdoc changes.
* Fix flutter_tools tests
Added a PluginRegistry to the new project template. The registry files will be automatically updated at build time to register the native plugins.
Fixes#7814.
This yak shave went as follows:
Fix https://github.com/flutter/flutter/issues/8795 by adding stocks to
the examples README.
Notice the layers entry in that README isn't quite right either.
Update that.
Check the layers/README file is worth pointing at.
Update the layers/README.
Let's run some of the layer tests to see if they still work.
Oops, need to update them to gradle.
Ok let's try running them again.
Oops, sector is broken.
Add a test for sector.
Fix sector. Find you need to add an assert to a const constructor.
Notice we need to turn const asserts on for the analyzer.
Notice the analysis_options files are out of sync with each other and
with the full list of lints.
Turn on the lints that should be on.
Fix the bugs that finds.
Xcode builds depend on the Python 'six' module. If not present, exit
immediately with a useful error message.
The six module is included in the system default Python installation. We
perform this check in case a custom Python install has higher priority
on $PATH; e.g., due to a Homebrew or MacPorts installation.
This extracts an existing doctor check to use it during the build step
as well.
The very first time `pod install` is invoked, CocoaPods downloads the master spec repository, which takes quite a while. Before this change, the build appeared to have stalled. With this change, at least the spinner is moving.
Added `pod setup` to the install instructions for CocoaPods, so the spec repo is downloaded while setting up Flutter, instead of during the first build.
Go through all packages brought in by pub, and write the name and path of every one that is a flutter plugin into .flutter-plugins.
In android/settings.gradle and ios/Podfile, read in .flutter-plugins, if that file exists. The Android / iOS code from the plugins is automatically added as dependencies of the native code of the app.
* Remove legacy .apk build.
Print out an error message telling the user to upgrade the project if
it's not Gradle-based. Removed all the obvious traces of the legacy
build.
Added support for Dart VM kernel snapshots in Gradle builds.
Fixed Android installs to verify that the app is actually installed, and
not just rely on the presence of the .sha1 file.
* Only run pod install if CocoaPods v1.0.0 or greater is installed.
Avoid issues with older versions of CocoaPods breaking the build. Users who genuinely use older versions of CocoaPods will have to run pod install manually when required.
1. Add `PortScanner` abstraction so that we don't do actual port scanning
in tests.
2. Don't change the real `cwd` of the isolate during tests, as it affects
all tests, not just the current running test.
Fixes#8761
* Revert "Eliminate CocoaPods install step (#8694)"
This reverts commit f4a13bc72b.
If the developer is relying on CocoaPods and hasn't done a pod install, we will do it for them. This is needed for a smooth native plugin experience, similar to what Gradle is doing on the Android side.
There's no hard dependency on CocoaPods. We only run pod install if the project uses CocoaPods, so developers are still free to use alternatives if they prefer (and if they don't want to use native plugins).
Fixes#8685Fixes#8657Fixes#8526
* Require CocoaPods 1.0.0 or newer.
And make sure we don't get a crash if running `pod install` fails.
* Address review feedback
If the developer is relying on CocoaPods and hasn't done a pod install,
they'll get a build failure indicating the issue.
This also avoids a hard dependency on CocoaPods in the tool and allows
developers to customize their Xcode steps to use alternatives such as
Carthage if they prefer.
As of the latest Xcode versions, the latest published libimobiledevice
is out-of-date and build from HEAD is required.
This fixes two bugs:
1. Update initial install instructions to add --HEAD flag.
2. Update uninstall, reinstall instructions to include
--ignore-dependencies flag, since other brew formulae depend on
libimobiledevice.
Since iOS builds are CocoaPods enabled by default, we should make sure to run `pod install` to get pods wired up before building the app.
Also added a check to `flutter doctor` to verify CocoaPods is installed.
I'm passing FLUTTER_FRAMEWORK_DIR to the `pod install` command, so we can have the app's Podfile link in Flutter.framework as a pod instead of having to copy it over in xcode_backend.sh.
Only implemented for Android devices for now. Compare the installed SHA1
to the latest build. If they match, there's no reason to reinstall the
build.
Fixes#8295
Do not attempt to deploy/debug wifi connected iOS devices. ios-deploy is
able to install over wifi, but we've had several bugs reporting failure
to run/debug once installation has completed when the device is also
connected via USB. Note that we don't currently support deploy/debug
over wifi since libimobiledevice (which is also required) requires a USB
connection.
* Teach flutter tools to find gradle
Flutter tools will now use Gradle from Android Studio, which is now found automatically.
flutter doctor will verify that Android Studio has been installed, and that the included Gradle is at least version 2.14.1.
It is still possible to manually configure the path to Android Studio (flutter config --android-studio-dir=XXX) or Gradle (flutter config --gradle-dir=XXX), but this should only be necessary if they're installed somewhere non-standard.
Only tested on Linux and macOS for now.
Fixes#8131
Artifacts are now located in a central place.
This will enable us to downlaod artifacts when we need them (instead of
downloading them all upfront).
This also makes replacing sky_snapshot with gen_snapshot easier.
If a developer has installed the Xcode command-line tools, then later
installs Xcode, they may need to run xcode-select to update the path of
the command-line tools to that of the new installation.
On flutter run, we update ios/Flutter/Generated.xcconfig with various
Flutter-specific settings required by xcode_backend.sh during a build
from Xcode. These settings need to be present at the time the project is
loaded since Xcode doesn't pick up live updates to these files.
Without these settings, Xcode fails to locate xcode_backend.sh itself,
causing the build to fail until the Xcode project has been closed and
re-opened. This also prevents Xcode's project updater from 'helpfully'
suggesting to clean up and delete the Generated.xcconfig file.
* Fix potential NPE in iOS doctor check
In case Xcode is not installed, the xcode-select path may be null.
* fixup! Fix potential NPE in iOS doctor check
If the developer has only installed the Xcode command-line tools,
xcode-select and some other tools may be present, but xcodebuild will be
missing. In this case, output a slightly improved message indicating
that the installation is incomplete rather than missing.
* Add doctor check for Python 'six' module
Required as part of Xcode lldb module. In all likelihood, if we
encounter this situation, the developer is using a custom Python install
(e.g., via MacPorts or Homebrew).
This removes direct file access from within flutter_tools
in favor of using `package:file` via a `FileSystem` that's
accessed via the `ApplicationContext`.
This lays the groundwork for us to be able to easily swap
out the underlying file system when running Flutter tools,
which will be used to provide a record/replay file system,
analogous to what we have for process invocations.
This argument will enable mocking of os-layer process invocations,
where the mock behavior will come from replaying a previously-
recorded set of invocations. At the point of process invocation,
the key metadata for the invocation will be looked up in the
recording's manifest, and iff a matching record exists in the
manifest, the process will be mocked out with data derived from
the corresponding recorded process (e.g. stdout, stderr, exit code).
This moves the various copies of port forwarding code in the Device subclasses into the ProtocolDiscovery class.
* move port forwarding to a common location
* throw exception if protocol Uri is not discovered or port forwarding fails
* cancel discovery protocol subscriptions on iOS launches (wasn't happening before)
* fix iOS port forwarding to match other implementations
* add tests
It is not enabled by default in the VM because applications not launched via the tools may try to connect with the debugger. This causes the debugger and the IDE to hang on Mac.
With this change, they're run via instance methods on an object
obtained through the context. This will allow us to substitute
that object in tests with replay/record versions to allow us to
mock out the os-layer in tests.
* Allow for application-specific log readers.
When running with prebuilt application binaries, those applications
aren't guaranteed to be named "Runner" (as it is when we build
the app locally in Flutter tools)
* convert pubGet to throw ToolExit on non-zero exit code
* convert commandValidator to throw ToolExit for non-zero exit code
* convert flutter commands to throw ToolExit for non-zero exit code
* use convenience method throwToolExit
* only show "if this problem persists" for unusual exceptions
It takes a little bit longer that 10 seconds to install, launch and
have the observatory be available on an iPod touch. Depending on the
size of the application, it could be a lot longer to transer the
bundle over the wire. The 60 seconds is arbitrary.
Previously, overall status was reported as good if Xcode and homebrew
were installed, even if there were issues with either of the tools. We
now report partial status if we detect issues with either installation.
libimobiledevice 1.2.0 is incompatible with Xcode >= 8.0. Since the
libimobiledevice tools don't include a --version flag, check for valid
exit code on running idevicename, if an iOS device is connected.
When using --use-application-binary:
- [x] Stop flutter run from checking for a pubspec.yaml in current directory
- [x] Stop flutter run from invoking pub get
- [x] Set 'shouldBuild' based on --use-application-binary
- [x] Stop requiring 'lib/main.dart' to be present before running.
- [x] Stop building an FLX when launching on Android
Ensure that we're monitoring logs before we launch the app on the
device. This prevents the case where we start monitoring after the ports
have already been logged.
Also update the regex to use an optional capture group.
With iOS 10, the syslog format changed from:
Sep 23 16:04:17 cbracken-iPhone Runner[6188] <Info>: .*
to:
Sep 23 16:04:17 cbracken-iPhone Runner(libsystem_asl.dylib)[6188] <Info>: .*
This updates the observatory port scraping to handle either case.
This change adds a top-level getBuildDirectory func and funcs for
android, aot, asset, ios build products.
Developers may now add a "build-dir" mapping to their
~/.flutter_settings (JSON format) config file. Output directory is
relative to the main flutter application directory.
This change also changes the default build directory for iOS builds to a
subdirectory of the configured build directory, 'build/ios' by default.
- [x] Refactor view support into separate classes.
- [x] Make the hot runner grab the main view and call runFromSource on it.
- [x] Remove Device.needsDevFS (because it is always true).
This patch changes the default `flutter create` command to generate a
HelloServices-style app on iOS. As a consequence, the Atom integration now
fully works with HelloServices.
The new `empty_statements` lint (in the next DEV roll), flags these empty statements. Harmless, except the one in `basic.dart` that fixes a real-life bug. :)
* working on making a faster flutter run restart
* clean up todos; fire events on isolate changes
* use the Flutter.FrameworkInitialization event
* review comments
* iOS tools cleanup
1) Fix `flutter install` on both device and simulator to refer to the actual
bundle and not just the .generated folder
2) Fix `flutter run` on device to actually run vs just installing
Still TODO:
1) Discovered that isAppInstalled on iOS simulator always reports true,
meaning it'll never actually try to install the app.
Fixes#3947Fixes#1823
This also fixes some related problems affecting "flutter run":
* FLXes built during AndroidDevice.startApp need to match the build mode
* APKs should always be rebuilt if the build mode uses AOT compilation
* rename service_protocol.dart to protocol_discovery.dart
* add a wrapper around the obs. protocol
* use json-rpc in run
* consolidate obs. code; implement flutter run --benchmark
* review comments
Host tools can be found in the artifact cache directory for the host platform.
If a developer wants to use a local engine build instead, then provide an
--engine-build flag that selects the specific engine build variant.
* more terse flutter upgrade
* fix an issue when updating and the engine update code has changed
* call flutter precache; add a --no-color hidden option
* fix a lint related to getters/setters
