Flutter uses `vswhere.exe` to find Visual Studio installations and determine if they satisfy Flutter's requirements. However, `vswhere.exe`'s JSON output is known to contain bad UTF-8. This change ignores bad UTF-8 as long as they affect JSON properties that are either unused, or, used only for display purposes by Flutter.
Fixes: https://github.com/flutter/flutter/issues/102451
Adds a bit more clarifying documentation to the implementation of the
outputFollowsExit case, and adds tests that verify the behaviour of
stderr, stdout of processes launched via FakeProcessManager.
Specifically:
* Verifies that stderr, stdout are not emitted immediately after process
exit if outputFollowsExit is true. They must be emitted at least one
turn through the event loop later.
* Verifies that ProcessResult.stderr, stdout have the type documented
according to the encoding passted to Process.run/runSync:
* List<int> if null is passed as the encoding.
* String (in the default system encoding) if no encoding is specified.
* String (in the specified encoding) if an encoding is specified.
This is additional testing relating to refactoring landed in:
https://github.com/flutter/flutter/pull/103947
Issue: https://github.com/flutter/flutter/issues/102451
`VisualStudio` calls `vswhere.exe` to find Visual Studio installations and determine if they satisfy Flutter's requirements. Previously, `VisualStudio` stored the JSON output from `vswhere.exe` as `Map`s, resulting in duplicated logic to read the JSON output (once to validate values, second to expose values). Also, `VisualStudio` stored two copies of the JSON output (the latest valid installation as well as the latest VS installation).
This change simplifies `VisualStudio` by introducing a new `VswhereDetails`. This type contains the logic to read `vswhere.exe`'s JSON output, and, understand whether an installation is usable by Flutter. In the future, this `VswhereDetails` type will be used to make Flutter doctor resilient to bad UTF-8 output from `vswhere.exe`.
Part of https://github.com/flutter/flutter/issues/102451.
* Use libraryFilters flag to speed up coverage collection
* Allow libraryNames to be null
* Unconditionally enable the reportLines flag
* Fix analysis errors
Because this class has some subtle behaviour with regards to control of
exit timing and when and how it streams data to stderr and stdout, it's
worth adding unit tests for this class directly, as well as (in a
followup patch) for FakeProcessManager.
This is additional testing relating to refactoring landed in:
https://github.com/flutter/flutter/pull/103947
Issue: https://github.com/flutter/flutter/issues/102451
`FakeProcessManager` is a test-oriented implementation of `ProcessManager`
that simulates launching processes and returning `ProcessResult` objects
whose `exitCode`, `stdout`, `stderr` can be used to write platform-portable,
hermetic tests that don't rely on actually launching processes from
executables on disk. Its `run` and `runSync` methods provide asynchronous and
synchronous variants of this functionality.
Previously, the behaviour of `run` and `runSync` were inconsistent with
regards to the treatment of the `stdoutEncoding` (similarly,
`stderrEncoding`) parameters:
`run`:
* if the encoding was null, `ProcessResult.stdout` was returned as a
String in UTF-8 encoding. This was incorrect. The behaviour as
specified in `ProcessResult.stdout` is that in this case, a raw
`List<int>` should be returned.
* If the encoding was unspecified, `ProcessResult.stdout` was returned as
a `String` in the `io.systemEncoding` encoding. This was correct.
* If the encoding was non-null, `ProcessResult.stdout` was returned as a
`String` in the specified encoding. This was correct.
`runSync`:
* if the encoding was null, `ProcessResult.stdout` was returned as a
`List<int>` in UTF-8 encoding. This was incorrect. The behaviour as
specified in `ProcessResult.stdout` is that in this case, a raw
`List<int>` should be returned.
* If the encoding was unspecified, `ProcessResult.stdout` was returned as
`List<int>` in UTF-8 encoding. This was incorrect. The behaviour as
specified in `ProcessResult.stdout` is that in this case, a String a
`String` in the `io.systemEncoding` encoding should be returned.
* if the encoding was non-null, `ProcessResult.stdout` was returned as a
`String` in unknown (but probably UTF-8) encoding. This was incorrect.
The behaviour as specified in `ProcessResult.stdout` is that in this
case, a `String` in the specified encoding should be returned.
`_FakeProcess`, from which we obtain the fake stdout and stderr values now
holds these fields as raw `List<int>` of bytes rather than as `String`s. It
is up to the user to supply values that can be decoded with the encoding
passed to `run`/`runAsync`.
`run` and `runAsync` have been updated to set stdout (likewise, stderr) as
specified in the `ProcessResult` documentation.
This is pre-factoring for #102451, in which the tool throws an exception
when processing the JSON output from stdout of the `vswhere.exe` tool,
whose output was found to include the `U+FFFD` Unicode replacement
character during UTF-8 decoding, which triggers a `toolExit` exception
when decoded using our [Utf8Decoder][decoder] configured with `reportErrors` =
true. Because `FakeProcessManager.runAsync` did not previously invoke
`utf8.decode` on its output (behaviour which differs from the non-fake
implementation), it was impossible to write tests to verify the fix.
Ref: https://api.flutter.dev/flutter/dart-io/ProcessResult/stdout.html
Issue: https://github.com/flutter/flutter/issues/102451
[decoder]: fd312f1ccf/packages/flutter_tools/lib/src/convert.dart (L51-L60)
Roll dependendencies
This rolls depdendencies to latest using
flutter update-packages --force-upgrade
This change includes three code changes:
* Removes charcode from the dependencies allowlist since it no longer
appears in the transitive closure of dependencies of the flutter,
flutter_test, flutter_driver, flutter_localizations, and
integration_test packages.
* Uses Resolver.create instead of the deprecated Resolver constructor.
The default Resolver constructor has been deprecated in favour of the
static Resolver.create() factory function, which unfortunately happens
to be async. Propagated the async-ness up the chain.
This change was partially reverted and the deprecation ignored in this
patch until package:coverage can be rolled internally at Google.
* Eliminates the use of the deprecated packagesPath parameter to
HitMap.parseJson. This parameter was deprecated and replaced with
packagePath in https://github.com/dart-lang/coverage/pull/370 which
was part of the overall deprecation of the .packages file in Dart
itself https://github.com/dart-lang/sdk/issues/48272. The overall goal
being that end-user code shouldn't need to know about implementation
details such as whether dependency information is stored in a
.packages file or a package_info.json file, but rather use the
package_config package to obtain the package metadata and perform
other functions such as resolving its dependencies to filesystem
paths. packagesPath was replaced by packagePath, which takes the path
to the package directory itself. Internally, package:coverage then
uses package_config to do the rest of the package/script URI
resolution to filesystem paths.
This change was partially reverted and the deprecation ignored in this
patch until package:coverage can be rolled internally at Google.
This is a pre-update prior to updating flutter_template_images in
https://github.com/flutter/flutter/pull/103739
Issue: https://github.com/flutter/flutter/issues/103371
Issue: https://github.com/flutter/flutter/issues/103775
Issue: https://github.com/flutter/flutter/issues/103830
When re-applying the partially-reverted changes to code coverage,
we'll need to patch host_entrypoint.dart internally to await the Future
that we'll be returning rather than a non-async value.
Point users of boolArgDeprecated to boolArg rather than to
boolArgDeprecated in order to avoid losing users to infinite loops or
stack overflows depending on their implementation.
* c4b899f63 Roll dart sdk to 7b24ff4d92e2d2136020fc5bedadfe7025861510 (flutter/engine#33309)
* 7956603cc [web] Migrate Flutter Web DOM usage to JS static interop - 12. (flutter/engine#33241)
* 1d0c2ae6f Roll Skia from 1e43dce386c9 to f6e31bf1dcfb (6 revisions) (flutter/engine#33320)
* ed5a2fef0 Roll Skia from f6e31bf1dcfb to 69fecd6c2d85 (1 revision) (flutter/engine#33322)
* 10e0e1534 Revert "[web] Migrate Flutter Web DOM usage to JS static interop - 12. (#33241)" (flutter/engine#33321)
* fix local tests
Co-authored-by: Jonah Williams <jonahwilliams@google.com>
* Provide flutter sdk kernel files to dwds launcher instead of dart ones
* Update log test to report all warnings
* Update licences for new files
* Addressed CR comments
* Addressed CR comments
* Implement trackpad gestures in framework
* Touch and Pan/Zoom pointers have separate IDs now
* Handle trackpad pointer device type
* Respect supportedDevices for pan/zoom events
* Update after rebase
* Fix check failures
* Avoid error with very short drags
* Address feedback
* Refactor drag event handler
* Address more feedback
* Add some missing punctuation
* Added a bool that allows us to limit debugProfileBuildsEnabled to user
created widgets.
* made it turned on by default
* switched to hashmap
* Cleaned everything up and added tests
* fixed an odd test where it wants to be able to add asserts and run in profile mode
* hixie feedback
* hixie2
* made it default to false
* updated docstring as per dans request
* Don't terminate Dart process pids from VM Service
These processes may be on another device, and in the case of attach the debugee should not be terminated anyway.
Previously, https://github.com/flutter/flutter/pull/100271 enabled
building universal macOS binaries by default, but included a bug causing
the arm64 App.framework to be built such that the TEXT section
containing the app instructions built by gen_snapshot incorrectly
contained x86_64 instructions rather than arm64 instructions.
When building macOS (and iOS) apps, Flutter builds them in three
components:
* The Runner application: built by Xcode
* The bundled App.framework: built from assembly code generated by
gen_snapshot from the application's Dart sources.
* The bundled FlutterMacOS.framework: built as part of the engine build
and packaged by copying the distributed binary framework from our
artifacts cache.
Building App.framework consists of the following steps:
* For each architecture, invoke gen_snapshot to generate
architecture-specific assembly code, which is then built to object
code and linked into an architecture-specific App.framework.
* Use the `lipo` tool to generate a universal binary that includes both
x86_64 and arm64 architectures.
Previously, we were building architecture specific App.framework
binaries. However, for all architectures we were (mistakenly) invoking
the general `gen_snapshot` tool (which emitted x64 instructions, and
which is now deprecated) instead of the architecture-specific
`gen_snapshot_x86` and `gen_snapshot_arm64` builds which emit
instructions for the correct architecture.
This change introduces a small refactoring, which is to split the
`getNameForDarwinArch` function into two functions:
* `getDartNameForDarwinArch`: the name for the specified architecture as
used in the Dart SDK, for example as the suffix of `gen_snapshot`.
* `getNameForDarwinArch`: the name for the specified architecture
as used in Apple tools, for example as an argument to `lipo`. For
consistency, and to match developer expectations on Darwin platforms,
this is also the name used in Flutter's build outputs.
Issue: https://github.com/flutter/flutter/issues/100348
This changes the default build architectures for Flutter macOS apps to
x86_64 and arm64. Previously, we manually excluded arm64 builds via the
EXCLUDE_ARCHS Xcode setting in Flutter's generated xcconfig file. This
eliminates setting EXCLUDE_ARCHS during the build and updates the
default architectures in the tool and in the macos_assemble.sh wrapper.
Issue: https://github.com/flutter/flutter/issues/97681
Umbrella issue: https://github.com/flutter/flutter/issues/60113
This patch adds an additional check to ensure the target length of a string is within the supported maximum string length prior to calling WideCharToMultiByte/MultiByteToWideChar in the Windows runner template.
This is to prevent resize() from failing if called with a count > std::string::max_size().
According to Win32 API docs (WideCharToMultiByte, MultiByteToWideChar) it's the caller responsibility to make sure the buffers are correctly allocated.
Authored by: Tomasz Gucio <tgucio@gmail.com>
This fixes a race condition in test/integration.shard/cache_test.dart which allowed the test to fail if the machine took a very long time to acquire a lock or launch a process.
* Starts using the `--source` flag to compile the dart registrant.
* updated general.shard tests
* Fixed the resident compiler flow
* added integration test
* made the integration test self contained
* renamed generated_main to dart_plugin_registrant
* cleaned up for review
* added task runner for ci
* added bringup and TESTOWNERS
* updated failure message
* Fixed order dependency and removed no-shuffle-tag in build_ios_framework_test.dart
* Removed trailing spaces in build_ios_framework_test.dart
* Removed unnecessary formatting
