Previously, the selection base and extent were stored internally as
iterators over text_. Since iterators must be treated as invalidated
whenever the underlying container changes, this requires that
selection_base_ and selection_extent_ be re-assigned after every change
to text_.
This is not currently particularly problematic, but once we add fields
to track the base and extent of the composing region for multi-step
input method support, as well as support for the sub-range within the
composing region to which edits/completions apply, we end up having to
regenerate a lot of iterators with each change, many of which are
logically unchanged in position.
A side benefit is that this simplifies inspection of these fields when
debugging.
Previously, TextInputModel's SetEditingState method was a 1:1 mapping of
the underlying protocol used on the text input channel between the
framework and the engine. This breaks it up into two methods, which
allows the selection to be updated independently of the text, and avoids
tying the API the the underlying protocol.
This will become more important when we add additional state to support
composing regions for multi-step input methods such as those used for
Japanese.
SetText resets the selection rather than making a best-efforts attempt
to preserve it. This choice was primarily to keep the code simple and
make the API easier to reason about. An alternative would have been to
make a best-effort attempt to preserve the selection, potentially
clamping one or both to the end of the new string. In all cases where an
embedder resets the string, it is expected that they also have the
selection, so can call SetSelection with an updated selection if needed.
Replaces the (temporary) compile-time option to pass engine switches
with the ability to pass them temporarily at runtime via environment
variables. This moves the recently-added code for doing this on Windows
to a shared location for use by all desktop embeddings.
This is enabled only for debug/profile to avoid potential issues with
tampering with released applications, but if there is a need for that in
the future it could be added (potentially with a whitelist, as is
currently used for Dart VM flags).
Temporarily adds a way to enable mirrors as a compile time option,
as is already provided in the Linux embedding, to provide a migration
path for the one remaining known need for compile-time options
that has been raised in flutter/flutter#38569.
When the EncodableValue implementation changed, the old version was
temporarily kept behind an #ifdef to allow temporarily using the old
version, so that the roll would not be blocked. All known existing
clients have migrated, so the legacy version is no longer necessary.
Cleans up header order/grouping for consistency: associated header, C/C++ system/standard library headers, library headers, platform-specific #includes.
Adds <cstring> where strlen, memcpy are being used: there are a bunch of places we use them transitively.
Applies linter-required cleanups. Disables linter on one file due to included RapidJson header. See https://github.com/flutter/flutter/issues/65676
This patch does not cover flutter/shell/platform/darwin. There's a separate, slightly more intensive cleanup for those in progress.
We currently use a mix of C standard includes (e.g. limits.h) and their
C++ variants (e.g. climits). This migrates to a consistent style for all
cases where the C++ variants are acceptable, but leaves the C
equivalents in place where they are required, such as in the embedder
API and other headers that may be used from C.
Add copyright headers in a few files where they were missing.
Trim trailing blank comment line where present, for consistency with
other engine code.
Use the standard libtxt copyright header in one file where it differed
(extra (C) and comma compared to other files in libtxt).
This also amends tools/const_finder/test/const_finder_test.dart to look
for a const an additional four lines down to account for the copyright
header added to the test fixture.
The C++ wrapper makes heavy use of templates to support arbitrary types
in the platform channel classes, but in practice EncodableValue is what
essentially all code will use. This defaults those template types to
reduce boilerplate in plugin code (e.g., allowing the use of
MethodChannel<> instead of MethodChannel<EncodableValue>).
The response APIs for method channels and event channels used pointers
for optional parameters; this kept the API surface simple, but meant
that they couldn't take rvalues. As a result, returning success values
or error details often took an extra line, declaring a variable for the
result just to have something to pass the address of.
This converts them to using references, with function overloading to
allow for optional parameters, so that values can be inlined.
For now the pointer versions are still present, so that conversion can
be done before it becomes a breaking change; they will be removed soon.
Part of https://github.com/flutter/flutter/issues/63975
Relands https://github.com/flutter/engine/pull/20399
Makes BinaryMessenger available from FlutterEngine, rather than just the plugin registrar. This allows for method channels directly in applications without building them as plugins, and matches the other platforms.
Requires some restructuring of code and GN targets in the client wrappers to make the internals in the shared section usable by the implementations of platform-specific parts of the wrappers. Also fixes a latent issue with EnableInputBlocking symbols being declared but not defined for Windows that came up during testing of the restructing.
Fixes https://github.com/flutter/flutter/issues/62871
Makes BinaryMessenger available from FlutterEngine, rather than just the plugin registrar. This allows for method channels directly in applications without building them as plugins, and matches the other platforms.
Requires some restructuring of code and GN targets in the client wrappers to make the internals in the shared section usable by the implementations of platform-specific parts of the wrappers. Also fixes a latent issue with EnableInputBlocking symbols being declared but not defined for Windows that came up during testing of the restructuring.
Fixes https://github.com/flutter/flutter/issues/62871
A recent refactoring broke the USE_LEGACY_ENCODABLE_VALUE codepath in
standard_codec.cc, which went unnoticed since it wasn't being compiled.
This fixes the breakage, and also adds a temporary minimal unit test
target that ensures that all the USE_LEGACY_ENCODABLE_VALUE paths are
being compiled.
__has_feature(cxx_rtti) is a clang extension, but clients of this code
are mostly using the VS toolchain. This makes the RTTI check work with
more compilers.
Standard*Codec allows for extensions to support arbitrary types; this
had not previously been implemented for the C++ version.
Overview of changes:
- EncodableValue's std::variant type now allows for a new CustomEncodableValue, which is a thin wrapper around std::any, to store arbitrary extension types.
- ByteBufferStream* has been split into an interface class and the buffer-based implementation, with the former now part of the public API surface to be used in standard codec extensions.
- They also gained utility methods for some common data types to simplify writing extensions.
- StandardCodecSerializer is now part of the public API surface, and is subclassable.
- StandardCodecSerializer's ReadValue has been split into ReadValue and ReadValueOfType to match the structure used when subclassing on the Dart side, for easier porting of custom extensions across languages.
- Standard*Codec now optionally accepts a non-default serializer in GetInstance, providing a shared instance using that serializer.
Fixes https://github.com/flutter/flutter/issues/31174
Changes the interaction between the view controller and engine in both the C API and
the engine API, so that there's always an engine (as on other platforms) rather than
the engine APIs being specific to headless mode.
While adjusting the C API, this does a large cleanup:
- Renames all methods to follow a `FlutterDesktop` (prefix) + "class" name + method-style name.
E.g., `FlutterDestkopViewControllerCreate` rather than `FlutterDesktopCreateViewController`.
This makes it easier to see what functions operate on which conceptual "object" in the API.
- Reorders and groups them by the object they operate on.
Fixes https://github.com/flutter/flutter/issues/61966
Instead of a hand-rolled discriminated union (originally used to avoid a C++17
dependency, which is no longer an issue), implement EncodableValue as a
std::variant. Rather than simply changing the internals, this makes EncodableValue
a minimal std::variant subclass with only a handful of added methodS, replacing
the old IsFoo/FooValue APIs with the standard std::holds_alternative/std::get,
so that plugin code will use a standard-based API rather than a Flutter-specific
API for wrapped values.
This is a breaking change for Windows and GLFW plugins. In the short
term USE_LEGACY_ENCODABLE_VALUE can be set in builds to use the old
version, to separate rolling from updating.
Fixes https://github.com/flutter/flutter/issues/61970
Tweak the primary flutter build rule so that fuchsia is more similar to
other platforms in how tests and the shell are built.
Only embedder_unittests and GLFW tests are disabled on Fuchsia now.
TEST: Ran unittests on host/fuchsia; workstation on fuchsia
BUG: fxb/53847, fxb/54056
The embedder.h API layer is an implementation detail of the desktop
embeddings, not part of the public API surface, so should not be part of
the public symbol list for those libraries.
The C++ text input model used by Windows and Linux currently uses UTF-32. The intention was to facilitate handling of arrow keys, backspace/delete, etc., however since part of what is synchronized with the engine is cursor+selection offsets, and those offsets are defined in terms of UTF-16 code units, this causes very bad interactions with the framework-side model.
This converts to using UTF-16, rather than UTF-32, so that the offsets align with the framework. It also adds surrogate pair handling to the operations that adjust indexes, to avoid breaking surrogate pairs. (Arbitrary grapheme cluster handling is out of scope for this PR; while definitely desirable in the long term, surrogate pair handling is much more critical since improper handling yields invalid UTF-16, which breaks the text field).
This partially fixes https://github.com/flutter/flutter/issues/55014. A framework-side fix is also necessary (since currently both the engine and the framework attempt to handle arrow keys, which is another out-of-scope-for-this-PR issue), but even without the framework fix this dramatically improves the cursor behavior on Windows when there are surrogate pairs somewhere in the string since at least the two sides agree on what indexes mean.
Includes minor plumbing changes to the text input plumbing on Windows so that we're not pointlessly converting from UTF-16 to UTF-32 and then back to UTF-16.
Fixes a few issues with Windows text input:
- Filters out ASCII control characters
- Filters out lead surrogates, which aren't valid UTF-16 on their own so will cause assertion failures if sent to Flutter
- Adds a bandaid fix for a crash due to mismatches in indexing in the C++ and Dart text models. (A better fix would be to use UTF-16 and add surrogate pair handling to deletion and forward/back; that will be a later PR since it has a larger scope.)
Fixes https://github.com/flutter/flutter/issues/54879
Makes InvokeMethod's reply a high-level response object, rather than
binary data, matching the abstraction level of the class (and the other
languages' implementations).
In support of that:
- Adds the logic to the codecs to decode response envelopes, which had
never been implemented.
- Adds a convience implementation of MethodResult that forwards to
lambdas, so that one-off invocation handlers are easier to write.
Also simplified BinaryMessenger's API so that subclasses only need to
implement one version of Send, rather than two almost-identical versions.
Fixes https://github.com/flutter/flutter/issues/53223
Relands #17489 with a fix for the unit test flake.
The previous unit test relied on the new instance not being created at the same memory address, which isn't guaranteed.
Seems to have triggered flaky failures on the Windows bot since landing.
Example failure:
[ RUN ] PluginRegistrarTest.ManagerRemovesOnDestruction
c:\b\s\w\ir\cache\builder\src\flutter\shell\platform\common\cpp\client_wrapper\plugin_registrar_unittests.cc(149): error: Expected: (manager->GetRegistrar<PluginRegistrar>(dummy_registrar_handle)) != (first_wrapper), actual: 000002400A90E3D0 vs 000002400A90E3D0
This reverts commit 478a7855943d81a58dd9e9037fb338d3a18bb294.
This makes two changes:
- Adds a way to register a callback for when a FlutterDesktopPluginRegistrarRef is destroyed, and implements the logic to call it in the Windows and Linux embeddings.
- Adds a class to the C++ wrapper that handles making a singleton owning PluginRegistrar wrappers, and destroying them when the underlying reference goes away, to avoid needing that boilerplate code in every plugin's source.
Fixes https://github.com/flutter/flutter/issues/53496
The JSON codec is awkward to use in the wrapper (since the client has to build and link one of the JSON libraries to do so). Since it would be very cumbersome to wrap in a C API, and there's essentially no reason to use it instead of the standard codec, this removes it from the wrapper entirely.
Since some system channels (internal to the engine) still use it, it's moved into common/cpp instead of being eliminated entirely. Internally we always use RapidJSON though, so the jsoncpp implementation is removed. Also adds some unit test coverage, since there wasn't any.
Fixes#30669
Currently every Linux runner has this code to allow relative resource paths; this moves it into the framework so that any embedder can get this behavior without that code needing to be in the template.
Rolls buildroot to pick up std::filesystem support in our libc++
MethodChannel and BasicMessageChannel in the C++ wrapper didn't have the
expected semantics that passing a null handler would remove any existing
handler. This was inconsistent with other platforms and with the
lower-level object APIs in this wrapper, and made unregistration in
cases where that's desirable more difficult due to needing to keep other
object references.
Adds tests for this functionality, and some backfill of missing tests
for basic handler behavior.
See https://github.com/flutter/flutter/issues/51207
Targeted suppression of some deprecation warnings that are build errors under
clang:
- Ignore the deprecation of codecvt's unicode conversion until we decide on
a replacement strategy.
- Allow the deprecated posix names of functions in third_party/txt.
Part of https://github.com/flutter/flutter/issues/16256
This was only necessary when the Engine had to build in multiple buildroots
where the sources where checked out at different paths relative to the
buildroot. This is no longer the case and there are already cases GN rules
have been written that mix and match variable usage with the direct
specification of the path to the Flutter sources relative to the sole buildroot.
* Support empty strings and vectors in standard codec
Fixes#41993
Currently an empty string or vector will call through to WriteBytes
which asserts that the number of bytes it is being asked to write is
strictly positive. Instead we should not call WriteBytes if the length
is zero.
Similarly, when we read, we don't need to call out if the length is
zero.
* fix typo in test name
* remove unnecessary length check in ReadValue for List
* we also don't need this check before calling read as memcpy can handle size 0
Makes the plugin registration structure consistent with macOS. This will
be used in generated plugin registrant files rather than a specific
implemenation class, so this helps unblock the creation of generated
registrants on Windows and Linux.
Significantly improves the behavior of non-ASCII text input on Windows. Correctly
processes incoming character events as UTF-16, and for now uses UTF-32 for
the text model so that the existing index-based logic will work much more often.
Future work is still needed, but this will handle far more cases correctly.
