*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].*
[C++, Objective-C, Java style guides]: https://github.com/flutter/engine/blob/main/CONTRIBUTING.md#style
Moves all backend-specific renderer configuration out of `EmbedderConfigBuilder` and into the backend-specific subclasses of `EmbedderTestContext`. `EmbedderTestContext` is already backend-specific and as of recent patches, also houses compositor configuration, making it the natural home of this code.
As a result, we no longer need backend-specific methods such as `SetSoftwareRendererConfig`, `SetMetalRendererConfig`, `SetOpenGLRendererConfig`, `SetVulkanRendererConfig`. Nor do we need manual backend initialisation in `EmbedderConfigBuilder`. Nor does that initialisation any longer require relying on internal backend-specific code within `EmbedderTestContext`, since we now do that initialisation in the `EmbedderTestContext` constructor.
Since the bulk of the work previously done by this method now occurs in the `EmbedderTestContext` constructor, the only work remaining in these methods is surface creation. Further, since this is all now implemented in backend-specific `EmbedderTestContext` subclasses, they have all been renamed to a single method: `SetSurface`.
Previously, all of these methods took a surface_size parameter, with two exceptions:
* `SetVulkanRendererConfig` also took an optional `FlutterVulkanInstanceProcAddressCallback` parameter. This has been extracted to a separate method `SetVulkanInstanceProcAddressCallback` on `EmbedderTestContextVulkan`.
* `SetSoftwareRendererConfig` defaulted the parameter to a size of (1, 1). For consistency, this is no longer defaulted, and all call sites have been updated, consistent with other backends.
Lastly, one nice benefit is that because the render config is initialised in the `EmbedderTestContext` constructor, there's no longer a requirement to call `Set*RendererConfig` prior modifying any specific properties on the config, nor is it problematic to call the (replacement) `SetSurface` method after modifying the config. Where the renderer config was being customised in embedder unit tests, I've pushed that customisation up to the top of the test where the rest of the test context is configured.
This eliminates nearly all remaining `#ifdef SHELL_ENABLE_$BACKEND` blocks in the EmbedderTest infrastructure.
Issue: https://github.com/flutter/flutter/issues/158998
[C++, Objective-C, Java style guides]: https://github.com/flutter/engine/blob/main/CONTRIBUTING.md#style
Without this if a channel method call was in progress and then the
handler was destroyed the call could attempt to access the destroyed
handler. This is possible as the handler is recreated on engine restart.
However, in this case the result is not used so no invalid memory access
would occur. Adding the cancellable fixes any future change where this
would no longer be the case.
Also updates the image_filter_test.dart expected color values for a blur unit test. The legacy blur algorithm would a) treat very small blur sigmas as the identity even though the effect was still perceptible and b) was very inaccurate for sigmas < 2. The unit test used a sigma of 1.0 so fell within the inaccurate range but the expected values appear to have been taken directly from the original Skia output. This updates them to match the output of the new algorithm (which is also more consistent between CPU and GPU).
[C++, Objective-C, Java style guides]: https://github.com/flutter/engine/blob/main/CONTRIBUTING.md#style
Fixes https://github.com/flutter/flutter/issues/158963
If the GLES version is at least 3, then we can attach a sync fence to the texture gles object. If this operation succeeds, then we can use gl.Flush instad of gl.Finish. Then, when binding the texture - if a sync fence is present we wait and then remove the fence.
Update to the latest definitions of emulator configurations. (except x86/api28 see b/379736755)
Specifically to past https://crrev.com/c/5908894 which added `gpu_mode: "swangle_indirect"` to the api 35 proto. This should help with emulator stability on versions past android 13.
Related to b/371020223
Research notes / my findings for how this system works at go/flutter-android-ci-emulator
Eliminates the opt-out of `-Wunguarded-availability-new` in the `ios_test_flutter` target.
Xcode 15 beta XCTest framework headers contained unguarded usage of iOS 17 API. This bug resulted in engine build failures.
This was fixed in a later XCTest framework release.
```
/Applications/Xcode_15-beta.app/Contents/Developer/Platforms/iPhoneSimulator.platform/Developer/Library/Frameworks/XCTest.framework/Headers/XCUIAccessibilityAuditTypes.h:19:30: error: 'XCUIAccessibilityAuditType' is only available on iOS 17.0 or newer [-Werror,-Wunguarded-availability-new]
19 | typedef NS_OPTIONS(uint64_t, XCUIAccessibilityAuditType) {
| ^
/Applications/Xcode_15-beta.app/Contents/Developer/Platforms/iPhoneSimulator.platform/Developer/Library/Frameworks/XCTest.framework/Headers/XCUIAccessibilityAuditTypes.h:19:30: note: 'XCUIAccessibilityAuditType' has been marked as being introduced in iOS 17.0 here, but the deployment target is iOS 13.0.0
/Applications/Xcode_15-beta.app/Contents/Developer/Platforms/iPhoneSimulator.platform/Developer/Library/Frameworks/XCTest.framework/Headers/XCUIAccessibilityAuditTypes.h:19:9: note: annotate anonymous enum with an availability attribute to silence this warning
19 | typedef NS_OPTIONS(uint64_t, XCUIAccessibilityAuditType) {
| ^
/Applications/Xcode_15-beta.app/Contents/Developer/Platforms/iPhoneSimulator.platform/Developer/SDKs/iPhoneSimulator17.0.sdk/System/Library/Frameworks/Foundation.framework/Headers/NSObjCRuntime.h:383:34: note: expanded from macro 'NS_OPTIONS'
383 | #define NS_OPTIONS(_type, _name) CF_OPTIONS(_type, _name)
| ^
/Applications/Xcode_15-beta.app/Contents/Developer/Platforms/iPhoneSimulator.platform/Developer/SDKs/iPhoneSimulator17.0.sdk/System/Library/Frameworks/CoreFoundation.framework/Headers/CFAvailability.h:155:96: note: expanded from macro 'CF_OPTIONS'
155 | #define CF_OPTIONS(_type, _name) __attribute__((availability(swift,unavailable))) _type _name; enum __CF_OPTIONS_ATTRIBUTES : _name
|
```
Issue: https://github.com/flutter/flutter/issues/128958
[C++, Objective-C, Java style guides]: https://github.com/flutter/engine/blob/main/CONTRIBUTING.md#style
This addresses https://github.com/flutter/flutter/issues/159088
If the image is rotated in its exif data, the coded width/height may differ from the display width/height. It's important to use the display width/height so that `texImage2D` does the right thing with the `VideoFrame`.
For consistency with the Test.*Context classes for other backends, which live in their own implementation file with their own header, extract TestEGLContext to its own header and TU so that in cases where only a TestEGLContext is required (e.g. EmbedderTestBackingStoreProducerGL), we don't need to include all the various test GL surface classes as well.
GetEGLError is used by both TestEGLContext and the TestGLSurface classes, so moves to its own utils file.
No tests because this is a refactoring with no semantic changes, and the code itself is test code.
Issue: https://github.com/flutter/flutter/issues/158998
[C++, Objective-C, Java style guides]: https://github.com/flutter/engine/blob/main/CONTRIBUTING.md#style
Enables the `-fobjc-arc` compiler flag for Objective-C and Objective-C++
translation units.
Eliminates the flutter_cflags_objc[c]_arc settings, since they're now
redundant.
All Obj-C/Obj-C++ code in our codebase has now been migrated to ARC.
Issue: https://github.com/flutter/flutter/issues/137801
Replace the top-level public EmbedderBackingStoreProducer::UserData struct with backend-specific UserData classes, and make them internal to those backends. UserData internals shouldn't be visible to/manipulated by unit tests and compositors, but instead constrained to the backing store producer itself.
Issue: https://github.com/flutter/flutter/issues/158998
[C++, Objective-C, Java style guides]: https://github.com/flutter/engine/blob/main/CONTRIBUTING.md#style
In modern Objective-C, `@property` directives automatically generate a backing ivar (property name prefixed with an underscore), the getter, and (for readwrite properties) the setter. `@synthesize` directives are generally only required if the backing ivar has a different name from the property.
Also updates the FlutterMetalLayer API to match CAMetalLayer:
* Adds API_AVAILABLE declaration to match that on CAMetalLayer.
* Eliminates wantsExtendedDynamicRangeContent property as it's also part of CALayer's interface and unused in our implementation.
Also eliminates unnecessary ivars where they're being synthesized by `@property` declarations.
Previously, we were overriding the behaviour of
UIViewController.prefersStatusBarHidden by synthesizing _flutterPrefersStatusBarHidden as a backing ivar. Since we're explicitly overriding the behaviour of a superclass, it's more idiomatic to synthesize a private property or explicitly declare an ivar then explicitly override the getter instead.
Further, this adds documentation to cases where `@synthesize` directives are required, such as:
* Creating a backing ivar for a readonly property.
* Creating a backing ivar for a property with a custom getter/setter.
* Synthesising the ivar, getter, and setter for a property declared in a protocol being implemented.
[C++, Objective-C, Java style guides]: https://github.com/flutter/engine/blob/main/CONTRIBUTING.md#style
In our vsync waiter and related tests, we hardcode the "CADisableMinimumFrameDurationOnPhone" key in several places. This pulls those into a constant kCADisableMinimumFrameDurationOnPhoneKey declared in the vsync waiter header, which is non-public, and uses that instead.
[C++, Objective-C, Java style guides]: https://github.com/flutter/engine/blob/main/CONTRIBUTING.md#style
When performing `isKindOfClass` checks, we were occasionally looking up the class in question using `NSClassFromString()`, instead we now check against the class directly, which has the added benefit of being type-safe, and not succeptible to string typos.
Moves the declaration of ForwardingGestureRecognizer and FlutterDelayingGestureRecognizer to the FlutterPlatfomViews_Internal.h header, which is non-public.
[C++, Objective-C, Java style guides]: https://github.com/flutter/engine/blob/main/CONTRIBUTING.md#style
This replaces a few ARC bridged retain casts to regular bridge casts to CoreFoundation types, which are then CFRetained via `sk_cfp::retain` calls.
This eliminates the last remaining unnecessary __bridge_retained casts in the codebase. The remaining casts have been manually vetted and are reasonable.
Issue: https://github.com/flutter/flutter/issues/155943
[C++, Objective-C, Java style guides]: https://github.com/flutter/engine/blob/main/CONTRIBUTING.md#style
A couple string constants had been incorrectly declared `const NSString*` rather than `NSString* const`. The former is a pointer to an immutable string, but (a) the pointer can be reassigned and (b) the default NSString implementation is immutable. The latter is an immutable pointer to a string, which is what we want, since we don't want developers reassinging our constants to other values.
Also updates an identifier name to pass lint rules.
No changes to tests since this is "tested" by the compiler.
[C++, Objective-C, Java style guides]: https://github.com/flutter/engine/blob/main/CONTRIBUTING.md#style
Previously, FlutterViewController.engine was declared as a weak readonly property, but we explicitly declared the `FlutterEngine* _engine` ivar as a strong reference in the implementation. This changes the property declaration to be strong and eliminates the now unnecessary ivar.
There is also no semantic change to FlutterViewController itself, since the `_engine` ivar had been manually declared as a strong reference.
There is no semantic change for users of FlutterViewController.engine since whether a user acquires a strong or weak reference to the engine is determined by whether they declare the pointer to which they assign it as strong or weak.
This also eliminates the need for the `engine` getter, which was only present to prevent a warning that the strong ivar didn't match the weak property declaration.
No changes to tests since this introduces no semantic changes.
Issue: https://github.com/flutter/flutter/issues/137801
[C++, Objective-C, Java style guides]: https://github.com/flutter/engine/blob/main/CONTRIBUTING.md#style
Three changes related to string constants:
1. Uses the kIOSurfaceColorSpace constant, which is a CFStringRef pointing to the string "IOSurfaceColorSpace"
2. Uses the kCVPixelFormatType_32BGRA enum value from the CoreVideo headers (which is equal to 'BGRA') in place of hardcoding the string.
From the headers:
```
kCVPixelFormatType_32BGRA = 'BGRA', /* 32 bit BGRA */
```
3. Declares kIOServicePlane as a `constexpr const char*` rather than `static const char*`, this ensures only a single instance is created, rather than one per translation unit into which the header is included. This string is part of IOKit, but see the comment at the top of the header as to why it's apparently needed.
No test changes since there are no semantic changes.
[C++, Objective-C, Java style guides]: https://github.com/flutter/engine/blob/main/CONTRIBUTING.md#style
FlutterEngineGroup keeps an array of all live FlutterEngine instances it has created such that after the first engine, it can spawn further engines using the first.
Previously we manually managed this array by adding engines to it upon creation, then having [FlutterEngine dealloc] emit a notification such that FlutterEngineGroup can listen for it, and remove instances from the array upon reception of the notification.
Instead, we now use an NSPointerArray of of weak pointers such that pointers are automatically nilled out by ARC after the last strong reference is collected. This eliminates the need for the manual notification during dealloc.
Unfortunately, NSPointerArray is "clever" and assumes that if the array has not been mutated to store a nil pointer since its last compact call, it must not contain a nil pointer and can thus skip compaction. Under ARC, weak pointers are automatically nilled out when the last strong reference is released, so the above heuristic is no longer valid. We work around the issue by storing a nil pointer before calling compact.
See http://www.openradar.me/15396578 for the radar tracking this bug.
I'm not thrilled with the fact that we're replacing one sort of TODO with another, but the code is much simpler and avoids relying on a trip through the notification center, which seems objectively better.
Issue: https://github.com/flutter/flutter/issues/155943
Issue: https://github.com/flutter/flutter/issues/137801
[C++, Objective-C, Java style guides]: https://github.com/flutter/engine/blob/main/CONTRIBUTING.md#style
