Summary:
changelog: [internal]
C++17 has implementation of shared_mutex in standard library. Let's use it instead of folly.
Reviewed By: cipolleschi
Differential Revision: D43275493
fbshipit-source-id: d766251226aa230110011aca94b4e697fe0d31a1
Summary:
changelog: [internal]
`useLayoutEffect` has two guarantees which React Native breaks:
- Layout metrics are ready.
- Updates triggered inside `useLayoutEffect` are applied before paint. State between first commit and update is not shown on the screen.
React Native breaks the first guarantee because it uses Background Executor. Background executor moves Yoga layout to another thread. If user core reads layout metrics in `useLayoutEffect` hook, it is a race. The information might be there, or it might not. They can even read partially update information. This diff does not affect this. We already have a way to turn off Background Executor. We haven't done this because it introduces regressions on one screen but we have a solution for that.
React Native breaks the second guarantee. After Fabric's commit phase, Fabric moves to mounting the changes right away. In this diff, we queue the mounting phase and give React a chance to change what is committed to the screen. To do that, we schedule a task with user blocking priority in `RuntimeScheduler`. React, if there is an update in `useLayoutEffect`, schedules a task in the scheduler with higher priority and stops the mounting phase.
We are not delaying mounting, this just gives React a chance to interrupt it.
Fabric commit phase may be triggered by different mechanisms. C++ state update, surface tear down, template update (not used atm), setNativeProps, to name a few. Fabric only needs to block paint if commit originates from React. Otherwise the scheduling is wrong and we will get into undefined behaviour land.
Rollout:
This change is gated behind `react_fabric:block_paint_for_use_layout_effect` and will be rolled out incrementally.
Reviewed By: javache
Differential Revision: D43083051
fbshipit-source-id: bb494cf56a11763e38dce7ba0093c4dafdd8bf43
Summary:
changelog: [internal]
Passing MountingCoordinator argument by value instead of reference. Using reference does not make sense since we eventually take ownership of shared_ptr anyway. This better communicates the intent.
Reviewed By: christophpurrer
Differential Revision: D43082955
fbshipit-source-id: 29e20abb9824c10a5f0d5e0ba1049ff6d67cee98
Summary:
Changelog: [Internal]
Caller needs to explicitly set commit options. This is for readability and making sure caller is aware of what are the options of the commit. This will be important in subsequent diff where we will add another commit option.
Reviewed By: christophpurrer
Differential Revision: D43082837
fbshipit-source-id: 1417205299c19430f902453c2b6d9bb9ca31707d
Summary:
changelog: [internal]
This is experimental implementation of `setNativeProps` in Fabric.
This diff brings `setNativeProps` to Fabric to make migration easier. I tried to stay as close as possible to Paper's behaviour, with all of its flaws (ready CAUTION section on https://reactnative.dev/docs/direct-manipulation)
State can't be stored in views because on iOS, eventually on Android, views are not the source of truth, shadow tree is. Fabric's implementation keeps state from setNativeProps in shadow node family in very inefficient data structure folly::dynamic. It is always reconciled with new prop updates. The performance cost is only paid in case node has used `setNativeProps` before.
Reviewed By: mdvacca
Differential Revision: D41875413
fbshipit-source-id: 453a5f7612a6f86a4cece269b13bd2ffd0c0e2d1
Summary:
changelog: Introduce setNativeProps to Fabric
Add support for `setNativeProps` in Fabric for backwards compatibility. It is still recommended to move away from `setNativeProps` because the API will not work with future features.
We can make step [Migrating off setNativeProps](https://reactnative.dev/docs/new-architecture-library-intro#migrating-off-setnativeprops) in migration guide optional.
Reviewed By: yungsters, mdvacca
Differential Revision: D41521523
fbshipit-source-id: 4d9bbd6304b8c5ee24a36b33039ed33ae1fc21f8
Summary:
changelog: [internal]
Add findShadowNodeByTag_DEPRECATED` to Fabric's UIManager.
This API makes it possible to get ShadowNode reference for given native tag.
We will leverage this API to make UIManager.measure* calls backwards compatible in the new architecture.
Reviewed By: yungsters
Differential Revision: D41550734
fbshipit-source-id: a45b6a76634d034eac70603bad8ac60f9135ff24
Summary:
changelog: [internal]
Add more clang tidy rules to prevent common class of bugs.
Reviewed By: javache
Differential Revision: D39245194
fbshipit-source-id: 5521c5c4653d7005b96ebba494e810ba5075afbc
Summary:
Fix todo and inconsistency across codebase. It's better to have just one way to refer to these types.
Changelog: [Internal]
Reviewed By: philIip
Differential Revision: D37653146
fbshipit-source-id: e82f09caa6cd6eec5512b78f413708d9c04a7a83
Summary:
This PR fixes RTTI (run-time type information) for ShadowNodeWrapper and ShadowNodeListWrapper classes, i.e., calls to dynamic_cast and dynamic_pointer_cast that are called via JSI's getHostObject calls.
The fix is simply to add a so-called "key function" in a form of virtual destructor. Key functions needs to be a virtual non-pure and non-inlined functions that points the compiler as to which library contains the vtable/type information for a given class (see https://itanium-cxx-abi.github.io/cxx-abi/abi.html#vague-vtable and https://developer.android.com/ndk/guides/common-problems#rttiexceptions_not_working_across_library_boundaries)
Without the "key function", calls to dynamic_cast for ShadowNodeWrapper instances won't work across library boundaries because the class will have separate definitions in each separate library, therefore objects created in one of those libraries won't be recognized as the same type by the other library. This has been a problem in reanimated and gesture-handler libraries where we call `object.getHostObject<ShadowNodeWrapper>(rt)` (this is a method from JSI) in order to access ShadowNode instance from a handle we have in JS. I think, this issue is going to be relevant to more libraries that cope with view instances. In this scenario, we have a separate library, say "libreanimated.so" that calls to `getHostObject` which is an inline function that calls `dynamic_cast` for the `ShadowNodeWrapper` class. On the other hand, the instances of `ShadowNodeWrapper` are created by the code from `libreact_render_uimanager.so`. Because of that `dynamic_cast` fails even though it is called on instance of `ShadowNodeWrapper` because the class has separate vtable/type info: one in `libreanimated.so` and one in `libreact_render_uimanager.so` (by "fails" I mean that it actually returns `nullptr`).
This problem has been documented here: https://developer.android.com/ndk/guides/common-problems#rttiexceptions_not_working_across_library_boundaries where the solution is for the class to have a so-called "key function". The key function makes it so that compiler sees that one of the implementation for a given class is missing and therefore can safely assume that a vtable/type info for a given class is embedded into some library we link to.
This change adds a virtual destructor that is declared in the header file but defined in file that gets compiled as a part of `libreact_render_uimanager`. As a result, the compiler only creates one vtable/type info and calls to dynamic_cast works as expected in all libraries for `ShadowNodeWrapper` and `ShadowNodeListWrapper` classes.
This issue would only surface on Android, because on iOS all libraries by default are bundled together via Pods, whereas on Android each library is loaded separately using dynamic loading.
## Changelog
[Fabric][Android specific] - Fix dynamic_cast (RTTI) for ShadowNodeWrapper and similar classes when accessed by third-party libraries.
Pull Request resolved: https://github.com/facebook/react-native/pull/33500
Test Plan:
1. In order to test this you need to add a library that'd include `<react/renderer/uimanager/primitives.h>` (i.e. use this branch of reanimated library: https://github.com/software-mansion/react-native-reanimated/tree/fabric)
2. After compiling the app inspect libreact_render_uimanager.so and libreanimated.so artifacts with `nm` tool
3. Notice that symbols like `vtable for facebook::react::ShadowNodeWrapper` and `typeinfo for facebook::react::ShadowNodeWrapper` are only present in the former and not in the latter library (before this change you'd see them both)
Reviewed By: ShikaSD
Differential Revision: D35143600
Pulled By: javache
fbshipit-source-id: 5fb25a02365b99a515edc81e5485a77017c56eb8
Summary:
changelog: [internal]
pass raw ShadowNode instead of shared_ptr. Ownership is not transferred, shared_ptr is misleading.
Reviewed By: javache
Differential Revision: D33917010
fbshipit-source-id: 4d9fdd4b4e0376149f1719ad160b957de4afdce3
Summary:
changelog: [internal]
You can read more about this rule on https://clang.llvm.org/extra/clang-tidy/checks/modernize-pass-by-value.html
# Isn't it wasteful to copy? Isn't reference more efficient?
This rule of thumb is no longer true since C++11 with move semantics. Let's look at some examples.
# Option one
```
class TextHolder
{
public:
TextBox(std::string const &text) : text_(text) {}
private:
std::string text_;
};
```
By using reference here, we prevent the caller from using rvalue to and avoiding copy. Regardless of what the caller passes in, copy always happens.
# Option two
```
class TextHolder
{
public:
TextBox(std::string const &text) : text_(text) {}
TextBox(std::string &&text) : text_(std::move(text)) {}
private:
std::string text_;
};
```
Here, we provide two constructors, one for const reference and one for rvalue reference. This gives the caller option to avoid copy. But now we have two constructors, which is not ideal.
# Option three (what we do in this diff)
```
class TextHolder
{
public:
TextBox(std::string text) : text_(std::move(text)) {}
private:
std::string text_;
};
```
Here, the caller has option to avoid copy and we only have single constructor.
Reviewed By: fkgozali, JoshuaGross
Differential Revision: D33276841
fbshipit-source-id: 619d5123d2e28937b22874650366629f24f20a63
Summary:
Renaming the `better` utilities to `butter`:
- to prevent claims that this library is superior to others - it really depends on use cases
- to indicate ease of use throughout the codebase, easily spread like butter
Changelog: [C++][Changed] Renaming C++ better util to butter, used by Fabric internals
Reviewed By: JoshuaGross
Differential Revision: D33242764
fbshipit-source-id: 26dc95d9597c61ce8e66708e44ed545e0fc5cff5
Summary:
changelog: [internal]
Just moving code that doesn't belong to UIManagerBinding out of the class.
Reviewed By: philIip
Differential Revision: D33060412
fbshipit-source-id: 2d54929072cef14fd1fa6b70bde382ae21ecff45
Summary:
Applies suggestions from default preset that make sense in our codebase
Changelog: [Internal]
Reviewed By: sammy-SC
Differential Revision: D33010231
fbshipit-source-id: 6bc9edf01fd9bd9938d211e3494dd1a127b24eaa
Summary:
The IDE warning suggests that passing folly::dynamic by value will create a copy on each call.
Changelog: [Internal]
Reviewed By: JoshuaGross
Differential Revision: D32978154
fbshipit-source-id: a47a60c332a9d299eb2110d3537dfab0bc2398b6
Summary:
changelog: [internal]
UIManager::animationTick is thread safe. Let's make it obvious by marking the method const.
Reviewed By: javache
Differential Revision: D32669102
fbshipit-source-id: 49e35d0f0a5c5d1b03baa1cbf9cdece082909e85
Summary:
Changelog: [internal]
Nothing was using `stop` parameter, let's get rid of it.
Reviewed By: philIip
Differential Revision: D32669018
fbshipit-source-id: dc2d52048a2f7dd3785dd959270087001c778962
Summary:
Unfortunately, parsing some props requires stateful context - namely, PlatformColor on Android. We explored several different options but they all seemed inferior to the approach of using ContextContainer, and most would require using global state.
By introducing this change everywhere as early as possible, we can avoid later pain. It is likely that some prop, on some platform, will require this mechanism. We'll be ready for it!
Because we can pass a constref of the ContextContainer through to all props and because the context and context data is never retained by prop parsers, perf and memory hit should be ~0.
This diff contains core changes only. Leaf changes to all props structs and conversions files will be in next diff(s).
Changelog: [Internal]
Reviewed By: mdvacca
Differential Revision: D29838789
fbshipit-source-id: f5090e7f02eb6e8fbe0ef4dd201e7d12104a3e3c
Summary:
Changelog: [internal]
Original commit changeset: 390e773283a7
Original diff didn't resolve crash T82119127. Let's back this out and investigate again.
Reviewed By: mdvacca
Differential Revision: D29909878
fbshipit-source-id: 54b054f4a5c97b8f6369f8790c2cf32a56c4fbd2
Summary:
I'm hunting down the source of a perf regression on a screen and think that having these systrace sections could be handy for this and future investigations.
Changelog: [internal]
Reviewed By: mdvacca
Differential Revision: D29802969
fbshipit-source-id: f4030261da8888ddeb32ae41b9cf2b25af6a5583
Summary:
Changelog: [internal]
[This assumption](https://fburl.com/diffusion/yeqtvxru) in `~Scheduler` might not be completely correct. There can be a race between JS thread and main thread. With main thread settings animationDelegate_ to nullptr and JS thread reading the value. Even though they always happen one after the other, JS thread can have the value of animationDelegate_ in its cache line and changing it from main thread might not invalidate it right away.
Reviewed By: JoshuaGross
Differential Revision: D29237290
fbshipit-source-id: 6cb898caf7c225cf8162e7560921b563dec514b1
Summary:
Changelog: [internal]
[This assumption](https://fburl.com/diffusion/yeqtvxru) in `~Scheduler` might not be completely correct. There can be a race between JS thread and main thread. With main thread settings delegate_ to nullptr and JS thread reading the value. Even though they always happen one after the other, JS thread can have the value of delegate_ in its cache line and changing it from main thread might not invalidate it right away.
Reviewed By: mdvacca
Differential Revision: D29233092
fbshipit-source-id: 390e773283a745742d78152b9921c9ce21148a4c
Summary:
PreAllocation currently always happens at revision 0 (after ShadowNode creation), and all CREATE mutations are triggered for ShadowNodes at revision 1 or higher (since CREATE mutations are generated by the differ, it means that all ShadowNodes have revision 1 or higher when CompleteRoot is called). This means that between PreAllocation and CREATE, we /always/ expect at least one clone.
It is possible for a node to be "non-view-forming" at revision 0, causing view preallocation to be skipped, and "view-forming" at revision 1 (causing the CREATE mutation to be thrown away, since all CREATE mutations of revision 0 or 1 are thrown away). This causes a crash. It is extremely marginal, but there are repros in the wild.
Thus, I'm introducing one new UIManager and Scheduler delegate method that allows the mounting layer to be notified of clones. If a clone from rev 0->1 results in a node going from non-view-forming to view-forming, we can preallocate then, as well.
This resolves this crash, and allows us to keep experimenting safely with this View PreAllocation optimization. I believe all edge-cases are accounted for.
Changelog: [Internal]
Reviewed By: sammy-SC
Differential Revision: D29043426
fbshipit-source-id: dff11d3140ded1cbb02a5518a3aeb52dc812cc50
Summary:
Changelog: [internal]
Extend LeakChecker so it is available on Android (or any other platform as it is completely in C++ now).
Reviewed By: JoshuaGross
Differential Revision: D28600243
fbshipit-source-id: c77a003e3ffc6171e61c998508c9f34f10bb65ca
Summary:
This diff extends startSurface and setSurfaceProps methods with the new parameter called displayMode
changelog: [internal] internal
Reviewed By: yungsters
Differential Revision: D27669847
fbshipit-source-id: c2ddb690ca897e46e00f07b491b91bb2bc8e847d
Summary:
This diff introduces a new fabric API called setSurfaceProps that will be used to call the new JS api "setSurfaceProps"
changelog: [internal] internal
Reviewed By: sammy-SC
Differential Revision: D27607588
fbshipit-source-id: 36a19f728af244d7e72687d9305b1c568e2b8ec6
Summary:
In the previous diff, we moved the destruction of a ShadowTree object from `UIManager::stopSurface` method. And the next logical step is to align `UIManager::startSurface` method with the symmetrical approach. Now `stopSurface` accepts a unique pointer to a ShadowTree and `stopSurface` returns it back to the caller; this way we can avoid returning a raw pointer from `startSurface`.
Changelog: [Internal] Fabric-specific internal change.
Reviewed By: sammy-SC
Differential Revision: D27010533
fbshipit-source-id: b9211fb7e67763cc190d6b8c86cb866d20c6693d
Summary:
Changelog: [internal]
Introducing LeakChecker. A tool that checks if all native components have been cleaned up when surface is stopped.
**Known shortcomings**:
- LeakChecker is only enabled in debug builds and the existence of leaks is logged to console.
- For now, Leak Checker looks at N-1 screen. This is intentional as there is a known limitation of React that doesn't free up all shadow nodes when surface is stopped. Because of this, the use of LeakChecker is not intuitive and I'll work with React team to try to work around this.
- It doesn't help locating the leak, it only informs that leak is present. I'll be looking into ways to help locate the leak.
Reviewed By: JoshuaGross, mdvacca
Differential Revision: D26727461
fbshipit-source-id: 8350190b99f24642f8e15a3c2e1d79cfaa810d3d
Summary:
The non-Fabric API has a `blockNativeResponder` param in setJSResponder. Make sure to pass that along in Fabric.
On Android this allows us to respect the flag and do the same thing non-Fabric was doing if `blockNativeResponder` is false. It's not clear yet what the impact is on iOS.
Changelog: [Internal]
Reviewed By: mdvacca
Differential Revision: D27058806
fbshipit-source-id: aa5074fa46191d78f5292a93d9040ab4bb58ca66
Summary:
As part of stoping a Surface, we have to commit an empty shadow node tree. And to avoid a deadlock we have to do it not under a `linkMutex_`. To do so, we need to move the part that commits an empty tree from UIManager to SurfaceHandler. And this diff does exactly this.
Changelog: [Internal] Fabric-specific internal change.
Reviewed By: sammy-SC
Differential Revision: D27010535
fbshipit-source-id: 60cc79b1c81d7340b550e78653737d2cc5bec24d
Summary:
See react_native_assert.{h,cpp}. Because of the BUCK+Android issue where NDEBUG is always defined, we use react_native_assert instead of assert to enable xplat asserts in debug/dev mode.
This migrates most of the codebase, but probably not 100%. The goal is to increase assertion coverage on Android, not to get to 100% (yet).
Changelog: [Internal]
Reviewed By: RSNara
Differential Revision: D26562866
fbshipit-source-id: a7bf2055b973e1d3650ed8d68a6d02d556604af9
Summary:
This is a Fabric-compliant implementation of `JSResponder` feature. To make it work e2e we also need to update FabricRenderer in React repository. But before we can do this, we need to ship the native changes.
Changelog: [Internal] Fabric-specific internal change.
Reviewed By: mdvacca
Differential Revision: D24630027
fbshipit-source-id: 70c30e1250b554d83862956b536714704093072f
Summary:
This implements `SurfaceHandler`, a new primitive for controlling the Surface life-cycle that ensures ownership, preserves state, maintains internal invariants, and simplifies surface manipulation from the application side.
For now, all this is an unused code. The coming diff will introduce an experiment that will route all surfaceId-based APIs to SurfaceHandler-based ones.
Changelog: [Internal] Fabric-specific internal change.
Reviewed By: sammy-SC
Differential Revision: D24290777
fbshipit-source-id: f2ff2b58e6d46e971a548f9f02113a1c783c4940
Summary:
This changes how we access UIManagerBinding in Fabric code. Instead of storing a pointer to it and accessing it, now we just request it from JavaScript runtime on demand. After a week of testing, I hope we will be able to simplify retaining and referencing logic between UIManager and UIManagerBinding. The change is gated just in case to have the ability to turn this new implementation off.
Changelog: [Internal] Fabric-specific internal change.
Reviewed By: sammy-SC
Differential Revision: D26082788
fbshipit-source-id: 211c551fe1aaba9b0bff18478e40e1e1f32e999c
Summary:
This is the core (Cxx) Fabric implementation of sendAccessibilityEvent. No platform support is added in this diff (Android and iOS do not build yet).
See following diffs.
Changelog: [Internal]
Reviewed By: mdvacca
Differential Revision: D25852412
fbshipit-source-id: 88b7d8e2e251183dfe7fb377a4cffbac8f458656
Summary:
While pulling in additional react-native code into react-native-windows to support fabric, I hit some bugs with the MSVC compiler. These minor changes work around those bugs and shouldn't have any functional effect on the code
## Changelog
[Internal] [Fixed] - Minor code tweaks to work around a compiler bug in MSVC.
Pull Request resolved: https://github.com/facebook/react-native/pull/30696
Test Plan: Code compiles in MSVC and clang.
Reviewed By: shergin
Differential Revision: D25830710
Pulled By: appden
fbshipit-source-id: 796bd8191ea62fe62e09dd16cc56b8edc092297f
Summary:
This replaces the internal core implementation of `setState` with the new `updateStateWithAutorepeat` which is now the only option.
In short, `updateStateWithAutorepeat` works as `setState` with the following features:
* The state update might be performed several times until it succeeds.
* The callback is being called on every retry with actual previous data provided (can be different on every call).
* In case of a static value is provided (simple case, not lambda, the only case on Android for now), the same *new*/provided value will be used for all state updates. In this case, the state update cannot fail.
* If a callback is provided, the update operation can be canceled via returning `nullptr` from the callback.
This diff removes all mentions of the previous state update approach from the core; some other leftovers will be removed separatly.
Changelog: [Internal] Fabric-specific internal change.
Reviewed By: sammy-SC
Differential Revision: D25695600
fbshipit-source-id: 14b3d4bad7ee69e024a9b0b9fc018f7d58bf060c
Summary:
Changelog: [internal]
The loop inside `UIManager::getNewestCloneOfShadowNode` is unnecessary as ancestorList already has index of the children.
Reviewed By: shergin
Differential Revision: D25681296
fbshipit-source-id: e8fab6c6d945064b0a7e63b6fd22697054fb7ea2
Summary:
The new UIManagerCommitHook can be used to implement commit-atering features without modifying the core.
E.g. State Reconciation seems can be implemented as a coomit hook but first we will use it for a new feature called Timeline (see the coming diffs).
Changelog: [Internal] Fabric-specific internal change.
Reviewed By: sammy-SC
Differential Revision: D25221312
fbshipit-source-id: dbe41b475bc8b36e1780d81447ab43b32758bdff
Summary:
With the change, a new delegate method allows a receiver to alter a new (proposed) shadow tree with another tree by returning the altered tree.
We will it use in future diffs to implement Commit Hooks.
Changelog: [Internal] Fabric-specific internal change.
Reviewed By: sammy-SC
Differential Revision: D25221313
fbshipit-source-id: 9f83577d862b713fff71fa365ce660cc1de87c84
Summary:
Changelog: [internal]
Background executor performs unnecessary operations when second `completeRoot` message from React arrives before first `completeRoot` was finished. This produces unnecessary `ShadowViewMutations`.
Mechanism:
Everytime `completeRoot` is received, before the call is dispatched on the background queue, `completeRootEventCounter_DO_NOT_USE_` is incremented.
Inside `ShadowTree::tryCommit` we check if the value has been incremented to determine if another `completeRoot` is queued.
Reviewed By: JoshuaGross
Differential Revision: D24419160
fbshipit-source-id: 11e19026feca01db6c8981b093a691a6b58a006f
Summary:
Changelog: [internal]
Fabric uses view commands instead of setNativeProps. This diff removes what's left of setNativeProps from the core.
Reviewed By: JoshuaGross
Differential Revision: D24309999
fbshipit-source-id: 70e54f0a984f8c36f77ba2cd59f59fc6923bc832
Summary:
Exceptions in C++ work quite differently from exceptions in other languages. To make exceptions actually work **correctly** all the code needs to be written with "exceptions in mind" (e.g., see https://www.stroustrup.com/except.pdf). In short, if the code is not "exceptions ready", throwing an exception causes memory leaks, dangling pointers, and invariant violations all over the place, which will probably cause another crashes down the road (which will be especially hard to investigate and attribute to the original issue).
Fabric Core (Layout, Props parsing, ShadowNodes management, and so on) does not use exceptions because in most (all?) the cases the exception is now recoverable. So, if a program detects some internal state invariant violation or missing some resource, *logically* it's fatal. We also don't want to pay code-size and performance tax for exception support, so that's why we don't use them. It's just not the right fit for Fabric Core.
This does not mean that exceptions don't happen though. C++ standard library can throw them... sometimes. And if our library is compiled with exceptions enabled (still the case, unfortunately), an exception can bubble to JavaScript code and losing all context down the road. And it's hard to investigate such crashes. To isolate those occasional exceptions inside C++ core we are marking all C++/JS boundaries with `noexcept` that stops the bubbling.
I hope that will give us much more informative crash reports.
Changelog: [Internal] Fabric-specific internal change.
Reviewed By: sammy-SC
Differential Revision: D23787492
fbshipit-source-id: 0822dbf36fc680c15b02b5cd0f2d87328296b642
Samuel Susla
Phillip Pan
Samuel Susla
Pieter De Baets
Krzysztof Magiera
Andres Suarez
Kevin Gozali
Andrei Shikov
Joshua Gross
David Vacca
Valentin Shergin
Andrew Coates