Summary:
Pull Request resolved: https://github.com/facebook/react-native/pull/45573
This makes this module now be fully in Kotlin instead of having mixed Java/Kotlin sources.
Changelog:
[Internal] [Changed] - Converted com.facebook.react.modules.dialog to Kotlin
Reviewed By: tdn120
Differential Revision: D60035771
fbshipit-source-id: b45fd099c0b353768ab6580eb6a4a3dccf68f07d
Summary:
Pull Request resolved: https://github.com/facebook/react-native/pull/45570
We do have several methods/classes that are `deprecated` in the JavaDoc but not
with an annotation. That's not correct as users will never get those deprecation otherwise
and we'll be forced to keep both implementation around for a longer time.
Changelog:
[Internal] [Changed] - Properly annotate with Deprecated methods that are just deprecated in JavaDoc
Reviewed By: javache
Differential Revision: D60036159
fbshipit-source-id: 466072d6a3fb4f1220e1dc3deaa51a46c714a388
Summary:
Pull Request resolved: https://github.com/facebook/react-native/pull/45567
Changelog: [internal]
(internal because our integration for Perfetto hasn't been released in OSS yet)
In our current React integration for Perfetto we're logging arbitrary time spans via `performance.measure` in specific tracks (that can be custom based on a naming scheme).
For a given track, Perfetto doesn't allow partially overlapping segments (as it's considered to always be a stack of time spans). When logging arbitrary time spans that partially overlap, Perfetto cuts the nested ones to make sure they fit into their suspected parent. This makes the logged data incorrect and makes it hard to understand the performance of an application using this data.
There's a fix for this problem: logging these arbitrary segments/time spans in separate tracks that only share the name. In this case, Perfetto groups the data in the UI but allows overlapping (as they're not really on the same track).
Reviewed By: sammy-SC
Differential Revision: D60010696
fbshipit-source-id: 378ea492c4fafbe55ef97fa91e4fa50bbc1893ae
Summary:
Pull Request resolved: https://github.com/facebook/react-native/pull/45564
Changelog: [internal]
(this is internal because the integration hasn't been enabled in OSS yet)
In our current React integration for Perfetto we're currently creating multiple custom tracks that are spread throughout the process section and it can be hard to identify the source of the information.
This adds a "Web Performance: " prefix to all custom tracks coming from JS to achieve 2 purposes:
* Group them together (in terms of order in the process)
* Clarify the source of the data
Reviewed By: sammy-SC
Differential Revision: D60010695
fbshipit-source-id: 081f5b6417d676c61005114337530a089142e7c6
Summary:
Pull Request resolved: https://github.com/facebook/react-native/pull/45539
Changelog: [internal]
This exposes several classes (interfaces in the spec lingo) related to the Performance API to the global scope, so users can access them directly to do things like refinements using `instanceof`. This also prevents the need from importing the modules from `react-native` directly, which would prevent code sharing with Web.
Reviewed By: rshest
Differential Revision: D59859654
fbshipit-source-id: e1f7afb0c98b394b1f97c3790db2e570e6ba0cd9
Summary:
Pull Request resolved: https://github.com/facebook/react-native/pull/45535
Changelog: [internal]
Small refactor to group things based on the spec where they're defined.
Reviewed By: rshest
Differential Revision: D59911334
fbshipit-source-id: 1c40d6bf82b6cc7be78bd81b652d6855c39a53eb
Summary:
Pull Request resolved: https://github.com/facebook/react-native/pull/45526
Changelog: [internal]
Just using the right interfaces so we can expose them in the global scope and do refinements as necessary using `instanceof`.
Reviewed By: rshest
Differential Revision: D59911144
fbshipit-source-id: 9779e3220f2c6f81955f54506f97142f0f4ffdd4
Summary:
Pull Request resolved: https://github.com/facebook/react-native/pull/45525
Changelog: [internal]
This makes several changes to the Performance API to align it closer with the spec:
* Makes fields of `PerformanceEntry` and subclasses read-only.
* Returns instances of the correct subclass of `PerformanceEntry` to observers.
* Renames `HighResTimeStamp` as `DOMHighResTimeStamp` for alignment with the spec and native
Additionally, I realized that the way we handle `performance.measure` is a bit problematic at the moment. When we call the function, we create a `PerformanceMeasure` instance with the data we receive, and return that value. In parallel, we notify the entry to native, which will in turn notify the observers. But the observers will not get those instances we just created, but new instances of `PerformanceEntry` (not even `PerformanceMeasure`) with the resolved values. At the same time, the `PerformanceMeasure` instance we return doesn't resolve its `startTime` and `duration` based on the indicated marks (when specified as strings). We need to fix this in the future by resolving the timing data synchronously when calling `performance.measure`.
Reviewed By: rshest
Differential Revision: D59911145
fbshipit-source-id: e0be0441f307cc9bdea8795ae88b6f390780fc7b
Summary:
Pull Request resolved: https://github.com/facebook/react-native/pull/45524
Changelog: [internal]
The `durationThreshold` option is only meant to be used with `event` entry types. `mark`, `measure`, `longtask`, etc. shouldn't take that option into account, as per the spec.
Reviewed By: mdvacca
Differential Revision: D59918519
fbshipit-source-id: 0553d46944cbe80a32712ff57140763f2514f734
Summary:
Pull Request resolved: https://github.com/facebook/react-native/pull/45545
ReactInstance creation in bridgeless returns a callback that requires dispatching on the UI thread to complete the init. If that thread is busy (eg with other startup tasks), we delay starting the initial surface unnecessarily.
This diff adds an experiment to return early from that startup task to allow follow-up tasks to be scheduled on the JS thread without waiting for lifecycle changes to have been applied.
Changelog: [Internal]
Reviewed By: markv
Differential Revision: D59961779
fbshipit-source-id: dd6a6fe093a32144ef6823bde5ac94a61d268fd2
Summary:
CircleCI was automatically cancelling an old run if a new commit was pushed on the branch.
GitHub does not have the same behavior enabled by default.
Keep running jobs in a pipeline when there is a new commit is usually wasteful of resources and cost money we can save.
## Changelog:
[Internal] - Cancel old jobs if a new commit is pushed
Pull Request resolved: https://github.com/facebook/react-native/pull/45568
Test Plan: Tested on GHA on this PR
Reviewed By: blakef
Differential Revision: D60035940
Pulled By: cipolleschi
fbshipit-source-id: 88b4dfc8bdd3eded6489a87db285e9544d3a1bcf
Summary:
In Node 20, the script to run unit tests in CI (`scripts/run-ci-javascript-tests.js`) will fail, even when all the Jest tests pass. This happens because one of the JS modules being tested is setting `process.exitCode` (see https://github.com/jestjs/jest/issues/9324#issuecomment-1808090455).
Changes:
- Modified the affected module to throw an exception when failing, instead of setting the exit code
- Adjusted the unit test for that module
## Changelog:
[General] [Fixed] - Remove setting of process.exitCode that breaks Jest tests
Pull Request resolved: https://github.com/facebook/react-native/pull/45562
Test Plan:
Before this change, running `node scripts/run-ci-javascript-tests.js` would fail with Node 20.
After this change, it succeeds.
Reviewed By: blakef
Differential Revision: D60033582
Pulled By: cipolleschi
fbshipit-source-id: 71b7f4495d414e719a9bd2d892bd1bc3045ddd5d
Summary:
Pull Request resolved: https://github.com/facebook/react-native/pull/45554
In this diff I'm extracting the creation of borderWidth into its own method, this is necessary for next diffs of the stack.
Nh behavior change is introduced here
changelog: [internal] internal
Reviewed By: NickGerleman
Differential Revision: D59942306
fbshipit-source-id: 85d39b64deaa4e8a8632d6f4aab72f626e594b5d
Summary:
Pull Request resolved: https://github.com/facebook/react-native/pull/45553
This diff introduces a "temporary" method called getDiffProps to calculate the difference between 2 props and serialize its result into a folly::dynamic map.
changelog: [internal] internal
Reviewed By: NickGerleman
Differential Revision: D59613245
fbshipit-source-id: 3e23cde0113ac2a3904c8daa48a1ca048cd0262d
Summary:
Pull Request resolved: https://github.com/facebook/react-native/pull/45556
I missed updating this, and the e2e test we had isn't running, so switching JS objects to camel case broke this.
Changelog: [internal]
Reviewed By: jorge-cab
Differential Revision: D60003747
fbshipit-source-id: 6b7b1138e3ebbfdb982f5a089804351cc4b198ba
Summary:
Pull Request resolved: https://github.com/facebook/react-native/pull/45550
`ConcurrentModificationException` is happening from `emitScrollEvent()`.
This usually happens if we modify the collection (add/remove) while accessing collection in a foreach loop.
Seems likely add/remove is called from a different thread while in the foreach loop.
Converting to list before we do the foreach as a quick fix.
Changelog: [Internal] - quick fix for exception
Issure reported here: https://fb.workplace.com/groups/rn.support/permalink/26557068097248454/
Reviewed By: mdvacca
Differential Revision: D59991739
fbshipit-source-id: a2fcc798430acaadd07561a5be871967cc8f2c3b
Summary:
Pull Request resolved: https://github.com/facebook/react-native/pull/45549
## Summary
Calling `getCurrentReactContext` on an unmounted `ReactRootView` could lead to a NPE. The method currently returns a nullable value so return type is the exact same. This change checks if the react instance manager is present and returns null early if not.
## Changelog:
[Android] [Fixed] - Adds a null check in react context getter
Reviewed By: zeyap
Differential Revision: D59982179
fbshipit-source-id: bac5c12e7dc4ee3296991063eb3746141b8446bc
Summary:
Pull Request resolved: https://github.com/facebook/react-native/pull/45544
## This diff now does 5 things:
1. removes the old way we used `actions/setup-node` to manage the cache itself.
2. it creates a new `update-node-modules-cache` workflow, which is the only job that will update the node modules cache
3. it create a `yarn-install-with-cache` action that should be used install of directly calling `yarn install --non-interactive`. This will load a cache against a hash of `package.json`.
4. updated the cache reaper to aggressively remove everything but the latest `npm-{{ hash('package.json') }}`.
5. removed a `cache-setup`, which couldn't be used (we're using artefacts now).
## Why are we doing this:
The various `node-cache-` keys for platforms and on various branches accounts for a very large proportion of the cache (10-20%).
We don't frequently change these dependencies, and even when we do running `yarn install` after loading the cache will resolve any issues.
Limiting the cache to `main` and aggressively pruning older cache entries will clean up a lot of "small win" caching.
Changelog: [Internal]
Reviewed By: cortinico
Differential Revision: D59917944
fbshipit-source-id: 4be6f1959e8fde642a4f208f7d19aceba2c3262f
Summary:
Pull Request resolved: https://github.com/facebook/react-native/pull/45530
Enable viewconfig validation for RNTester to catch where static and native viewconfigs are mismatched, during development or contribution time. This relies on `useNativeViewConfigsInBridgelessMode()` which is already enabled in `DefaultNewArchitectureEntryPoint` on Android, and seems to be in iOS AppDelegate `RCTRootViewFactory` as well.
We put it in an early place in the bundle before first render, since we don't have RNTester preludes right now, but I think it is still early enough?
Changelog: [Internal]
Reviewed By: javache
Differential Revision: D59946366
fbshipit-source-id: ae0c23b13a566489a88a7b4e408f61dce314b003
Summary:
This moves the `helloworld` app to build from the artifacts produced by build_npm_package so that we don't rebuild ReactNative Android from source 8 times.
It reduces build time of such jobs from 14mins to 4mins, resulting in 80mins of build time for every test_all run.
## Changelog:
[INTERNAL] - Move helloworld to build from artifacts on Android
Pull Request resolved: https://github.com/facebook/react-native/pull/45517
Test Plan: CI
Reviewed By: blakef
Differential Revision: D59957613
Pulled By: cortinico
fbshipit-source-id: b6c4adcf804af6c8d2661cf56549d037e09aa2c1
Summary:
Pull Request resolved: https://github.com/facebook/react-native/pull/45543
# Changelog:
[Internal]-
Introduced recently, this confuses viewconfig validation vs the vanilla Android view managers.
So this change effectively reverts the exposure of `ScrollView.scrollIndicatorInsets` to Android, achieving the goal in a different way (via the SVC injection workaround for the specific Android platform flavour).
Reviewed By: NickGerleman
Differential Revision: D59960782
fbshipit-source-id: 3b9a49f1466426d909e94bf4d33f1d09fbf822c2
Summary:
Pull Request resolved: https://github.com/facebook/react-native/pull/45542
As we do have several version numbers for external actions all across the codebase,
here I'm aligning all of them to just use the majors.
I'm doing it only for GitHub first party actions as we trust them,
so minor/patch changes can safely be pulled in without code changes.
Changelog:
[Internal] [Changed] - Align github/* action versions on major
Reviewed By: cipolleschi, blakef
Differential Revision: D59959978
fbshipit-source-id: bb07ce0dfd74d9502a2ac0ea90a2b32f55d6d655
Summary:
With the migration to GHA, we can remove all the duplicated jobs from CircleCI.
These are the only 4 jobs remained to migrate
## Changelog:
[Internal] - Remove all the jobs already migrated to GHA
Pull Request resolved: https://github.com/facebook/react-native/pull/45219
Test Plan: CCI is green
Reviewed By: cortinico
Differential Revision: D59156888
Pulled By: cipolleschi
fbshipit-source-id: 193f1f8fa7484154d5295ac36a63bb81a159da6e
Summary:
Pull Request resolved: https://github.com/facebook/react-native/pull/45541
This is another round of letting react-native-bot do the job that the
generic GitHub Action bot was doing.
Changelog:
[Internal] [Changed] - Act On label as react-native-bot
Reviewed By: cipolleschi
Differential Revision: D59959468
fbshipit-source-id: 8e0f7e2e90a40ed2aa265e637c8a809064e22747
Summary:
Nightly/Release workflow are currently broken due to a wrong path reference to a composite action. This fixes it.
## Changelog:
[INTERNAL] - Fix nightly/release workflow
Pull Request resolved: https://github.com/facebook/react-native/pull/45537
Test Plan: CI
Reviewed By: cipolleschi
Differential Revision: D59959185
Pulled By: cortinico
fbshipit-source-id: 02c556d86105eac35e152b4dc09705bc42c8031a
Summary:
Pull Request resolved: https://github.com/facebook/react-native/pull/45533
This input is unused and is causing a warning on the build pipeline.
I'm cleaning it up.
Changelog:
[Internal] [Changed] - Remove unused build-from-source input
Reviewed By: blakef
Differential Revision: D59958184
fbshipit-source-id: 23ba010da077342605afaaee122bc7ceabc89915
Summary:
Added space to $(inherited) string to avoid creation of wrong cpp flags in RCTAppDelegate podspec
<img width="1157" alt="Screenshot 2024-07-18 at 8 51 19 PM" src="https://github.com/user-attachments/assets/29d32d08-e81f-4c25-b8ee-5dccc0f620ea">
## Changelog:
[IOS] [FIXED] - Building of iOS project when RCTAppDelegate is used in the project
Pull Request resolved: https://github.com/facebook/react-native/pull/45520
Test Plan: To test this you can simply change in your node modules and run pod install, the build will now work successfully
Reviewed By: cipolleschi
Differential Revision: D59950906
Pulled By: arushikesarwani94
fbshipit-source-id: 0d58620aa0be7ac4fcbcd309f06df0eef7844016
Summary:
Pull Request resolved: https://github.com/facebook/react-native/pull/45529
`experimental_boxShadow` is not yet part of Android view managers, and when we enable it, we are likely to do a view manager at a time before moving to BaseViewManager.
This causes user-visible errors when viewconfig validation is turned on, since we have a static view config, but not yet a native view config.
This removes the static viewconfig for Android until we start adding setters to view manager.
It is kept in `ReactNativeStyleAttributes` (which I think can have members not in the native view-config, since it has component specific props like tintColor), and iOS base viewconfig. On Fabric iOS, this is part of BaseViewProps, and handled by RCTView, but it looks like the prop (and also `experimental_filter`, `experimental_mixBlendMode`) do not have entries in iOS RCTViewManager, which is fixed in next diff in the stack.
Changelog: [Internal]
Reviewed By: RSNara
Differential Revision: D59939866
fbshipit-source-id: 2781029a0c29ba111ed04edfe9940c6c72f4e5ac
Summary:
Viewconfig processors may still get called for nullish values I think. Most other processors explicitly handle these (but some don't??).
This returns an empty list, like on parse error, when we have a value, but the value is nullish.
Changelog: [Internal]
Reviewed By: joevilches
Differential Revision: D59933611
fbshipit-source-id: 3f1d89d21977bbe01a05e708aadf1a9451d88083
Summary:
Pull Request resolved: https://github.com/facebook/react-native/pull/45512
isBufferingEnabled_ can be read (by design) from multiple threads, but
it's not atomic. Make it so.
Changelog: [Internal]
Reviewed By: cortinico
Differential Revision: D59907026
fbshipit-source-id: ffce54a28404148b3e270fa90dfe850a334ca2f0
Summary:
This PR fixes a case where the user initializes react native in a directory that contains a space.
It was causing pod install to fail because the path to `create-dummy-hermes-xcframework.sh` script wasn't in a string.
## Changelog:
[IOS] [FIXED] - Hermes prepare_command fails with space in path
Pull Request resolved: https://github.com/facebook/react-native/pull/45316
Test Plan:
1. Create a directory with space in path
2. Initialize React Native inside
3. Install pods
4. Check if pod install doesn't fail
Reviewed By: dmytrorykun
Differential Revision: D59912979
Pulled By: cipolleschi
fbshipit-source-id: b2c08d5035a245f8b4d6bfaf562e46d9c5d127b5
Summary:
Pull Request resolved: https://github.com/facebook/react-native/pull/45501
changelog: [internal]
Implement caching for FabricUIManager::getColor. A call to FabricUIManager::getColor takes on average 0.4ms and there can be many of them. The arguments for the call keep repeating in majority of times. Employing a simple caching mechanism here to avoid unnecessary trips via JNI to reduce overhead.
The dependencies in graphics in BUCK are incorrect. I tried to separate this into multiple files and move implementation to .cpp file but this will require a bit of a more restructuring to make it possible.
Reviewed By: mdvacca, dmytrorykun
Differential Revision: D59859754
fbshipit-source-id: 748efce7f0b8c96001b6ac1a4b457b8c9d63fe9c
Summary:
Factor out the Build NPM package job in a separate action for code reuse
## Changelog:
[Internal] - Factor out the Build NPM package job in a separate action for code reuse
Pull Request resolved: https://github.com/facebook/react-native/pull/45493
Test Plan: GHA are green
Reviewed By: robhogan
Differential Revision: D59858572
Pulled By: cipolleschi
fbshipit-source-id: 561a215ba5812352034157aa254999db56fcd31e
Summary:
With the recent changes to the CI, we need to update the test-e2e-local to work with the new artifacts
## Changelog:
[Internal] - Update local-e2e-test to run with the new Android Artifacts
Pull Request resolved: https://github.com/facebook/react-native/pull/45499
Test Plan: Tested locally.
Reviewed By: blakef
Differential Revision: D59902087
Pulled By: cipolleschi
fbshipit-source-id: 84ef78e8dba222bf8a9e3620632fb2a9d286d42b
Summary:
Pull Request resolved: https://github.com/facebook/react-native/pull/45471
Changelog: [internal]
This is a React Native specific modification of the Long Tasks API that refines the logic to detect long tasks considering voluntary yielding checks.
In RN, as opposed to Web, we can have a very long task executing in the JS thread without causing any issues to the responsiveness of the app, as long as the task checks whether it should yield in short intervals. In this case, if the app always checks whether it should yield at least once every 50ms, the task will not be considered "long".
Check the new unit tests to see this behavior in practice.
Reviewed By: sammy-SC
Differential Revision: D55647992
fbshipit-source-id: 82ab41173d4d9deee65b8ade2268c40d7f58c6e2
Summary:
Pull Request resolved: https://github.com/facebook/react-native/pull/45473
This is a basic implementation of the Long Tasks API (https://w3c.github.io/longtasks/).
It detects and reports long tasks when using the Event Loop (in the modern RuntimeScheduler) when a new feature flag for this purpose is enabled.
This doesn't include attribution information at the moment.
Changelog: [internal]
Reviewed By: sammy-SC
Differential Revision: D55491870
fbshipit-source-id: e1ccad9cc6a35073b31230a8cf3a4660ab9a043d
Summary:
Pull Request resolved: https://github.com/facebook/react-native/pull/45492
Changelog: [internal]
When testing the changes in https://github.com/facebook/react-native/pull/45473 / D55491870, I saw that the reported time spans for long tasks didn't perfectly align with the long tasks themselves in traces (Perfetto).
Taking a closer look, I realized that I wasn't doing the conversion between times and durations from `chrono` and `DOMHighResTimeStamp` (a `double`) correctly, and we're doing this conversion very often.
This moves the definition of `DOMHighResTimeStamp` to its own library and adds conversion methods to make sure we don't make this mistake in the future.
Reviewed By: rshest
Differential Revision: D59820241
fbshipit-source-id: c123920de56336da384ddc484f6ac9d287724301
CHEN Xian-an
David Vacca
Nicola Corti
Rubén Norte
Samuel Susla
Pieter De Baets
Riccardo Cipolleschi
Douglas Lowder
Gabriel Donadel
Nick Gerleman
Alan Lee
Peter Abbondanzo
Blake Friedman
szymonrybczak
Ruslan Shestopalyuk
nik910
Alex Taylor (alta)
zhongwuzw
Alexander Blom
Oskar Kwaśniewski