Summary:
Bridgeless mode hasn't been able to load random images. I was able to repro this 100% with base64 images. Loading these images hits a particular flow in `RCTImageLoader` which relies on the bridge to access `RCTNetworking`. This diff uses the TM Lookup Delegate as a fallback.
Changelog: [iOS][Internal] Fix error in RCTImageLoader in bridgeless mode
Reviewed By: sammy-SC
Differential Revision: D19331467
fbshipit-source-id: 8239ee258425da4ed8cb9f6dcdcd7f37c162eb19
Summary:
When bridge is nil, RCTBlobManager crashes with `EXC_BAD_ACCESS`.
The fix: instead of using the bridge to find RCTNetworking, use TM lookup delegate.
Changelog: [iOS][Internal] Fix crash in RCTBlobManager in bridgeless mode
Reviewed By: mdvacca
Differential Revision: D19320136
fbshipit-source-id: 49aabb3ce53b2ec620fcc02be1c6c1b44066f440
Summary:
View should reset whether we are inside of a text or not. For example, inline images should only be rendered inside text, but if we have a view inside text, then it should render a regular image, not an inline image.
This logic *should* exist in native instead of in JS, but this is an easier change for now.
I'm sad to have to turn this back into a JS component instead of just being the string 'RCTView' as this will have performance implications on all surfaces, but this is how it always used to be so maybe it's fine.
This example previously crashed, and no longer does:
```
function PlaygroundContent(props: {}) {
return (
<View style={styles.container}>
<Text>
<View style={{width: 10, height: 10}}>
<Image source={fbicon.filled('chevron-down', 10)} />
</View>
</Text>
</View>
);
}
```
Changelog:
[General][Fixed] Fixes bug where <Text><View><Image> would crash.
Reviewed By: JoshuaGross
Differential Revision: D17564510
fbshipit-source-id: 0ecf49b3d466e7adf57a46a7a097dd3798c721a4
Summary:
We're keeping the perf logging flag, but the general instrumentation flag is no longer needed (the use case isn't significant to need separate flag).
Changelog: [Internal]
Reviewed By: JoshuaGross
Differential Revision: D19304789
fbshipit-source-id: 094f7079283e4f4bf3d546dfd0ae50eb641d562e
Summary:
Hand writing view configs for NativeImageViewComponent so that it'll work in bridgeless mode and won't fall back to the UIManager.
Changelog: [Internal]
Reviewed By: ejanzer
Differential Revision: D19217961
fbshipit-source-id: d5a123b35a75ba3e22c57b1dde18a47893681614
Summary:
It can be a pain to debug slow network issues, especially with the iOS simulator which doesn't have a network link conditioner. This makes it really easy and predictable by simply adding a `setTimeout` around calling `sendRequest`.
Changelog:
[General] [Added] - DEBUG_NETWORK_SEND_DELAY can be used to simulate slow network.
Reviewed By: PeteTheHeat
Differential Revision: D19236911
fbshipit-source-id: 14762c7e0c6408a8364aa569c482729a7a1fe740
Summary:
It is time to target SDK version 10.0+.
Changelog: [iOS] [Deprecated] - Deprecating support for iOS/tvOS SDK 9.x, 10.0+ is now required
Reviewed By: mdvacca
Differential Revision: D19265731
fbshipit-source-id: 93b6f9e8f61c5b36ff69e80d3f18256aa96cc2c0
Summary:
Changes the return type of StyleSheet.create to `$ReadOnly<S>`. At runtime, it just freezes the fields in dev mode and is the identity function otherwise. This cause 423 errors due to the wrong props being passed to different styles: P124372727
I went over errors at random with Eli to get confidence in the change.
Changelog:
[General][Fixed] Fix return type of StyleSheet.create
Reviewed By: zackargyle, TheSavior
Differential Revision: D19218817
fbshipit-source-id: b7aa10700f8333372c9f0bd61bfbef371caadf97
Summary:
In https://fb.workplace.com/groups/rn.support/permalink/3132785890103338/?comment_id=3132873900094537 we identified that when Metro restarts but the app keeps the main bundle alive, the next request for a split bundle will always produce an invalid bundle. This is because we rely on module ids being stable while Metro is running and when Metro is restarted, those module ids will likely be assigned to different modules.
The fix applied here simply forces a full reload when downloading (registering) a new split bundle if the connection to Metro disappeared.
Another way to mitigate this is to use stable module ids. That's a more involved change in Metro that I'm not planning on making right now.
Changelog: [Internal]
Reviewed By: rickhanlonii
Differential Revision: D19235555
fbshipit-source-id: b8ff10590655952a2af1379bbf4e930f9eb95c1b
Summary:
This diff removes the findShadowNodeByTag_DEPRECATED method from class UIManagerBinding. This method was created on D17175953 to implement findNodeHandle in Fabric. Recently we decided that we don't need to expose findNodeHandle anymore. This diff cleans up thi code.
Changelog: [internal]
Reviewed By: JoshuaGross
Differential Revision: D19216996
fbshipit-source-id: 07c16aeed28851afc09b0f5d6be338bb4440f813
Summary:
We currently use `UIManager.getViewManagerConfig()` in a few places to check if a view manager exists at runtime; right now these checks fail for Venice because our DummyUIManager returns null. We don't yet have a replacement for this API for Fabric (see T59411524) so for now let's just hard code the view managers that we know exist to return something truthy here. This is hacky, but it works ok for now.
Changelog: [Internal]
Reviewed By: PeteTheHeat
Differential Revision: D19185957
fbshipit-source-id: e3135b45db4c9ff27fd95362deaa3c1b4340bf48
Summary:
The logic in the method is only relevant if the sourceType is `UIImagePickerControllerSourceTypeCamera`. So we guard it here, to prevent NSInvalidArgumentException.
Changelog: [Internal]
Reviewed By: sammy-SC
Differential Revision: D19184818
fbshipit-source-id: 5c99dabfaed23629e22fdc888263bd09e8768c88
Summary:
This diff fixes two issues with LogBox in OSS.
Changelog: [Internal]
Reviewed By: mmmulani
Differential Revision: D19165696
fbshipit-source-id: 4dd9c480ea3d2b205f8e65cb32203949b1e32e8a
Summary:
When in debug mode, PlatformConstants doesn't have a `ServerHost` constant. This wasn't captured by `NativePlatformConstants`.
Changelog:
[Android][Fixed] - Make sure ServerHost is optional in NativePlatformConstants.js
Reviewed By: rickhanlonii
Differential Revision: D19091270
fbshipit-source-id: 4b3ac73a4ab6111b3e433ecca01cc769e8cdec6b
Summary:
`node` is passed to us from react devtools. We can call .measure on that directly instead of using findNodeHandle and UIManager. This makes this call compatible with Fabric.
Changelog:
[Internal]
Reviewed By: mdvacca
Differential Revision: D19152052
fbshipit-source-id: 05ab6d55f3a79221d99289f0341eb25352b96899
Summary:
We currently use `BatchedBridge.registerCallableModule` to call JS from platform code in a bunch of places (e.g. `RCTDeviceEventEmitter.emit()`), including some pretty essential ones - for example, Networking relies on `RCTDeviceEventEmitter` to emit events to JS for network requests and errors. In order to make the migration to bridgeless mode as easy as possible, it'd be best if we didn't have to update all the callsites in JS and platform code, at least for core JS modules like `RCTDeviceEventEmitter`. So we'd like to implement this callable module pattern for bridgeless mode as well.
In this diff, I'm installing a global variable, `RN$registerCallableModule`, which will store the JS module name and factory function in C++. This can then be called from platform code (Java/Obj-C) to invoke a JS function on a module using the existing APIs (`context.getJSModule()` on Android, `enqueueJSCall()` or similar on iOS).
Changelog: [Internal]
Reviewed By: PeteTheHeat
Differential Revision: D18798852
fbshipit-source-id: fbe9eaecaf6f5cab5173beec999d3a933b13375d
Summary:
We will need to operate on instances instead of react tags. This diff passes through the instance instead of the tag.
This does not make the inspector work in Fabric, but it is a necessary refactor in that direction.
Changelog:
[Internal]
Reviewed By: mdvacca
Differential Revision: D19149647
fbshipit-source-id: 53fa35a2886bf36a146d473a496409dffc114a80
Summary:
Some babel plugins add additional methods to `Function.prototype` (see https://github.com/MatAtBread/fast-async):
```js
Function.prototype.$asyncbind = function $asyncbind(self, catcher) {
...
```
Although undocumented, React Native allows functions to be passed to `StyleSheet.create()` for dynamic styling:
```js
const styles = StyleSheet.create({
animalItem: height => ({ height })
});
```
If there are additional custom methods on `Function.prototype`, React Native's `StyleSheetValidation.validateStyle` will loop through these properties and raise an error because those properties are not valid style keys, because it loops through _all_ properties, including inherited ones.
This PR modifies `StyleSheetValidation.validateStyle` to only loop through the style's own properties, instead of including inherited ones.
## Changelog
[General] [Fixed] - Fix stylesheet validation for functions with custom prototype methods.
Pull Request resolved: https://github.com/facebook/react-native/pull/27264
Test Plan: - Tested that non-function style properties are still validated
Differential Revision: D18694895
Pulled By: hramos
fbshipit-source-id: b36f4a62a435e7b6a689398de3bcc06d6bb14293
Summary:
In https://github.com/facebook/react-native/issues/25427, radex added initial support for running React Native projects on macOS via Catalyst. However, `RCTWebSocket` was disabled for that target because of some compilation issues. This meant that running projects via a connection to the packager wasn't possible: no live reload, and projects must be run in "Release" mode. It also meant making manual changes to Xcode projects deploying to macOS and scattering a number of conditional checks throughout the codebase.
In this change, I've implemented support for `RCTWebSocket` on the macOS target and re-enabled the affected features. Live reload and the inspector now work for macOS targets. Manual modifications of Xcode build settings are no longer necessary for react-native projects running on macOS.
### Limitations
There's no binding which displays the developer menu (since there's no shake event on macOS). We'll probably want to add one, perhaps to the menu bar.
I've chosen not to commit the modifications to RNTester which enable macOS support, since that would imply more "official" support for this target than I suspect you all would like to convey. I'm happy to add those chunks if it would be helpful.
## Changelog
[iOS] [Added] - Added web socket support for macOS (Catalyst), enabling debug builds and live reload
Pull Request resolved: https://github.com/facebook/react-native/pull/27469
Test Plan:
* Open RNTester/RNTester.xcodeproj with Xcode 11.2.1, run it like a normal iOS app -- make sure it compiles and runs correctly (no regression)
* Select "My Mac" as device target, and run. You may need to configure a valid development team to make signing work.
* RNTester should run fine with no additional configuration. Modify a file in RNTester, note that live reload is now working.
* Test the developer inspector. To display the developer menu, you'll need to manually show it; here's an example diff which does that:
```
diff --git a/RNTester/js/RNTesterApp.ios.js b/RNTester/js/RNTesterApp.ios.js
index 8245a68d12..a447ad3b1b 100644
--- a/RNTester/js/RNTesterApp.ios.js
+++ b/RNTester/js/RNTesterApp.ios.js
@@ -19,6 +19,8 @@ const React = require('react');
const SnapshotViewIOS = require('./examples/Snapshot/SnapshotViewIOS.ios');
const URIActionMap = require('./utils/URIActionMap');
+import NativeDevMenu from '../../Libraries/NativeModules/specs/NativeDevMenu';
+
const {
AppRegistry,
AsyncStorage,
@@ -143,6 +145,7 @@ class RNTesterApp extends React.Component<Props, RNTesterNavigationState> {
UNSAFE_componentWillMount() {
BackHandler.addEventListener('hardwareBackPress', this._handleBack);
+ NativeDevMenu.show();
}
componentDidMount() {
```
Reviewed By: sammy-SC
Differential Revision: D18945861
Pulled By: hramos
fbshipit-source-id: edcf02c5803742c89a845a3e5d72bc7dacae839f
Summary:
With a Picker we would like to allow accessibility labels to be passed as a prop for situations where we want go give more detail. For example if we have a number picker that will be used for a timer instead of just saying 3, we might want to say 3 hours.
## Changelog
[General] [Added] - Picker test with an accessibility label prop
[General] [Added] - Support for accessibility Label prop to the Picker component
Pull Request resolved: https://github.com/facebook/react-native/pull/27342
Test Plan: Test plan is testing in RNTester making sure the examples work
Differential Revision: D18770184
Pulled By: hramos
fbshipit-source-id: e6f8ab4a9c50f3fb46342198441ecc71394913d3
Summary:
Added basic hook to enable image instrumentation. The hook passes information to the existing image loader, where instrumentation is done, specific for each app, if any.
Changelog: [Internal]
Reviewed By: mdvacca
Differential Revision: D19047899
fbshipit-source-id: 6c013806cce98bcf1ea240d696a7ede9697e5cd9
Summary:
The `RCTImageURLLoaderWithAttribution` protocol historically only returns a cancellation block to cancel the url request. But for more complex instrumentation, we may need to associate a requestId for the specific URL request. To do this, the protocol now returns an object that has both the unique ID and the cancellation block, so that instrumentation logic can refer to the ID in the future.
Note that the `RCTImageURLLoader` protocol is unchanged, because the request ID is only relevant for instrumentation purpose.
Changelog: [Internal]
Reviewed By: mdvacca
Differential Revision: D19047900
fbshipit-source-id: cd029f2470c32fc7bffd674b09a5353fe1dbc80b
Summary:
`RCTImageLoader ` needs loaders & decoders. These can be passed in through init, or looked up at runtime using the bridge. In bridgeless mode, we need to pass them in through init.
This diff just moves the assert to actually protect `_bridge` usage, since there exist other codepaths which don't use the bridge.
Changelog: [iOS][Internal] Move RCTImageLoader asserts to protect bridge usage
Reviewed By: RSNara
Differential Revision: D18990760
fbshipit-source-id: 285008e7b3f473c45381ab472c3d86e5d8e218ae
Summary:
Refactors `Pressability` so that updates to the configuration are now explicitly committed using `configure()`.
Previously, the configuration was updated implicitly because `Pressability` accepted a series of functions whose closures capture values (e.g. `this.props`). Although these changes typically happen when component instances are "atomically" updated by React, it is not a guarantee. For example, arbitrary instance variables could be used to configure `Pressability`, and they could be muted at any time.
This change makes updates to the configuration of `Pressability` more predictable.
Changelog:
[Internal]
Reviewed By: TheSavior
Differential Revision: D18742620
fbshipit-source-id: d2e96dd1e3643289daab2177199a29f80d17b0bc
Summary:
`selectAll:` does not work for UITextView when it's being called inside UITextView's delegate methods.
This is a bug in UIKit and this quite ugly workaround is only a known solution for it.
Changelog: [Bug][iOS] Fixed bug in implementation of <TextInput>'s selectOnFocus prop
Reviewed By: sammy-SC
Differential Revision: D18966755
fbshipit-source-id: 2ba15fa94570d463a2ea3490331fb879611dc6b8
Summary:
Fixes https://github.com/facebook/react-native/issues/16067
The issue is due to a race between `onLayout` and `onContentSizeChange`, which in general should be fine because there is no expectation of ordering between the two, and only causes issues with certain configurations.
The bug can be triggered if `initialNumToRender` is smaller than needed to fill past the `onEndReachedThreshold` (say the default, 10, is only 580px tall, but it takes 15 to reach the threshold). This will cause an incrementally render of more items to try and fill the viewport. The problem is that if the `onLayout` comes back before the first `onContentSizeChange`, it will first do the state increment to render 20 items and then the stale `onContentSizeChange` callback from 10 items will fire and we'll think that the content size for 20 items is 580px when in fact it's 1160px (which is past the threshold). If those 20 items are also all of our available data, then we'll call `onEndReached` because we think we've rendered everything and are still within the `onEndReachedThreshold`.
The fundamental problem here is the system getting confused when a stale async `onContentSizeChange` comes in after increasing `state.last`. I wish there was a concrete timeframe, but Fabric will give us more flexibility to do things synchronously so hopefully we can avoid class of issues once that roles out.
The fix here simply adds a check to make sure `contentLength` has been set before adjusting the render window so it's not possible to increase the window size before the initial `onContentSizeChange` callback fires.
For completeness, there are a few user-code workarounds to avoid this issue entirely:
1) Provide the `getItemLayout` prop so the list doesn't have to rely on async layout data (you should do this whenever possible anyway for better perf). e.g. for the original snack example, you can just add `getItemLayout={(d, index) => ({length: 58, offset: 58 * index, index})}` since all the rows are height 58 and the issue will no longer repro. Note this is fragile and must be kept in sync with UI changes, a11y font scaling, etc - a more robust approach could be to render a single representative row offscreen and measure it with `onLayout` then use that value.
2) If `getItemLayout` is not feasible to compute for your UI, increase `initialNumToRender` to cover the `onEndReachedThreshold`.
3) And/or add your own logic to protect against extra calls to `onEndReached` as others have suggested.
Changelog:
[General][Fixed] - Fix sporadic issue with onEndReached called on load when not needed
# Test Plan
Adds a new jest test that fails without this fix and succeeds with it.
Reviewed By: TheSavior
Differential Revision: D18966721
fbshipit-source-id: de05d9f072e24a2faf351e7f5d60578a31def996
Summary:
Remove textColor font and textAlignment in RCTBackedTextInputViewProtocol since they're all wired up in defaultTextAttributes already
Changelog: [iOS] [Removed] Remove three properties: "textColor" "font" "textAlignment" from RCTBackedTextInputViewProtocol, unifying the usage into "defaultTextAttributes".
Reviewed By: PeteTheHeat
Differential Revision: D18954292
fbshipit-source-id: 17fd38a824d035843ae59b1b875dd9c48b3fcc9b
Summary:
Update LogBox on iOS to lazily initialize, using a synchronous RCTSurface, behind RCTSharedApplication checks.
This results in faster of LogBox, without keeping around a long lived window in the background, and only used when LogBox is used.
On Android, we still start the react app in the background but we create a dialog when it's shown and then destroy it when it's hidden. Once we have the sync APIs on android we can update it to use the same strategy.
Changelog: [Internal]
Reviewed By: fkgozali
Differential Revision: D18925538
fbshipit-source-id: 1a72c39aa0fc26c8ba657d36c7fa7bc0ae777eb9
Summary:
We can use the HostComponent type now instead of NativeComponent
Changelog:
[Internal]
Reviewed By: zackargyle, rickhanlonii
Differential Revision: D18871289
fbshipit-source-id: 3c70369c5848dedfc22ca6f6ccbb69d6d60a1330
Summary:
RCTBackedTextInputViewProtocol is a protocol that describes which capabilities must be exposed for some TextInput-like UIView to be compatible with the rest of TextInput infra. We use that in both implementations. In Classic React Native we use Objective-C runtime mechanisms to call many of those methods (via ViewManager's directives); that masked the problem that the protocol was incomplete.
In Fabric, we call all methods normally, so we need to declare all of them here.
Changelog: [Internal] Fabric-specific internal change.
Reviewed By: sammy-SC
Differential Revision: D18950432
fbshipit-source-id: 3cd5cb57a033f7c43fd5f2c10f9706408e8e6d0c
Summary:
Both RCTUITextField and RCTUITextView must maintain the same public interface that describes all possible features that TextInput needs. That features are declared via `RCTBackedTextInputViewProtocol`.
`scrollEnabled` is a part of this protocol. The idea behind this prop was borrowed from `UIScrollView` class (which `UITextView` extends) and implemented for `RCTUITextField` (yeah, implemented as no-op).
In this diff we change the implementation of this prop to be more consistent with original implementation in UIScrollView:
* Now the name of the getter is now `isScrollEnabled`;
* The object now retains the actual value of the prop (even if it ignores that).
We need all those features for Fabric-compatible implementation.
Changelog: [Internal] Fabric-specific internal change.
Reviewed By: sammy-SC
Differential Revision: D18950426
fbshipit-source-id: 50c399d0fbba1be31750dbe4f235a075f86e8c01
Summary:
We need to expose the `multiline` prop in View Manager class to enable Fabric component to use it.
Changelog: [Internal] Fabric-specific internal change.
Reviewed By: sammy-SC
Differential Revision: D18950429
fbshipit-source-id: 69b225b24b4512a1ea21381daacbf868a23ccf6d
Summary:
The general rule that we follow in RN is that the *exported* .h files must have only `<>` includes. That allows to use them outside of the library where they are defined. Those files are already exported technically, but without this change cannot be used as exported.
New Fabric text input implementation uses them, so we need to change those includes to have `<>` style.
Changelog: [Internal] Specific internal change.
Reviewed By: sammy-SC
Differential Revision: D18950427
fbshipit-source-id: 96cbe4cac9e28761c123bde8ac70464a0078ee6c
Summary:
This diff backports warning filter handling to yellow box. I also backported the skipped warning handling so that ignored patterns do not log to the console, same as LogBox.
Changelog: [Internal]
Reviewed By: gaearon
Differential Revision: D18573288
fbshipit-source-id: 5bf8e86f754adc808313d7ed02f98daaf65de98c
Summary:
It's possible for us to return no constants from the BlobModule. Therefore, I'm correcting the flow-type.
Changelog:
[Internal]
Reviewed By: fkgozali
Differential Revision: D18932328
fbshipit-source-id: 2b415d12effd16eda313d5591825c711a20f9ae3
Summary:
Outside of __DEV__ the app may still try to use LogBox. We're going to fix that but until we do, if it tries to use LogBox outside of dev render null.
Changelog: [Internal]
Reviewed By: motiz88
Differential Revision: D18932666
fbshipit-source-id: ef0f2542a295dc9332cae8b77faed78f8be350fe
Summary:
`JSDevSupport.onSuccess` is called in `JSDevSupportModule.getJSHierarchy`:
```
const JSDevSupportModule = {
getJSHierarchy: function(tag: number) {
try {
const {
computeComponentStackForErrorReporting,
} = ReactNative.__SECRET_INTERNALS_DO_NOT_USE_OR_YOU_WILL_BE_FIRED;
const componentStack = computeComponentStackForErrorReporting(tag);
if (!componentStack) {
JSDevSupport.onFailure(
JSDevSupport.ERROR_CODE_VIEW_NOT_FOUND,
"Component stack doesn't exist for tag " + tag,
);
} else {
JSDevSupport.onSuccess(componentStack);
}
} catch (e) {
JSDevSupport.onFailure(JSDevSupport.ERROR_CODE_EXCEPTION, e.message);
}
},
};
```
If you look at the implementation of `computeComponentStackForErrorReporting`, it returns a `string`. The Java NativeModule also accepts a `String` for the argument to `JSDevSupport.onSuccess`. So, I've changed the `NativeJSDevSupport.onSuccess` method signature to match the native implementation (i.e: accept a string).
Changelog:
[General] [Fixed] - Correct argument types of NativeJSDevSupport.onSuccess
Reviewed By: fkgozali
Differential Revision: D18908306
fbshipit-source-id: 1c9a5c6fe5b3a81b25baed520e586ebf7e2514f8
Summary:
This diff fixes an issue where symbolicated logs were getting stuck in a "pending" state if the timeout occurred before symbolication finished.
Changelog: [Internal]
Reviewed By: sahrens
Differential Revision: D18894154
fbshipit-source-id: ed225962468f67aef40e430aa798f8d426d31027
Peter Argany
Dan Abramov
Eli White
Christoph Nakazawa
Kevin Gozali
Jesse Katsumata
Poyan Nabati
Panagiotis Vekris
Spencer Ahrens
Jordan Brown
David Vacca
Samuel Susla
Emily Janzer
Jakub
Héctor Ramos
Rick Hanlon
Ramanpreet Nara
Freeman Latif
Andy Matuschak
Kevin Gonzales
Tim Yung
Valentin Shergin
Jimmy Zhang
