Summary:
public
NSJSONSerialization throws an exception when it encounters bad JSON data, including NaN values, which may not be a programming error.
This diff adds code to catch those exceptions and convert to an error. Also, if no error handling is in place, RCTJSONStringify will now display a redbox, and attempt to recover by sanitizing the JSON data and retrying.
Reviewed By: javache
Differential Revision: D2854778
fb-gh-sync-id: 18e6990af0d91083496d6a0b75c31a94ed9454a5
Summary:
I am using ReactNative in a hybrid App.
We have a setup like so:
Native Navigation Controller
Native Tab Controller
Native View Controller wrapping React
React Navigation Controller
React View Controller 1
React View Controller 2
Native View Controller 2.
When I pop Native View Controller 2 off the Navigation stack, I get a seg fault on this line:
NSUInteger indexOfFrom = [_currentViews indexOfObject:fromController.navItem];
I believe what's happening:
Your code is listening to Nav Controller transitions, assuming that they are all from React Native Nav Controllers.
You are catching this one instead, which is actually a Native Nav Controller transition.
You start trying to access the pushed/popped view controllers as if they were react native view controllers.
In this case, the view controllers are not react native -> no navItem field -> seg fault.
Solution: if we are catching this transition but it isn't from our react native nav controller, just
Closes https://github.com/facebook/react-native/pull/5495
Reviewed By: svcscm
Differential Revision: D2857473
Pulled By: nicklockwood
fb-gh-sync-id: cc7f0a16e2e0cea56ca9e49bcb87db4ebd3a0905
Summary:
public
This diff fixes a bug in RCTScreenScale() on iPhone6+. The issue was due to the fact that the iPhone6+'s virtual screen resolution of 414x736 points 3x resolution (1242x2208 pixels) does not match its actual screen resolution, which is 1080p (1080x1920 pixels).
I did not realize that `[UIScreen mainScreen].nativeBounds` reports the *actual* resolution, not the virtual resolution, and was dividing by the virtual scale (3.0) instead of the actual native scale factor (2.6).
This only affects iPhone6+, because for all other iOS devices, the virtual resolution matches the native resolution.
Reviewed By: milend
Differential Revision: D2854663
fb-gh-sync-id: bce8965a151e2f005a02a5f6b54f259d01b9ab12
Summary:
public
Although the feature itself is gated, once the user is on the experiment we want to make sure hot loading starts disabled up until the feature is enabled through the dev menu.
Reviewed By: javache
Differential Revision: D2850070
fb-gh-sync-id: 66e69e152806d3bb01985afe20827e3b9cffeb41
Summary:
Ok, so this started as fixing #5273 but ended up getting a little more complicated. 😄
Currently, AlertIOS has the following API:
* `alert(title, message, buttons, type)`
* `prompt(title, defaultValue, buttons, callback)`
I've changed the API to look like the following:
* `alert(title, message, callbackOrButtons)`
* `prompt(title, message, callbackOrButtons, type, defaultValue)`
I know that breaking changes are a big deal, but I find the current alert API to be fairly inconsistent and unnecessarily confusing. I'll try to justify my changes one by one:
1. Currently `type` is an optional parameter of `alert`. However, the only reason to change the alert type from the default is in order to create one of the input dialogs (text, password or username/password). So we're in a weird state where if you want a normal text input, you use `prompt`, but if you want a password input you use `alert` with the 'secure-text' type. I've moved `type` to `prompt` so all text input is now done with `pro
Closes https://github.com/facebook/react-native/pull/5286
Reviewed By: svcscm
Differential Revision: D2850400
Pulled By: androidtrunkagent
fb-gh-sync-id: 2986cfa2266225df7e4dcd703fce1e322c12b816
Summary:
public
Standardises the image decoding logic for all image sources, meaning we get the benefits of efficient downscaling of images from all sources, not just ALAssets.
Reviewed By: javache
Differential Revision: D2647083
fb-gh-sync-id: e41456f838e4c6ab709b1c1523f651a86ff6e623
Summary:
This solves https://github.com/facebook/react-native/issues/5090. Since 5b4e873c68 we had better reporting for when calls from native to JS fail. When trying to load an invalid bundle, this would now cause a stackoverflow, since RCTFatal would schedule a JS call to log, which would RCTFatal, which would ...
By invalidating the jsExecutor immediately after loading fails, we prevent any more attempts to log. We can't invalidate the whole bridge at this point since we still need the redbox module to actually display the error.
public
Reviewed By: majak
Differential Revision: D2834251
fb-gh-sync-id: a3e2ad425e40560beae4d3eacb93f66ace5341bf
Summary:
public
Promises are coming. And as part of it, we are standardizing the error objects that will be returned. This puts the code in place on the Android side to always send the proper error format.
It will be an error object like this
{
code : "E_SOME_ERROR_CODE_DEFINED_BY_MODULE", // Meant to be machine parseable
message : "Human readable message",
nativeError : {} // Some representation of the underlying error (Exception or NSError) , still figuring out exactly, but hopefully something with stack info
}
Reviewed By: nicklockwood
Differential Revision: D2840128
fb-gh-sync-id: 174d620e2beb53e1fc14161a10fd0479218d98a6
Summary:
This caused issues for me when I tried to provide a native module on init that was also KVO'd (and dynamically subclassed)
On closer inspection, it also seems highly inconsistent to register these classes in DEBUG mode but have them fail silently in production. Reducing the difference between debug and release seems like a safer option.
public
Reviewed By: nicklockwood
Differential Revision: D2819838
fb-gh-sync-id: 79ab72b1152c89eae38c965ff7724aba59a00949
Summary:
This implements #5073. It adds a static method `PixelRatio.pixel()` which returns the smallest drawable line width, primarily for use in styles.
It also updates the example apps to use the new function.
Closes https://github.com/facebook/react-native/pull/5076
Reviewed By: svcscm
Differential Revision: D2799849
Pulled By: nicklockwood
fb-gh-sync-id: b83a77790601fe882affbf65531114e7c5cf4bdf
Summary:
public
React Native currently exposes the iOS layer shadow properties more-or-less directly, however there are a number of problems with this:
1) Performance when using these properties is poor by default. That's because iOS calculates the shadow by getting the exact pixel mask of the view, including any tranlucent content, and all of its subviews, which is very CPU and GPU-intensive.
2) The iOS shadow properties do not match the syntax or semantics of the CSS box-shadow standard, and are unlikely to be possible to implement on Android.
3) We don't expose the `layer.shadowPath` property, which is crucial to getting good performance out of layer shadows.
This diff solves problem number 1) by implementing a default `shadowPath` that matches the view border for views with an opaque background. This improves the performance of shadows by optimizing for the common usage case. I've also reinstated background color propagation for views which have shadow props - this should help ensure that this best-case scenario occurs more often.
For views with an explicit transparent background, the shadow will continue to work as it did before ( `shadowPath` will be left unset, and the shadow will be derived exactly from the pixels of the view and its subviews). This is the worst-case path for performance, however, so you should avoid it unless absolutely necessary. **Support for this may be disabled by default in future, or dropped altogether.**
For translucent images, it is suggested that you bake the shadow into the image itself, or use another mechanism to pre-generate the shadow. For text shadows, you should use the textShadow properties, which work cross-platform and have much better performance.
Problem number 2) will be solved in a future diff, possibly by renaming the iOS shadowXXX properties to boxShadowXXX, and changing the syntax and semantics to match the CSS standards.
Problem number 3) is now mostly moot, since we generate the shadowPath automatically. In future, we may provide an iOS-specific prop to set the path explicitly if there's a demand for more precise control of the shadow.
Reviewed By: weicool
Differential Revision: D2827581
fb-gh-sync-id: 853aa018e1d61d5f88304c6fc1b78f9d7e739804
Summary:
public
This diff deprecates `scrollResponderScrollWithoutAnimationTo` and replaces it with an optional `animated` param in `scrollResponderScrollTo`. This is more consistent with our other APIs.
Using the old `ScrollResponder.scrollResponderScrollWithoutAnimationTo` or `ScrollView.scrollWithoutAnimationTo` functions will still work, but will trigger a warning.
Reviewed By: javache
Differential Revision: D2823479
fb-gh-sync-id: 259966512104ca7b3995c9586144812a91b8d3e9
Summary:
public
The fix for border smearing introduced a bug where borders + background would sometimes not be rendered if the view was created at a small size (e.g. zero) and then resized.
This diff fixes that by redrawing the border if the view size changes. There is some opportunity to optimize this in future by performing some logic up-front to detect if the redrawing is necessary, but I thought I'd keep it simple for this bug fix rather than risk introducing further bugs.
Reviewed By: jingc
Differential Revision: D2817365
fb-gh-sync-id: eca164e8ce03a66598677c9e05496791230b5210
Summary:
public
Blending semitransparent pixels against their background is fairly a fairly expensive operation on mobile GPUs. To reduce blending, React Native has a system called "background color propagation", where the background color of parent views is automatically inherited by child views unless explicitly overridden. This means that translucent pixels can be blended directly against a known background color, avoiding the need to do this dynamically on the GPU.
In practice, this is only useful for views that do their own drawing, which is basically just `<Image/>` and `<Text/>` components, and for image components it only really matters when the image has an alpha component.
The automatic background propagation is a bit of a hack, and often does the wrong thing - for example if a view overflows its bounds, or if it overlaps a sibling, the background color will often be incorrect and need to be manually disabled. Because the only place that it provides a significant performance benefit is for text, this diff disables the behavior for everything except `<Text/>` nodes. It might still be useful for `<Image/>` nodes too, but looking through the examples in UIExplorer, the number of places where it does the wrong thing for images outnumbers the cases where it provides significant reduction in blending.
Note that this diff does not prevent you from eliminating blending on image components by manually setting an opaque background color, nor does it stop you from disabling color propagation on text components by manually setting a transparent background.
Reviewed By: javache
Differential Revision: D2811031
fb-gh-sync-id: 2eb08918c9031c582a3dd2d40e04b27a663dac82
Summary:
public
The props argument of the `-[RCTComponentData createView:props:]` method was removed,
but the C function used to swizzle it in the profiler wasn't updated.
Reviewed By: majak
Differential Revision: D2811228
fb-gh-sync-id: 8896638c77370142e29913b5fb80e7fd748254b5
Summary:
public
Since we don't actually recreate our native modules every time (will fix in follow-up), we'd never update the reference after reloading the bridge, and all navigation would fail.
Reviewed By: majak
Differential Revision: D2811406
fb-gh-sync-id: 4f4fd73bbdecfe510e1e1554668b2354181f22a8
Summary:
public
The iOS border rendering code did not follow the CSS spec in cases where the sum of adjacent border radii was greater than the width of the view, resulting in drawing glitches such as pixel smear and borders appearing stretched or squashed.
This diff brings our implementation closer to spec-compliance in these cases. I also fixed a longstanding issue with ghostly diagonal lines appearing at the corners due to antialiasing rounding errors!
Fixes
https://github.com/facebook/react-native/issues/1572https://github.com/facebook/react-native/issues/2089https://github.com/facebook/react-native/issues/4604
Reviewed By: tadeuzagallo
Differential Revision: D2811249
fb-gh-sync-id: c3dd2721e0a01a432fa4dc78daa05680595edd08
Summary:
public
Attempting to load an undefined URL via XMLHttpRequest produced a confusing error deep within the network layer. This diff improves the networking stack to catch such errors earlier, and also adds a helpful error in the JS layer.
Fixes https://github.com/facebook/react-native/issues/4558
Reviewed By: javache
Differential Revision: D2811080
fb-gh-sync-id: 1837427e1080a0308f2c4f9a8a42bce2e041fb48
Summary:
public
Fixed a potential deadlock issue if code attempted to access a module via [bridge moduleForName/Class:] while it was being initialized.
Reviewed By: lry
Differential Revision: D2807827
fb-gh-sync-id: 58cafe9b92c094dde632d17245fb9b342a0fe9e0
Summary:
public
By doing this we fix 2 problems:
1. We use the same url, both the first time the simulator starts with Hot Loading disabled (no `hot` attribute), and after HL has been enabled and then disabled ('hot=false'). By doing so, the packager will rebuild more than one bundle as file changes. We could have ignored this attribute on the packager but I'd rather not contaminate the server with it and instead make the clients send only 2 types of URLs.
2. The code on `RCTBatchedBridge.m` that decides whether or not to enable HMR does so by looking at presence of the query string parameter `hot`. If the parameter is present, even when it's false, it will try to enable HL, which is wrong.
Reviewed By: nicklockwood
Differential Revision: D2807512
fb-gh-sync-id: 728b680c2383c328d8967d34c10e7a6288e455ac
Summary:
public
Android implement ViewManager methods via a dispatch method on UIManager, whereas iOS implements them by exposing the methods on the view manager modules directly.
This diff polyfills Android's implementation on top of the iOS implementation, allowing the same JS API to be used for both.
Reviewed By: javache
Differential Revision: D2803020
fb-gh-sync-id: 0da0544e593dc936467d16ce957a77f7ca41355b
Summary:
public
Unfortunately, it turns out that NSURLComponents.queryItems only works on iOS 8 and above. This diff re-implements the RCTGetURLQueryParam and RCTURLByReplacingQueryParam functions using functionality available in iOS 7.
Reviewed By: javache
Differential Revision: D2803679
fb-gh-sync-id: 56f10bef4894d16197975b6023b7aa5ab106d8cb
Summary:
public
The logic inside RCTBatchedBridge contained some race conditions that would occasionally cause an error if modules were loaded in the wrong order. This improves that logic and makes it safer by adding a lock to prevent concurrency.
Reviewed By: jspahrsummers
Differential Revision: D2802930
fb-gh-sync-id: d1ad25fa578649363dcaac029cb24dc3a453ae67
Summary:
public
This exposes a proper API for adding synchronous callbacks to JS, as an optional feature of the executor.
This is based on nicklockwood's work in D2764492, but avoids refactoring bridge/executor interactions for the time being, since we agree on this API and can move the actual callsites around later.
Reviewed By: nicklockwood
Differential Revision: D2799506
fb-gh-sync-id: af209d9a0be927f3404205feb16e59745cc37aec
Summary:
public
Expose JS hooks to create flow events in systrace (the nice arrows to show async work flow) +
add support to the showing all the work enqueued from the JS thread as added in D2743733
Depends on D2743733
Reviewed By: jspahrsummers
Differential Revision: D2773664
fb-gh-sync-id: 4a8854b17b4741b882f5f2cc425e4237a5e4b3eb
Summary:
Both iOS and Android currently support some sort of native pull to refresh control but the API was very different. I tried implementing a component based on PullToRefreshViewAndroid but that works on both platforms.
I liked the idea of wrapping the ListView or ScrollView with the PullToRefreshView component and allow styling the refresh view with platform specific props if needed. I also like the fact that 'refreshing' is a controlled prop so there is no need to keep a ref to the component or to the stopRefreshing function.
It is a pretty rough start so I'm looking for feedback and ideas to improve on the API before cleaning up everything.
On iOS we could probably deprecate the onRefreshStart property of the ScrollView and implement the native stuff in a PullToRefreshViewManager. We could then add props to customize the look of the UIRefreshControl (tintColor). We could also deprecate the Android only component and remove it later.
Closes https://github.com/facebook/react-native/pull/4915
Reviewed By: svcscm
Differential Revision: D2799246
Pulled By: nicklockwood
fb-gh-sync-id: 75872c12143ddbc05cc91900ab4612e477ca5765
Summary:
public
The implementation of the `methodQueue` lazy initializer in `RCTModuleData` could result in the queue being set twice, because calling `methodQueue` for a module that hasn't been instantiated would call `RCTModuleData.instance` to create the module, which itself calls `methodQueue`.
It's not clear if this was causing a bug, but it may be related to an occasional bug where the `RCTViewManager.methodQueue` returns nil.
Reviewed By: majak
Differential Revision: D2783320
fb-gh-sync-id: 9194da0fd7392f63825da1f5c450363dd300b635
Summary:
public
Until we support this fature on OSS, don't show the menu option.
Reviewed By: vjeux
Differential Revision: D2791198
fb-gh-sync-id: 11b66d467c1ab784bbf549b893d0a3abd69e2741
tuyou
Nick Lockwood
Satyajit Sahoo
DougBanksPersonal
Martín Bigio
Martin Kralik
Kyle Corbitt
tantan
Pieter De Baets
Dave Miller
Gaëtan Renaudeau
Tadeu Zagallo
Quentin Valmori
Justin Spahr-Summers
Fada Chen
Janic Duplessis
Jason Brown
Jed Lau
Mike Armstrong
