Summary:
Instead of just blindly retrying all ViewCommands if they fail - which could be dangerous, since it's arbitrary imperative commands we'd be executing twice, potentially with bad app state - we only retry if the ViewCommand throws a "RetryableMountingLayerException".
Changelog: [Internal] Optimization to ViewCommands
Reviewed By: mdvacca
Differential Revision: D20529985
fbshipit-source-id: 0217b43f4bf92442bcc7ca48c8ae2b9a9e543dc9
Summary:
Turns out that dispatchMountItems was reentrant, meaning that something (in particular, updateState) could cause mount items to be queued up which would then be executed synchronously, out-of-order, in the middle of the previous dispatchMountItems call.
We will continue to monitor this and see how often we're reentering: T63181639 and via any soft exceptions that are logged.
For context, there are currently three ways dispatchMountItems gets called: 1) On every UI Tick in the UI thread, in a loop; 2) via animations, via synchronouslyUpdateViewOnUIThread, which happens to fail a *lot* currently; 3) when there is a commit on the UI thread, like with a Java->C++ state update. We must account for reentrance and failure in all three cases and make sure the `mInDispatch` flag is reset after success or failure in all of those situations.
Changelog: [Internal]
Reviewed By: mdvacca
Differential Revision: D20170160
fbshipit-source-id: 834f1d9b65000caa7f2eea4849e5e7951d2b6be6
Summary:
In D18032458 we introduce getReactApplicationContextIfActiveOrWarn. In this diff, modules that access a JS or Native module through ReactApplicationContext need to check if the CatalystInstance is still alive before continuing.
Modules that don't derive from `ReactContextBaseJavaModule` manually check for the catalyst impl and log their own SoftExceptions.
In this diff we also introduce SoftExceptions that by contract never cause crashes, even in debug mode.
Changelog: [Internal]
Reviewed By: mdvacca
Differential Revision: D18112989
fbshipit-source-id: 868f01f388aa2db3518db9f873f2afc2a62eed45