We read the wrapper state during initial mount for server rendering support
but Fiber doesn't use it and we don't need it. We also can't because we
haven't yet completed the parent that has the selected.
I will need to remember to always insert children before setting the
selected value on the parent <select />. That way the DOM will deal with
the `selected` property of option properly.
Instead of passing around the owner and the internal fiber everywhere we
can set current owner during each commit. That way we have it available
globally where ever we need it.
That way we don't have to pass it as a DEV only argument nor expose the
internal representation to the host config.
This doesn't actually get the current owner yet. Will do that in a follow
up.
We'll stop tracking the input value tracking. We'll also stop uncaching the
node because we can just let the garbage collector take care of that. This
makes it easier to release components in trees that never mounted.
This also removes an invariant error which is covered by unit tests.
This is necessary to separate regardless because we don't want this to
fire for a component that was started but was thrown away because it never
mounted. We can come back to that later.
Basically, we don't need to remove the event listeners because we can just
check if something is still mounted when they fire instead. We'll rely on
garbage collection to clean them up.
We don't need to wait for the commit phase to start listening to events
since we have the node.
Next we'll stop unlistening to events too and instead just check isMounted.
Fiber manages children separately so we don't need to do it here.
However, we special case the text content children and
dangerouslySetInnerHTML.
This reveals a bug that we currently don't handle the case where we switch
from dangerouslySetInnerHTML or text children to element children, because
child insertions are handled before the parent updates. We could possibly
handle this case by removing all nodes before the first host child but
that is a bit unfortunate.
We have warned about not mutating styles already. At this point we can just
freeze the style object in DEV.
That way we can read it from the previous props object without storing
another copy of it.
Also delete the associated warning.
* Update ESLint to 3.10.2
Also pull in fbjs for extending properly, per @zpao. This also disables consistent-return, which has about 80 failing cases in React currently. If we'd like to turn this back on, we should do it separately and fix all the call sites properly (rather than just adding 'return undefined;' everywhere, which adds no value.
Fixes to all existing lint errors plus an update for yarn.lock to follow.
* Update yarn.lock after the eslint update.
* Fix all new eslint failures
Unfortunately I had to add three eslint-disable-next-line instances. All have explanations inline.
* Switch Travis to use yarn instead of npm
This traverses parent based on the type of internal instance it is passed.
If it is a Fiber it may have to traverse multiple steps until it finds a
HostComponent.
This will allow us to use the event system with Fiber.
This adds precaching to ReactDOMFiber. I.e. adding a handle from the DOM
node to the internal Fiber. This means that we need to expose an internal
handle to the reconciler.
We use duck typing to figure out if it is a Fiber or Stack instance.
The new failing tests are failing because this is now able to actually
fire events onto Fibers and then the result of those events are
incorrect where as they were ignored before.
* Reapply Check for event listener in props instead of bank (#8192)
This reverts the previous revert.
* Don't throw on falsy event listeners
Some code relies on passing null. This restores the earlier behavior.
This matches Stack behavior.
Had to rewrite a test that used Simulate because Fiber doesn't support events yet.
Also changed tests for componentDidUpdate() since Fiber intentionally doesn't pass prevContext argument.
We can avoid recreating the merged context object if the context provider work was reused. This is essential to avoid allocations for deep setState() calls.
As per discussion, this is better because then this code only runs for the class type of work. We will still need to fix bailouts for deep setState calls.
It is error-prone to push and pop context in begin or complete phases because of all the bailouts. Scheduler has enough knowledge to see if pushing is necessary because it knows when we go inside or outside the tree.
I'll plan to change all of our console.error and component-tree expects to expectDev. It's a little annoying that we need to make sure tests don't throw (see my change to normalizeCodeLocInfo) but any alternative would seem to require two separate test runs or a much more cumbersome syntax.
* Handles risky callbacks on setState. Fixes#8238
* Updates try-catch to cover `callback` when context is not present.
* Updates code to trapErrors instead of swallowing them.
* Fixes flow errors
* Incorporates review comments
* Traps only the first error.
Removes `callbackWasCalled` and updates fiber tests.
Sebastian Markbage
Tom Occhino
Ben Alpert
Sebastian Markbåge
Dan Abramov
Andrew Clark
Ben Alpert
Ankeet Maini