When a Suspense boundary suspends during initial mount, the primary
children's fibers are discarded because there is no current tree to
preserve them. If the suspended promise never resolves, the only way to
retry is something external like a context change. However, lazy context
propagation could not find the consumer fibers — they no longer exist in
the tree — so the Suspense boundary was never marked for retry and
remained stuck in fallback state indefinitely.
The fix teaches context propagation to conservatively mark suspended
Suspense boundaries for retry when a parent context changes, even when
the consumer fibers can't be found. This matches the existing
conservative approach used for dehydrated (SSR) Suspense boundaries.
Cleans up feature flags that do not have an active experiment and which
we don't currently plan to ship, one commit per flag. Notable removals:
* Automatic (inferred) effect dependencies / Fire: abandoned due to
early feedback. Shipped useEffectEvent which addresses some of the
use-cases.
* Inline JSX transform (experimented, not a consistent win)
* Context selectors (experimented, not a sufficient/consistent win given
the benefit the compiler already provides)
* Instruction Reordering (will try a different approach)
To decide which features to remove, I looked at Meta's internal repos as
well as eslint-pugin-react-hooks to see which flags were never
overridden anywhere. That gave a longer list of flags, from which I then
removed some features that I know are used in OSS.
When flushing the shell, stylesheets with precedence are emitted in the
`<head>` which blocks paint regardless. Outlining a boundary solely
because it has suspensey CSS provides no benefit during the shell flush
and causes a higher-level fallback to be shown unnecessarily (e.g.
"Middle Fallback" instead of "Inner Fallback").
This change passes a flushingInShell flag to hasSuspenseyContent so the
host config can skip stylesheet-only suspensey content when flushing the
shell. Suspensey images (used for ViewTransition animation reveals)
still trigger outlining during the shell since their motivation is
different.
When flushing streamed completions the behavior is unchanged — suspensey
CSS still causes outlining so the parent content can display sooner
while the stylesheet loads.
## Summary
Follow-up to https://github.com/vercel/next.js/pull/89823 with the
actual changes to React.
Replaces the `JSON.parse` reviver callback in `initializeModelChunk`
with a two-step approach: plain `JSON.parse()` followed by a recursive
`reviveModel()` post-process (same as in Flight Reply Server). This
yields a **~75% speedup** in RSC chunk deserialization.
| Payload | Original (ms) | Walk (ms) | Speedup |
|---------|---------------|-----------|---------|
| Small (2 elements, 142B) | 0.0024 | 0.0007 | **+72%** |
| Medium (~12 elements, 914B) | 0.0116 | 0.0031 | **+73%** |
| Large (~90 elements, 16.7KB) | 0.1836 | 0.0451 | **+75%** |
| XL (~200 elements, 25.7KB) | 0.3742 | 0.0913 | **+76%** |
| Table (1000 rows, 110KB) | 3.0862 | 0.6887 | **+78%** |
## Problem
`createFromJSONCallback` returns a reviver function passed as the second
argument to `JSON.parse()`. This reviver is called for **every key-value
pair** in the parsed JSON. While the logic inside the reviver is
lightweight, the dominant cost is the **C++ → JavaScript boundary
crossing** — V8's `JSON.parse` is implemented in C++, and calling back
into JavaScript for every node incurs significant overhead.
Even a trivial no-op reviver `(k, v) => v` makes `JSON.parse` **~4x
slower** than bare `JSON.parse` without a reviver:
```
108 KB payload:
Bare JSON.parse: 0.60 ms
Trivial reviver: 2.95 ms (+391%)
```
## Change
Replace the reviver with a two-step process:
1. `JSON.parse(resolvedModel)` — parse the entire payload in C++ with no
callbacks
2. `reviveModel` — recursively walk the resulting object in pure
JavaScript to apply RSC transformations
The `reviveModel` function includes additional optimizations over the
original reviver:
- **Short-circuits plain strings**: only calls `parseModelString` when
the string starts with `$`, skipping the vast majority of strings (class
names, text content, etc.)
- **Stays entirely in JavaScript** — no C++ boundary crossings during
the walk
## Results
You can find the related applications in the [Next.js PR
](https://github.com/vercel/next.js/pull/89823)as I've been testing this
on Next.js applications.
### Table as Server Component with 1000 items
Before:
```
"min": 13.782875000000786,
"max": 22.23400000000038,
"avg": 17.116868530000083,
"p50": 17.10766700000022,
"p75": 18.50787499999933,
"p95": 20.426249999998618,
"p99": 21.814125000000786
```
After:
```
"min": 10.963916999999128,
"max": 18.096083000000363,
"avg": 13.543286884999988,
"p50": 13.58350000000064,
"p75": 14.871791999999914,
"p95": 16.08429099999921,
"p99": 17.591458000000785
```
### Table as Client Component with 1000 items
Before:
```
"min": 3.888875000000553,
"max": 9.044959000000745,
"avg": 4.651271475000067,
"p50": 4.555749999999534,
"p75": 4.966624999999112,
"p95": 5.47754200000054,
"p99": 6.109499999998661
````
After:
```
"min": 3.5986250000005384,
"max": 5.374291000000085,
"avg": 4.142990245000046,
"p50": 4.10570799999914,
"p75": 4.392041999999492,
"p95": 4.740084000000934,
"p99": 5.1652500000000146
```
### Nested Suspense
Before:
```
Requests: 200
Min: 73ms
Max: 106ms
Avg: 78ms
P50: 77ms
P75: 80ms
P95: 85ms
P99: 94ms
```
After:
```
Requests: 200
Min: 56ms
Max: 67ms
Avg: 59ms
P50: 58ms
P75: 60ms
P95: 65ms
P99: 66ms
```
### Even more nested Suspense (double-level Suspense)
Before:
```
Requests: 200
Min: 159ms
Max: 208ms
Avg: 169ms
P50: 167ms
P75: 173ms
P95: 183ms
P99: 188ms
```
After:
```
Requests: 200
Min: 125ms
Max: 170ms
Avg: 134ms
P50: 132ms
P75: 138ms
P95: 148ms
P99: 160ms
```
## How did you test this change?
Ran it across many Next.js benchmark applications.
The entire Next.js test suite passes with this change.
---------
Co-authored-by: Hendrik Liebau <mail@hendrik-liebau.de>
## Summary
Add "RCTSelectableText" to the list of component names recognized as
being inside a text element, alongside "RCTText".
React Native's new text stack, tries to optimize and allows
differentiating between a custom TextView, with lower level control,
that can reuse the work performed during Fabric/Yoga layout, and a
native TextView, used for fidelity. On Android at least, the only place
we've needed native TextView for fidelity/native UX has been support for
`selectable` text, which has many unique UI interactions.
## How did you test this change?
When I patch this in, alongside
https://github.com/facebook/react-native/pull/55552, we no longer see
warnings when we render text inside of RCTSelectableText component.
---------
Co-authored-by: Eli White <github@eli-white.com>
Tracks locations for reactive scope dependencies, both on the deps and
portions of the path. The immediate need for this is a non-public
experiment where we're exploring type-directed compilation, and
sometimes look up the types of expressions by location. We need to
preserve locations accurately for that to work, including the locations
of the deps.
## Test Plan
Locations for dependencies are not easy to test: i manually spot-checked
the new fixture to ensure that the deps look right. This is fine as
best-effort since it doesn't impact any of our core compilation logic, i
may fix forward if there are issues and will think about how to test.
When the Flight Client resolves chunk references during model parsing,
it calls `transferReferencedDebugInfo` to propagate debug info entries
from referenced chunks to the parent chunk. Debug info on chunks is
later moved to their resolved values, where it is used by React DevTools
to show performance tracks and what a component was suspended by.
Debug chunks themselves (specifically `ReactComponentInfo`,
`ReactAsyncInfo`, `ReactIOInfo`, and their outlined references) are
metadata that is never rendered. They don't need debug info attached to
them. Without this fix, debug info entries accumulate on outlined debug
chunks via their references to other debug chunks (e.g. owner chains and
props deduplication paths). Since each outlined chunk's accumulated
entries are copied to every chunk that references it, this creates
exponential growth in deep component trees, which can cause the dev
server to hang and run out of memory.
This generalizes the existing skip of `transferReferencedDebugInfo` for
Element owner/stack references (which already recognizes that references
to debug chunks don't need debug info transferred) to all references
resolved during debug info resolution. It adds an
`isInitializingDebugInfo` flag set in `initializeDebugChunk` and
`resolveIOInfo`, which propagates through all nested
`initializeModelChunk` calls within the same synchronous stack. For the
async path, `waitForReference` captures the flag at call time into
`InitializationReference.isDebug`, so deferred fulfillments also skip
the transfer.
## Summary
ESLint v10.0.0 was released on February 7, 2026. The current
`peerDependencies` for `eslint-plugin-react-hooks` only allows up to
`^9.0.0`, which causes peer dependency warnings when installing with
ESLint v10.
This PR:
- Adds `^10.0.0` to the eslint peer dependency range
- Adds `eslint-v10` to devDependencies for testing
- Adds an `eslint-v10` e2e fixture (based on the existing `eslint-v9`
fixture)
ESLint v10's main breaking changes (removal of legacy eslintrc config,
deprecated context methods) don't affect this plugin - flat config is
already supported since v7.0.0, and the deprecated APIs already have
fallbacks in place.
## How did you test this change?
Ran the existing unit test suite:
```
cd packages/eslint-plugin-react-hooks && yarn test
```
All 5082 tests passed.
## Overview
While building the RSC sandboxes I notice error messages like:
> An error occurred in the `<Offscreen>` component
This is an internal component so it should show either:
> An error occurred in the `<Suspense>` component.
> An error occurred in the `<Activity>` component.
It should only happen when there's a lazy in the direct child position
of a `<Suspense>` or `<Activity>` component.
Follow up to https://github.com/facebook/react/pull/35630
We don't currently have any operations that depend on the updating of
text nodes added or removed after Fragment mount. But for the sake of
completeness and extending the ability to any other host configs, this
change calls `commitNewChildToFragmentInstance` and
`deleteChildFromFragmentInstance` on HostText fibers.
Both DOM and Fabric configs early return because we cannot attach event
listeners or observers to text. In the future, there could be some
stateful Fragment feature that uses text that could extend this.
After https://github.com/facebook/react/pull/34089, when updating
(possibly, mounting) inside disconnected subtree, we don't record this
as an operation. This only happens during reconnect. The issue is that
`recordProfilingDurations()` can be called, which diffs tree base
duration and reports it to the Frontend:
65db1000b9/packages/react-devtools-shared/src/backend/fiber/renderer.js (L4506-L4521)
This operation can be recorded before the "Add" operation, and it will
not be resolved properly on the Frontend side.
Before the fix:
```
commit tree › Suspense › should handle transitioning from fallback back to content during profiling
Could not clone the node: commit tree does not contain fiber "5". This is a bug in React DevTools.
162 | const existingNode = nodes.get(id);
163 | if (existingNode == null) {
> 164 | throw new Error(
| ^
165 | `Could not clone the node: commit tree does not contain fiber "${id}". This is a bug in React DevTools.`,
166 | );
167 | }
at getClonedNode (packages/react-devtools-shared/src/devtools/views/Profiler/CommitTreeBuilder.js:164:13)
at updateTree (packages/react-devtools-shared/src/devtools/views/Profiler/CommitTreeBuilder.js:348:24)
at getCommitTree (packages/react-devtools-shared/src/devtools/views/Profiler/CommitTreeBuilder.js:112:20)
at ProfilingCache.getCommitTree (packages/react-devtools-shared/src/devtools/ProfilingCache.js:40:46)
at Object.<anonymous> (packages/react-devtools-shared/src/__tests__/profilingCommitTreeBuilder-test.js:257:44)
```
Fixes https://github.com/facebook/react/issues/33423,
https://github.com/facebook/react/issues/35245,
https://github.com/facebook/react/issues/19732.
As demoed
[here](https://github.com/facebook/react/issues/33423#issuecomment-2970750588),
React DevTools incorrectly highlights re-renders for descendants of
filtered-out nodes that didn't actually render.
There were multiple fixes suggesting changes in `didFiberRender()`
function, but these doesn't seem right, because this function is used in
a context of whether the Fiber actually rendered something (updated),
not re-rendered compared to the previous Fiber.
Instead, this PR adds additional validation at callsites that either
used for highlighting re-renders or capturing tree base durations and
are relying on `didFiberRender`. I've also added a few tests that
reproduce the failure scenario. Without the changes, the tests are
failing.
When using a partial prerender stream, i.e. a prerender that is
intentionally aborted before all I/O has resolved, consumers of
`createFromReadableStream` would need to keep the stream unclosed to
prevent React Flight from erroring on unresolved chunks. However, some
browsers (e.g. Chrome, Firefox) keep unclosed ReadableStreams with
pending reads as native GC roots, retaining the entire Flight response.
With this PR we're adding an `unstable_allowPartialStream` option, that
allows consumers to close the stream normally. The Flight Client's
`close()` function then transitions pending chunks to halted instead of
erroring them. Halted chunks keep Suspense fallbacks showing (i.e. they
never resolve), and their `.then()` is a no-op so no new listeners
accumulate. Inner stream chunks (ReadableStream/AsyncIterable) are
closed gracefully, and `getChunk()` returns halted chunks for new IDs
that are accessed after closing the response. Blocked chunks are left
alone because they may be waiting on client-side async operations like
module loading, or on forward references to chunks that appeared later
in the stream, both of which resolve independently of closing.
`encodeReply` throws "React Element cannot be passed to Server Functions
from the Client without a temporary reference set" when a React element
is the root value of a `serializeModel` call (either passed directly or
resolved from a promise), even when a temporary reference set is
provided.
The cause is that `resolveToJSON` hits the `REACT_ELEMENT_TYPE` switch
case before reaching the `existingReference`/`modelRoot` check that
regular objects benefit from. The synthetic JSON root created by
`JSON.stringify` is never tracked in `writtenObjects`, so
`parentReference` is `undefined` and the code falls through to the
throw. This adds a `modelRoot` check in the `REACT_ELEMENT_TYPE` case,
following the same pattern used for promises and plain objects.
The added `JSX as root model` test also uncovered a pre-existing crash
in the Flight Client: when the JSX element round-trips back, it arrives
as a frozen object (client-created elements are frozen in DEV), and
`Object.defineProperty` for `_debugInfo` fails because frozen objects
are non-configurable. The same crash can occur with JSX exported as a
client reference. For now, we're adding `!Object.isFrozen()` guards in
`moveDebugInfoFromChunkToInnerValue` and `addAsyncInfo` to prevent the
crash, which means debug info is silently dropped for frozen elements.
The proper fix would likely be to clone the element so each rendering
context gets its own mutable copy with correct debug info.
closes#34984closes#35690
## Summary
- Fixes the `createRequest` call in `renderToPipeableStream` to pass
`debugChannelReadable !== undefined` instead of `debugChannel !==
undefined` in the turbopack, esm, and parcel Node.js server
implementations
- The webpack version already had the correct check; this brings the
other bundler implementations in line
The bug was introduced in #33754. With `debugChannel !== undefined`, the
server could signal that debug info should be emitted even when only a
write-only debug channel is provided (no readable side), potentially
causing the client to block forever waiting for debug data that never
arrives.
There is an existing issue with serialisation logic for the traces from
Profiler panel.
I've discovered that `TREE_OPERATION_UPDATE_TREE_BASE_DURATION`
operation for some reason appears earlier in a sequence of operations,
before the `TREE_OPERATION_ADD` that registers the new node. It ends up
cloning non-existent node, which just creates an empty object and adds
it to the map of nodes.
This change only adds additional layer of validation to cloning logic,
so we don't swallow the error, if we attempt to clone non-existent node.
Small optimization for useEffectEvent. Not sure we even need a flag for
it, but it will be a nice killswitch.
As an added benefit, it fixes a bug when `enableViewTransition` is on,
where we were not updating the useEffectEvent callback when a tree went
from hidden to visible.
Related to https://github.com/facebook/react/pull/35548,
`enableViewTransition` fixes a bug where `getSnapshotBeforeUpdate` was
running in hidden trees when it shouldn't (`componentWillUpdate` won't
run for hidden updates, and when it becomes visible it will be called
with `componentWillMount`).
## Overview
Adds a feature flag `enableParallelTransitions` to experiment with
engantling transitions less often.
## Motivation
Currently we over-entangle transition lanes.
It's a common misunderstanding that React entangles all transitions,
always. We actually will complete transitions independently in many
cases. For example, [this
codepen](https://codepen.io/GabbeV/pen/pvyKBrM) from
[@gabbev](https://bsky.app/profile/gabbev.bsky.social/post/3m6uq2abihk2x)
shows transitions completing independently.
However, in many cases we entangle when we don't need to, instead of
letting the independent transitons complete independently. We still want
to entangle for updates that happen on the same queue.
## Example
As an example of what this flag would change, consider two independent
counter components:
```js
function Counter({ label }) {
const [count, setCount] = useState(0);
return (
<div>
<span>{use(readCache(`${label} ${count}`))} </span>
<Button
action={() => {
setCount((c) => c + 1);
}}
>
Next {label}
</Button>
</div>
);
}
```
```js
export default function App() {
return (
<>
<Counter label="A" />
<Counter label="B" />
</>
);
}
```
### Before
The behavior today is to entange them, meaning they always commit
together:
https://github.com/user-attachments/assets/adead60e-8a98-4a20-a440-1efdf85b2142
### After
In this experiment, they will complete independently (if they don't
depend on each other):
https://github.com/user-attachments/assets/181632b5-3c92-4a29-a571-3637f3fab8cd
## Early Research
This change is in early research, and is not in the experimental
channel. We're going to experiment with this at Meta to understand how
much of a breaking change, and how beneficial it is before commiting to
shipping it in experimental and beyond.
Database libraries like Gel/EdgeDB can create very long linear chains of
async sequences through temporal async sequencing in connection pools.
The recursive traversal of `node.previous` chains in `visitAsyncNode`
causes stack overflow on these deep chains.
The fix converts the `previous` chain traversal from recursive to
iterative. We collect the chain into an array, then process from deepest
to shallowest.
The `awaited` traversal remains recursive since its depth is bounded by
promise dependency depth, not by the number of event loop turns. Each
`awaited` branch still benefits from the iterative `previous` handling
within its own traversal.
I've verified that this fixes the
[repro](https://github.com/jere-co/next-debug) provided in #35246.
closes#35246
Summary:
I noticed there's a bug where the lint will recognize the type on a cast
annotation as a missing dependency;
```
function MyComponent() {
type ColumnKey = 'id' | 'name';
type Item = {id: string, name: string};
const columns = useMemo(
() => [
{
type: 'text',
key: 'id',
} as TextColumn<ColumnKey, Item>,
^^^^^^^^ here
],
[],
);
}
```
This is due to the AST of AsExpressions being something like:
AsExpression
└── typeAnnotation: GenericTypeAnnotation
└── typeParameters: TypeParameterInstantiation
└── params[0]: GenericTypeAnnotation
└── id: Identifier (name: "ColumnKey")
Where `ColumnKey` never has a TypeParameter Annotation. So we need to
consider it to be a flow type due to it belonging to a
GenericTypeAnnotation
Test Plan:
Added unit tests
Before:
```
Test Suites: 1 failed, 2 passed, 3 total
Tests: 2 failed, 5065 passed, 5067 total
Snapshots: 0 total
Time: 16.517 s
Ran all test suites.
error Command failed with exit code 1.
```
After:
```
PASS __tests__/ReactCompilerRuleTypescript-test.ts
PASS __tests__/ESLintRulesOfHooks-test.js (6.192 s)
PASS __tests__/ESLintRuleExhaustiveDeps-test.js (9.97 s)
Test Suites: 3 passed, 3 total
Tests: 5067 passed, 5067 total
Snapshots: 0 total
Time: 10.21 s, estimated 11 s
Ran all test suites.
✨ Done in 12.66s.
```
## Summary
Fix react-hooks/set-state-in-effect false negatives when Hooks are
called via a namespace import (e.g. `import * as React from 'react'` and
`React.useEffect(...))`. The validation now checks the MethodCall
property (the actual hook function) instead of the receiver object.
Issue: Bug: #35377
## How did you test this change?
Added a regression fixture;
Ran tests and verified it reports `EffectSetState` and matches the
expected output.
<img width="461" height="116" alt="Screenshot 2025-12-27 at 14 13 38"
src="https://github.com/user-attachments/assets/fff5aab9-0f2c-40e9-a6a5-b864c3fa6fbd"
/>
* A few new minimization strategies, removing function params and
array/object pattern elements
* Ensure that we preserve the same set of errors based on not just
category+reason but also description.
More snap improvements for use with agents:
* `yarn snap compile [--debug] <path>` for compiling any file,
optionally with debug logs
* `yarn snap minimize <path>` now accepts path as a positional param for
consistency w 'compile' command
* Both compile/minimize commands properly handle paths relative to the
compiler/ directory. When using `yarn snap` the current working
directory is compiler/packages/snap, but you're generally running it
from the compiler directory so this matches expectations of callers
better.
This is a combination of a) a subagent for investigating compiler errors
and b) testing that agent by fixing bugs with for loops within
try/catch. My recent diffs to support maybe-throw within value blocks
was incomplete and handled many cases, like optionals/logicals/etc
within try/catch. However, the handling for for loops was making more
assumptions and needed additional fixes.
Key changes:
* `maybe-throw` terminal `handler` is now nullable. PruneMaybeThrows
nulls the handler for blocks that cannot throw, rather than changing to
a `goto`. This preserves more information, and makes it easier for
BuildReactiveFunction's visitValueBlock() to reconstruct the value
blocks
* Updates BuildReactiveFunction's handling of `for` init/test/update
(and similar for `for..of` and `for..in`) to correctly extract value
blocks. The previous logic made assumptions about the shape of the
SequenceExpression which were incorrect in some cases within try/catch.
The new helper extracts a flattened SequenceExpression.
Supporting changes:
* The agent itself (tested via this diff)
* Updated the script for invoking snap to keep `compiler/` as the
working directory, allowing relative paths to work more easily
* Add an `--update` (`-u`) flag to `yarn snap minimize`, which updates
the fixture in place w the minimized version
