Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you

• Open a pull request for review
• Mark a draft as ready
• Comment "@codex review".

If Codex has suggestions, it will comment; otherwise it will react with 👍.

Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".

packages/react/runtime/__test__/snapshot/mtf-execid-churn.test.jsx (1)

12-16: Make helper failures explicit instead of a TypeError.
If call or call[1].data is missing, the current code throws a vague error. Adding explicit assertions makes failures easier to diagnose.

Suggested change

 function getSnapshotPatchFromPatchUpdateCall(call) {
-  const obj = call[1];
-  const parsed = JSON.parse(obj.data);
+  expect(call, 'expected a patch update call').toBeTruthy();
+  const obj = call[1];
+  expect(obj?.data, 'expected patch payload').toBeTypeOf('string');
+  const parsed = JSON.parse(obj.data);
   return parsed.patchList?.[0]?.snapshotPatch;
 }

Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you

• Open a pull request for review
• Mark a draft as ready
• Comment "@codex review".

If Codex has suggestions, it will comment; otherwise it will react with 👍.

Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".

packages/react/runtime/__test__/snapshot/mtf-execid-churn.test.jsx (2)

166-172: Nit: toJSON destructured binding is unused.

Minor readability nit — rename to _ to signal intent (the goal is only to strip toJSON from rest).

Proposed tweak

-      toJSON() {
-        const { toJSON, ...rest } = this;
+      toJSON() {
+        const { toJSON: _toJSON, ...rest } = this;
         return {
           ...rest,
           __isSerialized: true,
         };
       },

🤖 Prompt for AI Agents

Verify each finding against the current code and only fix it if needed.

In `@packages/react/runtime/__test__/snapshot/mtf-execid-churn.test.jsx` around
lines 166 - 172, The destructured binding named toJSON inside the toJSON method
is unused and hurts readability; change the destructuring from `{ toJSON,
...rest } = this` to use a throwaway identifier (e.g. `{ _: _, ...rest } = this`
or simply `{ _: , ...rest } = this` depending on style) so it's obvious the
intent is only to remove the toJSON property and keep the remaining fields;
update the toJSON method in the test (the toJSON method on the test
object/class) to destructure the property into a clearly unused name like `_`
and return `{ ...rest, __isSerialized: true }`.

---

49-215: Optional: extract the shared render/hydrate/rerender harness.

The three it blocks share near-identical main-thread-render → background-render → hydrate → rerender scaffolding; only the component and the assertion target differ. Extracting a helper like runStableRerenderCase(Comp, triggerRerender) would cut ~60 lines and make future churn tests a one-liner. Deferable — current form is readable as-is.

🤖 Prompt for AI Agents

Verify each finding against the current code and only fix it if needed.

In `@packages/react/runtime/__test__/snapshot/mtf-execid-churn.test.jsx` around
lines 49 - 215, Tests duplicate the same main-thread render → background render
→ hydrate → rerender scaffold across three it blocks; extract a helper (e.g.,
runStableRerenderCase) to remove duplication. Implement
runStableRerenderCase(Comp, triggerRerender) that performs: set __root.__jsx and
renderPage() for main-thread render; switch to background and call render(<Comp
/>, __root); run hydrate by invoking
lynxCoreInject.tt.OnLifecycleEvent(...__OnLifecycleEvent.mock.calls[0]) then
switch to main thread and invoke the recorded callLepusMethod; for rerender
switch to background, clear mocks
(lynx.getNativeApp().callLepusMethod.mockClear()), call triggerRerender (this
can be bump_ or rendering with new props), await waitSchedule(), switch to main
thread and return the rLynxChange so callers can assert
getSnapshotPatchFromPatchUpdateCall(rLynxChange) is undefined; update the three
tests to call runStableRerenderCase with their Comp and appropriate trigger
function.

packages/react/runtime/__test__/snapshot/gesture.test.jsx (1)

1148-1155: Consider also covering user-provided gesture attr preservation.

This regression covers retaining clip-radius/flatten, but the PR objective also calls out not clobbering unrelated user-provided gesture attrs. A small variant with both gesture and main-thread:gesture, then removing only main-thread:gesture, would lock that behavior down too.

Also applies to: 1196-1199

🤖 Prompt for AI Agents

Verify each finding against the current code and only fix it if needed.

In `@packages/react/runtime/__test__/snapshot/gesture.test.jsx` around lines 1148
- 1155, Add a test variant that ensures user-provided 'gesture' attributes are
not clobbered when only 'main-thread:gesture' is removed: update the existing
props construction (the variable props used with keepGesture and _gesture) to
include a case where both 'gesture' and 'main-thread:gesture' are present, then
simulate removal of only 'main-thread:gesture' and assert that 'gesture' remains
unchanged; mirror the same change for the other related case referenced (the
analogous code around the other instance of props). Ensure assertions reference
the same props/_gesture symbols so the test verifies preservation of
user-provided 'gesture' attributes.

packages/react/runtime/__test__/snapshot/gesture.test.jsx (1)

1215-1221: Optional: fixture toJSON patterns are inconsistent across this file.

Most updated fixtures now use toJSON: function() { return { ...this, __isSerialized: true }; }, but the two test-scoped createGesture helpers (Lines 1215-1221, 1309-1315) still strip toJSON via const { toJSON, ...rest } = this. Both work, but the mixed style makes it harder to scan; consider unifying to one pattern. Purely cosmetic.

Also applies to: 1309-1315

🤖 Prompt for AI Agents

Verify each finding against the current code and only fix it if needed.

In `@packages/react/runtime/__test__/snapshot/gesture.test.jsx` around lines 1215
- 1221, Update the test-scoped createGesture helpers so their toJSON
implementations use the unified pattern used elsewhere: replace the method that
destructures off toJSON (currently written as toJSON() { const { toJSON, ...rest
} = this; return { ...rest, __isSerialized: true }; }) with the concise form
toJSON: function() { return { ...this, __isSerialized: true }; } in both
createGesture helpers (the helper that defines toJSON at the top and the second
test-scoped createGesture), ensuring both fixtures use the same style for
consistency.

packages/react/runtime/src/snapshot/gesture/processGestureBagkround.ts (1)

49-56: Optional: document the call-site invariant on attachCommittedSerializer.

Object.assign(gesture, { serialize, toJSON }) mutates the input. Current callers (Lines 72 and 93) always pass a freshly spread copy, so this is safe — but nothing in the type signature enforces it, and a future caller could accidentally pass the original and leak the committed serializer onto the background-cached gesture, reintroducing the very mutation churn this PR is trying to eliminate. A one-line comment like // callers MUST pass a fresh copy; this mutates the argument would harden the contract without changing behavior.

🤖 Prompt for AI Agents

Verify each finding against the current code and only fix it if needed.

In `@packages/react/runtime/src/snapshot/gesture/processGestureBagkround.ts`
around lines 49 - 56, Add a one-line invariant comment above
attachCommittedSerializer documenting that it mutates its argument and therefore
callers MUST pass a fresh/shallow-copied gesture; reference
attachCommittedSerializer and serializeCommittedGesture in the comment to make
it clear this function assigns serialize/toJSON onto the passed-in gesture and
that callers must not pass shared background-cached gesture objects.