[iOS] Give hidden semantics nodes in scrollables a non-empty accessib… by uditswaroopa09 · Pull Request #193723 · flutter/flutter

uditswaroopa09 · 2026-10-02T19:37:55Z

Since iOS 27, VoiceOver no longer treats elements whose accessibilityFrame
is CGRectZero as candidates for the next focusable element. Flutter's
scrollables flag off-screen children as isHidden and report CGRectZero
for their frame, causing single-finger swipe to stop at the last visible
item instead of triggering accessibilityScrollToVisible and scrolling
implicitly.

Hidden nodes inside a FlutterScrollableSemanticsObject now report a 1pt
strip at the leading or trailing edge of the scrollable's viewport (depending
on which side the node lives on). VoiceOver treats these like clipped elements
and fires the existing showOnScreen path to scroll them into view.

Vertical and horizontal scrollables are both handled. Hidden nodes with no
scrollable ancestor retain their CGRectZero frame — no behaviour change for
those.

Related Issues

Fixes #192774

Tests

Three new XCTest cases in SemanticsObjectTest.mm:

testHiddenNodeInVerticalScrollableHasNonEmptyFrameAtViewportEdge
testHiddenNodeInHorizontalScrollableHasNonEmptyFrameAtViewportEdge
testHiddenNodeOutsideScrollableKeepsEmptyFrame

Checklist

I read the Contributor Guide and followed the process outlined there for submitting PRs
I read the Tree Hygiene wiki page, which explains my responsibilities
I read and followed the Flutter Style Guide, including Features we expect every widget to implement
I signed the CLA
I listed at least one issue that this PR fixes in the description above
I updated/added relevant documentation
I added new tests to check the change I am making, or this PR is test-exempt
All existing and new tests are passing

…ility frame Since iOS 27, VoiceOver no longer treats elements whose accessibilityFrame is CGRectZero as candidates for the next element. Off-screen children of a scrollable are flagged isHidden and reported CGRectZero, so a one-finger swipe stopped at the last visible item instead of triggering accessibilityScrollToVisible and an implicit scroll. Hidden nodes inside a FlutterScrollableSemanticsObject now report a 1pt strip at the leading or trailing edge of the scrollable's viewport, depending on which side the node lives on. VoiceOver handles them like clipped elements, and the existing showOnScreen path scrolls them in. Fixes flutter#192774

gemini-code-assist

Code Review

This pull request updates SemanticsObject.mm to calculate and return a non-empty frame for hidden nodes inside scrollable viewports, preventing VoiceOver from skipping them, and adds corresponding tests in SemanticsObjectTest.mm. The review feedback highlights memory leaks in the new tests due to manual heap allocation of MockAccessibilityBridge and suggests using std::unique_ptr for lifetime management. It also recommends replacing XCTAssertTrue with XCTAssertEqual in assertions to provide clearer failure messages.

gemini-code-assist · 2026-10-02T19:40:37Z

+  fml::WeakPtrFactory<flutter::testing::MockAccessibilityBridge> factory(
+      new flutter::testing::MockAccessibilityBridge());
+  fml::WeakPtr<flutter::testing::MockAccessibilityBridge> bridge = factory.GetWeakPtr();


The test allocates MockAccessibilityBridge on the heap using new but never deletes it, leading to a memory leak. Since fml::WeakPtrFactory does not take ownership of the pointer, we should manage its lifetime using std::unique_ptr to prevent memory leaks.

auto bridge_impl = std::make_unique<flutter::testing::MockAccessibilityBridge>();\n fml::WeakPtrFactory<flutter::testing::MockAccessibilityBridge> factory(bridge_impl.get());\n fml::WeakPtr<flutter::testing::MockAccessibilityBridge> bridge = factory.GetWeakPtr();

References

C++ code should follow the Google C++ Style Guide, which recommends using std::unique_ptr to manage ownership of dynamically allocated objects. ^(link)

gemini-code-assist · 2026-10-02T19:40:37Z

+  XCTAssertTrue(bridge->observations.size() == 1);
+  XCTAssertTrue(bridge->observations[0].id == 3);
+  XCTAssertTrue(bridge->observations[0].action == flutter::SemanticsAction::kShowOnScreen);


Using XCTAssertEqual instead of XCTAssertTrue with an equality operator provides much better failure messages because it prints both the expected and actual values when the assertion fails.

XCTAssertEqual(bridge->observations.size(), 1U);\n XCTAssertEqual(bridge->observations[0].id, 3);\n XCTAssertEqual(bridge->observations[0].action, flutter::SemanticsAction::kShowOnScreen);

gemini-code-assist · 2026-10-02T19:40:37Z

+  fml::WeakPtrFactory<flutter::AccessibilityBridgeIos> factory(
+      new flutter::testing::MockAccessibilityBridge());
+  fml::WeakPtr<flutter::AccessibilityBridgeIos> bridge = factory.GetWeakPtr();


The test allocates MockAccessibilityBridge on the heap using new but never deletes it, leading to a memory leak. Since fml::WeakPtrFactory does not take ownership of the pointer, we should manage its lifetime using std::unique_ptr to prevent memory leaks.

auto bridge_impl = std::make_unique<flutter::testing::MockAccessibilityBridge>();\n fml::WeakPtrFactory<flutter::AccessibilityBridgeIos> factory(bridge_impl.get());\n fml::WeakPtr<flutter::AccessibilityBridgeIos> bridge = factory.GetWeakPtr();

References

C++ code should follow the Google C++ Style Guide, which recommends using std::unique_ptr to manage ownership of dynamically allocated objects. ^(link)

gemini-code-assist · 2026-10-02T19:40:37Z

+  fml::WeakPtrFactory<flutter::AccessibilityBridgeIos> factory(
+      new flutter::testing::MockAccessibilityBridge());
+  fml::WeakPtr<flutter::AccessibilityBridgeIos> bridge = factory.GetWeakPtr();


The test allocates MockAccessibilityBridge on the heap using new but never deletes it, leading to a memory leak. Since fml::WeakPtrFactory does not take ownership of the pointer, we should manage its lifetime using std::unique_ptr to prevent memory leaks.

auto bridge_impl = std::make_unique<flutter::testing::MockAccessibilityBridge>();\n fml::WeakPtrFactory<flutter::AccessibilityBridgeIos> factory(bridge_impl.get());\n fml::WeakPtr<flutter::AccessibilityBridgeIos> bridge = factory.GetWeakPtr();

References

C++ code should follow the Google C++ Style Guide, which recommends using std::unique_ptr to manage ownership of dynamically allocated objects. ^(link)

uditswaroopa09 · 2026-10-02T19:47:59Z

Code Review

This pull request updates SemanticsObject.mm to calculate and return a non-empty frame for hidden nodes inside scrollable viewports, preventing VoiceOver from skipping them, and adds corresponding tests in SemanticsObjectTest.mm. The review feedback highlights memory leaks in the new tests due to manual heap allocation of MockAccessibilityBridge and suggests using std::unique_ptr for lifetime management. It also recommends replacing XCTAssertTrue with XCTAssertEqual in assertions to provide clearer failure messages.

Thanks. These follow the existing convention in SemanticsObjectTest.mm —
MockAccessibilityBridge is constructed with new at ~57 existing call
sites in this file, and XCTAssertTrue(... == ...) is the prevailing assert
style here. I kept the new tests consistent with surrounding code rather
than introducing a mixed pattern. Happy to switch all of them in a separate
cleanup if the team prefers std::unique_ptr / XCTAssertEqual broadly.

FrancescoAbbondo · 2026-10-09T10:17:50Z

Hi @LongCatIsLooong @LouiseHsu Do you have an estimated timeline for approving this pull request? We urgently need this fix. Thanks!

LongCatIsLooong · 2026-10-09T21:57:20Z

Hi @LongCatIsLooong @LouiseHsu Do you have an estimated timeline for approving this pull request? We urgently need this fix. Thanks!

I looked at this PR but I'm still debugging the issue to figure out if there's a way to fix the root cause. The plan is to put a fix in the upcoming 3.50 release. If you need a fix right now the fastest way is to apply the patch locally (https://github.com/flutter/flutter/blob/master/docs/releases/Flutter-build-release-channels.md#will-a-particular-bug-fix-be-provided-in-a-hotfix-release).

john-madaras · 2026-10-20T23:44:31Z

Hi, any news on this ticket? This is causing significant issues for impaired iOS 27 users who leverage Voiceover. Thanks!

LongCatIsLooong · 2026-11-12T23:47:40Z

(keeping this open in case the other PR isn't a good cherry-pick candidate).

uditswaroopa09 requested a review from a team as a code owner October 2, 2026 19:37

github-actions Bot added platform-ios iOS applications specifically engine flutter/engine related. See also e: labels. team-ios Owned by iOS platform team labels Oct 2, 2026

gemini-code-assist Bot reviewed Oct 2, 2026

View reviewed changes

LouiseHsu requested review from LongCatIsLooong and LouiseHsu October 8, 2026 21:32

davefaliskie mentioned this pull request Oct 27, 2026

[iOS 27] VoiceOver's implicit-scroll-into-view no longer works - single-finger swipe never advances past the visible boundary of a Scrollable #192774

Closed

Navigation Menu

Repository navigation

[iOS] Give hidden semantics nodes in scrollables a non-empty accessib… - #193723

[iOS] Give hidden semantics nodes in scrollables a non-empty accessib…#193723
uditswaroopa09 wants to merge 1 commit into
flutter:masterfrom
uditswaroopa09:ios27-voiceover-scroll-hidden-frame

uditswaroopa09 commented Oct 2, 2026

Uh oh!

gemini-code-assist Bot left a comment

Uh oh!

gemini-code-assist Bot Oct 2, 2026

Uh oh!

gemini-code-assist Bot Oct 2, 2026

Uh oh!

gemini-code-assist Bot Oct 2, 2026

Uh oh!

gemini-code-assist Bot Oct 2, 2026

Uh oh!

uditswaroopa09 commented Oct 2, 2026

Code Review

Uh oh!

FrancescoAbbondo commented Oct 9, 2026

Uh oh!

LongCatIsLooong commented Oct 9, 2026

Uh oh!

john-madaras commented Oct 20, 2026

Uh oh!

LongCatIsLooong commented Nov 12, 2026

Uh oh!

Reviewers

Uh oh!

Assignees

Labels

Projects

Milestone

Development

Uh oh!

4 participants

Conversation

uditswaroopa09 commented Oct 2, 2026

Related Issues

Tests

Checklist

Uh oh!

gemini-code-assist Bot left a comment

Choose a reason for hiding this comment

Code Review

Uh oh!

gemini-code-assist Bot Oct 2, 2026

Choose a reason for hiding this comment

Uh oh!

gemini-code-assist Bot Oct 2, 2026

Choose a reason for hiding this comment

Uh oh!

gemini-code-assist Bot Oct 2, 2026

Choose a reason for hiding this comment

Uh oh!

gemini-code-assist Bot Oct 2, 2026

Choose a reason for hiding this comment

Uh oh!

uditswaroopa09 commented Oct 2, 2026

Code Review

Uh oh!

FrancescoAbbondo commented Oct 9, 2026

Uh oh!

LongCatIsLooong commented Oct 9, 2026

Uh oh!

john-madaras commented Oct 20, 2026

Uh oh!

LongCatIsLooong commented Nov 12, 2026

Uh oh!

Reviewers

Uh oh!

Assignees

Labels

Projects

Milestone

Development

Uh oh!

4 participants