fix(ui-dialog): cancel the pending focus region activation on close by matyasf · Pull Request #2697 · instructure/instructure-ui

matyasf · 2026-08-25T12:40:24Z

Summary

Dialog.close() cancels any still-scheduled requestAnimationFrame region activation and clears
_focusRegion after blurring. Without it, a Dialog closed before that frame ran (e.g. a Tray
opened and closed within one frame, which happens when rAF callbacks land late under load)
activated a region nothing ever blurred — its document keydown listener then ran scopeTab on
an unrendered element, preventDefaulting every later Tab press.
Adds a Dialog regression test that flushes the activation frame manually after the close and
asserts Tab still moves focus outside the Dialog.

Test Plan

This was the cause of the flaky Tray should handle focus properly in complex cases browser
test. To reproduce the old failure, stub window.requestAnimationFrame to fire ~150ms late and
run packages/ui-tray/src/Tray/__tests__/Tray.test.tsx — it fails on the first userEvent.tab()
without this fix and passes with it.
Worth a manual keyboard pass on Modal/Popover/Tray/Menu: open, close, then Tab around to confirm
focus return and tab order are unchanged.

🤖 Generated with Claude Code

github-actions · 2026-08-25T12:45:55Z

PR Preview Action v1.8.1
Preview removed because the pull request was closed.
2026-09-01 08:52 UTC

github-actions · 2026-08-25T12:49:31Z

Visual regression report

Cypress suite: ✅ Passing

Visual diff:⚠️Changes detected.

Status	Count
Unchanged	95
Changed	1
New	0
Removed	0

Accessibility (axe): ✅ No violations.

📊 View full report — click a screenshot's ⚠ badge to see each violation boxed on the image, with the offending element named and contrast failures shown as color swatches.

Diff images (1)

badge-canvas.png — 1573 pixels differ

_{Baselines come from the visual-baselines branch. They refresh on every merge to master. The Cypress suite line covers the a11y and console-error assertions — a ❌ there means the suite found real issues even if the visual diff is clean.}

balzss

although the code looks good, the 2 visual regression test changes look interesting. the table one seems unrelated which is even weirder, but the menu one might has something to do with the changes. it looks like the dark theme now not highlights the first menu option? also the other themes are not consistent with highlighting/not-highlighting the first item. do you think it's related to this change?

matyasf · 2026-08-27T20:36:37Z

@balzss the 2 visual regression test changes look interesting

I've re-ran the VRT and the changes are gone..

Dialog activates its FocusRegion in a requestAnimationFrame callback. When the Dialog closed before that frame ran (e.g. a Tray that is opened and closed within the same frame, which happens on a loaded machine where rAF callbacks land late), close() found no region to blur and left the frame scheduled. The callback then activated a region for an already closed Dialog, which nothing ever blurred: componentWillUnmount only closes while open. The leaked region kept a document keydown listener that scoped every later tab press to an element that is not rendered anymore, so scopeTab called preventDefault and Tab stopped working. close() now cancels any scheduled activation and clears _focusRegion after blurring. This is what made the Tray "should handle focus properly in complex cases" browser test flaky on CI. 🤖 Generated with [Claude Code](https://claude.com/claude-code) Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>

matyasf · 2026-08-28T11:54:38Z

VRT is flaky, now I run it again, and now it shows a change in Badge

matyasf · 2026-08-25T12:40:24Z

Summary

Dialog.close() cancels any still-scheduled requestAnimationFrame region activation and clears
_focusRegion after blurring. Without it, a Dialog closed before that frame ran (e.g. a Tray
opened and closed within one frame, which happens when rAF callbacks land late under load)
activated a region nothing ever blurred — its document keydown listener then ran scopeTab on
an unrendered element, preventDefaulting every later Tab press.
Adds a Dialog regression test that flushes the activation frame manually after the close and
asserts Tab still moves focus outside the Dialog.

Test Plan

This was the cause of the flaky Tray should handle focus properly in complex cases browser
test. To reproduce the old failure, stub window.requestAnimationFrame to fire ~150ms late and
run packages/ui-tray/src/Tray/__tests__/Tray.test.tsx — it fails on the first userEvent.tab()
without this fix and passes with it.
Worth a manual keyboard pass on Modal/Popover/Tray/Menu: open, close, then Tab around to confirm
focus return and tab order are unchanged.

🤖 Generated with Claude Code

github-actions · 2026-08-25T12:45:55Z

PR Preview Action v1.8.1
Preview removed because the pull request was closed.
2026-09-01 08:52 UTC

github-actions · 2026-08-25T12:49:31Z

Visual regression report

Cypress suite: ✅ Passing

Visual diff:⚠️Changes detected.

Status	Count
Unchanged	95
Changed	1
New	0
Removed	0

Accessibility (axe): ✅ No violations.

📊 View full report — click a screenshot's ⚠ badge to see each violation boxed on the image, with the offending element named and contrast failures shown as color swatches.

Diff images (1)

badge-canvas.png — 1573 pixels differ

_{Baselines come from the visual-baselines branch. They refresh on every merge to master. The Cypress suite line covers the a11y and console-error assertions — a ❌ there means the suite found real issues even if the visual diff is clean.}

balzss

although the code looks good, the 2 visual regression test changes look interesting. the table one seems unrelated which is even weirder, but the menu one might has something to do with the changes. it looks like the dark theme now not highlights the first menu option? also the other themes are not consistent with highlighting/not-highlighting the first item. do you think it's related to this change?

matyasf · 2026-08-27T20:36:37Z

@balzss the 2 visual regression test changes look interesting

I've re-ran the VRT and the changes are gone..

Dialog activates its FocusRegion in a requestAnimationFrame callback. When the Dialog closed before that frame ran (e.g. a Tray that is opened and closed within the same frame, which happens on a loaded machine where rAF callbacks land late), close() found no region to blur and left the frame scheduled. The callback then activated a region for an already closed Dialog, which nothing ever blurred: componentWillUnmount only closes while open. The leaked region kept a document keydown listener that scoped every later tab press to an element that is not rendered anymore, so scopeTab called preventDefault and Tab stopped working. close() now cancels any scheduled activation and clears _focusRegion after blurring. This is what made the Tray "should handle focus properly in complex cases" browser test flaky on CI. 🤖 Generated with [Claude Code](https://claude.com/claude-code) Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>

matyasf · 2026-08-28T11:54:38Z

VRT is flaky, now I run it again, and now it shows a change in Badge

matyasf · 2026-08-25T12:40:24Z

Summary

Dialog.close() cancels any still-scheduled requestAnimationFrame region activation and clears
_focusRegion after blurring. Without it, a Dialog closed before that frame ran (e.g. a Tray
opened and closed within one frame, which happens when rAF callbacks land late under load)
activated a region nothing ever blurred — its document keydown listener then ran scopeTab on
an unrendered element, preventDefaulting every later Tab press.
Adds a Dialog regression test that flushes the activation frame manually after the close and
asserts Tab still moves focus outside the Dialog.

Test Plan

This was the cause of the flaky Tray should handle focus properly in complex cases browser
test. To reproduce the old failure, stub window.requestAnimationFrame to fire ~150ms late and
run packages/ui-tray/src/Tray/__tests__/Tray.test.tsx — it fails on the first userEvent.tab()
without this fix and passes with it.
Worth a manual keyboard pass on Modal/Popover/Tray/Menu: open, close, then Tab around to confirm
focus return and tab order are unchanged.

🤖 Generated with Claude Code

github-actions · 2026-08-25T12:45:55Z

PR Preview Action v1.8.1
Preview removed because the pull request was closed.
2026-09-01 08:52 UTC

github-actions · 2026-08-25T12:49:31Z

Visual regression report

Cypress suite: ✅ Passing

Visual diff:⚠️Changes detected.

Status	Count
Unchanged	95
Changed	1
New	0
Removed	0

Accessibility (axe): ✅ No violations.

📊 View full report — click a screenshot's ⚠ badge to see each violation boxed on the image, with the offending element named and contrast failures shown as color swatches.

Diff images (1)

badge-canvas.png — 1573 pixels differ

_{Baselines come from the visual-baselines branch. They refresh on every merge to master. The Cypress suite line covers the a11y and console-error assertions — a ❌ there means the suite found real issues even if the visual diff is clean.}

balzss

although the code looks good, the 2 visual regression test changes look interesting. the table one seems unrelated which is even weirder, but the menu one might has something to do with the changes. it looks like the dark theme now not highlights the first menu option? also the other themes are not consistent with highlighting/not-highlighting the first item. do you think it's related to this change?

matyasf · 2026-08-27T20:36:37Z

@balzss the 2 visual regression test changes look interesting

I've re-ran the VRT and the changes are gone..

Dialog activates its FocusRegion in a requestAnimationFrame callback. When the Dialog closed before that frame ran (e.g. a Tray that is opened and closed within the same frame, which happens on a loaded machine where rAF callbacks land late), close() found no region to blur and left the frame scheduled. The callback then activated a region for an already closed Dialog, which nothing ever blurred: componentWillUnmount only closes while open. The leaked region kept a document keydown listener that scoped every later tab press to an element that is not rendered anymore, so scopeTab called preventDefault and Tab stopped working. close() now cancels any scheduled activation and clears _focusRegion after blurring. This is what made the Tray "should handle focus properly in complex cases" browser test flaky on CI. 🤖 Generated with [Claude Code](https://claude.com/claude-code) Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>

matyasf · 2026-08-28T11:54:38Z

VRT is flaky, now I run it again, and now it shows a change in Badge

matyasf · 2026-08-25T12:40:24Z

Summary

Dialog.close() cancels any still-scheduled requestAnimationFrame region activation and clears
_focusRegion after blurring. Without it, a Dialog closed before that frame ran (e.g. a Tray
opened and closed within one frame, which happens when rAF callbacks land late under load)
activated a region nothing ever blurred — its document keydown listener then ran scopeTab on
an unrendered element, preventDefaulting every later Tab press.
Adds a Dialog regression test that flushes the activation frame manually after the close and
asserts Tab still moves focus outside the Dialog.

Test Plan

This was the cause of the flaky Tray should handle focus properly in complex cases browser
test. To reproduce the old failure, stub window.requestAnimationFrame to fire ~150ms late and
run packages/ui-tray/src/Tray/__tests__/Tray.test.tsx — it fails on the first userEvent.tab()
without this fix and passes with it.
Worth a manual keyboard pass on Modal/Popover/Tray/Menu: open, close, then Tab around to confirm
focus return and tab order are unchanged.

🤖 Generated with Claude Code

github-actions · 2026-08-25T12:45:55Z

PR Preview Action v1.8.1
Preview removed because the pull request was closed.
2026-09-01 08:52 UTC

github-actions · 2026-08-25T12:49:31Z

Visual regression report

Cypress suite: ✅ Passing

Visual diff:⚠️Changes detected.

Status	Count
Unchanged	95
Changed	1
New	0
Removed	0

Accessibility (axe): ✅ No violations.

📊 View full report — click a screenshot's ⚠ badge to see each violation boxed on the image, with the offending element named and contrast failures shown as color swatches.

Diff images (1)

badge-canvas.png — 1573 pixels differ

_{Baselines come from the visual-baselines branch. They refresh on every merge to master. The Cypress suite line covers the a11y and console-error assertions — a ❌ there means the suite found real issues even if the visual diff is clean.}

balzss

although the code looks good, the 2 visual regression test changes look interesting. the table one seems unrelated which is even weirder, but the menu one might has something to do with the changes. it looks like the dark theme now not highlights the first menu option? also the other themes are not consistent with highlighting/not-highlighting the first item. do you think it's related to this change?

matyasf · 2026-08-27T20:36:37Z

@balzss the 2 visual regression test changes look interesting

I've re-ran the VRT and the changes are gone..

Dialog activates its FocusRegion in a requestAnimationFrame callback. When the Dialog closed before that frame ran (e.g. a Tray that is opened and closed within the same frame, which happens on a loaded machine where rAF callbacks land late), close() found no region to blur and left the frame scheduled. The callback then activated a region for an already closed Dialog, which nothing ever blurred: componentWillUnmount only closes while open. The leaked region kept a document keydown listener that scoped every later tab press to an element that is not rendered anymore, so scopeTab called preventDefault and Tab stopped working. close() now cancels any scheduled activation and clears _focusRegion after blurring. This is what made the Tray "should handle focus properly in complex cases" browser test flaky on CI. 🤖 Generated with [Claude Code](https://claude.com/claude-code) Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>

matyasf · 2026-08-28T11:54:38Z

VRT is flaky, now I run it again, and now it shows a change in Badge

matyasf · 2026-08-25T12:40:24Z

Summary

Dialog.close() cancels any still-scheduled requestAnimationFrame region activation and clears
_focusRegion after blurring. Without it, a Dialog closed before that frame ran (e.g. a Tray
opened and closed within one frame, which happens when rAF callbacks land late under load)
activated a region nothing ever blurred — its document keydown listener then ran scopeTab on
an unrendered element, preventDefaulting every later Tab press.
Adds a Dialog regression test that flushes the activation frame manually after the close and
asserts Tab still moves focus outside the Dialog.

Test Plan

This was the cause of the flaky Tray should handle focus properly in complex cases browser
test. To reproduce the old failure, stub window.requestAnimationFrame to fire ~150ms late and
run packages/ui-tray/src/Tray/__tests__/Tray.test.tsx — it fails on the first userEvent.tab()
without this fix and passes with it.
Worth a manual keyboard pass on Modal/Popover/Tray/Menu: open, close, then Tab around to confirm
focus return and tab order are unchanged.

🤖 Generated with Claude Code

github-actions · 2026-08-25T12:45:55Z

PR Preview Action v1.8.1
Preview removed because the pull request was closed.
2026-09-01 08:52 UTC

github-actions · 2026-08-25T12:49:31Z

Visual regression report

Cypress suite: ✅ Passing

Visual diff:⚠️Changes detected.

Status	Count
Unchanged	95
Changed	1
New	0
Removed	0

Accessibility (axe): ✅ No violations.

📊 View full report — click a screenshot's ⚠ badge to see each violation boxed on the image, with the offending element named and contrast failures shown as color swatches.

Diff images (1)

badge-canvas.png — 1573 pixels differ

_{Baselines come from the visual-baselines branch. They refresh on every merge to master. The Cypress suite line covers the a11y and console-error assertions — a ❌ there means the suite found real issues even if the visual diff is clean.}

balzss

although the code looks good, the 2 visual regression test changes look interesting. the table one seems unrelated which is even weirder, but the menu one might has something to do with the changes. it looks like the dark theme now not highlights the first menu option? also the other themes are not consistent with highlighting/not-highlighting the first item. do you think it's related to this change?

matyasf · 2026-08-27T20:36:37Z

@balzss the 2 visual regression test changes look interesting

I've re-ran the VRT and the changes are gone..

Dialog activates its FocusRegion in a requestAnimationFrame callback. When the Dialog closed before that frame ran (e.g. a Tray that is opened and closed within the same frame, which happens on a loaded machine where rAF callbacks land late), close() found no region to blur and left the frame scheduled. The callback then activated a region for an already closed Dialog, which nothing ever blurred: componentWillUnmount only closes while open. The leaked region kept a document keydown listener that scoped every later tab press to an element that is not rendered anymore, so scopeTab called preventDefault and Tab stopped working. close() now cancels any scheduled activation and clears _focusRegion after blurring. This is what made the Tray "should handle focus properly in complex cases" browser test flaky on CI. 🤖 Generated with [Claude Code](https://claude.com/claude-code) Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>

matyasf · 2026-08-28T11:54:38Z

VRT is flaky, now I run it again, and now it shows a change in Badge

matyasf · 2026-08-25T12:40:24Z

Summary

Dialog.close() cancels any still-scheduled requestAnimationFrame region activation and clears
_focusRegion after blurring. Without it, a Dialog closed before that frame ran (e.g. a Tray
opened and closed within one frame, which happens when rAF callbacks land late under load)
activated a region nothing ever blurred — its document keydown listener then ran scopeTab on
an unrendered element, preventDefaulting every later Tab press.
Adds a Dialog regression test that flushes the activation frame manually after the close and
asserts Tab still moves focus outside the Dialog.

Test Plan

This was the cause of the flaky Tray should handle focus properly in complex cases browser
test. To reproduce the old failure, stub window.requestAnimationFrame to fire ~150ms late and
run packages/ui-tray/src/Tray/__tests__/Tray.test.tsx — it fails on the first userEvent.tab()
without this fix and passes with it.
Worth a manual keyboard pass on Modal/Popover/Tray/Menu: open, close, then Tab around to confirm
focus return and tab order are unchanged.

🤖 Generated with Claude Code

github-actions · 2026-08-25T12:45:55Z

PR Preview Action v1.8.1
Preview removed because the pull request was closed.
2026-09-01 08:52 UTC

github-actions · 2026-08-25T12:49:31Z

Visual regression report

Cypress suite: ✅ Passing

Visual diff:⚠️Changes detected.

Status	Count
Unchanged	95
Changed	1
New	0
Removed	0

Accessibility (axe): ✅ No violations.

📊 View full report — click a screenshot's ⚠ badge to see each violation boxed on the image, with the offending element named and contrast failures shown as color swatches.

Diff images (1)

badge-canvas.png — 1573 pixels differ

_{Baselines come from the visual-baselines branch. They refresh on every merge to master. The Cypress suite line covers the a11y and console-error assertions — a ❌ there means the suite found real issues even if the visual diff is clean.}

balzss

although the code looks good, the 2 visual regression test changes look interesting. the table one seems unrelated which is even weirder, but the menu one might has something to do with the changes. it looks like the dark theme now not highlights the first menu option? also the other themes are not consistent with highlighting/not-highlighting the first item. do you think it's related to this change?

matyasf · 2026-08-27T20:36:37Z

@balzss the 2 visual regression test changes look interesting

I've re-ran the VRT and the changes are gone..

Dialog activates its FocusRegion in a requestAnimationFrame callback. When the Dialog closed before that frame ran (e.g. a Tray that is opened and closed within the same frame, which happens on a loaded machine where rAF callbacks land late), close() found no region to blur and left the frame scheduled. The callback then activated a region for an already closed Dialog, which nothing ever blurred: componentWillUnmount only closes while open. The leaked region kept a document keydown listener that scoped every later tab press to an element that is not rendered anymore, so scopeTab called preventDefault and Tab stopped working. close() now cancels any scheduled activation and clears _focusRegion after blurring. This is what made the Tray "should handle focus properly in complex cases" browser test flaky on CI. 🤖 Generated with [Claude Code](https://claude.com/claude-code) Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>

matyasf · 2026-08-28T11:54:38Z

VRT is flaky, now I run it again, and now it shows a change in Badge

matyasf · 2026-08-25T12:40:24Z

Summary

Dialog.close() cancels any still-scheduled requestAnimationFrame region activation and clears
_focusRegion after blurring. Without it, a Dialog closed before that frame ran (e.g. a Tray
opened and closed within one frame, which happens when rAF callbacks land late under load)
activated a region nothing ever blurred — its document keydown listener then ran scopeTab on
an unrendered element, preventDefaulting every later Tab press.
Adds a Dialog regression test that flushes the activation frame manually after the close and
asserts Tab still moves focus outside the Dialog.

Test Plan

This was the cause of the flaky Tray should handle focus properly in complex cases browser
test. To reproduce the old failure, stub window.requestAnimationFrame to fire ~150ms late and
run packages/ui-tray/src/Tray/__tests__/Tray.test.tsx — it fails on the first userEvent.tab()
without this fix and passes with it.
Worth a manual keyboard pass on Modal/Popover/Tray/Menu: open, close, then Tab around to confirm
focus return and tab order are unchanged.

🤖 Generated with Claude Code

github-actions · 2026-08-25T12:45:55Z

PR Preview Action v1.8.1
Preview removed because the pull request was closed.
2026-09-01 08:52 UTC

github-actions · 2026-08-25T12:49:31Z

Visual regression report

Cypress suite: ✅ Passing

Visual diff:⚠️Changes detected.

Status	Count
Unchanged	95
Changed	1
New	0
Removed	0

Accessibility (axe): ✅ No violations.

📊 View full report — click a screenshot's ⚠ badge to see each violation boxed on the image, with the offending element named and contrast failures shown as color swatches.

Diff images (1)

badge-canvas.png — 1573 pixels differ

_{Baselines come from the visual-baselines branch. They refresh on every merge to master. The Cypress suite line covers the a11y and console-error assertions — a ❌ there means the suite found real issues even if the visual diff is clean.}

balzss

although the code looks good, the 2 visual regression test changes look interesting. the table one seems unrelated which is even weirder, but the menu one might has something to do with the changes. it looks like the dark theme now not highlights the first menu option? also the other themes are not consistent with highlighting/not-highlighting the first item. do you think it's related to this change?

matyasf · 2026-08-27T20:36:37Z

@balzss the 2 visual regression test changes look interesting

I've re-ran the VRT and the changes are gone..

Dialog activates its FocusRegion in a requestAnimationFrame callback. When the Dialog closed before that frame ran (e.g. a Tray that is opened and closed within the same frame, which happens on a loaded machine where rAF callbacks land late), close() found no region to blur and left the frame scheduled. The callback then activated a region for an already closed Dialog, which nothing ever blurred: componentWillUnmount only closes while open. The leaked region kept a document keydown listener that scoped every later tab press to an element that is not rendered anymore, so scopeTab called preventDefault and Tab stopped working. close() now cancels any scheduled activation and clears _focusRegion after blurring. This is what made the Tray "should handle focus properly in complex cases" browser test flaky on CI. 🤖 Generated with [Claude Code](https://claude.com/claude-code) Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>

matyasf · 2026-08-28T11:54:38Z

VRT is flaky, now I run it again, and now it shows a change in Badge

matyasf · 2026-08-25T12:40:24Z

Summary

Dialog.close() cancels any still-scheduled requestAnimationFrame region activation and clears
_focusRegion after blurring. Without it, a Dialog closed before that frame ran (e.g. a Tray
opened and closed within one frame, which happens when rAF callbacks land late under load)
activated a region nothing ever blurred — its document keydown listener then ran scopeTab on
an unrendered element, preventDefaulting every later Tab press.
Adds a Dialog regression test that flushes the activation frame manually after the close and
asserts Tab still moves focus outside the Dialog.

Test Plan

This was the cause of the flaky Tray should handle focus properly in complex cases browser
test. To reproduce the old failure, stub window.requestAnimationFrame to fire ~150ms late and
run packages/ui-tray/src/Tray/__tests__/Tray.test.tsx — it fails on the first userEvent.tab()
without this fix and passes with it.
Worth a manual keyboard pass on Modal/Popover/Tray/Menu: open, close, then Tab around to confirm
focus return and tab order are unchanged.

🤖 Generated with Claude Code

github-actions · 2026-08-25T12:45:55Z

PR Preview Action v1.8.1
Preview removed because the pull request was closed.
2026-09-01 08:52 UTC

github-actions · 2026-08-25T12:49:31Z

Visual regression report

Cypress suite: ✅ Passing

Visual diff:⚠️Changes detected.

Status	Count
Unchanged	95
Changed	1
New	0
Removed	0

Accessibility (axe): ✅ No violations.

📊 View full report — click a screenshot's ⚠ badge to see each violation boxed on the image, with the offending element named and contrast failures shown as color swatches.

Diff images (1)

badge-canvas.png — 1573 pixels differ

_{Baselines come from the visual-baselines branch. They refresh on every merge to master. The Cypress suite line covers the a11y and console-error assertions — a ❌ there means the suite found real issues even if the visual diff is clean.}

balzss

although the code looks good, the 2 visual regression test changes look interesting. the table one seems unrelated which is even weirder, but the menu one might has something to do with the changes. it looks like the dark theme now not highlights the first menu option? also the other themes are not consistent with highlighting/not-highlighting the first item. do you think it's related to this change?

matyasf · 2026-08-27T20:36:37Z

@balzss the 2 visual regression test changes look interesting

I've re-ran the VRT and the changes are gone..

Dialog activates its FocusRegion in a requestAnimationFrame callback. When the Dialog closed before that frame ran (e.g. a Tray that is opened and closed within the same frame, which happens on a loaded machine where rAF callbacks land late), close() found no region to blur and left the frame scheduled. The callback then activated a region for an already closed Dialog, which nothing ever blurred: componentWillUnmount only closes while open. The leaked region kept a document keydown listener that scoped every later tab press to an element that is not rendered anymore, so scopeTab called preventDefault and Tab stopped working. close() now cancels any scheduled activation and clears _focusRegion after blurring. This is what made the Tray "should handle focus properly in complex cases" browser test flaky on CI. 🤖 Generated with [Claude Code](https://claude.com/claude-code) Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>

matyasf · 2026-08-28T11:54:38Z

VRT is flaky, now I run it again, and now it shows a change in Badge

matyasf self-assigned this Aug 25, 2026

matyasf requested review from balzss and joyenjoyer August 25, 2026 12:41

github-actionsBot pushed a commit that referenced this pull request Aug 25, 2026
visual-regression: PR#2697(90bfe1e)90bfe1e
ad24329

joyenjoyer approved these changes Aug 26, 2026
View reviewed changes

balzss reviewed Aug 27, 2026
View reviewed changes

github-actionsBot pushed a commit that referenced this pull request Aug 27, 2026
visual-regression: PR#2697(90bfe1e)90bfe1e
3b02dca

matyasf force-pushed the fix_dialog_test branch from 48622de to 0fbf026Compare August 28, 2026 09:03

github-actionsBot pushed a commit that referenced this pull request Aug 28, 2026
visual-regression: PR#2697(4abe212)4abe212
1ed9dce

matyasf requested a review from balzss August 28, 2026 11:54

balzss approved these changes Aug 28, 2026
View reviewed changes

matyasf merged commit 9b0467d into masterSep 1, 2026
11 of 13 checks passed

matyasf deleted the fix_dialog_test branch September 1, 2026 08:52

github-actionsBot added a commit that referenced this pull request Sep 1, 2026
visual-regression: remove report for closed PR#2697
3a1ee5f