fix(frames): reject context waits when page closes by microsoft-playwright-automation[bot] · Pull Request #43650 · microsoft/playwright

microsoft-playwright-automation · 2026-10-26T16:28:32Z

An uncommitted frame can wait indefinitely for an execution context after its page closes. Race context acquisition against the page's close scope and propagate target-closed errors without retrying or replacing them.

Extend the existing page-close promise-cancellation test with an unanswered iframe navigation. Verified on Linux across Chromium, Firefox and WebKit: 232 tests passed, 2 existing skips; the regression also passed in separate driver-process mode. An external DevTools target-close variant hangs without the fix and passes with it. Build and flint passed.

Fixes: #43647

_{Prepared by the triage workflow.}

Dmitry Gozman (dgozman) · 2026-10-27T11:21:19Z

    if (this._page.delegate.noUtilityWorld?.())
      world = 'main';
-    return this._contextData.get(world)!.contextPromise.then(contextOrDestroyedReason => {
+    return this._page.openScope.race(this._contextData.get(world)!.contextPromise.then(contextOrDestroyedReason => {


I think this is the wrong place. Instead, when frame detaches, it should reject all contextPomise instances on itself. Similarly, when the page closes, it should detach all frames (including the main frame) and thus invalidate the context.

github-actions · 2026-10-27T11:23:26Z

Hi, I'm the Playwright bot. I started to address the review comments on Linux: agent run

#43650 (comment): Reuse frame-detach context invalidation for all frames on page disposal, preserve the close reason, and remove the context-acquisition race and its retry exception.

github-actions · 2026-10-27T12:54:33Z

🔴 The Firefox close-reason failure is caused by this PR

Hi, I'm the Playwright bot. I compared the failing test on the CI merge and its base: 5/5 repetitions failed with this PR; 5/5 passed without it.

Details

The latest merged report has one failed test and one flaky test. The Firefox failure is a regression in the revised frame-disposal change; the video flake is pre-existing.

Caused by this PR

[firefox-page] › page/expect-misc.spec.ts:679 › EX-243 should report page close reason when page closes during expect — the assertion error omits Bye bye. on all four CI attempts. The new dispose() call to _onDetached(error.message) closes frame/context waits with plain Errors. The detached frame stops retrying, but Frame.expect() only recognizes target/session-closed errors when reporting the close reason, so it drops that reason.

On Linux with Firefox 158, the unchanged test failed all five repetitions, including every retry, on CI merge 124de408b289 (PR head 0b1a2e04b385). The same command on its base, c409f8c83ec0, passed all five: npm run test -- tests/page/expect-misc.spec.ts --project=firefox-page --grep EX-243 --workers=1 --repeat-each=5 --reporter=dot.

The refreshed CI database does show the same missing-reason error outside this PR under tracing-firefox, including a failure then pass on main. That establishes an existing flake, but the controlled comparison above ties the repeatable standard-Firefox failure to this change.

Pre-existing flake / infra

[chromium-library] › library/video.spec.ts:762 › screencast › VI-030 should work with video+trace — passed on retry here. The same test, project, Linux bot, and Cannot read properties of undefined (reading 'file') error occur on unrelated main SHA c409f8c83ec0 in run 40425277344.

Two MCP jobs are still running; this verdict covers the reported failures and completed checks.

_{Triaged by the Playwright bot - agent run}

Dmitry Gozman (dgozman)

This is now failing a test, please check.

Dmitry Gozman (dgozman) · 2026-10-28T15:07:23Z

+  if (contentOrError instanceof Error)
+    expect(contentOrError.message).toContain(kTargetClosedErrorMessage);
+  else
+    expect(contentOrError).toBe('<html><head></head><body></body></html>');


Why can this resolve with content? The iframe never loads, so this is very strange.

github-actions · 2026-10-28T15:07:57Z

Hi, I'm the Playwright bot. I started to address the review comments on Linux: agent run

github-actions · 2026-10-29T19:34:05Z

Hi, I'm the Playwright bot. I started to address the review comments on Linux: agent run

github-actions · 2026-10-29T20:29:29Z

Hi, I'm the Playwright bot. I started to address the review comments on Linux: agent run

#43650 (comment): Keep detaching all frames on page disposal and preserve the original error during context invalidation. #43650 (review): Fix the reproduced Firefox EX-243 failure by preserving TargetClosedError so assertions report the custom page close reason. #43650 (comment): Firefox can read the initial about:blank document before navigation commits; use a never-resolving iframe evaluation and require a target-closed error on every browser.

Dmitry Gozman (dgozman)

This needs a rebase now.

github-actions · 2026-10-30T14:32:15Z

Hi, I'm the Playwright bot. I started to address the review comments on Linux: agent run

#43650 (review): Rebased on 1d5572d; resolved the conflict by keeping detached-request cleanup on frame removal rather than page disposal, preserving page-close request errors. #43650 (comment): Already addressed; retained frame detachment and context invalidation on page disposal. #43650 (review): Retained original close errors; EX-243 and PA-051 pass across all three browsers after the rebase correction. #43650 (comment): Already addressed; retained the pending iframe evaluation and required target-closed error.

github-actions · 2026-10-30T15:18:23Z

Test results for "MCP"

5194 passed, 616 skipped

Merge workflow run.

github-actions · 2026-10-30T15:18:46Z

Test results for "tests 1"

1 flaky

⚠️ [chromium-library] › library/video.spec.ts:762 › screencast › VI-030 should work with video+trace `@chromium-playwright-x64-ubuntu24-64-core-node22`

23598 passed, 705 skipped

Merge workflow run.

github-actions Bot mentioned this pull request Oct 26, 2026

[Bug]: frame.content() on a frame whose document never arrives hangs forever, even after the page is closed #43647

Closed

This comment has been minimized.

Sign in to view

Dmitry Gozman (dgozman) reviewed Oct 27, 2026

View reviewed changes

Dmitry Gozman (dgozman) added the needs-revision label Oct 27, 2026

This comment has been minimized.

Sign in to view

Dmitry Gozman (dgozman) reviewed Oct 28, 2026

View reviewed changes

Dmitry Gozman (dgozman) added needs-revision and removed needs-revision labels Oct 28, 2026

Dmitry Gozman (dgozman) added needs-revision and removed needs-revision labels Oct 29, 2026

Pavel Feldman (pavelfeldman) added needs-revision and removed needs-revision labels Oct 29, 2026

This comment has been minimized.

Sign in to view

Dmitry Gozman (dgozman) approved these changes Oct 30, 2026

View reviewed changes

Dmitry Gozman (dgozman) reviewed Oct 30, 2026

View reviewed changes

Dmitry Gozman (dgozman) added needs-revision and removed needs-revision labels Oct 30, 2026

microsoft-playwright-automation Bot force-pushed the triage/43647/40348362900-1 branch from e4c04ea to ff487bf Compare October 30, 2026 14:43

Dmitry Gozman (dgozman) marked this pull request as ready for review October 30, 2026 16:12

Pavel Feldman (pavelfeldman) merged commit f0f066e into main Oct 30, 2026
24 checks passed

Pavel Feldman (pavelfeldman) deleted the triage/43647/40348362900-1 branch October 30, 2026 16:14

microsoft-playwright-automation Bot mentioned this pull request Oct 30, 2026

chore: roll Playwright to 1.65.0-next (eac236da28) microsoft/playwright-python#3305

Merged

Conversation

microsoft-playwright-automation Bot commented Oct 26, 2026

Uh oh!

This comment has been minimized.

This comment has been minimized.

Dmitry Gozman (dgozman) Oct 27, 2026

Choose a reason for hiding this comment

Uh oh!

github-actions Bot commented Oct 27, 2026

Uh oh!

This comment has been minimized.

github-actions Bot commented Oct 27, 2026

🔴 The Firefox close-reason failure is caused by this PR

Uh oh!

This comment has been minimized.

Dmitry Gozman (dgozman) left a comment

Choose a reason for hiding this comment

Uh oh!

Dmitry Gozman (dgozman) Oct 28, 2026

Choose a reason for hiding this comment

Uh oh!

github-actions Bot commented Oct 28, 2026

Uh oh!

github-actions Bot commented Oct 29, 2026

Uh oh!

github-actions Bot commented Oct 29, 2026

Uh oh!

This comment has been minimized.

This comment has been minimized.

Dmitry Gozman (dgozman) left a comment

Choose a reason for hiding this comment

Uh oh!

github-actions Bot commented Oct 30, 2026

Uh oh!

github-actions Bot commented Oct 30, 2026

Test results for "MCP"

Uh oh!

github-actions Bot commented Oct 30, 2026

Test results for "tests 1"

Uh oh!

Uh oh!

Reviewers

Uh oh!

Assignees

Labels

Projects

Milestone

Development

Uh oh!

2 participants