test(webkit): screenshot webgl canvas after drawing buffer resize

[Yury Semikhatsky (yury-s)]

Summary

• Regression test for a blank WebGL canvas in screenshots after its drawing buffer is resized (Windows WebKit).
• Needs the WebKit fix from https://github.com/microsoft/playwright-browsers/pull/2564 (build 2371) to pass on Windows.

Fixes #42885

[github-actions]

Test results for "tests 1"

1 failed
❌ [playwright-test] › global-setup.spec.ts:313 › globalSetup should work for auth @windows-latest-node22

6 flaky

⚠️ [chromium-library] › library/browsercontext-fetch.spec.ts:1377 › should abort requests when browser context closes `@chromium-ubuntu-22.04-arm-node20`
⚠️ [chromium-library] › library/video.spec.ts:356 › screencast › should work for popups `@chromium-ubuntu-22.04-arm-node20`
⚠️ [chromium-library] › library/video.spec.ts:762 › screencast › should work with video+trace `@chromium-ubuntu-22.04-arm-node20`
⚠️ [chromium-library] › library/video.spec.ts:762 › screencast › should work with video+trace `@chromium-ubuntu-22.04-node20`
⚠️ [firefox-library] › library/browsercontext-cookies-third-party.spec.ts:257 › third party 'Partitioned;' cookies `@firefox-ubuntu-22.04-node20`
⚠️ [firefox-library] › library/browsercontext-cookies-third-party.spec.ts:470 › top level 'Partitioned;' cookie and same origin iframe `@firefox-ubuntu-22.04-node20`

52492 passed, 1268 skipped

---

Merge workflow run.

[github-actions]

Hi, I'm the Playwright bot and I took a first look at the CI failures.

🟡 The remaining failure is likely unrelated to this PR, but is not a confirmed flake

global-setup.spec.ts:313 timed out on Windows while its nested auth test exited with code 1. This PR does not change global setup, authentication, the test runner, or the Chromium paths exercised by that test, but I could not find the same test failing on another SHA.

Details

Uncertain

• [playwright-test] › global-setup.spec.ts:313 › globalSetup should work for auth — the test timed out after 30 seconds on windows-latest-node22; the nested run returned exit code 1 instead of 0. The PR changes a WebGL screenshot regression test and MCP download-error handling, neither of which reaches this test's global-setup auth flow. However, the same test passed in all 39 other recent matching Windows shard runs checked, including main at 56cbd437cef484769ba023114b0becf393a924e3, so there is no prior failure that would satisfy the flake criterion. A rerun would distinguish a one-off failure from a reproducible PR interaction.

_{Triaged by the Playwright bot - agent run}

[github-actions]

Test results for "MCP"

8937 passed, 1489 skipped

---

Merge workflow run.