NEW 222840
REGRESSION (r272842): security/contentSecurityPolicy/video-with-data-url-allowed-by-media-src-star.html is consistently failing
https://bugs.webkit.org/show_bug.cgi?id=222840
Summary REGRESSION (r272842): security/contentSecurityPolicy/video-with-data-url-allo...
Ryan Haddad
Reported 2021-03-05 16:29:42 PST
After Media In GPUProcess was enabled for tests in r272842, security/contentSecurityPolicy/video-with-data-url-allowed-by-media-src-star.html is consistently failing on iOS https://results.webkit.org/?suite=layout-tests&test=security%2FcontentSecurityPolicy%2Fvideo-with-data-url-allowed-by-media-src-star.html https://build.webkit.org/results/Apple-iOS-14-Simulator-Release-WK2-Tests/r273999%20(371)/security/contentSecurityPolicy/video-with-data-url-allowed-by-media-src-star-diffs.html The test says "This tests that loading a video with a data URL is allowed when the page has Content Security Policy "media-src *". This test PASSED if you see a square with a green-to-gray horizontal gradient. Otherwise, it FAILED." The "expected" image doesn't show the green-to-gray gradient, but the "actual" image does.
Attachments
actual image (69.87 KB, image/png)
2021-03-05 16:29 PST, Ryan Haddad
no flags
expected image (68.63 KB, image/png)
2021-03-05 16:30 PST, Ryan Haddad
no flags
Test gardening (1.45 KB, patch)
2021-03-22 17:12 PDT, Peng Liu
no flags
Ryan Haddad
Comment 1 2021-03-05 16:29:55 PST
Created attachment 422451 [details] actual image
Ryan Haddad
Comment 2 2021-03-05 16:30:07 PST
Created attachment 422452 [details] expected image
Radar WebKit Bug Importer
Comment 3 2021-03-05 16:30:30 PST
Peng Liu
Comment 4 2021-03-22 11:07:01 PDT
This test has been flaky on both macOS and iOS-device for quite a while (bug 155196). WKTR loads video-with-data-url-allowed-by-media-src-star.html and video-with-data-url-allowed-by-media-src-star-expected.html, and compares the rendering results. This test expects the rendering results be the same, but looks like WebKit loads video-with-data-url-allowed-by-media-src-star.html faster, so the rendering results are not the same for a while and WKTR reports the failure. However, the rendering results will become the same eventually.
Peng Liu
Comment 5 2021-03-22 12:15:51 PDT
Based on Jer's suggestion, we may need to delay changing the readyState of a video element to “HAVE_FUTURE_DATA” until we can guarantee that we won’t return an empty image if asked to paint.
Peng Liu
Comment 6 2021-03-22 17:12:59 PDT
Created attachment 423966 [details] Test gardening
Peng Liu
Comment 7 2021-03-22 21:15:45 PDT
Comment on attachment 423966 [details] Test gardening Clearing flags on attachment: 423966 Committed r274840 (235630@main): <https://commits.webkit.org/235630@main>
Daniel Bates
Comment 8 2021-03-22 21:17:51 PDT
Good that these test no longer fail. Consider removing the tests and the setting altogether. These tests were for a setting added in bug #157005. That setting was for bin compat and guarded by linked-on-or-after. It's been 5 years since then and I don't think we need to support such an old iOS.
Peng Liu
Comment 9 2021-03-22 21:23:13 PDT
Thanks for sharing the important information! It would be great if we can remove the settings and tests.
Peng Liu
Comment 10 2021-04-12 20:47:32 PDT
https://ews-build.s3-us-west-2.amazonaws.com/macOS-AppleSilicon-Big-Sur-Debug-WK2-Tests-EWS/r425796-1952/results.html Some tests in security/contentSecurityPolicy are flaky crash on AS-debug bots.
Note You need to log in before you can comment on or make changes to this bug.