This test doesn't clean up cookies, so they show up in subsequent tests, and in particular break third-party-cookie-relaxing.html, because it depends on whether there are already cookies set for the particular first party. Bots hide this misbehavior, because they re-run failing tests, this time without set-cookie-on-redirect.html. Sergio, can you please look into this? I think that the patch will need to be rolled out if this cannot be fixed soon.
(In reply to comment #0) > This test doesn't clean up cookies, so they show up in subsequent tests, and in particular break third-party-cookie-relaxing.html, because it depends on whether there are already cookies set for the particular first party. > > Bots hide this misbehavior, because they re-run failing tests, this time without set-cookie-on-redirect.html. > > Sergio, can you please look into this? I think that the patch will need to be rolled out if this cannot be fixed soon. Sure. It's true that the test I added does not clean up cookies, but in any case, shouldn't third-party-cookie-relaxing.html call clearAllCookies() at the beginning as the other tests do?
Possibly. Cookie testing machinery is extremely fragile, and I'm not sure if that would help.
Created attachment 190237 [details] Patch
Comment on attachment 190237 [details] Patch Clearing flags on attachment: 190237 Committed r144045: <http://trac.webkit.org/changeset/144045>
All reviewed patches have been landed. Closing bug.
Confirmed that the patch helped on Mac. Thank you!
More reports in https://bugs.webkit.org/show_bug.cgi?id=111650