Using Epiphany 3.25.2 (with WebKit2 Gtk) and loading https://google.com results in a failure of the "notifications bubble" in the top right of the page. I can see the list of notifications available, but clicking the "popover" fails with an error in the console like: https://notifications.google.com/accounts/SetOSID?... Failed to load resource: Too many redirections If there is a better component to use for filing these sort of bugs, I'd be happy to set that on future bug reports.
WebKitGTK+ was the best component to pick since it ensures we see it. This is a longstanding issue. Good to know it's the cause of the Google notifications problem too. *** This bug has been marked as a duplicate of bug 109576 ***
Actually this doesn't seem like a duplicate. It's caused by blocking third-party cookies. If you muck with your cookie settings in Epiphany's preferences dialog, it should start working again. I noticed it when testing another redirection loop that I thought was an Epiphany bug, but which I could reproduce in Firefox. Firefox displayed a warning that sometimes loops can occur due to blocking cookies. So that's a bit surprising. Not sure what to do here. I guess it'll just have to remain broken. We want to be tightening cookie policy going forward, not loosening it.
You are right, changing cookie policy fix the issue.
As it works in Chrome and Firefox, I guess WebKit2.CookieAcceptPolicy.NO_THIRD_PARTY should check base domain name instead of netloc. If I visit plus.google.com, it should accept any google.com cookie request. Does it?
(In reply to Cédric Bellegarde from comment #4) > As it works in Chrome and Firefox, I guess > WebKit2.CookieAcceptPolicy.NO_THIRD_PARTY should check base domain name > instead of netloc. What's netloc? Anyway no, the cookie is only allowed if the domain matches exactly, see soup_cookie_jar_add_cookie_with_first_party() which calls soup_cookie_domain_matches() which calls soup_host_matches_host(). > If I visit plus.google.com, it should accept any google.com cookie request. > Does it? Should it really? Is that how Safari works? We should match whatever Safari does.
My girlfrien is using Safari on her Mac and notifications are working, but not sure third party cookies are blocked, will give a look this evening. Here, the issue is that for example, while opening notifications popover from plus.google.com, it does a request for a cookie from notifications.google.com and so WebKit blocks it.
Moved to GNOME Bugzilla: https://bugzilla.gnome.org/show_bug.cgi?id=792130
Unlikely coincidence that this happened today... but after working on bug #181231, I now have a layout test that I want to mark against this bug: http/tests/resourceLoadStatistics/grandfathering.html Reopening since we need an open WebKit bug for the test expectations.
(In reply to Michael Catanzaro from comment #8) > Unlikely coincidence that this happened today... but after working on bug > #181231, I now have a layout test that I want to mark against this bug: > > http/tests/resourceLoadStatistics/grandfathering.html > > Reopening since we need an open WebKit bug for the test expectations. That's wrong, the problem in that test is unrelated. I'll reassign the expectation to some other bug.
(In reply to Michael Catanzaro from comment #7) > Moved to GNOME Bugzilla: https://bugzilla.gnome.org/show_bug.cgi?id=792130 Should be fixed in libsoup. Please reopen if you have trouble with updated libsoup.