Something in the range of r240117-240121 has caused layout test imported/w3c/web-platform-tests/service-workers/service-worker/websocket-in-service-worker.https.html to time out for WPE/GTK. And the only likely suspect there is r240117 "Stop using NetworkStorageSession::storageSession in WebCore". Alex, any suspicions here? --- /home/slave/webkitgtk/gtk-linux-64-release-tests/build/layout-test-results/imported/w3c/web-platform-tests/service-workers/service-worker/websocket-in-service-worker.https-expected.txt +++ /home/slave/webkitgtk/gtk-linux-64-release-tests/build/layout-test-results/imported/w3c/web-platform-tests/service-workers/service-worker/websocket-in-service-worker.https-actual.txt @@ -1,3 +1,5 @@ -FAIL Verify WebSockets can be created in a Service Worker assert_equals: expected "PASS" but got "FAIL: Got an error event" +Harness Error (TIMEOUT), message = null +TIMEOUT Verify WebSockets can be created in a Service Worker Test timed out +
If I were looking into this I would first try the revision before and after that to verify that was the change that broke it. Then I'd look at the soup-specific code in SocketStreamHandleImpl::create and see if nullptr is being passed. My guess is before there was some kind of fallback to "the default" storage and now it needs to be explicit. I'd add a new fallback in the soup code if that's what you want, or even better: find a way to pass the proper session to the socket stream.
SocketStreamHandleImplSoup is going to die soon (hopefully, see bug #200162). I'll check this again after the switch to the libsoup API.
Your bug may have been fixed by https://bugs.webkit.org/show_bug.cgi?id=199906
Passing in last runs.