|Summary:||[WPE] WebExtension API test /webkit/WebKitWebView/web-process-crashed is flaky failing|
|Product:||WebKit||Reporter:||Lauro Moura <lmoura>|
|Component:||WPE WebKit||Assignee:||Lauro Moura <lmoura>|
|Version:||WebKit Nightly Build|
Description Lauro Moura 2021-08-12 20:30:03 PDT
/WPE/TestWebExtensions /webkit/WebKitWebView/web-process-crashed Output: /webkit/WebKitWebView/web-process-crashed: FAIL WPE-FDO-FATAL-WARNING: Instance::dispatchFrameCallbacks(): Cannot find surface with bridgeId 1 in view backend map. Probably the associated surface is gone due to a premature exit in the client side (/app/webkit/Tools/glib/../../WebKitBuild/Release/bin/TestWebKitAPI/WPE/TestWebExtensions:778): WPE-FDO-WARNING **: 06:40:11.689: Instance::dispatchFrameCallbacks(): Cannot find surface with bridgeId 1 in view backend map. Probably the associated surface is gone due to a premature exit in the client side In the release bot, the first failure was in the build #2602 for identifier 240345@main. Since this revision, it failed 31 out of 41 runs. Locally: Laptop: 4 fails in 20 runs Desktop: 7 fails in 20 runs In a manual build outside Flatpak, no errors in 10 runs.
Comment 1 Carlos Garcia Campos 2021-08-12 23:12:01 PDT
I guess that's expected to happen in case of web process crash, so maybe we can just make warnings non fatal at that point or not use g_warning in wpe-fdo
Comment 3 Lauro Moura 2021-08-13 05:27:08 PDT
(In reply to Carlos Garcia Campos from comment #1) > I guess that's expected to happen in case of web process crash, so maybe we > can just make warnings non fatal at that point or not use g_warning in > wpe-fdo Indeed. With this patch, it's consistently passing.
Comment 4 Lauro Moura 2021-08-13 05:59:20 PDT
Created attachment 435478 [details] Patch with updated expectations
Comment 5 Carlos Garcia Campos 2021-08-13 07:20:48 PDT
Comment on attachment 435478 [details] Patch with updated expectations View in context: https://bugs.webkit.org/attachment.cgi?id=435478&action=review > Tools/TestWebKitAPI/Tests/WebKitGLib/TestWebExtensions.cpp:166 > + Test::removeLogFatalFlag(G_LOG_LEVEL_WARNING); > g_main_loop_run(test->m_mainLoop); > + Test::addLogFatalFlag(G_LOG_LEVEL_WARNING); Ok, I think we should still check if we can remove the g_warning in wpe-fdo, because thi might hide other warnings happening.