Issue metadata
Sign in to add a comment
|
KioskUpdateTest.LaunchOfflineEnabledAppNoUpdate flaky on linux_chromium_chromeos_rel_ng |
||||||||||||||||||||||||
Issue descriptionThis test's quite flaky on linux_chromium_chromeos_rel_ng, causing CQ jobs to be retried unacceptably. Examples: https://ci.chromium.org/buildbot/tryserver.chromium.linux/linux_chromium_chromeos_rel_ng/543869 https://ci.chromium.org/buildbot/tryserver.chromium.linux/linux_chromium_chromeos_rel_ng/543839 https://ci.chromium.org/buildbot/tryserver.chromium.linux/linux_chromium_chromeos_rel_ng/543812 It looks like the test's actually skipped but that a PRE_ step is running and asserting sometimes. [1025/1787] KioskUpdateTest.LaunchOfflineEnabledAppNoUpdate (SKIPPED) [ RUN ] KioskUpdateTest.PRE_LaunchOfflineEnabledAppNoUpdate ... [6685:6685:1103/142417.625835:FATAL:extension_function.cc(505)] Check failed: !browser_client || browser_client->IsShuttingDown() || did_respond() || ignore_all_did_respond_for_testing_do_not_use. app.window.create #0 0x000003b6858c base::debug::StackTrace::StackTrace() #1 0x000003b8247c logging::LogMessage::~LogMessage() #2 0x000002af8be4 UIThreadExtensionFunction::~UIThreadExtensionFunction() #3 0x0000017d61de extensions::ChromeosInfoPrivateSetFunction::~ChromeosInfoPrivateSetFunction() #4 0x000002b97f9b _ZN4base8internal9BindStateIM17ExtensionFunctionFvNSt3__110unique_ptrINS2_19ResponseValueObjectENS3_14default_deleteIS5_EEEEEJ13scoped_refptrIN10extensions23AppWindowCreateFunctionEENS0_13PassedWrapperIS8_EEEE7DestroyEPKNS0_13BindStateBaseE #5 0x000002ad999e extensions::AppWindow::~AppWindow() #6 0x000002ad9aee extensions::AppWindow::~AppWindow() #7 0x000002ada733 extensions::AppWindow::OnNativeClose() #8 0x000004a27775 views::Widget::OnNativeWidgetDestroyed() #9 0x000004a32d0b views::NativeWidgetAura::OnWindowDestroyed() #10 0x0000058e0456 aura::Window::~Window() #11 0x0000058e0ace aura::Window::~Window() #12 0x0000006ce997 _ZN4base8internal7InvokerINS0_9BindStateIMN12_GLOBAL__N_116SimpleHttpServer10ConnectionEFvvEJNS_7WeakPtrIS5_EEEEEFvvEE7RunOnceEPNS0_13BindStateBaseE #13 0x000003b68d55 base::debug::TaskAnnotator::RunTask() #14 0x000003c1eee9 base::internal::IncomingTaskQueue::RunTask() #15 0x000003b8a078 base::MessageLoop::RunTask() #16 0x000003b8a3da base::MessageLoop::DeferOrRunPendingTask() #17 0x000003b8a666 base::MessageLoop::DoWork() #18 0x000003b8c6f9 base::MessagePumpLibevent::Run() #19 0x000003b899c9 base::MessageLoop::Run() #20 0x000003bb3478 base::RunLoop::Run() #21 0x0000041d75c3 content::RunMessageLoop() #22 0x0000013144ab chromeos::KioskTest::WaitForAppLaunchWithOptions() #23 0x00000130de8f chromeos::KioskUpdateTest::PreCacheAndLaunchApp() #24 0x00000130dd81 chromeos::KioskUpdateTest_PRE_LaunchOfflineEnabledAppNoUpdate_Test::RunTestOnMainThread() #25 0x00000419ae13 content::BrowserTestBase::ProxyRunTestOnMainThreadLoop() #26 0x000003c98445 ChromeBrowserMainParts::PreMainMessageLoopRunImpl() #27 0x000003c9726d ChromeBrowserMainParts::PreMainMessageLoopRun() #28 0x0000017a10ac chromeos::ChromeBrowserMainPartsChromeos::PreMainMessageLoopRun() #29 0x000002456da1 content::BrowserMainLoop::PreMainMessageLoopRun() #30 0x000002878b55 content::StartupTaskRunner::RunAllTasksNow() #31 0x00000245524f content::BrowserMainLoop::CreateStartupTasks() #32 0x00000245a2d8 content::BrowserMainRunnerImpl::Initialize() #33 0x000002452dd2 content::BrowserMain() #34 0x000003b4a4e1 content::ContentMainRunnerImpl::Run() #35 0x000005a67afb service_manager::Main() ... This needs to be investigated and/or disabled right away.
,
Nov 3 2017
+xiyuan I have to leave early today, so I can't investigate right now. If it's flaky feel free to disable it. A quick look through git history didn't show any obvious kiosk changes, so maybe it's something else.
,
Nov 3 2017
It is probably a racing caused by how the test is written. The test probably should wait for the app to close itself instead of closing it from C++ code, which caused the chrome.app.window.create function DCHECK because it has not yet responded to invoke the callback.
,
Nov 3 2017
Let's disable the test for now since I would not get the fix ready today. I can look at it early next week.
,
Nov 7 2017
Looks like related to browser side navigation that is turned on recently, which makes app.window.create to response after navigation is committed [1]. And the test C++ code closes the window before that happens. https://cs.chromium.org/chromium/src/extensions/browser/api/app_window/app_window_api.cc?rcl=a94d3aebb880652968e32153ced650b21a5757d2&l=434
,
Nov 13 2017
|
|||||||||||||||||||||||||
►
Sign in to add a comment |
|||||||||||||||||||||||||
Comment 1 by kbr@chromium.org
, Nov 3 2017