browser_tests 's ResourceCoordinatorRenderProcessProbeBrowserTest.TrackAndMeasureActiveRenderProcesses fails on mac clang tot bots |
||||
Issue descriptionStarted here: https://ci.chromium.org/buildbot/chromium.clang/ToTMacASan/1089 Which contains https://chromium-review.googlesource.com/1060127 [ RUN ] ResourceCoordinatorRenderProcessProbeBrowserTest.TrackAndMeasureActiveRenderProcesses [22655:10499:0516/173010.214682:WARNING:notification_platform_bridge_mac.mm(510)] AlertNotificationService: XPC connection invalidated. 2018-05-16 17:30:10.674 browser_tests[22655:381264] *** Owner supplied to -[NSTrackingArea initWithRect:options:owner:userInfo:] referenced a deallocating object. Tracking area behavior is undefined. Break on NSTrackingAreaDeallocatingOwnerError to debug. [22655:67391:0516/173014.004214:WARNING:embedded_test_server.cc(229)] Request not handled. Returning 404: /favicon.ico ../../chrome/browser/resource_coordinator/resource_coordinator_render_process_probe_browsertest.cc:174: Failure Value of: probe.AllMeasurementsAreAtCurrentCycle() Actual: false Expected: true Stack trace: 0 browser_tests 0x0000000104603012 testing::internal::UnitTestImpl::CurrentOsStackTraceExceptTop(int) + 242 1 browser_tests 0x0000000104601bf7 testing::internal::AssertHelper::operator=(testing::Message const&) const + 359 2 browser_tests 0x0000000101c36a6b resource_coordinator::ResourceCoordinatorRenderProcessProbeBrowserTest_TrackAndMeasureActiveRenderProcesses_Test::RunTestOnMainThread() + 12427 3 browser_tests 0x000000010bc13543 content::BrowserTestBase::ProxyRunTestOnMainThreadLoop() + 1475 4 browser_tests 0x000000010acaf410 ChromeBrowserMainParts::PreMainMessageLoopRunImpl() + 9888 5 browser_tests 0x000000010acac91e ChromeBrowserMainParts::PreMainMessageLoopRun() + 270 6 browser_tests 0x0000000106a41044 content::BrowserMainLoop::PreMainMessageLoopRun() + 356 7 browser_tests 0x000000010771ed94 content::StartupTaskRunner::RunAllTasksNow() + 132 8 browser_tests 0x0000000106a3cbd6 content::BrowserMainLoop::CreateStartupTasks() + 1798 9 browser_tests 0x0000000106a48070 content::BrowserMainRunnerImpl::Initialize(content::MainFunctionParams const&) + 448 10 browser_tests 0x0000000106a35e9e content::BrowserMain(content::MainFunctionParams const&) + 574 11 browser_tests 0x000000010a75730e content::ContentMainRunnerImpl::Run() + 1006 12 browser_tests 0x00000001102b5184 service_manager::Main(service_manager::MainParams const&) + 4708 13 browser_tests 0x000000010a755431 content::ContentMain(content::ContentMainParams const&) + 353 14 browser_tests 0x000000010bc1222a content::BrowserTestBase::SetUp() + 5578 15 browser_tests 0x000000010ab3a212 InProcessBrowserTest::SetUp() + 1410 Seems to not happen on the main waterfall asan builder: https://ci.chromium.org/buildbot/chromium.memory/Mac%20ASan%2064%20Tests%20%281%29/40737 siggi, does the stack ring a bell? Could this be a code bug? Alternatively it's possible that something changed for the worse in the trunk asan runtime.
,
May 18 2018
Fails on non-asan mac tot too: https://ci.chromium.org/buildbot/chromium.clang/ToTMac/1636
,
May 18 2018
On Windows too: https://ci.chromium.org/buildbot/chromium.clang/ToTWin64/1486 ...but passed in the next build.
,
May 18 2018
So, we think the test is sensitive to ASan's slowdown, i.e. it is flaky?
,
May 22 2018
Huh, that's weird, must be a race - looking. The failure in #3 is different, this happens because CPU time doesn't necessarily strictly increase from measurement to measurement.
,
May 22 2018
This may happen if I ever get a zero cumulative CPU time measurement for a process. I'm going to use a negative TimeDelta for a sentinel instead of zero, as that don't wash.
,
May 23 2018
The following revision refers to this bug: https://chromium.googlesource.com/chromium/src.git/+/ce246a382d9d150fea4a649023da10d3ac0491c3 commit ce246a382d9d150fea4a649023da10d3ac0491c3 Author: Sigurdur Asgeirsson <siggi@chromium.org> Date: Wed May 23 13:05:41 2018 Render Process Probe: Fix a unittest flake. As-is a CPU usage measurement of zero will be conflated with the sentinel value for a new entry, so use a TimeDelta(-1) instead. Bug: 844388 , 755840 Change-Id: Ia8478dd7e7735aab4335426c3010fb94ee58c832 Reviewed-on: https://chromium-review.googlesource.com/1067863 Reviewed-by: François Doray <fdoray@chromium.org> Commit-Queue: Sigurður Ásgeirsson <siggi@chromium.org> Cr-Commit-Position: refs/heads/master@{#561051} [modify] https://crrev.com/ce246a382d9d150fea4a649023da10d3ac0491c3/chrome/browser/resource_coordinator/resource_coordinator_render_process_probe.cc [modify] https://crrev.com/ce246a382d9d150fea4a649023da10d3ac0491c3/chrome/browser/resource_coordinator/resource_coordinator_render_process_probe.h [modify] https://crrev.com/ce246a382d9d150fea4a649023da10d3ac0491c3/chrome/browser/resource_coordinator/resource_coordinator_render_process_probe_browsertest.cc
,
May 23 2018
That should fix it, I think. |
||||
►
Sign in to add a comment |
||||
Comment 1 by thakis@chromium.org
, May 18 2018