DomSerializerBrowserTests are flaky |
|
Issue descriptionThe tests DomSerializerTests.SerializeHTMLDOMWithNoMetaCharsetInOriginalDoc and DomSerializerTests.SerializeHTMLDOMWithDocType have been flaking on several bots, including Linux ChromiumOS Ozone Tests, Cast Linux and Win x64 GN. See example failures: https://build.chromium.org/p/chromium.win/builders/Win%20x64%20GN/builds/11615 https://build.chromium.org/p/chromium.chromiumos/builders/Linux%20ChromiumOS%20Ozone%20Tests%20%281%29/builds/27422 https://build.chromium.org/p/chromium.chromiumos/builders/Linux%20ChromiumOS%20Ozone%20Tests%20%281%29/builds/27403 https://build.chromium.org/p/chromium.linux/builders/Cast%20Linux/builds/20717 Sample log (Linux ChromeOS): [ RUN ] DomSerializerTests.SerializeHTMLDOMWithNoMetaCharsetInOriginalDoc [27027:27027:0309/093419:5826679027:ERROR:proxy_service.cc(1512)] ProxyConfigService for ChromeOS should be created in profile_io_data.cc::CreateProxyConfigService and this should be used only for examples. [27027:27051:0309/093419:5826700654:WARNING:histograms.cc(40)] Started multiple compositor clients (Browser, Renderer) in one process. Some metrics will be disabled. [644/644] DomSerializerTests.SerializeHTMLDOMWithNoMetaCharsetInOriginalDoc (224 ms) ... 1 test failed: DomSerializerTests.SerializeHTMLDOMWithNoMetaCharsetInOriginalDoc (../../content/renderer/dom_serializer_browsertest.cc:737) Additional test environment: CHROME_DEVEL_SANDBOX=/opt/chromium/chrome_sandbox LANG=en_US.UTF-8 Sample log (Win x64): [ RUN ] DomSerializerTests.SerializeHTMLDOMWithNoMetaCharsetInOriginalDoc [4664:3712:0309/195435:2249082:WARNING:histograms.cc(40)] Started multiple compositor clients (Browser, Renderer) in one process. Some metrics will be disabled. Backtrace: v8::internal::GlobalHandles::Node::DecreaseBlockUses [0x00000001410C309F+31] blink::WindowProxy::~WindowProxy [0x00000001414B2B16+46] blink::HeapObjectHeader::finalize [0x00000001413FE7FA+62] blink::NormalPage::sweep [0x00000001413FF666+154] blink::BaseArena::sweepUnsweptPage [0x00000001413FF768+60] blink::BaseArena::completeSweep [0x00000001413FE63E+50] blink::ThreadState::completeSweep [0x0000000141401969+213] blink::ThreadState::detachMainThread [0x0000000141401C16+30] blink::shutdownWithoutV8 [0x000000014144AA09+101] content::RenderThreadImpl::Shutdown [0x000000014248663E+1038] content::ChildProcess::~ChildProcess [0x00000001425822E8+60] content::RenderProcessImpl::`scalar deleting destructor' [0x000000014247E788+20] content::InProcessRendererThread::CleanUp [0x0000000142439119+37] base::Thread::ThreadMain [0x0000000142863286+406] base::`anonymous namespace'::ThreadFunc [0x000000014282C6D7+343] BaseThreadInitThunk [0x00000000774A59ED+13] RtlUserThreadStart [0x00000000775DC541+33] [1247/1249] DomSerializerTests.SerializeHTMLDOMWithNoMetaCharsetInOriginalDoc (225 ms) I will follow-up with a patch to disable these tests shortly. CCing some recent editors of dom_serializer_browsertest.cc.
,
Mar 10 2016
The following revision refers to this bug: https://chromium.googlesource.com/chromium/src.git/+/97b442d48fca108bb345f80df4c1db21f6948577 commit 97b442d48fca108bb345f80df4c1db21f6948577 Author: tsergeant <tsergeant@chromium.org> Date: Thu Mar 10 05:46:04 2016 Disable flaky DomSerializerBrowserTests on all platforms Disables the following two tests: DomSerializerTests.SerializeHTMLDOMWithNoMetaCharsetInOriginalDoc DomSerializerTests.SerializeHTMLDOMWithDocType These tests have been flaky across Windows and Linux builders BUG=593575 TBR=esprehn@chromium.org Review URL: https://codereview.chromium.org/1780143002 Cr-Commit-Position: refs/heads/master@{#380348} [modify] https://crrev.com/97b442d48fca108bb345f80df4c1db21f6948577/content/renderer/dom_serializer_browsertest.cc
,
Mar 10 2016
Thanks for disabling the tests. FWIW, the callstack in the topmost comment doesn't seem related to the things the test is trying to exercise. Would it be a cop-out if I tried to blame the flakiness on --single-process (and ignored this bug because of this)? BTW, I also compared with the list of tests I reenabled in https://crrev.com/1575173002 and these are different tests from the ones that are flaky right now...
,
Mar 11 2016
If there's nothing that can be changed in the test to improve it, and there's a separate bug filed for improving --single-process, that seems like an okay thing to do. |
|
►
Sign in to add a comment |
|
Comment 1 by tsergeant@chromium.org
, Mar 10 2016