virtual/.../window-focus-self.html in webkit_layout_tests failing on chromium.webkit/WebKit Win10 |
||||||||||||
Issue descriptionFiled by sheriff-o-matic@appspot.gserviceaccount.com on behalf of kpaulhamus@chromium.org virtual/.../window-focus-self.html in webkit_layout_tests failing on chromium.webkit/WebKit Win10 Builders failed on: - WebKit Win10: https://ci.chromium.org/buildbot/chromium.webkit/WebKit%20Win10 Increasing flakiness (9 out of the last 24 runs). Failing with "FAIL: Window was focused"
,
Jul 12
,
Jul 12
The following revision refers to this bug: https://chromium.googlesource.com/chromium/src.git/+/a7be9cc2e39f872aeb8e03afb7d0aabf92ffde03 commit a7be9cc2e39f872aeb8e03afb7d0aabf92ffde03 Author: Kim Paulhamus <kpaulhamus@chromium.org> Date: Thu Jul 12 16:49:02 2018 Mark window-focus-self as flaky on Win10 Tbr: tkent@chromium.org Bug: 863067 Change-Id: I9f043eec8363dbb26ccaa8d901321e0d579f1a4f Reviewed-on: https://chromium-review.googlesource.com/1135433 Commit-Queue: Kim Paulhamus <kpaulhamus@chromium.org> Reviewed-by: Kim Paulhamus <kpaulhamus@chromium.org> Cr-Commit-Position: refs/heads/master@{#574600} [modify] https://crrev.com/a7be9cc2e39f872aeb8e03afb7d0aabf92ffde03/third_party/WebKit/LayoutTests/TestExpectations
,
Jul 12
,
Jul 12
Non-virtual version is very stable. This is specific to user-activation-v2.
,
Jul 12
,
Jul 13
The following revision refers to this bug: https://chromium.googlesource.com/chromium/src.git/+/5427394bb739114e0a03c80594fcf4271fa461af commit 5427394bb739114e0a03c80594fcf4271fa461af Author: Kim Paulhamus <kpaulhamus@chromium.org> Date: Fri Jul 13 00:31:23 2018 Disabling the virtual window-focus-self test, not the non-virtual Tbr: tkent@chromium.org Bug: 863067 Change-Id: I913685da055480d9cddea105808a16fd7e70facd Reviewed-on: https://chromium-review.googlesource.com/1135982 Reviewed-by: Kim Paulhamus <kpaulhamus@chromium.org> Commit-Queue: Kim Paulhamus <kpaulhamus@chromium.org> Cr-Commit-Position: refs/heads/master@{#574790} [modify] https://crrev.com/5427394bb739114e0a03c80594fcf4271fa461af/third_party/WebKit/LayoutTests/TestExpectations
,
Jul 13
,
Jul 13
kpaulhamus@: Wondering how to access the failure logs. The link in #c1 doesn't seem to provide details unless I am missing something. For records: the test has been flaky on Mac even w/o UAv2 enabled: Issue 816766 .
,
Jul 13
Nm, found a log it: https://test-results.appspot.com/data/layout_results/WebKit_Win10/37746/layout-test-results/results.html Looks like the windows gets "blur" and then "focus" events, while it expected only the first one. We get the exact same outcome in mac ( Issue 816766 ): https://test-results.appspot.com/data/layout_results/mac_chromium_rel_ng/659500/layout-test-results/results.html
,
Jul 13
Here is a relevant MDN note on Chrome vs Firefox: https://developer.mozilla.org/en-US/docs/Web/API/Window/open#Note_on_use_of_window_open
,
Jul 13
Thanks for looking! I'll direct this towards the same component. Feel free to unassign yourself.
,
Jul 16
,
Jul 17
kochi@, I remember you have recently investigated a similar issue. Could you triage?
,
Jul 17
,
Jul 17
This seems to be a different problem than issue 859605, as the other issue is about Element.focus(), while window.focus() runs completely different code path. window.focus() involves renderer<->browser IPC, so checking focus/blur event on window against window.focus() can't be completely synchronous as opposed to Element.focus(). The test in question relies on the event is dispatched (or not dispatched) immediately using `setTimeout(, 0)`, which seems to me the source of flakiness. > window.open('javascript:window.focus()', '_self', ''); > setTimeout(function() { testRunner.notifyDone(); }, 0);
,
Jul 18
The following revision refers to this bug: https://chromium.googlesource.com/chromium/src.git/+/cea59c5edf0cf821af87ee665a9448d19a07f9df commit cea59c5edf0cf821af87ee665a9448d19a07f9df Author: Christian Dullweber <dullweber@chromium.org> Date: Wed Jul 18 14:03:45 2018 Disable window-focus-self.html on Windows fast/dom/Window/window-focus-self.html is failing on Windows. https://ci.chromium.org/buildbot/chromium.webkit/WebKit%20Win10/38018 https://ci.chromium.org/buildbot/chromium.webkit/WebKit%20Win10/38017 Tbr: kochi@chromium.org Bug: 816766 , 863067 Change-Id: Ie54d5722b6c9c76d548d43a190ab827f051cdeb9 Reviewed-on: https://chromium-review.googlesource.com/1141869 Commit-Queue: Christian Dullweber <dullweber@chromium.org> Reviewed-by: Christian Dullweber <dullweber@chromium.org> Cr-Commit-Position: refs/heads/master@{#576029} [modify] https://crrev.com/cea59c5edf0cf821af87ee665a9448d19a07f9df/third_party/WebKit/LayoutTests/TestExpectations
,
Jul 19
My analysis in comment 16 looks wrong. The original intention of this test when it was introduced: https://chromiumcodereview.appspot.com/14735005 checks opening a popup window against "_self" explicitly. Probably the code was lost at some point - will investigate.
,
Jul 19
,
Jul 23
During initialization, the code to focus main frame (for delivering input events etc.) sends IPC to focus, but the actual handling of this IPC could be delayed after the test function is run, and occasionally "focus" event handler is unexpectedly called for it, not the focus event originating from the script. https://cs.chromium.org/chromium/src/content/browser/renderer_host/render_widget_host_impl.cc?l=1031&rcl=27d9265aae1ced02ca4e722a6585319010a0f5e9 is the place where the IPC is sent. I'll make a change to the test to delay the execution of test function after initial rendering.
,
Jul 23
The following revision refers to this bug: https://chromium.googlesource.com/chromium/src.git/+/44bc78241ecd613537c76e0df0c3a495ab01be01 commit 44bc78241ecd613537c76e0df0c3a495ab01be01 Author: Takayoshi Kochi <kochi@chromium.org> Date: Mon Jul 23 08:44:31 2018 Delay the test execution to avoid flakiness The frame's initialization code focuses the content from browser process via an IPC (browser->renderer), but its handling in the renderer can be delayed until after the test code script execution is started. This caused the test to have a race condition, that a focus event is delivered where it is not expected, and resulted the flakiness. The test can fail on the initial use of content_shell process. Delay the execution of test function so that it will not be affected by the frame initial focus event. Bug: 863067 Change-Id: I475d912b0e9baffa3917212b2fc0735189381cb8 Reviewed-on: https://chromium-review.googlesource.com/1146523 Commit-Queue: Takayoshi Kochi <kochi@chromium.org> Reviewed-by: Kent Tamura <tkent@chromium.org> Cr-Commit-Position: refs/heads/master@{#577124} [modify] https://crrev.com/44bc78241ecd613537c76e0df0c3a495ab01be01/third_party/WebKit/LayoutTests/TestExpectations [modify] https://crrev.com/44bc78241ecd613537c76e0df0c3a495ab01be01/third_party/WebKit/LayoutTests/fast/dom/Window/window-focus-self.html
,
Jul 23
,
Jul 24
It seems there's no flakiness/timeouts so far. https://test-results.appspot.com/dashboards/flakiness_dashboard.html#testType=webkit_layout_tests&tests=Window%2Fwindow-focus-self.html |
||||||||||||
►
Sign in to add a comment |
||||||||||||
Comment 1 by kpaulhamus@chromium.org
, Jul 12