mash: enabling tablet mode or fullscreen causes loss of responsiveness in web contents |
||||||||||
Issue descriptionenabling tablet mode causes the web contents to stop responding to input. Using --ash-debug-shortcuts and ctrl+shift+alt+T, the page will cease rendering (particularly easy to notice if browser window was restored, rather than maximized). From that point, you get in a funky state where cursors and click targets seem to be recognized rendering isn't updated until you do something else that triggers it, like pressing ctrl+shift+alt+R to rotate the screen.
,
Oct 1
also, when attempting to interact with the page (e.g. mouse wheel), I often get crashes like [1:1:1001/140935.810148:FATAL:find_paint_offset_and_visual_rect_needing_update.h(72)] Check failed: context.tree_builder_context_actually_needed_. #0 0x7eff517d096c base::debug::StackTrace::StackTrace() #1 0x7eff517049db logging::LogMessage::~LogMessage() #2 0x7eff4777ea3b blink::PaintInvalidator::InvalidatePaint() #3 0x7eff477cf39a blink::PrePaintTreeWalk::WalkInternal() #4 0x7eff477ce1f8 blink::PrePaintTreeWalk::Walk() #5 0x7eff477ce25b blink::PrePaintTreeWalk::Walk() #6 0x7eff477ce25b blink::PrePaintTreeWalk::Walk() #7 0x7eff477ce25b blink::PrePaintTreeWalk::Walk() #8 0x7eff477ce25b blink::PrePaintTreeWalk::Walk() #9 0x7eff477ce25b blink::PrePaintTreeWalk::Walk() #10 0x7eff477ce25b blink::PrePaintTreeWalk::Walk() #11 0x7eff477ce25b blink::PrePaintTreeWalk::Walk() #12 0x7eff477ce25b blink::PrePaintTreeWalk::Walk() #13 0x7eff477cdc29 blink::PrePaintTreeWalk::Walk() #14 0x7eff477cd243 blink::PrePaintTreeWalk::WalkTree() #15 0x7eff47113a91 blink::LocalFrameView::RunPrePaintLifecyclePhase() #16 0x7eff47113186 blink::LocalFrameView::UpdateLifecyclePhasesInternal() #17 0x7eff47112260 blink::LocalFrameView::UpdateLifecyclePhases() #18 0x7eff471125a4 blink::LocalFrameView::UpdateAllLifecyclePhasesExceptPaint() #19 0x7eff475644c9 blink::LayoutView::HitTest() #20 0x7eff46df77b2 blink::Document::PerformMouseEventHitTest() #21 0x7eff4732735f blink::EventHandlingUtil::PerformMouseEventHitTest() #22 0x7eff47320452 blink::EventHandler::HandleMouseMoveOrLeaveEvent() #23 0x7eff4731feaa blink::EventHandler::HandleMouseMoveEvent() #24 0x7eff476f0c07 blink::PageWidgetEventHandler::HandleMouseMove() #25 0x7eff476f0a74 blink::PageWidgetDelegate::HandleInputEvent() #26 0x7eff4705de19 blink::WebViewImpl::HandleInputEvent() #27 0x7eff471779eb blink::WebViewFrameWidget::HandleInputEvent() #28 0x7eff4ef4f77b content::RenderWidgetInputHandler::HandleInputEvent() #29 0x7eff4f0d7899 content::RenderWidget::HandleInputEvent() #30 0x7eff4ef4d8f0 content::QueuedWebInputEvent::Dispatch() #31 0x7eff4ef4d3ec content::MainThreadEventQueue::DispatchRafAlignedInput() #32 0x7eff4f0d7f1b content::RenderWidget::BeginMainFrame() #33 0x7eff4d211c67 cc::ProxyMain::BeginMainFrame()
,
Oct 1
The same loss of responsiveness occurs when entering fullscreen via js [in console, type document.documentElement.requestFullscreen(); ]
,
Oct 1
,
Oct 3
Seeing this issue in the FilesApp ZipFile tests in chrome-os. The tests were disabled, issue 891116 , due a FindIt report [1]. Various crash stacks in that report contain: [1:1:1001/155952.046875:FATAL:find_paint_offset_and_visual_rect_needing_update.h(72)] Check failed: context.tree_builder_context_actually_needed_. Note: these ZipFile browser tests do not run in tablet mode or fullscreen mode. [1] https://findit-for-me.appspot.com/flake/occurrences?key=ag9zfmZpbmRpdC1mb3ItbWVyQwsSBUZsYWtlIjhjaHJvbWl1bUBicm93c2VyX3Rlc3RzQFppcEZpbGVzL0ZpbGVzQXBwQnJvd3NlclRlc3QuVGVzdAw
,
Oct 3
,
Oct 3
,
Oct 4
Bumping to Pri-1 since this is blocking Pri-1 bugs. This is the top cause of test flakiness right now according to Ranked Flaky Tests: https://findit-for-me.appspot.com/ranked-flakes
,
Oct 4
This bug has morphed into two separate issues. estade originally filed this against mash specific issues (see comment #1), which I am fixing. I don't believe the original issue is causing any test flake at this point. Issue #1 is *not* the same as the stack mentioned in comment #2, 5 and 8. Those test failures are unrelated to the original issue. I believe the ZipFile flake is filed at issue 889703 . I'm going to lower this back to a 2 and will up the priority of issue 889703 .
,
Oct 5
The following revision refers to this bug: https://chromium.googlesource.com/chromium/src.git/+/12ef71460252b042f52da8eddc26adf641e2d50b commit 12ef71460252b042f52da8eddc26adf641e2d50b Author: Scott Violet <sky@chromium.org> Date: Fri Oct 05 14:51:43 2018 aura: update ClientSurfaceEmbedder and server on LocalSurfaceId change If Window::UpdateLocalSurfaceIdFromEmbeddedClient() is called, then the ClientSurfaceEmbedder needs to be updated as well as the server. BUG= 890969 TEST=covered by test Change-Id: Ic7fc94166cc5a3f495b01bdd63e41be16092b0d7 Reviewed-on: https://chromium-review.googlesource.com/c/1257365 Reviewed-by: Fady Samuel <fsamuel@chromium.org> Commit-Queue: Scott Violet <sky@chromium.org> Cr-Commit-Position: refs/heads/master@{#597117} [modify] https://crrev.com/12ef71460252b042f52da8eddc26adf641e2d50b/services/ws/client_root.cc [modify] https://crrev.com/12ef71460252b042f52da8eddc26adf641e2d50b/ui/aura/mus/client_surface_embedder.cc [modify] https://crrev.com/12ef71460252b042f52da8eddc26adf641e2d50b/ui/aura/mus/client_surface_embedder.h [modify] https://crrev.com/12ef71460252b042f52da8eddc26adf641e2d50b/ui/aura/mus/window_port_mus.cc [modify] https://crrev.com/12ef71460252b042f52da8eddc26adf641e2d50b/ui/aura/mus/window_port_mus.h [modify] https://crrev.com/12ef71460252b042f52da8eddc26adf641e2d50b/ui/aura/mus/window_port_mus_unittest.cc [modify] https://crrev.com/12ef71460252b042f52da8eddc26adf641e2d50b/ui/aura/mus/window_tree_client.cc [modify] https://crrev.com/12ef71460252b042f52da8eddc26adf641e2d50b/ui/aura/test/mus/window_port_mus_test_helper.cc [modify] https://crrev.com/12ef71460252b042f52da8eddc26adf641e2d50b/ui/aura/test/mus/window_port_mus_test_helper.h
,
Oct 5
,
Oct 8
,
Oct 8
|
||||||||||
►
Sign in to add a comment |
||||||||||
Comment 1 by est...@chromium.org
, Oct 1