[Webkit-unassigned] [Bug 160450] New: [GTK][Threaded Compositor] Several flaky tests due to differences in scrollbars

[bugzilla-daemon at webkit.org]

https://bugs.webkit.org/show_bug.cgi?id=160450

            Bug ID: 160450
           Summary: [GTK][Threaded Compositor] Several flaky tests due to
                    differences in scrollbars
    Classification: Unclassified
           Product: WebKit
           Version: WebKit Local Build
          Hardware: Unspecified
                OS: Unspecified
            Status: NEW
          Keywords: Gtk, LayoutTestFailure
          Severity: Normal
          Priority: P2
         Component: WebKit Gtk
          Assignee: webkit-unassigned at lists.webkit.org
          Reporter: cgarcia at igalia.com
                CC: bugs-noreply at webkitgtk.org, yoon at igalia.com,
                    zan at falconsigh.net

We still have a lot of flaky tests since we switched to the threaded compositor. I can't reproduce them though. I've seen several reftests failing because there's a part of the vertical scrollbar that is not rendered. It turns out that the part not rendered is the second tile of the scrollbar layer, that for some reason it seems to be rendered later. So, my guess is that we force a repaint that when finishes hasn't actually painted all the tiles. Then the UI process takes the screenshot too early, before the second tile of the scrollbar layer is actually painted. I took advantage of the DRI3 bug, to see how things happen when everything is slow, and that's exactly what happens, everything is rendered in one pass and then the second tile of the scrollbar layer. See:

http://people.igalia.com/cgarcia/webkitgtk-scrollbar-layer.mp4

I'm not sure how to fix it, because I don't know why that happens and I can't really reproduce the flakiness, but we could try to wait for more damage events in the UI process before taking the screenshot.

-- 
You are receiving this mail because:
You are the assignee for the bug.
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://lists.webkit.org/pipermail/webkit-unassigned/attachments/20160802/0e58a198/attachment.html>