REGRESSION (r164133): Selection doesn't paint when scrolling some pages
Created attachment 230140 [details] Patch
Comment on attachment 230140 [details] Patch View in context: https://bugs.webkit.org/attachment.cgi?id=230140&action=review r=me > Source/WebCore/editing/FrameSelection.cpp:393 > +static void clearRenderViewSelection(Node& node) Shouldn't this still be const?
Comment on attachment 230140 [details] Patch View in context: https://bugs.webkit.org/attachment.cgi?id=230140&action=review >> Source/WebCore/editing/FrameSelection.cpp:393 >> +static void clearRenderViewSelection(Node& node) > > Shouldn't this still be const? It’d do no harm to mark it const, but not much good either. This is a function with side effects, so it’s not like we are just asking the Node a question. It’s not easy to answer the question of what you should and should not do with a const Node. But a side effect of a style update, which can do things like trigger loads, does not seem to be a “const” kind of thing to me.
Committed r167845: <http://trac.webkit.org/changeset/167845>
This caused an assertion failure in svg/custom/bug79798.html: http://webkit-test-results.appspot.com/dashboards/flakiness_dashboard.html#showAllRuns=true&tests=svg%2Fcustom%2Fbug79798.html Seems like it is causing us to leave layout with dirty layout.
(In reply to comment #5) > This caused an assertion failure in svg/custom/bug79798.html: http://webkit-test-results.appspot.com/dashboards/flakiness_dashboard.html#showAllRuns=true&tests=svg%2Fcustom%2Fbug79798.html > Seems like it is causing us to leave layout with dirty layout. I filed https://bugs.webkit.org/show_bug.cgi?id=132297.
Comment on attachment 230140 [details] Patch View in context: https://bugs.webkit.org/attachment.cgi?id=230140&action=review > Source/WebCore/editing/FrameSelection.cpp:473 > + renderView->setNeedsLayout(); See bug 178651