A invalid region, part of a dependency cycle, should not attempt to repaint the content from the associated named flow.
Created attachment 219900 [details] Patch
Comment on attachment 219900 [details] Patch View in context: https://bugs.webkit.org/attachment.cgi?id=219900&action=review This looks sane to me. Feel free to have a domain expert look over this patch. > LayoutTests/fast/regions/repaint/invalid-region-repaint-crash.html:4 > + <script src="../../repaint/resources/text-based-repaint.js" type="text/javascript"></script> Please remove the HTML attribute type as it's unnecessary given that this document is an HTML5 document. > Source/WebCore/rendering/RenderLayer.cpp:6278 > + RenderNamedFlowFragment* region = toRenderBlockFlow(&renderer())->renderNamedFlowFragment(); Notice that we generate a toRenderBlockFlow()-variant that takes and returns a reference. For you consideration, I suggest we use this variant to remove the address-of operator and to document the "non-nullness" of the return value of toRenderBlockFlow() since renderer() is guaranteed to be non-null. RenderNamedFlowFragment* region = toRenderBlockFlow(renderer()).renderNamedFlowFragment();
Created attachment 219958 [details] Patch for landing
Comment on attachment 219958 [details] Patch for landing Clearing flags on attachment: 219958 Committed r161054: <http://trac.webkit.org/changeset/161054>
All reviewed patches have been landed. Closing bug.