run-webkit-tests editing/selection/fake-drag.html editing/selection/paragraph-granularity.htmlresults is the above assertion failure. Backtrace: 0 com.apple.WebKit 0x0105ab14 -[WebHTMLView(WebPrivate) _topHTMLView] + 320 (WebHTMLView.m:946) 1 com.apple.WebKit 0x0105ab70 -[WebHTMLView(WebPrivate) _isTopHTMLView] + 60 (WebHTMLView.m:952) 2 com.apple.WebKit 0x01074324 -[WebHTMLView(WebDocumentInternalProtocols) draggingCancelledWithDraggingInfo:] + 64 (WebHTMLView.m:5952) 3 com.apple.WebKit 0x0109cf74 -[WebView _delegateDragOperationForDraggingInfo:] + 216 (WebView.mm:2338) 4 com.apple.WebKit 0x0109d2f8 -[WebView draggingUpdated:] + 68 (WebView.mm:2379) 5 DumpRenderTree 0x00005d60 -[EventSendingController mouseMoveToX:Y:] + 1160 (EventSendingController.m:221) The problem is that EventSendingController calls WebView's draggingUpdated, passing nil. Down in _delegateDragOperationForDraggingInfo:, windowPoint is assigned garbage and newDraggingView gets assigned randomly (with either 0 or the WebView). As fras as I can tell, the regression is just the result of "different garbage" (I observed that windowPoint is different). I think DumpRenderTree is at fault, as EventSendingController shouldn't call draggingUpdated if draggingInfo is 0 (nor draggedImage:movedTo:). Maybe WebView's behavior in this case can improve, by asserting and returning immediately from draggingUpdated if passed nil.
The first sentence should say:run-webkit-tests editing/selection/fake-drag.html editing/selection/paragraph-granularity.html results is the above assertion failure.
Created attachment 11230 [details] patch Fix based on mitz's comments.
Comment on attachment 11230 [details] patch r=me
Landed by ggaren in r17347.