mozilla-central/view
Name Description Size
crashtests
moz.build 599
nsView.cpp 37482
nsView.h nsView's serve two main purposes: 1) a bridge between nsIFrame's and nsIWidget's, 2) linking the frame tree of a(n) (in-process) subdocument with its parent document frame tree. Historically views were used for more things, but their role has been reduced, and could be reduced to nothing in the future (bug 337801 tracks removing views). Views are generally associated with a frame. A view that does not have a frame is called an anonymous view. Some frames also have associated widgets (think os level windows). If a frame has a widget it must also have a view, but not all frames with views will have widgets. Only four types of frames can have a view: root frames (ViewportFrame), subdocument frames (nsSubDocumentFrame), menu popup frames (nsMenuPopupFrame), and list control frames (nsListControlFrame). Root frames and subdocument frames have views to link the two documents together (the frame trees do not link up otherwise). Menu popup frames, and list control frames have views because they (sometimes) need to create widgets. Menu popup frames handles xul popups, which is anything where we need content to go over top the main window at an os level. List control frames handle select popups/dropdowns in non-e10s mode. The term "root view" refers to the root view of a document. Just like root frames, root views can have parent views. Only the root view of the root document in the process will not have a parent. All views are created by their frames except root views. Root views are special. Root views are created in nsDocumentViewer::MakeWindow before the root frame is created, so the root view will not have a frame very early in document creation. Subdocument frames have an anonymous (no frame associated with it) inner view that is a child of their "outer" view. On a subdocument frame the inner view serves as the parent of the root view of the subdocument. The outer and inner view of the subdocument frame belong to the subdocument frame and hence to the parent document. The root view of the subdocument belongs to the subdocument. nsLayoutUtils::GetCrossDocParentFrame and nsPresContext::GetParentPresContext depend on this view structure and are the main way that we traverse across the document boundary in layout. When the load of a new document is started in the subdocument, the creation of the new subdocument and destruction of the old subdocument are not linked. (This creation and destruction is handled in nsDocumentViewer.cpp.) This means that the old and new document will both exist at the same time during the loading of the new document. During this period the inner view of the subdocument parent will be the parent of two root views. This means that during this period there is a choice for which subdocument we draw, nsSubDocumentFrame::GetSubdocumentPresShellForPainting is what makes that choice. Note that this might not be a totally free choice, ie there might be hidden dependencies and bugs if the way we choose is changed. One thing that is special about the root view of a chrome window is that instead of creating a widget for the view, they can "attach" to the existing widget that was created by appshell code or something else. (see nsDocumentViewer::ShouldAttachToTopLevel) 21322
nsViewManager.cpp XXX TODO XXX DeCOMify newly private methods Optimize view storage 31224
nsViewManager.h Initialize the ViewManager Note: this instance does not hold a reference to the presshell because it holds a reference to this instance. @result The result of the initialization, NS_OK if no errors 15949