Since r178414 we don't resize the redirected XComposite window until we enter accelerated compositing mode, but after leaving it the redirected window keeps its size. We should resize to 1x1 again and free the XPixmap and cairo surface to save that memory.
Created attachment 245440 [details] Patch
Thanks for the patch. If this patch contains new public API please make sure it follows the guidelines for new WebKit2 GTK+ API. See http://trac.webkit.org/wiki/WebKitGTK/AddingNewWebKit2API
Comment on attachment 245440 [details] Patch View in context: https://bugs.webkit.org/attachment.cgi?id=245440&action=review Awesome fix. > Source/WebKit2/UIProcess/gtk/RedirectedXCompositeWindow.cpp:219 > + XResizeWindow(m_display, m_window, std::max(1, size.width()), std::max(1, size.height())); Perhaps a comment explaining that 0x0 windows are not allowed by X... > Source/WebKit2/UIProcess/gtk/RedirectedXCompositeWindow.cpp:259 > + std::max(1, m_size.width()), std::max(1, m_size.height()))); Ditto. Orthogonal to this we should probably assert in ::surface() if we aren't in AC.
(In reply to comment #3) > Comment on attachment 245440 [details] > Patch > > View in context: > https://bugs.webkit.org/attachment.cgi?id=245440&action=review > > Awesome fix. Thanks. > > Source/WebKit2/UIProcess/gtk/RedirectedXCompositeWindow.cpp:219 > > + XResizeWindow(m_display, m_window, std::max(1, size.width()), std::max(1, size.height())); > > Perhaps a comment explaining that 0x0 windows are not allowed by X... Sure. > > Source/WebKit2/UIProcess/gtk/RedirectedXCompositeWindow.cpp:259 > > + std::max(1, m_size.width()), std::max(1, m_size.height()))); > > Ditto. I've realized we don't need this, since we are going to assert for empty sizes here. > Orthogonal to this we should probably assert in ::surface() if we aren't in > AC. Sure.
Committed r179397: <http://trac.webkit.org/changeset/179397>