Use WindowProxy in DOMWindow.idl to match the specification more closely.
Created attachment 338430 [details] Patch
Comment on attachment 338430 [details] Patch View in context: https://bugs.webkit.org/attachment.cgi?id=338430&action=review > Source/WebCore/page/DOMWindow.cpp:2305 > +WindowProxy* DOMWindow::open(DOMWindow& activeWindow, DOMWindow& firstWindow, const String& urlString, const AtomicString& frameName, const String& windowFeaturesString) I'm a little worried about this losing it's RefPtr'ness. I think giving WindowProxy ref() and deref() methods, that just ref/deref the underlying AbstractFrame and making this return a RefPtr<WindowProxy> would be safer. What do you think?
(In reply to Sam Weinig from comment #2) > Comment on attachment 338430 [details] > Patch > > View in context: > https://bugs.webkit.org/attachment.cgi?id=338430&action=review > > > Source/WebCore/page/DOMWindow.cpp:2305 > > +WindowProxy* DOMWindow::open(DOMWindow& activeWindow, DOMWindow& firstWindow, const String& urlString, const AtomicString& frameName, const String& windowFeaturesString) > > I'm a little worried about this losing it's RefPtr'ness. I think giving > WindowProxy ref() and deref() methods, that just ref/deref the underlying > AbstractFrame and making this return a RefPtr<WindowProxy> would be safer. > What do you think? Oh, I had similar concerns but did not want to have WindowProxy subclass RefCounted. Your idea sounds good.
Created attachment 338463 [details] Patch
Comment on attachment 338463 [details] Patch Clearing flags on attachment: 338463 Committed r230867: <https://trac.webkit.org/changeset/230867>
All reviewed patches have been landed. Closing bug.
<rdar://problem/39613880>