This was the cause of bug 743325.

The main issue was that xpc_qsUnwrapArgImpl didn't unwrap security wrappers before checking mozilla::dom::binding::instanceIsProxy.

Comment on attachment 631104 [details] [diff] [review] v1 Would it make sense to test for IsDOMClass before InstanceIsProxy in both getWrapper and castNative? Which do we think will be more common? r=me in either case.

Comment on attachment 631104 [details] [diff] [review] v1 Review of attachment 631104 [details] [diff] [review]: ----------------------------------------------------------------- ::: js/xpconnect/src/XPCQuickStubs.cpp @@ +697,5 @@ > *tearoff = nsnull; > > + js::Class* clasp = js::GetObjectClass(obj); > + if (mozilla::dom::binding::instanceIsProxy(obj) || > + mozilla::dom::IsDOMClass(clasp)) { You can drop the 'mozilla::' here @@ +754,5 @@ > + if (mozilla::dom::binding::instanceIsProxy(cur)) { > + native = static_cast<nsISupports*>(js::GetProxyPrivate(cur).toPrivate()); > + entries = nsnull; > + } else if (mozilla::dom::IsDOMClass(clasp)) { > + native = mozilla::dom::UnwrapDOMObject<nsISupports>(cur); And here @@ +761,4 @@ > native = static_cast<nsISupports*>(xpc_GetJSPrivate(cur)); > entries = GetOffsetsFromSlimWrapper(cur); > } else { > + NS_NOTREACHED("what kind of wrapper is this?"); MOZ_NOT_REACHED?