Comment on attachment 8402820 [details] [diff] [review] v1 Review of attachment 8402820 [details] [diff] [review]: ----------------------------------------------------------------- ::: js/xpconnect/src/nsXPConnect.cpp @@ +406,5 @@ > + // create ends up there. > + JSAutoCompartment ac(aJSContext, aGlobal); > + if (!(aFlags & nsIXPConnect::OMIT_COMPONENTS_OBJECT)) { > + // XPCCallContext gives us an active request needed to save/restore. > + if (!GetCompartmentPrivate(aGlobal)->scope->AttachComponentsObject(aJSContext) || Preexisting, I guess, but do we know that the private already exists, or should this use EnsureCompartmentPrivate? @@ +414,5 @@ > + } > + > + // Stuff coming through this path always ends up as a DOM global. > + // XXX Someone who knows why we can assert this should re-check > + // (after bug 720580). Remove this comment while you're here, bholley confirmed that the assertion is correct back in the day, but I never got around to removing it again.

Comment on attachment 8402820 [details] [diff] [review] v1 Review of attachment 8402820 [details] [diff] [review]: ----------------------------------------------------------------- r=bholley ::: js/xpconnect/src/nsXPConnect.cpp @@ +406,5 @@ > + // create ends up there. > + JSAutoCompartment ac(aJSContext, aGlobal); > + if (!(aFlags & nsIXPConnect::OMIT_COMPONENTS_OBJECT)) { > + // XPCCallContext gives us an active request needed to save/restore. > + if (!GetCompartmentPrivate(aGlobal)->scope->AttachComponentsObject(aJSContext) || We do know. XPCWrappedNativeScope creates it in the constructor.