As a convenience, nsTreeBodyFrame currently implements nsITreeBoxObject. But really there's no need for this... the only caller is nsTreeBoxObject, and it can just as easily call methods directly on nsTreeBodyFrame. This gets rid of some virtual overhead as well as a vtable pointer, and makes bug 396185 easier/possible. There's a fair bit of additional deCOM (removing nsresults, direct return values) that could happen, but I don't know whether it's worth it and I was trying to keep the size of this patch to a minimum. I can file a followup if you think it's worth pursuing.

Comment on attachment 344478 [details] [diff] [review] Make nsTreeBodyFrame not implement nsITreeBoxObject, rev. 1 Found a few things that need fixup first.

This fixes up two more cases where we were trying to QI the wrong type to nsITreeBoxObject.

Comment on attachment 344672 [details] [diff] [review] Make nsTreeBodyFrame not implement nsITreeBoxObject, rev. 1.1 +#ifndef nsTreeBodyFrame_h__ +#define nsTreeBodyFrame_h__ + Technically this shouldn't have two trailing underscores, since those identifiers are reserved for the platform/compiler.