> I'm afraid simply adding the the undo boundary around newline() > will break yank(), which sets its own boundary and calls newline() > among other changes. If we want this undo stuff, then we probably > should add checks such that none of these functions set boundaries > if they were disabled (by some other function) to start with. What > do you think? > > Hmm. You are right. I considered this problem once before (and could swear I remember solving it). Let me scour my drive for notes. A disable refcount might do the trick, ignoring re-enables until all nested boundary calls are done.
will look at the rest of the diff shortly. thanks! -kj
