On Tue, 20 Apr 2010 06:27:02 +0100, David Edmondson <dme@dme.org> wrote: > The second chunk was intended to cover a similar case (len == 0), but > becomes unnecessary after the first chunk. At least, that's what I > convinced myself after the conversation with Anthony Towns > (id:h2y87b3a4191004060117v5421db8ejbe3030d0626e7440@mail.gmail.com). Thanks for the clarification. And I really appreciated seeing a reference to the original discussion that led to this patch. You'll notice there that Anthony's proposed commit had the same patch content that you had in your tree, but with a more detailed commit message, ("and always return a newly talloced array"). That was exactly the kind of explanation I was looking for but couldn't find in the commit I first reviewed. It wasn't really a question of whether the code was correct. The problem was that there was a code change that wasn't described in the commit message. I don't want that even if the change is correct. Anyway, thanks AJ and David. I've now pushed my version of these changes up through this point. I'm currently working on the make-emacs-use-JSON patch, (it's got some confusion about "body visible" vs. "message visible" that I want to fix before pushing). -Carl