Problem with ESS

Not sure if it's really a bug, that's why I didn't use M-x report-xemacs-bug. Something between 21.5-b6 and 21.5-b12 broke the ESS (Emacs Speaks Statistics) package for me. It works in a similar way to ilisp -- you can edit code and send functions to be evaluated in an interactive process buffer. When I eval a function with 21.5-b6 I get this in the process buffer: + x; + } with the point being at the end of the output, after the final ">", which is the correct behavior. The same thing in 21.5-b12 leaves the point at the _beginning_ of the function and leaves it to me to skip (M->) to the end and press RET to actually evaluate it. That's on Linux, BTW. Any hints as to what could have caused that? Or perhaps the ESS package needs to be fixed? Or do I have to binary-search? --J.