Jussi Piitulainen writes in a private email to John Cowan:
(cond-expand ...) entry (page 15): "Otherwise, the cond-expand has no effect." I think it would be better for it to "expand to nothing", except the sample implementation seems to be a syntax error, which surely is an effect.
SRFI 0 clearly requires that an error be signaled if the cond-expand falls off the end. Unfortunately, when I created CondExpandCowan, on which people actually voted, I forgot to specify anything for this case. In Chibi it has no effect. Changed editorially to "the behavior is unspecified".
Also, "spliced into the current context as if by begin" -> "expanded to a begin" and let that begin have the splicing semantics if the context is appropriate.
Changed editorially.