Glen Choo <chooglen@xxxxxxxxxx> writes: > + /* > + * FIXME for whatever reason, do_event passes the _previous_ event, so > + * in order for our callback to receive the error event, we have to call > + * do_event twice > + */ > + do_event(cs, CONFIG_EVENT_ERROR, &event_data); > + do_event(cs, CONFIG_EVENT_ERROR, &event_data); > + return -1; > } I think this is because do_event() uses the current position in the config source as the end_offset and emits the previous event if there is one (that is, if do_event() has been called before). It can't emit the current event because it still does not know what the current event's end_offset is. I also noticed some more difficult-to-understand things, like _EOF and _ERROR are never emitted as of current master (i.e. without this patch set) because they are always the last events to be passed to do_event(). A refactoring to have separate "do" and "flush" would make things clearer.
