HTTP read rule assumes that the parser consumes all of the readable_region.

[sadjad]

https://github.com/keithw/mush/blob/99b598a4366bb10cfe4326703b3af39925a58a12/src/examples/fun.cc#L126

With the new changes, it is possible that the HTTPResponseParser::parse doesn't consume all of the readable_region (for example, in case of a partial header in the buffer). That will lead this rule to not emptying the readable_region, and leads the EventPoll to repeatedly calling this rule.

[keithw]

Hmmm.... yes, you're right. This zero-copy stuff is tricky. What's the right condition here...? Maybe the right condition is something more like "cumulative # of bytes pushed to the RingBuffer is > than what the HTTPMessageSequence has scanned" ?