I'm seeing an interesting hang scenario on rio, and I can't remember if I posted it to this list.
The scenario is simple. I drawterm in to a qemu or lguest plan 9 guest. make a window. cpu -h somehost. It hangs for whatever reason. tcpdump shows it hanging on 17010 -- cpu sending to 17010, remote host not responding. Do any rio operation. It doesn't matter what -- middle, right are best. Just one op. make a window, whatever. It's the last op you will do with rio for that session of drawterm. I can drawterm in again and get another session. It's fine. I can kill the hanging cpu from that session. At that point, the hung rio unhangs. Interesting, eh? ron
