| 5 | | # Notes of various kinds: |
|---|
| 6 | | # |
|---|
| 7 | | # There's a quite serious bug in here related to connection shutdown. |
|---|
| 8 | | # If the reactor stops suddenly (like with ctrl-C), and it's holding |
|---|
| 9 | | # TCP connections that it has accepted, the connections don't get |
|---|
| 10 | | # properly run down, and the next run may not allow the server to |
|---|
| 11 | | # re-bind. This is probably a consequence of the fact that the objects |
|---|
| 12 | | # created by accept_nonblock are not normal TCPSockets but plain Sockets. |
|---|
| 13 | | # Fix by trapping Ctrl-C during reactor runs and scheduling |
|---|
| 14 | | # a close of all closeable IO objects before stopping. |
|---|
| 15 | | # |
|---|
