Better handling of WS client frames, contributed in the comments of #212:
* Reject unmasked frames as per RFC 6455
* Avoid unnecessary data copy from/to evbuffer
* Remove conditions on has_mask
2 new tests cover this change:
* minimal ping-pong with masked client frame, unmasked response
* rejected unmasked client frame
1. Introduce ws_client struct
2. Handle all communications from websocket.c for WS clients
3. Always use a dedicated Redis connection for WS clients
4. Add rbuf & wbuf evbuffers for incoming & outgoing WS data
5. Use event_base_once to control R/W events
6. WS test: make sure to read complete HTTP response
Add new pub-sub test using WebSockets, disabled by default due to
message ordering not matching what is expected.
Enable the test with `PUBSUB=1 ./ws-tests.py`
The overall rate was incorrect when using non-default intervals. Also
simplify the code using plain nanosecond values, and fix the rate for
the last event which was based on the number of sleeps instead of the
actual time elapsed.
1. Generate random WS connection key
2. Read headers returned in WS connection (upgrade) response
3. Validate Sec-WebSocket-Accept return value from webdis
Tested by introducing an artificial error in webdis, reported as
expected.
* Use pure.css for a basic grid
* Detect disconnections, update UI accordingly
* Make GET/SET commands configurable and interactive
* Add button to clear logs
* Test with current branch
1. Switch to evbuffer for correct handling of partial writes
2. Implement WS state machine in each worker
3. Clean up debug logging
4. Add detailed network log messages to help find WS issues
5. Switch to getopt_long
Current test suite pass with python-msgpack v0.2 but fails with python-msgpack
v0.3 and later due to changes in the library. Modern distributions ship
python-msgpack > 0.3 for very long time ago.