fix(websocket): lock conflict between emit and poll

[SSebo]

No description provided.

[codecov]

Codecov Report

Merging #213 (f3dc134) into main (31996cf) will decrease coverage by 0.38%.
The diff coverage is 87.80%.

@@            Coverage Diff             @@
##             main     #213      +/-   ##
==========================================
- Coverage   86.77%   86.38%   -0.39%     
==========================================
  Files          30       30              
  Lines        2351     2358       +7     
==========================================
- Hits         2040     2037       -3     
- Misses        311      321      +10     

Impacted Files	Coverage Δ
...asynchronous/async_transports/websocket_general.rs	76.36% <78.57%> (-11.45%)	⬇️
engineio/src/transports/websocket.rs	92.85% <90.90%> (-2.15%)	⬇️
engineio/src/transports/websocket_secure.rs	93.10% <90.90%> (-2.06%)	⬇️
...eio/src/asynchronous/async_transports/websocket.rs	89.18% <100.00%> (+0.30%)	⬆️
.../asynchronous/async_transports/websocket_secure.rs	88.00% <100.00%> (+0.32%)	⬆️
socketio/src/socket.rs	76.08% <100.00%> (-0.76%)	⬇️

Help us with your feedback. Take ten seconds to tell us how you rate us. Have a feature suggestion? Share it here.

[1c3t3a]

Hi @SSebo, first of all thanks for the contribution and for finding this bug! That is great :) For your PR I am not so sure about the implementation. The async transport never actually locks at the same time, so the only problem is the sync transport. A minimal inversiv and nice change would be to find a way for removing the lock from the sync transport. I assume this should be possible as all methods from the Transport interface only take &self instead of .&mut self. Your solution would work, but basically gives access to another clone of the sender which is harder to read and more difficult to reason about.

[SSebo]

@1c3t3a I agree! I found the &mut self is because of the StreamExt api, so manual write a poll_next function can get ride of the &mut self.

[SSebo]

@1c3t3a the code is more reasonable now, PTAL

[1c3t3a]

LGTM, thanks a lot!