Serious problem where our clients are able to crash our server

[akjt]

We got a websocket server . We run the server using the run_async method and it all works fine.

Occasionally, the client closes the connection (and is likely not doing it in a friendly way) , our server crashes with following message:

[Error ]: Worker Crash: An uncaught exception occured: remote endpoint: Transport endpoint is not connected

I can see from the code that Crow does handle such exception by just logging them and then continue , so why is it crashing al of a sudden.

I have discovered a few things. Firstly, it seems the problem can be reproduced by having the client disabling the wifi and enabling it again. This causes an uncleanly disconnection. What I am not sure about is why Crow's onclose function only is triggered when the client wifi is enabled - why cant crow detect the client is offline?

But that is not the real problem, the problem is the call conn.get_remote_ip()... That seems to behave very strangely on such scenarios. It throws and then segfaults. Not sure why your catch statements are not catching this.

[Maximsiv1410]

I have encountered the same behavior with onerror websocket handler which was called right after onclose.
Probably crow::websocket::connection object is not valid at that point...