after_this_websocket is not called if the client disconnects?

[willstott101]

I'm trying to use after_this_websocket to cleanup some long-running tasks that go alongside websocket connections. It appears this isn't called reliably enough for me to use it as a cleanup. My server has a memory leak right now and we think it's due to this.

Environment:

• Python version: Tox 310 311
• Quart version: 0.19.5 (current git tip)

I have reproduced this behaviour in the Quart tests: https://github.com/willstott101/quart/pull/1 I hope that's enough info to quickly confirm if this is a bug or accepted behaviour. The docs recommend we always re-raise the cancelled error in websockets which makes it impossible for us to use after_this_websocket for cleanup.

[pgjones]

I think this is expected and teardown_websocket is what you are looking for

[willstott101]

There doesn't seem to be a teardown_request. Would it be more intuitive for both after_this_request and after_this_websocket to be usable for cleanup. I don't understand the benefit of teardown_websocket and after_this_websocket being different things.

I can experiment with teardown_websocket - but would a PR which combines teardown_websocket and after_this_websocket be considered?