QuicListener AcceptConnectionAsync and OperationCanceledException

[pepone]

The QuicListener can throw OperationCanceledException when the internal TLS handshake timeout expires, this is unexpected and not documented behavior.

See

runtime/src/libraries/System.Net.Quic/src/System/Net/Quic/QuicListener.PendingConnection.cs

Line 58 in 1366574

_cancellationTokenSource.CancelAfter(s_handshakeTimeout);

It would be better to wrap the OperationCanceled exception from the internal timeout in a QuicException with the ConnectionTimeout error code

You should consider fixing this along #75184

Tagging subscribers to this area: @dotnet/ncl
See info in area-owners.md if you want to be subscribed.

Issue Details

---

The QuicListener can throw OperationCanceledException when the internal TLS handshake timeout expires, this is unexpected and not documented behavior.

See

runtime/src/libraries/System.Net.Quic/src/System/Net/Quic/QuicListener.PendingConnection.cs

Line 58 in 1366574

_cancellationTokenSource.CancelAfter(s_handshakeTimeout);

It would be better to wrap the OperationCanceled exception from the internal timeout in a QuicException with the ConnectionTimeout error code

You should consider fixing this along #75184

Author:	pepone
Assignees:	-
Labels:	area-System.Net.Quic
Milestone:	-

[ManickaP]

Triage: we should fix this, seems easy enough and should be addressed in 8.0.

[rzikm]

@ManickaP I see in the code that this scenario should throw QuicException with QuicError.ConnectionTimeout. I think we can close this issue now.

[ManickaP]

Depends where we decide to go with the exceptions. I left this open on purpose so we can discuss this.

In some cases, it looks we're deciding to go the way of different exception types, e.g. throwing SocketException for IP related errors. But here we're doing QuicException for a timeout. So I think that the exception type here is still something we should include in the whole "exception discussion".

[karelz]

@wfurt this should also be addressed by your exception changes, shouldn't it?

[wfurt]

no, we did not touch this one AFAIK. We should investigate separately and write some tests.