The most "appropriate" way to match libp2p::TransportError::Other?

[gnattu]

I'm trying to make error handling logic for SwarmEvent::OutgoingConnectionError. But the libp2p::TransportError::Other is extremely hard to match because it comes from all over the place and I don't know how to match them exactly.

For example, the timeout Error is printed like Other(Custom { kind: Other, error: Timeout }), and connection refused is printed like Other(Custom { kind: Other, error: Other(Left(Left(Os { code: 61, kind: ConnectionRefused, message: "Connection refused" }))) })

I currently use format! to turn them into string and check if the string contains anything like Timeout or ConnectionRefused, but is there any practical way that I can do this better? I know that we can use get_ref() but I don't know which type should I down_cast into after that.

[thomaseizinger]

Walking the source chain is likely the best option. Most errors will be io::Errors. If you are looking for something specific, then it depends on what transports you are using, e.g. if you are using QUIC, then the error may also be a libp2p::quic::Error.

But before you go down that rabbit hole, can you explain why you need to specific error that occurred? Maybe your problem can be solved in a different way.

[gnattu]

I want to check if the outgoing connection error is due to our(the dialer) side or is it really the remote peer is not reachable. If it's the former I will schedule a re-dial a few seconds later, if it's the latter I will just remove that peer from the routing table. Only knowing it is an IO Error is not enough because we may have a case where we already have an address up but that address is not able to connect to the Internet yet, if that is the case, I will get an NetworkUnreachable Error so I know that the networking is not ready on our side. But if it is ConnectionRefused or Timeout, then it is probably the remote side.

[thomaseizinger]

In that case, I think your best bet is to walk the source chain of the error, try to downcast each one to an IO error and checking its ErrorKind.

In general, this seems like a very fragile way of going about it though. Also, note that we will automatically remove every address that generated a dial error from the routing table already:

rust-libp2p/protocols/kad/src/behaviour.rs

Lines 2001 to 2005 in 1139f1c

	if let DialError::Transport(addresses) = error {
	for (addr, _) in addresses {
	self.address_failed(peer_id, addr)
	}
	}