chore: update deps

[Stebalien]

Mostly boxo + libp2p.

[guillaumemichel]

go-libp2p-kad-dht/query_test.go

Lines 42 to 65 in 25d299e

	// close both hosts so query fails
	require.NoError(t, d1.host.Close())
	require.NoError(t, d2.host.Close())
	// peers will still be in the RT because we have decoupled membership from connectivity
	require.NoError(t, tu.WaitFor(ctx, func() error {
	if !checkRoutingTable(d1, d2) {
	return fmt.Errorf("should have routes")
	}
	return nil
	}))
	// failed queries should remove the peers from the RT
	_, err := d1.GetClosestPeers(ctx, "test")
	require.NoError(t, err)
	_, err = d2.GetClosestPeers(ctx, "test")
	require.NoError(t, err)
	require.NoError(t, tu.WaitFor(ctx, func() error {
	if checkRoutingTable(d1, d2) {
	return fmt.Errorf("should not have routes")
	}
	return nil
	}))

It seems that with go-libp2p v0.33 and previous, host.Connect() after host is closed returns an error (failed to dial: swarm closed). But after v0.34, host.Connect() doesn't return an error and actually connect to the node, even after the host.Close() is called. This is the reason why TestRTEvictionOnFailedQuery is failing.

I haven't been able to find where this change of behavior comes from in libp2p.

cc: @sukunrt @MarcoPolo

[aschmahmann]

Some bisecting shows that the issue with TestRTEvictionOnFailedQuery was introduced in libp2p/go-libp2p#2118.

The culprit seems to be libp2p/go-libp2p@17ba661

[guillaumemichel]

I have updated TestRTEvictionOnFailedQuery. The goal of closing the hosts was that the next query fails, but clearing the addresses of the other node in the peerstore has the same effect and also makes the query fail.

[sukunrt]

@guillaumemichel You can also close the swarm if you want. But the behavior change in go-libp2p is intentional.

[Stebalien]

Let's continue the discussion in the go-libp2p issue.

[sukunrt]

I've removed 4667e81 and updated go-libp2p.

[guillaumemichel]

Connectedness issues will be addressed in #976 if necessary