lmxopcua/tests/Server/ZB.MOM.WW.OtOpcUa.Runtime.Tests/Drivers/DriverInstanceActorConnectionLossSignalTests.cs

using Akka.Actor;
using Shouldly;
using Xunit;
using ZB.MOM.WW.OtOpcUa.Core.Abstractions;
using ZB.MOM.WW.OtOpcUa.Runtime.Drivers;
using ZB.MOM.WW.OtOpcUa.Runtime.Tests.Harness;

namespace ZB.MOM.WW.OtOpcUa.Runtime.Tests.Drivers;

/// <summary>
///     Gitea #676 — a driver that KNOWS its connection is gone (<see cref="IDriverConnectionLossSource"/>) gets
///     re-initialised at once instead of after the #527 watchdog's silence window (~90 s in production). The watchdog is
///     configured at an hour here, so anything that reconnects within seconds can only be the new signal.
/// </summary>
public sealed class DriverInstanceActorConnectionLossSignalTests : RuntimeActorTestBase
{
    private static readonly TimeSpan FastPoll = TimeSpan.FromMilliseconds(50);

    /// <summary>A permanently Healthy (fresh-read) stub — the watchdog can never fire for it — that can raise
    /// <see cref="IDriverConnectionLossSource.ConnectionLost"/> on demand and counts its attached handlers.</summary>
    private sealed class LossSourceStubDriver : StubDriver, IDriverConnectionLossSource
    {
        private EventHandler<DriverConnectionLossEventArgs>? _handlers;
        private int _handlerCount;

        private int _healthReads;

        public int HandlerCount => Volatile.Read(ref _handlerCount);

        /// <summary>Health-poll observations — the positive control for an absence assertion.</summary>
        public int HealthReadCount => Volatile.Read(ref _healthReads);

        public override DriverHealth GetHealth()
        {
            Interlocked.Increment(ref _healthReads);
            return base.GetHealth();
        }

        public event EventHandler<DriverConnectionLossEventArgs>? ConnectionLost
        {
            add { _handlers += value; Interlocked.Increment(ref _handlerCount); }
            remove { _handlers -= value; Interlocked.Decrement(ref _handlerCount); }
        }

        /// <summary>Raised from a pool thread, as a real driver's keep-alive callback would.</summary>
        public Task RaiseAsync(string reason) =>
            Task.Run(() => _handlers?.Invoke(this, new DriverConnectionLossEventArgs(reason)));
    }

    private (IActorRef Actor, Akka.TestKit.TestProbe Parent) Connect(LossSourceStubDriver driver)
    {
        var parent = CreateTestProbe();
        var actor = parent.ChildActorOf(DriverInstanceActor.Props(
            driver,
            reconnectInterval: TimeSpan.FromHours(1),
            healthPollInterval: FastPoll,
            commsLossTimeout: TimeSpan.FromHours(1)));
        actor.Tell(new DriverInstanceActor.InitializeRequested("{}"));
        parent.ExpectMsg<DriverInstanceActor.ConnectivityChanged>(TimeSpan.FromSeconds(3)).Connected.ShouldBeTrue();
        return (actor, parent);
    }

    [Fact]
    public async Task A_connection_loss_from_a_connected_driver_reconnects_at_once_not_after_the_watchdog()
    {
        var driver = new LossSourceStubDriver();
        var (_, parent) = Connect(driver);
        var started = DateTime.UtcNow;

        await driver.RaiseAsync("keep-alive reported BadServerHalted");

        parent.ExpectMsg<DriverInstanceActor.ConnectivityChanged>(TimeSpan.FromSeconds(3)).Connected.ShouldBeFalse();
        (DateTime.UtcNow - started).ShouldBeLessThan(TimeSpan.FromSeconds(3),
            "the watchdog is configured at an hour, so only the connection-loss signal can have done this");
    }

    /// <summary>
    ///     The falsification control: the same driver and actor with the event suppressed stays Connected — so the
    ///     Reconnecting above is the signal's doing, not something the setup would have done anyway.
    /// </summary>
    [Fact]
    public void Without_a_raise_the_same_driver_stays_connected()
    {
        var driver = new LossSourceStubDriver();
        var (_, parent) = Connect(driver);

        var ticksAtConnect = driver.HealthReadCount;
        parent.ExpectNoMsg(TimeSpan.FromMilliseconds(600));
        (driver.HealthReadCount - ticksAtConnect).ShouldBeGreaterThanOrEqualTo(5,
            "health-poll ticks did not run — the absence assertion above would be vacuous");
        driver.HandlerCount.ShouldBe(1, "attached, so the absence above is not merely an unwired event");
        driver.InitializeCount.ShouldBe(1, "and nothing re-initialised the driver");
    }

    /// <summary>
    ///     A raise while NOT Connected is ignored, not deferred: here it lands while the initial connect is still in
    ///     flight. Acting on it later would tear down the connection that very attempt is building.
    /// </summary>
    [Fact]
    public async Task A_raise_while_connecting_is_ignored()
    {
        var driver = new LossSourceStubDriver();
        var release = new TaskCompletionSource();
        var initStarted = new TaskCompletionSource();
        driver.InitBehavior = async _ =>
        {
            initStarted.TrySetResult();
            await release.Task;
        };
        var parent = CreateTestProbe();
        var actor = parent.ChildActorOf(DriverInstanceActor.Props(
            driver,
            reconnectInterval: TimeSpan.FromHours(1),
            healthPollInterval: FastPoll,
            commsLossTimeout: TimeSpan.FromHours(1)));
        actor.Tell(new DriverInstanceActor.InitializeRequested("{}"));
        await initStarted.Task.WaitAsync(TimeSpan.FromSeconds(3));

        // The raise is enqueued before the connect can complete (RaiseAsync returns after the Tell), so the mailbox
        // delivers it while the actor is still Connecting.
        await driver.RaiseAsync("a loss reported mid-connect");
        release.SetResult();

        parent.ExpectMsg<DriverInstanceActor.ConnectivityChanged>(TimeSpan.FromSeconds(3)).Connected.ShouldBeTrue();
        parent.ExpectNoMsg(TimeSpan.FromMilliseconds(600));
        driver.InitializeCount.ShouldBe(1);
    }

    [Fact]
    public void The_handler_is_attached_once_and_detached_when_the_actor_stops()
    {
        var driver = new LossSourceStubDriver();
        var (actor, parent) = Connect(driver);
        driver.HandlerCount.ShouldBe(1);

        parent.Watch(actor);
        Sys.Stop(actor);
        // PostStop announces comms loss (03/S15) before the Terminated arrives.
        parent.FishForMessage(m => m is Terminated t && t.ActorRef.Equals(actor), TimeSpan.FromSeconds(3));

        driver.HandlerCount.ShouldBe(0, "the IDriver outlives the actor on a respawn — a leaked handler holds a dead Self");
    }

    /// <summary>A driver that does not implement the interface is simply never subscribed — the signal is opt-in.</summary>
    [Fact]
    public void A_driver_without_the_capability_connects_as_before()
    {
        var driver = new StubDriver();
        var parent = CreateTestProbe();
        var actor = parent.ChildActorOf(DriverInstanceActor.Props(
            driver, reconnectInterval: TimeSpan.FromHours(1), healthPollInterval: FastPoll,
            commsLossTimeout: TimeSpan.FromHours(1)));
        actor.Tell(new DriverInstanceActor.InitializeRequested("{}"));
        parent.ExpectMsg<DriverInstanceActor.ConnectivityChanged>(TimeSpan.FromSeconds(3)).Connected.ShouldBeTrue();
        parent.ExpectNoMsg(TimeSpan.FromMilliseconds(300));
    }
}