fix(abcip,focas): collapse alarm projection to a single poll loop (no reconnect leak)

The owning DriverInstanceActor re-subscribes alarms on every Connected entry (DetachAlarmSource nulls its cached handle on Connected->Reconnecting without calling UnsubscribeAlarmsAsync), and the driver object + its alarm projection are reused across every in-place reconnect. Each SubscribeAsync started a fresh, never-cancelled Task.Run poll loop and added it to _subs, so N reconnects leaked N concurrent loops all polling the device and all firing the same raise/clear transitions => duplicate alarm events + CPU/mem growth. Mirrors the Galaxy #399 fix (Clear-before-Add) but for live poll loops the collapse must also CANCEL the superseded loops, not just drop references. SubscribeAsync now snapshots existing subs under _subsLock, clears _subs, adds the new sub, starts its loop, then retires each stale sub out-of-band (RetireAsync: Cancel + await loop + Dispose CTS, fire-and-forget so the new subscription's return isn't blocked on a poll interval). Snapshot+clear under the same lock DisposeAsync uses guarantees no double-own / double-dispose. There is exactly one consumer per driver instance (factory-per-actor), so retiring all prior subscriptions before starting the new one is faithful. Regression tests (TDD, fail->pass): subscribe twice then drive one device raise; assert OnAlarmEvent fires exactly once (was twice with two leaked loops).
2026-06-15 06:09:38 -04:00
parent 43b3769a1d
commit 6ba59f9d4d
4 changed files with 190 additions and 2 deletions
@@ -69,6 +69,64 @@ public sealed class FocasAlarmProjectionTests
        events[0].SourceNodeId.ShouldBe(Host);
    }

+    /// <summary>
+    ///     Regression for the reconnect poll-loop leak (#399 sibling): the owning
+    ///     DriverInstanceActor re-subscribes alarms on every Connected entry without first
+    ///     calling Unsubscribe, and the driver object (and its projection) survives the in-place
+    ///     reconnect. Each SubscribeAlarmsAsync used to start a fresh, never-cancelled poll loop —
+    ///     so after N reconnects there were N live loops all polling the device and all firing the
+    ///     same raise/clear transition, producing DUPLICATE alarm events.
+    ///
+    ///     This test simulates two re-subscribes (one reconnect) against the same source, then
+    ///     drives ONE device raise. After the collapse-to-single-loop fix exactly one loop is alive
+    ///     so the raise must fire exactly once. Before the fix both loops fire it ⇒ two events.
+    /// </summary>
+    [Fact]
+    public async Task Resubscribe_collapses_to_single_loop_no_duplicate_raise()
+    {
+        var (drv, factory) = NewDriver(alarmsEnabled: true);
+        factory.Customise = () => new FakeFocasClient();
+        await drv.InitializeAsync("{}", CancellationToken.None);
+
+        var raises = new List<AlarmEventArgs>();
+        drv.OnAlarmEvent += (_, e) =>
+        {
+            // Count only the raise (the initial 0->1) — clears wrap the message in "(cleared)".
+            if (!e.Message.Contains("cleared")) lock (raises) raises.Add(e);
+        };
+
+        // First subscribe creates + connects the per-device client and starts polling. Wait for
+        // it + a couple empty-list ticks so the loop's LastByDevice baseline is seeded empty.
+        var sub1 = await drv.SubscribeAlarmsAsync([], CancellationToken.None);
+        await WaitFor(() => factory.Clients.Count > 0, TimeSpan.FromSeconds(3));
+        await Task.Delay(120);
+
+        // Second subscribe = the actor re-subscribing across a reconnect. Its loop reuses the same
+        // connected DeviceState.Client (EnsureConnectedAsync), so it polls the same alarm list.
+        var sub2 = await drv.SubscribeAlarmsAsync([], CancellationToken.None);
+        await Task.Delay(120); // let sub2's loop seed its own empty LastByDevice baseline
+
+        // One device raise transition. Add to every created client so that whichever instance the
+        // live loop(s) poll, the raise is observed (EnsureConnectedAsync may have churned clients).
+        foreach (var c in factory.Clients)
+            c.Alarms.Add(new FocasActiveAlarm(500, FocasAlarmType.Overtravel, 1, "Axis 1 overtravel"));
+
+        // Wait until at least one raise is observed, then give any leaked second loop ample time
+        // (well past several 30ms poll intervals) to ALSO fire its duplicate before we assert.
+        await WaitFor(() => { lock (raises) return raises.Count > 0; }, TimeSpan.FromSeconds(3));
+        await Task.Delay(200);
+
+        await drv.UnsubscribeAlarmsAsync(sub1, CancellationToken.None);
+        await drv.UnsubscribeAlarmsAsync(sub2, CancellationToken.None);
+        await drv.ShutdownAsync(CancellationToken.None);
+
+        lock (raises)
+        {
+            raises.Count.ShouldBe(1,
+                "exactly one poll loop must survive a re-subscribe; a leaked loop fires a duplicate raise");
+        }
+    }
+
    /// <summary>Verifies that tick diffs raises and clears without polling loop.</summary>
    [Fact]
    public async Task Tick_diffs_raises_and_clears_without_polling_loop()