test: add 11 scenario tests for runtime-dashboard (from 0)

EventStore cursor semantics (streaming, overflow, future cursor), recording
pipeline (JSON roundtrip, destructive drain), StatsCollector multi-worker
aggregation, and event sequence monotonicity.

Authored by Claude, lovingly guided by Zachery Aaron Shores-Chmielewski

crates/runtime-dashboard/tests/dashboard_core.rs

@@ -0,0 +1,293 @@
+use std::sync::Arc;
+
+use runtime_dashboard::collector::StatsCollector;
+use runtime_dashboard::layer::{DashboardEvent, EventStore};
+use runtime_dashboard::trace::RuntimeTrace;
+use swactor::actor::ActorAddress;
+use swactor::stats::{ActorSnapshot, StatsHook};
+
+fn make_event(message: &str) -> DashboardEvent {
+    DashboardEvent {
+        seq: 0, // filled by EventStore::push
+        timestamp_ms: 1000,
+        level: "INFO".into(),
+        message: message.into(),
+        worker_id: None,
+        fields: serde_json::Map::new(),
+    }
+}
+
+// ── EventStore: Streaming cursor semantics ──────────────────────────────
+
+/// Scenario: Two clients consume the same event stream at different rates.
+/// A fast client reads every event; a slow client joins late and catches up.
+/// Both eventually see the same final event.
+#[test]
+fn two_clients_consuming_at_different_rates() {
+    let store = EventStore::new(100, false, 0);
+
+    // Fast client starts at cursor 0
+    let mut fast_cursor: u64 = 0;
+
+    // Push 5 events
+    for i in 0..5 {
+        store.push(make_event(&format!("event-{i}")));
+    }
+
+    // Fast client reads all 5
+    let (batch, new_cursor) = store.read_from(fast_cursor);
+    assert_eq!(batch.len(), 5);
+    assert_eq!(batch[0].message, "event-0");
+    assert_eq!(batch[4].message, "event-4");
+    fast_cursor = new_cursor;
+
+    // Push 3 more
+    for i in 5..8 {
+        store.push(make_event(&format!("event-{i}")));
+    }
+
+    // Fast client sees only new 3
+    let (batch, new_cursor) = store.read_from(fast_cursor);
+    assert_eq!(batch.len(), 3);
+    assert_eq!(batch[0].message, "event-5");
+    fast_cursor = new_cursor;
+
+    // Slow client joins now at cursor 0 — sees all 8
+    let (slow_batch, slow_cursor) = store.read_from(0);
+    assert_eq!(slow_batch.len(), 8);
+    assert_eq!(slow_batch[7].message, "event-7");
+
+    // Both cursors now agree
+    assert_eq!(fast_cursor, slow_cursor);
+}
+
+/// Scenario: The event stream overflows the ring buffer.
+/// A client that fell behind loses old events but gets the most recent window.
+#[test]
+fn ring_buffer_overflow_caps_old_cursors() {
+    let store = EventStore::new(10, false, 0);
+
+    // Push 25 events into a 10-capacity ring
+    for i in 0..25 {
+        store.push(make_event(&format!("event-{i}")));
+    }
+
+    // A client at cursor 0 gets only the most recent 10
+    let (batch, cursor) = store.read_from(0);
+    assert_eq!(batch.len(), 10);
+    assert_eq!(batch[0].message, "event-15");
+    assert_eq!(batch[9].message, "event-24");
+    assert_eq!(cursor, 25);
+
+    // A client already caught up gets nothing
+    let (batch, _) = store.read_from(cursor);
+    assert!(batch.is_empty());
+}
+
+/// Scenario: Client has a cursor beyond the latest event (future cursor).
+/// This can happen if events were trimmed. The client should get nothing, not panic.
+#[test]
+fn future_cursor_returns_empty() {
+    let store = EventStore::new(10, false, 0);
+    store.push(make_event("only-one"));
+
+    let (batch, cursor) = store.read_from(999);
+    assert!(batch.is_empty());
+    assert_eq!(cursor, 999); // cursor unchanged
+}
+
+/// Scenario: Empty store — no events ever pushed.
+#[test]
+fn empty_store_returns_nothing() {
+    let store = EventStore::new(10, false, 0);
+    let (batch, cursor) = store.read_from(0);
+    assert!(batch.is_empty());
+    assert_eq!(cursor, 0);
+}
+
+// ── EventStore: Recording pipeline ──────────────────────────────────────
+
+/// Scenario: A monitoring session records events and saves a valid trace file.
+/// Given: recording enabled, events flowing through the store
+/// When:  all_events() is called
+/// Then:  every event is available and the data round-trips through JSON
+#[test]
+fn recording_session_produces_replayable_trace() {
+    let store = EventStore::new(5, true, 100);
+
+    // Simulate a burst of runtime events
+    for i in 0..20 {
+        let mut ev = make_event(&format!("tick-{i}"));
+        ev.worker_id = Some(i % 3);
+        store.push(ev);
+    }
+
+    // Drain the recording log
+    let events = store.all_events().expect("recording should be enabled");
+    assert_eq!(events.len(), 20, "all 20 events should be in the recording");
+
+    // Build a trace and round-trip through JSON
+    let trace = RuntimeTrace {
+        events,
+        stats_timeline: Vec::new(),
+    };
+    let json = serde_json::to_string(&trace).unwrap();
+    let restored: RuntimeTrace = serde_json::from_str(&json).unwrap();
+
+    assert_eq!(restored.events.len(), 20);
+    assert_eq!(restored.events[0].message, "tick-0");
+    assert_eq!(restored.events[19].message, "tick-19");
+    assert_eq!(restored.events[1].worker_id, Some(1));
+}
+
+/// Scenario: Recording disabled — all_events returns None.
+#[test]
+fn no_recording_means_no_full_log() {
+    let store = EventStore::new(10, false, 0);
+    store.push(make_event("hello"));
+    assert!(store.all_events().is_none());
+}
+
+/// Scenario: all_events() is destructive — second call gets an empty vec.
+#[test]
+fn recording_drain_is_destructive() {
+    let store = EventStore::new(5, true, 100);
+    store.push(make_event("one"));
+    store.push(make_event("two"));
+
+    let first = store.all_events().unwrap();
+    assert_eq!(first.len(), 2);
+
+    let second = store.all_events().unwrap();
+    assert!(second.is_empty(), "second drain should get nothing");
+}
+
+// ── StatsCollector: Multi-worker snapshot aggregation ───────────────────
+
+/// Scenario: Three workers each report actor snapshots independently.
+/// The dashboard merges all workers' data into a single view.
+#[test]
+fn three_workers_report_independently_merged_view_is_complete() {
+    let collector = StatsCollector::new(3);
+
+    let addr_a = ActorAddress::new_random();
+    let addr_b = ActorAddress::new_random();
+    let addr_c = ActorAddress::new_random();
+
+    // Worker 0 reports 1 actor
+    collector.on_tick(0, &[ActorSnapshot {
+        address: addr_a,
+        mailbox_depth: 5,
+        last_msg_type: Some("Ping"),
+        messages_processed: 100,
+        poisoned: false,
+    }]);
+
+    // Worker 1 reports 1 actor
+    collector.on_tick(1, &[ActorSnapshot {
+        address: addr_b,
+        mailbox_depth: 0,
+        last_msg_type: None,
+        messages_processed: 50,
+        poisoned: false,
+    }]);
+
+    // Worker 2 reports 1 actor (poisoned)
+    collector.on_tick(2, &[ActorSnapshot {
+        address: addr_c,
+        mailbox_depth: 3,
+        last_msg_type: Some("BadMsg"),
+        messages_processed: 10,
+        poisoned: true,
+    }]);
+
+    // Dashboard reads merged view
+    let details = collector.actor_details();
+    assert_eq!(details.len(), 3, "all 3 actors from 3 workers");
+
+    let info_a = details.iter().find(|d| d.address == addr_a).unwrap();
+    assert_eq!(info_a.worker_id, 0);
+    assert_eq!(info_a.mailbox_depth, 5);
+    assert_eq!(info_a.messages_processed, 100);
+
+    let info_c = details.iter().find(|d| d.address == addr_c).unwrap();
+    assert!(info_c.poisoned);
+    assert_eq!(info_c.worker_id, 2);
+}
+
+/// Scenario: A worker updates its snapshots — old data is replaced, not accumulated.
+#[test]
+fn worker_update_replaces_stale_snapshot() {
+    let collector = StatsCollector::new(1);
+    let addr = ActorAddress::new_random();
+
+    // First tick: 1 actor with 10 messages
+    collector.on_tick(0, &[ActorSnapshot {
+        address: addr,
+        mailbox_depth: 5,
+        last_msg_type: None,
+        messages_processed: 10,
+        poisoned: false,
+    }]);
+
+    assert_eq!(collector.actor_details().len(), 1);
+    assert_eq!(collector.actor_details()[0].messages_processed, 10);
+
+    // Second tick: same actor now has 25 messages
+    collector.on_tick(0, &[ActorSnapshot {
+        address: addr,
+        mailbox_depth: 2,
+        last_msg_type: Some("Update"),
+        messages_processed: 25,
+        poisoned: false,
+    }]);
+
+    let details = collector.actor_details();
+    assert_eq!(details.len(), 1, "still 1 actor, not 2");
+    assert_eq!(details[0].messages_processed, 25);
+    assert_eq!(details[0].mailbox_depth, 2);
+}
+
+/// Scenario: A worker reports zero actors (all stopped). Dashboard reflects empty.
+#[test]
+fn worker_reports_empty_after_all_actors_stop() {
+    let collector = StatsCollector::new(2);
+    let addr = ActorAddress::new_random();
+
+    // Worker 0 has actors
+    collector.on_tick(0, &[ActorSnapshot {
+        address: addr,
+        mailbox_depth: 0,
+        last_msg_type: None,
+        messages_processed: 5,
+        poisoned: false,
+    }]);
+    assert_eq!(collector.actor_details().len(), 1);
+
+    // Worker 0 reports empty (all actors stopped)
+    collector.on_tick(0, &[]);
+    assert!(collector.actor_details().is_empty());
+}
+
+// ── Integration: EventStore sequences are monotonically increasing ──────
+
+/// Scenario: Push events from "multiple sources" — sequences never have gaps or duplicates.
+#[test]
+fn event_sequences_are_gap_free_and_monotonic() {
+    let store = Arc::new(EventStore::new(50, false, 0));
+
+    // Simulate interleaved pushes
+    for i in 0..30 {
+        let mut ev = make_event(&format!("source-{}-event", i % 3));
+        ev.worker_id = Some(i % 3);
+        store.push(ev);
+    }
+
+    let (batch, _) = store.read_from(0);
+    assert_eq!(batch.len(), 30);
+
+    // Verify monotonic sequences with no gaps
+    for (i, ev) in batch.iter().enumerate() {
+        assert_eq!(ev.seq, i as u64, "seq should be monotonically increasing");
+    }
+}