App v0.2.26 and gateway 0.1.20

Client: seek controls, Bazarr subtitle download and cast panel in the
player; MDBList ratings strip; episode and schedule detail pages; series
pace estimate; what's new panel; install-permission onboarding step;
synced per-profile preferences; Emby outage banner.

Gateway: rebuilt admin console (one fragment per page), preference
history and restore, merged Continue Watching, Emby health probe,
subtitle selection and Bazarr download, structured request logging with
per-request identity, and embedded build version.

Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>

server/internal/api/emby_health_test.go

@@ -0,0 +1,95 @@
+package api
+
+import (
+	"testing"
+	"time"
+)
+
+func TestEmbyHealthWaitsForTheThresholdBeforeReportingAnOutage(t *testing.T) {
+	var health embyHealth
+	start := time.Now().UTC()
+	health.begin(60*time.Second, start)
+
+	// One failed probe is a hiccup — a restart, a slow scan. A red bar across somebody's
+	// film for that is worse than saying nothing.
+	health.record(false, start.Add(time.Minute))
+	if !health.get().reachable {
+		t.Fatal("a single failure raised an outage")
+	}
+
+	health.record(false, start.Add(2*time.Minute))
+	state := health.get()
+	if state.reachable {
+		t.Fatal("two consecutive failures did not raise an outage")
+	}
+	if !state.since.Equal(start.Add(2 * time.Minute)) {
+		t.Errorf("since = %v, want the probe that crossed the threshold", state.since)
+	}
+}
+
+// The bar clears itself. One good answer is enough — the failure mode of clearing early
+// is a bar that comes back, which is far better than one that outlives the outage.
+func TestEmbyHealthRecoversOnTheFirstSuccess(t *testing.T) {
+	var health embyHealth
+	start := time.Now().UTC()
+	health.begin(60*time.Second, start)
+	health.record(false, start.Add(time.Minute))
+	health.record(false, start.Add(2*time.Minute))
+
+	health.record(true, start.Add(3*time.Minute))
+	state := health.get()
+	if !state.reachable {
+		t.Fatal("a successful probe did not clear the outage")
+	}
+	if !state.since.Equal(start.Add(3 * time.Minute)) {
+		t.Errorf("since = %v, want the recovery time", state.since)
+	}
+}
+
+// A failure that does not reach the threshold must not accumulate across a recovery, or a
+// server with one flaky probe an hour eventually reports an outage that never happened.
+func TestEmbyHealthForgetsIsolatedFailures(t *testing.T) {
+	var health embyHealth
+	start := time.Now().UTC()
+	health.begin(60*time.Second, start)
+	for minute := 1; minute <= 10; minute++ {
+		health.record(minute%2 == 0, start.Add(time.Duration(minute)*time.Minute))
+	}
+	if !health.get().reachable {
+		t.Fatal("alternating failures were reported as an outage")
+	}
+}
+
+// With the probe switched off there is nothing to report, and the client must be told to
+// ignore the field rather than trust a value nothing updates.
+func TestEmbyHealthPayloadIsReachableWhenUnmonitored(t *testing.T) {
+	payload := embyHealthFor(embyHealthState{})
+	if payload.Monitored {
+		t.Error("an unstarted probe reported itself as monitored")
+	}
+	if !payload.Reachable {
+		t.Error("an unmonitored server was reported unreachable")
+	}
+	if payload.Since != "" || payload.CheckedAt != "" {
+		t.Error("an unstarted probe reported timestamps it never took")
+	}
+}
+
+func TestEmbyHealthPayloadCarriesTheRetryInterval(t *testing.T) {
+	var health embyHealth
+	start := time.Now().UTC()
+	health.begin(60*time.Second, start)
+	health.record(false, start.Add(time.Minute))
+	health.record(false, start.Add(2*time.Minute))
+
+	payload := embyHealthFor(health.get())
+	if payload.RetrySeconds != 60 {
+		t.Errorf("retrySeconds = %d, want 60", payload.RetrySeconds)
+	}
+	if payload.Reachable {
+		t.Error("payload reported a reachable server during an outage")
+	}
+	if payload.CheckedAt == "" {
+		t.Error("payload omitted the last probe time the bar counts down from")
+	}
+}