fix(logging): name the antenna disagreement, and confirm an outbound send

Two silences from one operator's log.

The SteppIR is not blocked — it is being overruled. Every tune is commanded,
acknowledged on the next poll with the frequency asked for, then replaced on
the poll after by a different one: 21075 asked, 21075 confirmed, 21050
reported, again and again, with the operator tuning once more each time. Two
minutes of that reads as a dead link. It now says what it is, once, when the
motors have stopped and the reported frequency is more than 10 kHz from the
commanded one — a controller quantising to its own grid is not a fault, one
sitting somewhere else entirely is.

UDP outbound had no positive signal at all: only failures were logged. When
an operator sets up an ADIF message to a second logger and nothing arrives,
"OpsLog never sent it" and "the other program did not take it" looked exactly
the same from here — and the first is the common case, because a row created
as an inbound ADIF listener instead of an outbound ADIF message matches
nothing and emits nothing. Successful sends are logged with their destination
and service, and a QSO logged with no outbound row says so once a session.

internal/steppir/drift_test.go

@@ -0,0 +1,50 @@
+package steppir
+
+import "testing"
+
+// An operator's log read as "the antenna stops responding": every tune was
+// commanded, acknowledged with the requested frequency on the next poll, then
+// replaced on the poll after by a different one — 21075 asked, 21075 confirmed,
+// 21050 reported — with the operator tuning again each time. Nothing in the log
+// named that, so it looked like a dead link.
+func TestCheckDriftReportsOnceAndRecovers(t *testing.T) {
+	c := &Client{}
+	c.lastSetKHz = 21075
+
+	// Where it was asked to go: nothing to say.
+	c.checkDrift(&Status{Frequency: 21075})
+	if c.lastDriftKHz != 0 {
+		t.Errorf("a controller on frequency was reported as drifting (%d)", c.lastDriftKHz)
+	}
+	// Its own grid is not a fault.
+	c.checkDrift(&Status{Frequency: 21070})
+	if c.lastDriftKHz != 0 {
+		t.Errorf("%d kHz is within tolerance and was reported", 21070)
+	}
+	// Somewhere else entirely: recorded, so it is said once.
+	c.checkDrift(&Status{Frequency: 21050})
+	if c.lastDriftKHz != 21050 {
+		t.Errorf("lastDriftKHz = %d, want 21050", c.lastDriftKHz)
+	}
+	// Still there on the next poll: nothing new to say, and the marker stands.
+	c.checkDrift(&Status{Frequency: 21050})
+	if c.lastDriftKHz != 21050 {
+		t.Errorf("lastDriftKHz = %d after a repeat, want it unchanged", c.lastDriftKHz)
+	}
+	// Back where it belongs: the marker clears, so the next divergence is said.
+	c.checkDrift(&Status{Frequency: 21075})
+	if c.lastDriftKHz != 0 {
+		t.Errorf("lastDriftKHz = %d after recovery, want 0", c.lastDriftKHz)
+	}
+}
+
+// A frequency read while the elements are travelling is not a disagreement, it
+// is an antenna on its way — reporting it would cry wolf on every single tune.
+func TestCheckDriftIgnoresMovingMotors(t *testing.T) {
+	c := &Client{}
+	c.lastSetKHz = 21075
+	c.checkDrift(&Status{Frequency: 21000, MotorsMoving: 1})
+	if c.lastDriftKHz != 0 {
+		t.Errorf("a moving antenna was reported as drifting (%d)", c.lastDriftKHz)
+	}
+}

internal/steppir/steppir.go

@@ -102,6 +102,10 @@ type Client struct {
 	statusMu   sync.RWMutex
 	lastStatus *Status
 	lastSetKHz int
+	// lastDriftKHz is the frequency last reported for a controller that had gone
+	// somewhere other than where it was told, so the disagreement is stated once
+	// and not on every poll. Zero when it is where it should be.
+	lastDriftKHz int
 
 	// lastRaw holds the previous raw status frame so we only log a status line
 	// when the controller's reply actually changes — enough to diagnose a stuck
@@ -269,10 +273,61 @@ func (c *Client) pollLoop() {
 			c.applyPendingDir(st)
 			c.lastStatus = st
 			c.statusMu.Unlock()
+			c.checkDrift(st)
 		}
 	}
 }
 
+// driftToleranceKHz is how far the controller's reported frequency may sit from
+// the one we commanded before it is worth saying so. A SteppIR quantises to its
+// own grid, so a few kHz is normal and must not be reported as a fault.
+const driftToleranceKHz = 10
+
+// checkDrift reports a controller that settled somewhere other than where it was
+// told to go.
+//
+// This exists because of a log that read as "the antenna stops responding": every
+// tune was commanded, acknowledged on the next poll with the requested frequency,
+// and then REPLACED on the poll after by a different one — 21075 asked, 21075
+// confirmed, 21050 reported, over and over, with the operator tuning again each
+// time. Nothing in the log named that, so it looked like a dead link rather than
+// a controller with a mind of its own (its own radio interface tracking the rig,
+// or a front-panel mode that overrides the host).
+//
+// Only when the motors have stopped: a frequency read mid-travel is not a
+// disagreement, it is an antenna on its way. And only when the value changes, so
+// a controller parked somewhere else does not fill the log.
+func (c *Client) checkDrift(st *Status) {
+	if st.MotorsMoving != 0 {
+		return
+	}
+	c.statusMu.Lock()
+	want := c.lastSetKHz
+	last := c.lastDriftKHz
+	if want <= 0 || st.Frequency <= 0 {
+		c.statusMu.Unlock()
+		return
+	}
+	diff := st.Frequency - want
+	if diff < 0 {
+		diff = -diff
+	}
+	if diff <= driftToleranceKHz {
+		c.lastDriftKHz = 0 // back where it was asked to be
+		c.statusMu.Unlock()
+		return
+	}
+	if last == st.Frequency {
+		c.statusMu.Unlock()
+		return // already said, and it has not moved since
+	}
+	c.lastDriftKHz = st.Frequency
+	c.statusMu.Unlock()
+	log.Printf("steppir: commanded %d kHz but the controller settled on %d kHz — "+
+		"something else is driving it (its own radio interface, or a front-panel mode overriding the host)",
+		want, st.Frequency)
+}
+
 // applyPendingDir replaces a freshly polled direction with the one the operator
 // last commanded, until the controller confirms it (or the hold expires). The
 // caller holds statusMu.