Merge pull request 'power: its loop and boot tests run on a fake clock, so a busy build node cannot fail them (issue 1126)' (#551) from fix/power-loop-flake into main

modules/power/cmd/power/draw.go

@@ -271,13 +271,12 @@ func (m *Machine) Publish(r Reading) error {
 
 // RunDraw samples and publishes until the context ends.
 func RunDraw(ctx context.Context, m *Machine) error {
-	return runDraw(ctx, m, SampleEvery, m.Publish)
+	return runDraw(ctx, m, SampleEvery, sleepFor(ctx), m.Publish)
 }
 
-// runDraw publishes one reading every `every`. A battery reading is instant, so it waits the interval
-// itself; a counter reading waits inside its sample. It returns as soon as the context ends.
-func runDraw(ctx context.Context, m *Machine, every time.Duration, publish func(Reading) error) error {
-	wait := func(d time.Duration) bool {
+// sleepFor waits d on the real clock, or less when ctx ends first; it reports whether the full wait passed.
+func sleepFor(ctx context.Context) func(time.Duration) bool {
+	return func(d time.Duration) bool {
 		select {
 		case <-ctx.Done():
 			return false
@@ -285,6 +284,12 @@ func runDraw(ctx context.Context, m *Machine, every time.Duration, publish func(
 			return true
 		}
 	}
+}
+
+// runDraw publishes one reading every `every`. A battery reading is instant, so it waits the interval
+// itself; a counter reading waits inside its sample. The wait is injected so a test drives the clock;
+// it returns false when the loop must stop. It returns as soon as the context ends.
+func runDraw(ctx context.Context, m *Machine, every time.Duration, wait func(time.Duration) bool, publish func(Reading) error) error {
 	for ctx.Err() == nil {
 		d := m.Sample(ctx, every, wait)
 		if ctx.Err() != nil {

modules/power/cmd/power/draw_test.go

@@ -187,24 +187,25 @@ func TestTheLoopPublishesEveryIntervalAndStopsWhenCancelled(t *testing.T) {
 		"/sys/class/power_supply/BAT0/status":    "Discharging",
 		"/sys/class/power_supply/BAT0/power_now": "12000000",
 	})
+	// A fake clock: each wait returns at once; the tenth cancels the context, as the end of the run.
 	ctx, cancel := context.WithCancel(context.Background())
+	defer cancel()
 	var lines []string
-	done := make(chan error)
-	go func() {
-		done <- runDraw(ctx, m, 20*time.Millisecond, func(r Reading) error { lines = append(lines, r.Text); return nil })
-	}()
-	time.Sleep(210 * time.Millisecond)
-	cancel()
-	select {
-	case err := <-done:
-		if err != nil {
-			t.Fatal(err)
+	waits := 0
+	wait := func(d time.Duration) bool {
+		if d != 20*time.Millisecond {
+			t.Errorf("waited %v, want the 20 ms interval", d)
 		}
-	case <-time.After(time.Second):
-		t.Fatal("the loop did not stop when cancelled")
+		if waits++; waits == 10 {
+			cancel()
+		}
+		return ctx.Err() == nil
 	}
-	// One at once, then one every 20 ms: about eleven in 210 ms, never a burst, never a stall.
-	if len(lines) < 6 || len(lines) > 12 || lines[0] != "12.0 W machine" {
+	if err := runDraw(ctx, m, 20*time.Millisecond, wait, func(r Reading) error { lines = append(lines, r.Text); return nil }); err != nil {
+		t.Fatal(err)
+	}
+	// One at once, then one after each of the nine waits that passed; the tenth wait ended the loop.
+	if len(lines) != 10 || lines[0] != "12.0 W machine" {
 		t.Fatalf("%d lines: %v", len(lines), lines)
 	}
 }

modules/power/cmd/power/watcher_test.go

@@ -213,11 +213,24 @@ func TestAMachineWithoutMainsSaysNothingAboutSupplies(t *testing.T) {
 func TestRunWithoutLogindStillSaysTheBoot(t *testing.T) {
 	m := machine(t, map[string]string{"/proc/sys/kernel/random/boot_id": "xyz\n"})
 	b := &bus{}
-	w := NewWatcher(m, b.emit, func() (Login1, error) { return nil, errors.New("no system bus") })
-	w.state = filepath.Join(t.TempDir(), "b")
-	w.poll = 10 * time.Millisecond
-	ctx, cancel := context.WithTimeout(context.Background(), 100*time.Millisecond)
+	// No timeout racing the loop: the run ends once logind was tried and the boot was published,
+	// both of which happen in Run's own goroutine.
+	ctx, cancel := context.WithCancel(context.Background())
 	defer cancel()
+	dialed, emitted := false, false
+	stopWhenBoth := func() {
+		if dialed && emitted {
+			cancel()
+		}
+	}
+	w := NewWatcher(m, func(typ string, body any) error {
+		err := b.emit(typ, body)
+		emitted = true
+		stopWhenBoth()
+		return err
+	}, func() (Login1, error) { dialed = true; defer stopWhenBoth(); return nil, errors.New("no system bus") })
+	w.state = filepath.Join(t.TempDir(), "b")
+	w.poll = time.Millisecond
 	w.Run(ctx)
 	if got := b.seen(); len(got) == 0 || got[0] != Booted || w.Snapshot().Problem == "" {
 		t.Fatalf("%v %+v", got, w.Snapshot())