Helper: replace the last sleeping loop with the consistency helper its only caller needs (#8183)

WaitForDeploymentReplicaCountChange was the one loop in tests/helper still
timed by time.Sleep after #8068 and #8085. It polled a deployment for a fixed
number of iterations and returned the replica count at the point it first
changed, or the last count seen when the iterations ran out.

Its only remaining caller, azure_workload_identity_user_assigned, compares
that count to 0 after 30 x 1s to check that a ScaledObject with the wrong
identity does not scale the deployment out. That is a "stays at 0 for the
window" assertion, and AssertReplicaCountNotChangeDuringTimePeriod already
expresses exactly that on top of KedaConsistently: it observes the count on
every attempt inside the window instead of only reporting the first change,
and it does not fold a failed read into a 0 that looks like a scale to zero.
The caller now uses it with the same 30s window; the message that claimed a
minute was describing the loop wrongly.

With no callers left, the helper is removed rather than converted. Its name
suggested a wait for a change, while the two callers it had left both used it
to assert that nothing changed, and #8110 already had to fix the other one,
which called it with the arguments transposed into a no-op.
tests/helper/helper.go no longer contains a time.Sleep.

Relates to #7991

Signed-off-by: Mikhail Fedosin <mfedosin@redhat.com>

Author: Fedosin

tests/helper/helper.go

@@ -730,36 +730,6 @@ func WaitForReplicaSetReplicaReadyCount(t *testing.T, kc *kubernetes.Clientset,
 	})
 }
 
-// Waits for number of iterations and returns replica count.
-func WaitForDeploymentReplicaCountChange(t *testing.T, kc *kubernetes.Clientset, name, namespace string, iterations, intervalSeconds int) int {
-	t.Log("Waiting for some time to see if deployment replica count changes")
-	var replicas, prevReplicas int32
-	prevReplicas = -1
-
-	for i := 0; i < iterations; i++ {
-		deployment, err := kc.AppsV1().Deployments(namespace).Get(context.Background(), name, metav1.GetOptions{})
-		if err != nil {
-			// Reading on would report 0 replicas, which the caller cannot tell apart from a
-			// genuine scale to zero.
-			t.Logf("cannot get deployment %s/%s - %s", namespace, name, err)
-		} else {
-			replicas = deployment.Status.Replicas
-
-			t.Logf("Deployment - %s, Current  - %d", name, replicas)
-
-			if replicas != prevReplicas && prevReplicas != -1 {
-				break
-			}
-
-			prevReplicas = replicas
-		}
-
-		time.Sleep(time.Duration(intervalSeconds) * time.Second)
-	}
-
-	return int(replicas)
-}
-
 // Waits some time to ensure that the replica count doesn't change.
 func AssertReplicaCountNotChangeDuringTimePeriod(t *testing.T, kc *kubernetes.Clientset, name, namespace string, target, intervalSeconds int) {
 	t.Logf("Waiting for some time to ensure deployment replica count doesn't change from %d", target)

tests/secret-providers/azure_workload_identity_user_assigned/azure_workload_identity_user_assigned_test.go

@@ -188,9 +188,10 @@ func testScaleOutWithIncorrectIdentity(t *testing.T, kc *kubernetes.Clientset, c
 		}, nil)
 	}
 
-	// scale out should fail as we are using the incorrect identity
-	assert.True(t, WaitForDeploymentReplicaCountChange(t, kc, deploymentName, testNamespace, 30, 1) == 0,
-		"replica count should be 0 after 1 minute")
+	// Scale out should fail as we are using the incorrect identity, so the replica count has to stay
+	// at 0 for the whole window rather than merely be 0 at the end of it. Thirty seconds is what the
+	// previous 30 x 1s wait allowed.
+	AssertReplicaCountNotChangeDuringTimePeriod(t, kc, deploymentName, testNamespace, 0, 30)
 }
 
 func testScaleOutWithCorrectIdentity(t *testing.T, kc *kubernetes.Clientset, data templateData) {