Add initial integration test for dag tektoncd#168 🤓

This tests DAG functionality by defining a pipeline with both fan in and
fan out. The idea is that each Task echoes the current time, so after
the pipeline compeletes, we can look at which Task echoes which which
time to make sure they run in order. The tasks are also declared in the
Pipeline in the wrong order on purpose, to make sure that the order of
declaration doesn't affect how they are run (the opposite of the current
functionality)

Three caveats to this integration test:
- It was created before tektoncd#320, so the resource binding is not correct
- It was created before tektoncd#387, so it relies on the log PVC which will no
longer exist (can work around this by mounting a PVC explicitly in the
test and writing directly to it instead of echoing?)
- It doesn't exercise `runAfter` functionality

test/crd.go

@@ -50,15 +50,15 @@ const (
 	buildOutput     = "Build successful"
 )
 
-func getHelloWorldValidationPod(namespace, volumeClaimName string) *corev1.Pod {
+func getLogFetcherPod(namespace, volumeClaimName, podName string) *corev1.Pod {
 	return &corev1.Pod{
 		ObjectMeta: metav1.ObjectMeta{
 			Namespace: namespace,
-			Name:      hwValidationPodName,
+			Name:      podName,
 		},
 		Spec: corev1.PodSpec{
 			Containers: []corev1.Container{{
-				Name:  hwValidationPodName,
+				Name:  podName,
 				Image: "busybox",
 				Command: []string{
 					"cat",
@@ -118,29 +118,29 @@ func getHelloWorldTaskRun(namespace string) *v1alpha1.TaskRun {
 	}
 }
 
-func getBuildOutputFromVolume(logger *logging.BaseLogger, c *clients, namespace, testStr string) (string, error) {
+func getBuildOutputFromVolume(logger *logging.BaseLogger, c *clients, namespace, volumeClaimName, podName string) (string, error) {
 	// Create Validation Pod
 	pods := c.KubeClient.Kube.CoreV1().Pods(namespace)
 
 	// Volume created for Task should have the same name as the Task
-	if _, err := pods.Create(getHelloWorldValidationPod(namespace, hwTaskRunName)); err != nil {
-		return "", fmt.Errorf("failed to create Validation pod to mount volume `%s`: %s", hwTaskRunName, err)
+	if _, err := pods.Create(getLogFetcherPod(namespace, volumeClaimName, podName)); err != nil {
+		return "", fmt.Errorf("failed to create Validation pod to mount volume `%s`: %s", volumeClaimName, err)
 	}
 
-	logger.Infof("Waiting for pod with test volume %s to come up so we can read logs from it", hwTaskRunName)
-	if err := WaitForPodState(c, hwValidationPodName, namespace, func(p *corev1.Pod) (bool, error) {
+	logger.Infof("Waiting for pod with test volume %s to come up so we can read logs from it", volumeClaimName)
+	if err := WaitForPodState(c, podName, namespace, func(p *corev1.Pod) (bool, error) {
 		// the "Running" status is used as "Succeeded" caused issues as the pod succeeds and restarts quickly
 		// there might be a race condition here and possibly a better way of handling this, perhaps using a Job or different state validation
 		if p.Status.Phase == corev1.PodRunning {
 			return true, nil
 		}
 		return false, nil
 	}, "ValidationPodCompleted"); err != nil {
-		return "", fmt.Errorf("error waiting for Pod %s to finish: %s", hwValidationPodName, err)
+		return "", fmt.Errorf("error waiting for Pod %s to finish: %s", podName, err)
 	}
 
 	// Get validation pod logs and verify that the build executed a container w/ desired output
-	req := pods.GetLogs(hwValidationPodName, &corev1.PodLogOptions{})
+	req := pods.GetLogs(podName, &corev1.PodLogOptions{})
 	readCloser, err := req.Stream()
 	if err != nil {
 		return "", fmt.Errorf("failed to open stream to read: %v", err)