fix failure evaluation

There were two problems with gdt-kube's evaluation. First, I was
improperly passing a `*interface{}` instead of a `interface{}` to the
`newAssertions()` constructor. This mean the type-casting within
methods like `assertions.lenOK()` was failing.

Secondly, I was not calling `t.Error()` appropriately when there were
failures collected from the assertions.

Signed-off-by: Jay Pipes <jaypipes@gmail.com>

assertions.go

@@ -340,7 +340,7 @@ func (a *assertions) notFoundOK() bool {
 // lenOK returns true if the subject matches the Len condition, false otherwise
 func (a *assertions) lenOK() bool {
 	exp := a.exp
-	if exp.Len != nil {
+	if exp.Len != nil && a.hasSubject() {
 		// if the supplied resp is a list of objects returned by the dynamic
 		// client check its length
 		list, ok := a.r.(*unstructured.UnstructuredList)

eval.go

@@ -50,6 +50,7 @@ func (s *Spec) Eval(ctx context.Context, t *testing.T) *result.Result {
 	ticker := backoff.NewTicker(bo)
 	attempts := 0
 	start := time.Now().UTC()
+	success := false
 	for tick := range ticker.C {
 		attempts++
 		after := tick.Sub(start)
@@ -64,8 +65,8 @@ func (s *Spec) Eval(ctx context.Context, t *testing.T) *result.Result {
 				return result.New(result.WithRuntimeError(err))
 			}
 		}
-		a = newAssertions(s.Assert, err, &out)
-		success := a.OK()
+		a = newAssertions(s.Assert, err, out)
+		success = a.OK()
 		debug.Println(
 			ctx, t, "%s (try %d after %s) ok: %v",
 			s.Title(), attempts, after, success,
@@ -81,5 +82,10 @@ func (s *Spec) Eval(ctx context.Context, t *testing.T) *result.Result {
 			)
 		}
 	}
+	if !success {
+		for _, fail := range a.Failures() {
+			t.Error(fail)
+		}
+	}
 	return result.New(result.WithFailures(a.Failures()...))
 }