Fix Config Daemon node selector

When node selectors are added then removed via "configDaemonNodeSelector"
via the sriovoperatorconfigs CRD, certain combinations will not trigger
an update due to the ordering of arguments into "DeepDerivative" from
the reflect library.

This is an example combination of setting "configDaemonNodeSelector" that
would make it such that the sriov-config-daemon daemon set's nodeSelector
to become out of sync with the original "DeepDerivative" argument order:

Step 1) Create 3 labels in node selector:
configDaemonNodeSelector = {"labelA": "", "labelB": "", "labelC": ""}
config-daemon DS NodeSelector = {"labelA": "", "labelB": "", "labelC": ""}

Step 2) Remove 1 label in node selector without making changes to the other labels:
configDaemonNodeSelector = {"labelA": "", "labelB": ""}
config-daemon DS NodeSelector = {"labelA": "", "labelB": "", "labelC": ""} (Out of Sync)

"DeepDerivative" assumes that the left argument is the original (v1) and
the right argument is the updated (v2). For maps it only checks if v1 >
v2 in length:
    case reflect.Map:
        ...
        if v1.Len() > v2.Len() {
            return false
        }
        ...

This commit just reverts changes from commit:
"661a65b8e1aee6339037948732f75d06ceb91611"
Use DeepDerivative to compare the kube object content
Such that we react to changes to node selectors properly.

Signed-off-by: William Zhao <[email protected]>

pkg/apply/apply.go

@@ -90,11 +90,11 @@ func ApplyObject(ctx context.Context, client k8sclient.Client, obj *uns.Unstruct
 	if err := MergeObjectForUpdate(existing, obj); err != nil {
 		return errors.Wrapf(err, "could not merge object %s with existing", objDesc)
 	}
-	if !equality.Semantic.DeepDerivative(obj, existing) {
+	if !equality.Semantic.DeepEqual(existing, obj) {
 		if err := client.Update(ctx, obj); err != nil {
 			return errors.Wrapf(err, "could not update object %s", objDesc)
 		} else {
-			log.Printf("update was successful")
+			log.Printf("update was successful %s", objDesc)
 		}
 	}