fix(controller): Do not panic of woc.orig in not hydrated. Fixes argo…

…proj#3118

workflow/controller/operator.go

@@ -543,16 +543,17 @@ func (woc *wfOperationCtx) persistWorkflowSizeLimitErr(wfClient v1alpha1.Workflo
 	woc.markWorkflowError(err, true)
 	_, err = wfClient.Update(woc.wf)
 	if err != nil {
-		woc.log.Warnf("Error updating workflow: %v", err)
+		woc.log.Warnf("Error updating workflow with size error: %v", err)
 	}
 }
 
 // reapplyUpdate GETs the latest version of the workflow, re-applies the updates and
 // retries the UPDATE multiple times. For reasoning behind this technique, see:
 // https://github.com/kubernetes/community/blob/master/contributors/devel/api-conventions.md#concurrency-control-and-consistency
 func (woc *wfOperationCtx) reapplyUpdate(wfClient v1alpha1.WorkflowInterface, nodes wfv1.Nodes) (*wfv1.Workflow, error) {
-	if !woc.controller.hydrator.IsHydrated(woc.orig) {
-		panic("original workflow is not hydrated")
+	err := woc.controller.hydrator.Hydrate(woc.orig)
+	if err != nil {
+		return nil, err
 	}
 	// First generate the patch
 	oldData, err := json.Marshal(woc.orig)