Workflow hangs when 'when' clause references skipped step outputs

[KowalczykBartek]

Pre-requisites

• I have double-checked my configuration
• I have tested with the :latest image tag (i.e. quay.io/argoproj/workflow-controller:latest) and can confirm the issue still exists on :latest. If not, I have explained why, in detail, in my description below.
• I have searched existing issues and could not find a match for this bug
• I'd like to contribute the fix myself (see contributing guide) - ( I can try but needs to know if that is really an issue/bug)

What happened? What did you expect to happen?

Up to version 3.7.1

- - name: job2-error-handler
    arguments:
      parameters:
        - name: job-name
          value: "error-handler"
    templateRef:
      name: simple-job
      template: run-job
    when: "\"{{steps.job2.outputs.parameters.status}}\" != \"SUCCESS\" && \"{{steps.job1.outputs.parameters.status}}\" == \"SUCCESS\""

this step was correctly skipped,

steps.job2.outputs.parameters.status

was treat as an empty, after upgrading entire workflow hangs

STEP                       TEMPLATE            PODNAME                                               DURATION  MESSAGE
 ● requeue-bug-demo-wqcnr  main                                                                                                                          
 ├───✔ job1                simple-job/run-job  requeue-bug-demo-wqcnr-simple-job/run-job-1716469494  6s                                                  
 ├───○ job2                simple-job/run-job                                                                  when 'SUCCESS == FAILED' evaluated false  

Question is, is that a bug or previously it worked by accident ?

Version(s)

v3.7.12, dbb6c34

Paste a minimal workflow that reproduces the issue. We must be able to run the workflow; don't enter a workflow that uses private images.

---
# Simple template that returns a status
apiVersion: argoproj.io/v1alpha1
kind: WorkflowTemplate
metadata:
  name: simple-job
  namespace: argo
spec:
  templates:
    - name: run-job
      inputs:
        parameters:
          - name: job-name
      outputs:
        parameters:
          - name: status
            valueFrom:
              path: /tmp/status.txt
      script:
        image: python:alpine3.6
        command: [python]
        source: |
          with open('/tmp/status.txt', 'w') as f:
            f.write('SUCCESS')

---
apiVersion: argoproj.io/v1alpha1
kind: WorkflowTemplate
metadata:
  name: requeue-bug-demo
  namespace: argo
spec:
  entrypoint: main
  templates:
    - name: main
      steps:
        - - name: job1
            arguments:
              parameters:
                - name: job-name
                  value: "first-job"
            templateRef:
              name: simple-job
              template: run-job
        
        - - name: job2
            arguments:
              parameters:
                - name: job-name
                  value: "second-job"
            templateRef:
              name: simple-job
              template: run-job
            when: "{{steps.job1.outputs.parameters.status}} == FAILED"
        
        # CAUSES THE BUG
        - - name: job2-error-handler
            arguments:
              parameters:
                - name: job-name
                  value: "error-handler"
            templateRef:
              name: simple-job
              template: run-job
            when: "\"{{steps.job2.outputs.parameters.status}}\" != \"SUCCESS\" && \"{{steps.job1.outputs.parameters.status}}\" == \"SUCCESS\""

Logs from the workflow controller

[controller] (running)  time="2026-04-06T19:33:13.321Z" level=error msg="Failed to resolve references" error=requeue namespace=argo stepName=job2-error-handler workflow=requeue-bug-demo-7rbjj

Logs from in your workflow's wait container

there is no

[coderabbitai]

⚠️ Possible Duplicate Issue(s)

• Bug: Using an Output from a skipped stage as a Parameter blocks workflows in v4.0.3 #15837

🔗 Related PRs

#15240 - fix: use input.defaults for suspend templates [merged]
#15442 - fix: requeue workflow if expected variables are missing. Fixes #15513 [merged]
#15736 - fix: add stepgroup and taskgroup to scope. Fixes #15737 [merged]

---

📝 Issue Planner

_{Check the box below or use the @coderabbitai plan command to generate an implementation plan and prompts that you can use with your favorite coding assistant.}

• Create Plan

---

🧪 Issue enrichment is currently in open beta.

You can configure auto-planning by selecting labels in the issue_enrichment configuration.

To disable automatic issue enrichment, add the following to your .coderabbit.yaml:

issue_enrichment:
  auto_enrich:
    enabled: false

💬 Have feedback or questions? Drop into our discord!

[isubasinghe]

This is a known issue that has been addressed in 3.7.13

[KowalczykBartek]

@isubasinghe thanks for answer, but looks reproducer from my original post still hangs because of

[controller] (running)  time="2026-04-08T11:53:30.654Z" level=error msg="Failed to resolve references" error=requeue namespace=argo stepName=job2-error-handler workflow=requeue-bug-demo-qbp66

I tried with 8204ae1 - it still happens

STEP                       TEMPLATE            PODNAME                                               DURATION  MESSAGE
 ● requeue-bug-demo-qbp66  main                                                                                                                          
 ├───✔ job1                simple-job/run-job  requeue-bug-demo-qbp66-simple-job/run-job-1716469494  18s                                                 
 ├───○ job2                simple-job/run-job                                                                  when 'SUCCESS == FAILED' evaluated false  

something like this make my case happy

+++ b/workflow/controller/steps.go
@@ -161,6 +161,23 @@ func (woc *wfOperationCtx) executeSteps(ctx context.Context, nodeName string, tm
                                woc.buildLocalScope(stepsCtx.scope, prefix, sgNode)
                        } else {
                                woc.buildLocalScope(stepsCtx.scope, prefix, childNode)
+
+                               if (childNode.Phase == wfv1.NodeSkipped || childNode.Phase == wfv1.NodeOmitted) && childNode.Outputs == nil {
+                                       _, tmpl, _, err := stepsCtx.tmplCtx.ResolveTemplate(&step)
+                                       if err == nil && tmpl != nil {
+                                               for _, param := range tmpl.Outputs.Parameters {
+                                                       key := fmt.Sprintf("xoxoxoxo %s.outputs.parameters.%s", prefix, param.Name)
+                                                       stepsCtx.scope.addParamToScope(key, "")
+                                               }
+                                               if tmpl.Outputs.Result != nil {
+                                                       stepsCtx.scope.addParamToScope(fmt.Sprintf("%s.outputs.result", prefix), "")
+                                               }
+                                       }
+                               }
                        }
                }
        }

in case it will actually make sense I opened pr #15884