When Expressions not scoped to the task when results is missing too

[RafaeLeal]

Expected Behavior

Not sure if this is the expected behavior, I just thought that this was pretty confusing for my scenario, let me know what you think.
I have PipelineRun that runs something like this:

  ┌──────────┐  ┌─────────┐
  │build-app1├─►│push-app1├──┐
  └──────────┘  └─────────┘  │ ┌──────┐
                             │►│deploy│
  ┌──────────┐  ┌─────────┐  │ └──────┘
  │build-app2├─►│push-app2├──┘
  └──────────┘  └─────────┘

This PipelineRun is generated and quite big/complex, so I tried to simplify as much as possible, to something like this:

apiVersion: tekton.dev/v1beta1
kind: PipelineRun
metadata:
  name: testing-when-expressions
  namespace: pipelines
spec:
  pipelineSpec:
    tasks:
      - name: build-app1
        taskSpec:
          results:
            - name: app1-image-tarball
              type: string
          steps: [...]
        when:
          - input: "app1"
            operator: in
            values: [app1, both]

      - name: push-app1
        params:
          - name: app1-image-tarball
            value: $(tasks.build-app1.results.app1-image-tarball)
        taskSpec:
          params:
            - name: app1-image-tarball
              type: string
          steps: [...]
        when:
          - input: "app1"
            operator: in
            values: [app1, both]

      - name: build-app2
        taskSpec:
          results:
            - name: app2-image-tarball
              type: string
          steps: [...]
        when:
          - input: "app1"
            operator: in
            values: [app2, both]

      - name: push-app2
        params:
          - name: app2-image-tarball
            value: $(tasks.build-app2.results.app2-image-tarball)
        taskSpec:
          params:
            - name: app2-image-tarball
              type: string
          steps: [...]
        when:
          - input: "app1"
            operator: in
            values: [app2, both]
      - name: deploy
        runAfter:
          - push-app1
          - push-app2
        taskSpec:
          steps: [ ... ]

What do I expected:

1. build-app1 runs
2. build-app2 is skipped
3. push-app1 runs
4. push-app2 is skipped
5. deploy runs

Actual Behavior

What actually happened:

1. ✅ build-app1 runs
2. ✅ build-app2 is skipped
3. ✅ push-app1 runs
4. 🟡 push-app2 is skipped
5. ❌ deploy runs

Debugging what happened, I first noticed in the status:

  provenance:
    featureFlags:
      ...
      ScopeWhenExpressionsToTask: false

This is confusing since this feature-flag used to have default "true" value and I didn't change it. But AFAIK this is just a visibility problem since this feature flag doesn't actually exist anymore.

Looking at the skipped tasks:

skippedTasks:
   - name: build-app2
     reason: When Expressions evaluated to false
     whenExpressions: [...]
   - name: deploy
     reason: Parent Tasks were skipped
   - name: push-app2
     reason: Results were missing
     whenExpressions: [...]

So here it lies the problem: push-app2 was skipped due to results being missing, instead of when expressions.. So the deploy was skipped too, since the parent was skipped not due to when expressions.

From what I could gather, this happens because we evaluate first the results references then the when expressions here

So what happened was that push-app2 was skipped for the "wrong" reason so deploy was unexpectedly skipped.

Steps to Reproduce the Problem

1. Create a PipelineRun as described above

Additional Info

• Kubernetes version:

  Output of kubectl version -o yaml | yq .serverVersion:

buildDate: "2023-06-16T17:32:40Z"
compiler: gc
gitCommit: 78c8293d1c65e8a153bf3c03802ab9358c0e1a14
gitTreeState: clean
gitVersion: v1.27.3-eks-a5565ad
goVersion: go1.20.5
major: "1"
minor: 27+
platform: linux/amd64

• Tekton Pipeline version:

  Output of tkn version or kubectl get pods -n tekton-pipelines -l app=tekton-pipelines-controller -o=jsonpath='{.items[0].metadata.labels.version}'

v0.48.0

[tekton-robot]

Issues go stale after 90d of inactivity.
Mark the issue as fresh with /remove-lifecycle stale with a justification.
Stale issues rot after an additional 30d of inactivity and eventually close.
If this issue is safe to close now please do so with /close with a justification.
If this issue should be exempted, mark the issue as frozen with /lifecycle frozen with a justification.

/lifecycle stale

Send feedback to tektoncd/plumbing.

[tommyb82]

We have just encountered this issue, Tekton pipelines v0.56.4.

• Task A - Skip when conditional = false. If executed, results are calculated and used in Task B
• Task B - Skip on same when conditional as Task A. If executed, uses results from Task A. But skipped reason is "Results were missing"
• Task C - Should always execute (no when block) but is skipped if Task B skips due to "Results were missing".

Task C has no dependency on Task A's results, and if we remove declaration/use of results between Task A and B by hardcoding some values temporarily, then Task C executes regardless of whether Task A or B are skipped (conclusion: feature ScopeWhenExpressionsToTask is defaulting to true as expected, despite being reported as false in the PR yaml as observed by OP).