Updating yaml pipeline variables

Question

Hi I have a yml pipeline in azure devops.

variables:
 - name: dependencyChanged
    value: false

-stage: stage1
 jobs:
 - task: Bash@3
   inputs:
       targetType: 'inline'
       script: |
           CHANGED_FILES=$(git diff HEAD HEAD~ --name-only)
           DEPENDENCY_CHANGED=$(echo "$CHANGED_FILES" | grep -q "aml/ml_service/docker/" && echo "true" || echo "false")
           echo "##vso[task.setvariable variable=dependencyChanged=true;isOutput=true]]$DEPENDENCY_CHANGED"
           echo "Debug: adfChanged value is '${{ variables['dependencyChanged'] }}'"

-stage: stage2
 condition: and(succeeded('stage1'), eq(variables['dependencyChanged'], 'true'))
 jobs:
 .... some jobs here

The problem is: DEPENDENCY_CHANGED is 'true' and dependency_change is also true becasue it is set from DEPENDENCY_CHANGED

but stage2 never gets executed, not sure why. even if if do echo "##vso[task.setvariable variable=dependencyChanged=true;isOutput=true]]'true'"

In summary, I want to update the pipeline variables with in different stages and access them in multiple stages.

Answer 1

There are some syntax errors in your YAML pipeline for setting and using output variables between stages. You can reference below sample to correct your YAML pipeline.

1. The pipeline main YAML.

   # azure-pipelines.yml
   
    stages:
    - stage: stage1
      jobs:
      - job: job1
        steps:
        - bash: |
            CHANGED_FILES=$(git diff HEAD HEAD~ --name-only)
            DEPENDENCY_CHANGED=$(echo "$CHANGED_FILES" | grep -q "aml/ml_service/docker/" && echo "true" || echo "false")
            echo "##vso[task.setvariable variable=dependencyChanged;isOutput=true]$DEPENDENCY_CHANGED"
          name: setOutVar
          displayName: 'Set Output Variable'
   
        # Use output variable within the same job.
        - bash: |
            echo "dependencyChanged = $(setOutVar.dependencyChanged)"
          displayName: 'Print Output Variable'
   
      # Use output variable from a different job within the same stage.
      # Use the output variable on job-level condition.
      # Map the output variable as a job-level variable.
      - job: job2
        dependsOn: job1
        condition: and(succeeded('job1'), eq(dependencies.job1.outputs['setOutVar.dependencyChanged'], 'true'))
        variables:
          dependencyChanged: $[ dependencies.job1.outputs['setOutVar.dependencyChanged'] ]
        steps:
        - bash: |
            echo "dependencyChanged = $(dependencyChanged)"
          displayName: 'Print job-level Variable'
   
    # Use output variable from a different stage.
    # Use the output variable on stage-level condition.
    # Map the output variable as a stage-level variable.
    - stage: stage2
      condition: and(succeeded('stage1'), eq(stageDependencies.stage1.outputs['job1.setOutVar.dependencyChanged'], 'true'))
      variables:
        dependencyChanged: $[ stageDependencies.stage1.job1.outputs['setOutVar.dependencyChanged'] ]
      jobs:
      - job: job3
        steps:
        - bash: |
            echo "dependencyChanged = $(dependencyChanged)"
          displayName: 'Print stage-level Variable'
   

2. the results.

   • When dependencyChanged=true, job2 and stage2 will run due to the conditions.

     

   • When dependencyChanged=false, job2 and stage2 will get skipped due to the conditions.

     

Related documentations:

• Use output variables from tasks
• Dependency syntax overview

---

Answer 2

I think the task.setvariable line should read

echo "##vso[task.setvariable variable=dependencyChanged;isOutput=true]$DEPENDENCY_CHANGED"

also if you want to reference it in another stage you should use

condition: and(succeeded('stage1'), eq(dependencies.stage1.outputs['PUTYOURTASKNAMEHERE.dependencyChanged'], 'true'))

And replace PUTYOURTASKNAMEHERE appropriately