Why is a code coverage report published as an Artifact when I have publishJUnitResults set to false?

Question

The reason I want to disable it is that it adds about an extra 30 seconds to my pipeline when I don't even need it. All I want is for the report to be published to Sonarqube, which is working fine.

My pipeline YAML file:

trigger: none
pool:
  name: <REDACTED>
  demands:
  - agent.name -equals  <REDACTED>
  
stages:
- stage: InstallJava
  jobs:
    - job:
      steps:
      - task: CmdLine@2
        inputs:
            script: |
              echo Write your commands here
              echo Hello world
              pwd
              dir
              mvn -version
              java -version
              mvn clean compile
              export JAVA_HOME='/usr/lib/jvm/java-11-openjdk-11.0.21.0.9-1.el7_9.x86_64'
      - task: MavenAuthenticate@0
        inputs:
          artifactsFeeds: '<REDACTED>'
      - task: SonarQubePrepare@5
        inputs:
          SonarQube: '<REDACTED> SonarQube'
          scannerMode: 'Other'
      - task: Maven@4
        inputs:
          mavenPomFile: 'pom.xml'
          goals: 'clean install'
          publishJUnitResults: false
          javaHomeOption: 'path'
          codeCoverageToolOption: 'JaCoCo'
          jdkDirectory: '/usr/lib/jvm/java-11-openjdk'
          mavenVersionOption: 'Default'
          mavenAuthenticateFeed: false
          effectivePomSkip: false
          sonarQubeRunAnalysis: true
      - task: SonarQubePublish@5
        inputs:
          pollingTimeoutSec: '300'

The code coverage report being generated as an Artifact:

Answer 1

To answer the question WHY, this unfortunately appears to be by design. The Maven task has the following logic, lifted from the source code of the Maven@4 task.

 // Maven task orchestration occurs as follows:
 // 1. Check that Maven exists by executing it to retrieve its version.
 // 2. Apply any goals for static code analysis tools selected by the user.
 // 3. Run Maven. Compilation or test errors will cause this to fail.
 //    In case the build has failed, the analysis will still succeed but the report will have less data. 
 // 4. Attempt to collate and upload static code analysis build summaries and artifacts.
 // 5. Always publish test results even if tests fail, causing this task to fail.
 // 6. If #3 or #4 above failed, exit with an error code to mark the entire step as failed.

Step #5 has some logic to fail the build if the coverage is empty, which appears to be coupled with the implementation of publishing code-coverage artifacts.

I'm assuming that you want both code coverage and unit-test results to be included in the data set published to SonarQube. If they are not needed at all, you can prevent publishing by setting the codeCoverageToolOption to none, which will set the isCodeCoverageOpted variable (shown above) to false.