kubernetes-novolume - container step state is not copied before deletion

[kbntx]

Summary

Hello, I am currently experimenting with the new kubernetes-novolume mode and while evaluating our different use cases, I noticed that there is no logic to copy the final workspace state of a container step before deleting the pod running it.

Is it an expected breaking change with this new approach ? In the v0.7.0, the updates done by a container step to the workspace are available to the following steps running the main job pod.

Example

jobs:
  job1:
    runs-on: self-hosted
    container:
      image: xxx

    steps:
      - name: Checkout repository
        uses: actions/checkout@v4

      # Add a foo.txt file in the workspace folder
      - name: Do something in container
        uses: docker://xxx

      # Read the foo.txt file > file not found
      - name: Read file foo.txt
        run: |
          cat foo.txt
        shell: bash

Investigation

I found the following code during my investigation. Is it expected that after running execPodStep, the pod is directly deleted for the container step case ?

// run-container-step.ts

l90 => await execCpFromPod(getJobPodName(), `/__w`, `${runnerWorkspace}`)
l92 => writeContainerStepScript()
l100 => execCpToPod(podName, `${runnerWorkspace}/__w`, '/__w')
l106 => execPodStep(
        ['/__e/sh', '-e', containerPath],
        pod.metadata.name,
        JOB_CONTAINER_NAME
      )
l122 => await deletePod(podName)

// run-script-step.ts

l27 => await execCpToPod(state.jobPod, runnerTemp, containerTemp)
l33 => await execPodStep(
      [args.entryPoint, ...args.entryPointArgs],
      state.jobPod,
      JOB_CONTAINER_NAME
    )
l54 => await execCpFromPod(state.jobPod, containerTemp, workdir)

Thank you in advance for your support!