Add automatic stdout, stderr and exitcode outputs per step

[halberom]

Describe the enhancement
Add automatic stdout, stderr and exitcode outputs per run step.

This would make it very useful for things like generating changes in one step/job, and then adding them to a PR comment or sending them in a Slack notify in another.

Code Snippet

steps:
  - uses: actions/checkout@v2
  - name: get python version
    id: python_version
    run: python -V
  - name: output python version
    run: echo ${{ steps.python_version.outputs.stdout }}

Additional information
See also this community post for another example - https://xhtogithub.hejiu.icumunity/t/suggest-automatic-output-id-for-stdout-stderr/17419

[TingluoHuang]

I don't think we should do this since the stdout/stderr could be huge for certain step, we need to make sure we don't blow the size of memory the runner use.

We want the customer to be explicit when pass info between steps.

[ma7dev]

This workaround works for me. https://xhtogithub.hejiu.icumunity/t/how-can-i-set-an-expression-as-an-environment-variable-at-workflow-level/16516/9

Here is an example of my use case:

name: Automated Experiments
on:
  push:
    branches: [ main ]
jobs:
  build_on:
    runs-on: [ server ]
    defaults:
      run:
        shell: bash -l {0}
    env:
      EXPERIMENTS: "XXX"
      CONDA: "YYY"
      CUDA_HOME: "/usr/local/apps/cuda/cuda-10.1"
      CMAKE_HOME: "/usr/local/apps/cmake/3.14.5"
    steps:
      - uses: actions/checkout@v2
      - name: set environment variables
        run: |
          echo "---------------set environment variables---------------"
          export DATE=$(date "+%Y-%m-%d-%H-%M-%S")
          export SCRATCH=/scratch/$USER
          export WORKSPACE=$SCRATCH/actions-runner/_work/project/project
          export CONFIG="$WORKSPACE/experiments/cfgs/runner.yml"
          export LOGGER=$SCRATCH/logs
          export PYTHON=$SCRATCH/miniconda3/envs/$CONDA/bin/python
          echo "---------------add to GITHUB ENV---------------"
          echo "DATE=$DATE" >> $GITHUB_ENV
          echo "SCRATCH=$SCRATCH" >> $GITHUB_ENV
          echo "WORKSPACE=$WORKSPACE" >> $GITHUB_ENV
          echo "CONFIG=$CONFIG" >> $GITHUB_ENV
          echo "LOGGER=$LOGGER" >> $GITHUB_ENV
          echo "PYTHON=$PYTHON" >> $GITHUB_ENV
      - name: start the job
        run: |
          echo x-x-x-x-x-x-x-x-x-x-x ${{ env.DATE }} x-x-x-x-x-x-x-x-x-x-x
          
          echo "---------------source bash---------------"
          . ~/.bashrc
          echo "---------------CUDA---------------"
          export PATH="${{ env.CUDA_HOME }}/bin:$PATH"
          export LD_LIBRARY_PATH="${{ env.CUDA_HOME }}/lib64:$LD_LIBRARY_PATH"
          echo "---------------CMAKE---------------"
          export PATH="${{ env.CMAKE_HOME }}/bin:$PATH"
          export LD_LIBRARY_PATH="${{ env.CMAKE_HOME }}/lib64:$LD_LIBRARY_PATH"
          echo "---------------source conda env---------------"
          conda activate ${{ env.CONDA }}
          echo "---------------start experiments---------------"
          cd ${{ env.WORKSPACE }}/experiments/scripts/
          ${{ env.PYTHON }} experiments.py --config=${{ env.CONFIG }} --selected="${{ env.EXPERIMENTS }}" --logger="${{ env.LOGGER }}/${{ env.DATE }}"

Edit: My use case is more complicated where I get access to the environment variables in multiple runs steps

[AlissonRS]

@halberom @ma7dev @TingluoHuang I started a discussion about this here: https://github.com/orgs/community/discussions/83102

People are already doing this by appending stuff to GITHUB_OUTPUT, so if the runner was to have memory issues, it would already have it. Also we can make this feature optional and turned off by default, thus the user must explicitly enable it only for the steps they need it.