#6540 Windows CI doesn't report a failure of any commands before the last if calling executables Issue created by visch

Why is Install Dependencies not failing

Found the issue, it's still perplexing but atleast I understand more about what is happening

- name: Build tap on Python ${{ matrix.python-version }} 
        run: |
          poetry env use purposelyfail
          poetry debug

You would expect to fail, because

purposelyfail

is an invalid environment. I have a test repo setup. When you run this on these platforms here's what you get • Ubuntu - Fails as expected • Windows - Succeeds Even weirder is on Windows if you run

- name: Build tap on Python ${{ matrix.python-version }} 
        run: |
          poetry env use purposelyfail

The action does fail as you would expect. Searching around I found a few issues written up but could only find them cached on google it looks like discussions at least in this area did not index properly https://webcache.googleusercontent.com/search?q=cache:c48kgTB1SeMJ:https://github.community/t/multiline-commands-on-windows-do-not-fail-if-individual-commands-fail/16753+&cd=1&hl=en&ct=clnk&gl=us Other resources: • https://stackoverflow.com/questions/69615233/githubactions-on-windows-host-powershell-exit-code-of-previous-lines-being-i • https://stackoverflow.com/questions/9948517/how-to-stop-a-powershell-script-on-the-first-error

Error action preference only stops for a failed cmdlet. A failed external command (e.g. npm process) does not stop the script from proceeding.

An an alternative, please use bash/cmd shell.

Github documentation is here on the

pwsh

shell https://docs.github.com/en/actions/using-workflows/workflow-syntax-for-github-actions#exit-codes-and-error-action-preference From these docs

We append if ((Test-Path -LiteralPath variable:\LASTEXITCODE)) { exit $LASTEXITCODE } to powershell scripts so action statuses reflect the script's last exit code.

Is problematic for us because

poetry debug

returns exit code 0. I'd think that

$ErrorActionPreference = 'stop'

would do what we need here and fail at the first failure, but this doesn't happen probably due to calling an external application. Still thinking but we're much closer to an answer here. There's plenty of ways forward as well now that we know what's happening Put in https://github.com/orgs/community/discussions/27923 maybe we'll get some pointers there Originally posted by @visch in #6479 (comment) meltano/meltano