Workflow dispatch vs using the scripts directly

[joao-paulo-parity]

We can envision at least two ways for how the scripts can be used on GitHub

• Use them directly in other repositories' workflows, i.e. git clone --depth=1 releng-scripts; ./releng-scripts/foo ...
• Host the scripts in a specific repository and trigger them via workflow_dispatch, i.e. they'd not be used directly

With the workflow_dispatch approach we'd be able to set up secrets in a single repository, which would be dispatch <ed to from other repositories. Otherwise, if we use the scripts directly we'll need to set up secrets on each repository that they are used, which seems worse.

The workflow_dispatch approach's downside is that we can't transfer files from one repository's workflow to another via workflow_dispatch, so we'd have to

1. Upload the artifacts to GitHub in repository A (actions/upload-artifact@v3)
2. Trigger workflow_dispatch to repository B
3. Download the artifacts in the dispatched workflow from repository B

Using the scripts directly seems more convenient in this regard since we can use the files directly, without having to download them after workflow_dispatch.

cc @chevdor

[chevdor]

Using workflow dispatch is where we want to get at but we are missing a few things before we can cleanly get there.
This is why we started building the scripts you can find at https://github.com/paritytech/release-engineering/tree/master/scripts

The idea of there is:

• get the lib in a state where it can handle all current releases properly. I think we are now very close to a usable state
• extract the lib as a "package" (install script, docker image, etc...) that can easily be installed
• Package the above in a GHA so we it is easier to fix issues in one place for all repos
• Use the new GHA

[joao-paulo-parity]

@chevdor regarding this point specifically

> - Package the above in a GHA so we it is easier to fix issues in one place for all repos

I'm not understanding why we'd need to wrap the script in a GHA when executing it is as simply as git clone repository && ./repository/script. That approach also has the "fix issues in one place for all repos" aspect. I'm failing to understand why we need to put the scripts in a GHA in order to use them, since they can be used directly. Could you please explain this?

EDIT: I see that this was answered in https://github.com/paritytech/release-engineering/issues/102#issuecomment-1308414160, thanks