Is: About the future of this package. Was: Unit testing (and Container Structure Tests)

[simenandre]

This image is used and has been used in hundreds of pipelines (this is what makes it production-ready, IMHO). However, as this has become an Open Source project I think it's time to add some unit testing to it.

I addition we'd have to do some form of integration testing. I'm going to work on this probably this week.

[simenandre]

I've had some time thinking about it.

We could easily test our Dockerfile. However, the real threat is the builder utility. Given that as it stands it mostly battle tested as is, even with the new changes bringing some flexibility to it, it is still a dumb shell script.

We should consider if it's better to decouple the builder utility and make a NPM package who deals with it and (or just) rewrite the script into Javascript. I'd love to write it some other language, but we need Node.js runtime for semantic-release. Probably make sense.

[simenandre]

I've since had some time to see how this package has evolved in our own workspace. Frankly, this only serves a few of our projects. I've had some time to think about why and how we can improve it, to make it more valuable for our own stack.

The why: Just a few months after we released this package, we moved over to Github instead of Bitbucket. Since then, tools like Github Actions has been amazing to us, and the tools available for that is likewise amazing. Generally, we don't need this utility anymore, other than for our legacy repositories. However, we could easily utilize this package for Github Actions same way we did for Bitbucket Pipelines – however, it's too specific. Meaning, it serves a specific use-case (where you need semantic-release and are using GCP to some extent). If you don't, why struggle with this library when there are awesome other tools.

The how: Obviously, we can make this more lightweight. We can improve it by separating the functionality into different parts so that you can use it however you like – but right now, I don't see the point.

Right now, things point towards depreciating it. Our efforts are better spent on improving the tools that are available, such as semantic-release docker image(s) or Github Actions for Semantic Release etc.

I'll leave this comment for a few months to make sure this is the way to go.

[simenandre]

We have decided to depreciate this library. Want to take it over? #7