Ok, I found this merged pull-request which discusses exactly this. It looks like this was intentional.

Hmm...

@ralph Hi Ralph, git-blame points to you here.

I am trying to understand how this is supposed to work in practice.

Let's say I have a product-scenario where I test 2 product-builds. The first one is RHEL, and the second is let's say Ansible Tower.

How should the artifacts list look like for these 2 product-builds?

Thanks!

if the artifacts list is empty, the sha256 id of the product build will be always the same and thus not unique for different products/builds. So using the id to track it back to the original subject won't work (?)

@msrb good point with the no artifacts :) I guess no good solution here, I see 2 options:

1. do not accept empty artifacts
   or
2. define how artifact.id is counted if artifacts array is empty

I would opt for 1. @Ralph what was the use case for empty artifacts?

I see that there is also nvrfield in the schema. I guess it can be used to identify the test subject even if the artifacts list is empty.

@mvadkert @msrb

Actual idea of getting artifact.id value when the artifacts list is empty is that you actually get sha256sum of empty string ('') which would end up in this string:

    "id": "sha256:e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855",

Use case for empty artifacts would be cases where something goes wrong and artifacts are not produced, for example, my build produces artifacts (jars) but is broken and so I cannot really came up with forged artifacts, empty list is the result and product-build object will be part of build.error. That could be example..

I guess nvr would also be an option instead of empty artifacts...