Improve CI checks in pull requests

[eeberhard]

• Skip build-test jobs if only documentation changed or in case of duplicate commits

• Check contributions by looking for a diff between the head and base of a PR in VERSION and CHANGELOG files

---

Isn't it annoying when you forget to update the CHANGELOG? And then you have to wait for the CI to rebuild just because of that one line changed in the .md file that has no effect on the pipeline? Or, how about forgetting to update the CHANGELOG or VERSION in the first place? Lots of room for minor inconveniences.

This PR adds a check to ensure the CHANGELOG and VERSION have both been modified. It also uses a special action to check if a CI job is skippable - that is, if only a certain type of file (.md) has been changed.

[domire8]

Does that work? Imagine the following: You have several commits changing the source and the VERSION, then you open the PR, it fails because you didn't change the CHANGELOG. Then you change the CHANGELOG and it still doesn't run because you only changed an .md file, no? Or do I miss a detail?
I'm not against the idea, but not running the CI if the CHANGELOG is not changed is maybe a bit too much, not running it when you only change md files might be enough no?

[eeberhard]

not running the CI if the CHANGELOG is not changed

The CI build-test jobs still run even if the "Check if changelog and version have been updated" check fails. So in the first instance where you make your code changes but not the changelog, you would still pass 5/6 checks. Then you update the changelog, pass the 6th test, and the other 5 would still count as passed. In theory, if all the checks have passed and you just change the changelog, it won't trigger a rebuild but still passes.

I will try that now by making a single change to the .md file.

[eeberhard]

If you look at the test result history for commits 66c1496, b2656b2 and fc0f78f, you'll see the desired behaviour in action

[domire8]

I see now, thats very cool thanks

[eeberhard]

@buschbapti do you agree with this approach to help make contributions more standardised?

[buschbapti]

Yes sorry for the delay. I like it a lot and thanks for the detailed explanation yesterday.