I was thinking we could use a label on the PR, something like 'trivial' or 'no-changelog'. Then the CI test could query the GH API to see if the PR had that label and if so, it wouldn't require that change to be listed.
One other thing: it's usually helpful that PRs are linked to LP bugs, so that might be another good test to run. OTOH, some changes don't need an LP bug. We could probably handle both cases for a 'trivial' label. So if the 'trivial' label is added, then it would be okay to omit the d/changelog entry, or to have a d/changelog entry with no LP: bug link. Without a 'trivial' tag, a d/changelog entry with an LP: link would be required. (Bonus if the LP bug number is chased to ensure it's valid with an ubuntu-image upstream bugtask.)
The one case this wouldn't handle is a 'trivial' tag where we still want a d/changelog entry w/o an LP link, but that is missing from the PR. I think that's fine; we'll get a false-ok on the CI test, but we can request the d/changelog entry in a PR review.
I was thinking we could use a label on the PR, something like 'trivial' or 'no-changelog'. Then the CI test could query the GH API to see if the PR had that label and if so, it wouldn't require that change to be listed.
One other thing: it's usually helpful that PRs are linked to LP bugs, so that might be another good test to run. OTOH, some changes don't need an LP bug. We could probably handle both cases for a 'trivial' label. So if the 'trivial' label is added, then it would be okay to omit the d/changelog entry, or to have a d/changelog entry with no LP: bug link. Without a 'trivial' tag, a d/changelog entry with an LP: link would be required. (Bonus if the LP bug number is chased to ensure it's valid with an ubuntu-image upstream bugtask.)
The one case this wouldn't handle is a 'trivial' tag where we still want a d/changelog entry w/o an LP link, but that is missing from the PR. I think that's fine; we'll get a false-ok on the CI test, but we can request the d/changelog entry in a PR review.