I am confused about the back port process wrt to changelogs. Should we add the entry as unreleased in all versions and only when actually releasing the lowest version one we remove the unreleased entry in the higher version?
What if we we release the higher version first?
Since these are different release lines, should we have more than one release entry?

@thanethomson do you have any thoughts on ^? I've made a mistake in merging a changelog entry from another PR, but I don't think that's an issue @lasarojc is talking about.

My understanding is the following. Let $t$ be the moment the release of minor version v0.y.0 is tagged.

• for any time after $t$ we consider any change to the v0.y.x branch (and also the changelog) independent from main and any future release branch
• for any time before $t$, we act as though v0.y.x and main were still the same branch.

Examples:

• if you fix a bug on main and backport it to v0.y.x before $t$, its changelog should only appear on v0.y.x (not on main)
• if you fix a bug on main and backport it to v0.y.x after $t$, the changelog should appear in both branches

This means that the preparation work for tagging version v0.y.0 is different from that required on v0.y.z (for z > 0), as there are some extra checks we need to do in the backlog because of cases like the first example above.

For 1.x and later we might need to adapt this... I haven't given it yet the proper thought 🙏