auto version skipRelease behaviour

[Yeti-or]

I'm trying to release feature branch;

I have several pr-s in these branch, some of them are minors, some patches and some ci commits that should be skipped;
I merged itб but auto decided to skip my release

Please help me understand this behaviour

My labels:

prLabels [
  [ 'plasma-infra', 'skip-release' ],
  [ 'skip-release' ],
  [ 'skip-release' ],
  [ 'skip-release' ],
  [ 'skip-release' ],
  [ 'minor' ],
  [ 'minor' ],
  [ 'plasma-infra', 'Hotfix', 'skip-release' ],
  [ 'patch' ],
  [ 'plasma-infra', 'skip-release' ],
  [ 'skip-release' ],
  [ 'skip-release' ],
  [ 'plasma-infra', 'skip-release' ],
  [ 'skip-release' ],
  [ 'skip-release' ],
  [ 'skip-release' ],
  [ 'minor' ],
  [ 'minor' ],
  [ 'skip-release' ]
]

yes the last pr to feature branch is CI related, but I don't understand why we should skip release if only last label is 'skip-release' ?

the code I am asking about:

https://github.com/intuit/auto/blob/main/packages/core/src/semver.ts#L221-L223

  const skipRelease = onlyPublishWithReleaseLabel
    ? !lastMergedCommitLabels.some((label) => releaseLabels.includes(label))
    : lastMergedCommitLabels.some((label) => skipReleaseLabels.includes(label));

I don't understand why lastMergedCommit is more important then all other commits

Environment Information:

"auto" version: v10.36.6
"git"  version: v2.32.1 (Apple Git-133)
"node" version: v14.21.2

Project Information:

✔ Repository:      salute-developers/plasma
✔ Author Name:     Vasiliy Loginevskiy
✔ Author Email:    [email protected]
✖ Current Version: v
✔ Latest Release:  @salutejs/[email protected]

✖ Labels configured on GitHub project (Try running "auto create-labels")

GitHub Token Information:

✔ Token:            [Token starting with 3b3d]
✖ Repo Permission:  read
✔ User:            Eva
✔ API:
✔ Enabled Scopes:   gist, notifications, repo, user, workflow, write:discussion, write:packages
✔ Rate Limit:       4447/5000

Additional context

I use conventional-commits plugin to calc version bump

[hipstersmoothie]

I'm pretty sure the behavior is to skip the release if the head commit has skip-release on it. Without that a release should have gone out

[restfulhead]

I got stumped by this as well. I'm trying to write a simple plugin that filters out commits. If no commits are in scope, the version command should return empty (SEMVER.noVersion).

This seems to be impossible: If I use omitCommit and all commits are filtered out, auto uses the default label (patch) instead of noVersion. So I thought ok then let me use parseCommit and slap a skip-release label on those commits instead. But then I run into the same problem as stated here, where the input like labels: [ [ 'skip-release' ], [ 'skip-release' ], [ 'release-patch' ], [] ], causes the SEMVER.noVersion instead of patch.

[hipstersmoothie]

I haven't been in the code here for awhile. but we may have a default label? I'd be open to a tiny pr to enable your use case

[restfulhead]

@hipstersmoothie I've made my use case work in my filter-by-workspace-path plugin by overwriting the release.getSemverBump function and just returning SEMVER.noVersion if there are no commits. I don't think this really solves @Yeti-or's issue though. However, if you think it's a useful patch, I'd be happy to open a PR with this change.

[hipstersmoothie]

That's creative! Would have never thought to do that. Kinda cool you can completely control that

Will have to think about if the patch makes sense

[restfulhead]

That's creative! Would have never thought to do that. Kinda cool you can completely control that

Yeah, it does feel a bit hacky, but it works, hahah.

Btw, off-topic: I've just created an unrelated issue and PR for improving calcNextVersion, which I solved in a similar way. Would be nice if this could be merged, so that I could remove this workaround.