-
-
Notifications
You must be signed in to change notification settings - Fork 5.8k
Stop auto labelers because there are too many inacculate behaviours to make confusing #27439
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
Conversation
If our intend is to capture PR intend in the labels, then yes I agree, the labeler is unsuited. |
Labeler is only suitable for "subsystem" style labels, like |
Overall I think the labeler is more beneficial than it hurts. Take a look at recent PRs, they do have good labels. Maybe we should disable the |
i.e. The auto labelers for #23894 is totally wrong and annoying. And I also encounter many times like this. |
It correctly labeled what we told it to. Big refactors like that are likely to trigger many labels. I think we should probably just remove these labels from the config: |
It's a kind of laziness to let a bot process the unclear labels. I do not think the labeler is really useful at the moment. The mergers should should pay enough attention to the PR itself (title, description, code change) and the labels, instead of just clicking the "merge" button. And the label system is not well-designed, for example: what is When there is a problem, it's not right to say "oh, just add a bot to deal with it, nobody would really spend time on it". |
If anyone really spent time on the label system, really ever maintained the labels, and then introduce the labeler carefully, then I could agree with them. |
I can guarantee that if we remove the labeler, many PRs will merge unlabeled because many do not care about those labels in reality. I'm fine with merging unlabeled PRs too, I don't need them. But the changelog tool might. Another option might be to require a prefix on commit message to categorize the PRs and remove most if not all of the labels. |
Default/no-op is better than wrong.
I can understand such requirement. IMO the solution for such "label problem" should start from the changelog tool. Define which kinds of labels are really needed and necessary first, maintain the current labels, remove dead ones, clarify how to use the labels correctly, then fine tune the labeler bot. |
That's somewhat related. One of the problems is that the "changelog" rules are not clear. For example, why there is a separate "API" section? A change should be either a feature/enhancement/bugfix/.../misc. If a PR does a bugfix(enhancment) for both web and api, why it should appear in the API section? The historical unclear rules are the root problem. |
I wouldn't mind if we just set a optional single category label manually like feature,enhancement,fix,refactor. If we retain subsystem labels like |
I agree that we need to rework our label system before we start applying labels automatically. I would suggest:
I think that would make labels more logical and useful for everyone. And the section part is the only one we can automatically apply according to the files changed. |
Alternative PR: #27502 |
Alternative to #27439. Removes a few spammy labels, and disables `sync-labels` which make it never remove labels (which is default behaviour).
#27502 was merged, pending further adjustments. |
It's very difficult to get to know the aim of PR creator according changed files. A file maybe changed project files but the PR maybe a bugfix/a refactor/a comment added and etc.
Revert #26962