Workflow: Close Stale Issues#37429
Conversation
|
|
I'd say nack and it's a bad idea in general, see: https://nostalebots.xyz/ For PRs (though more on that below) I would say it's fine but not for issues. For the last point about being the "bad guy", nothing changes. It's just bot getting downvoted for closing a valid issues and making maintainers reopen them over an over again or having the same issue opened again and again instead of just keeping the old one open. What would be needed is a better triage, which this just isn't. Also the description feels LLM made and not being read prior to posting. For instance:
Correct me if I'm wrong but this doesn't even touch pull requests?
It's a straight up lie, as it closes ANY issue without validating if it's valid, junk or relevant. What it does is it shrinks the list down to the issues which someone bothers to post "not stale" on over and over again.
Which needs-feedback label already solves. |
I recently closed about 100 issues with AI assistance, and the false-positive rate was about 25%, but these mostly came from the AI not fully reading the linked PRs or only partially reading the issues. It's an exhaustive job, even for an AI. |
|
On topic: Blanket-closing all stale issues seems too crude of a measure. What I could see it filter the activity on the |
+1 |
|
related old discussion: #27447 looks less people like this mechanism. you can filter isssue which you interesting by options like lables althougth many issue exists. |
|
@sebastianertz can you check to limit this closing to only issues with label |
They are already handled by GiteaBot #37429 (comment) |
|
I think we'll reject this for now. It was tried previously but did not work out and will not work out now. We prefer a accurate list of issues, this is important when searching for open issues about topics. I estimate that > ~80% of issues currently open are valid. |
Why manage stale issues?
In an active project, hundreds of issues accumulate. Many of these are "one-hit wonders": Someone reports a bug but then never responds to follow-up questions. Without automation, these dead issues remain open indefinitely. Stale filters these out so you and your collaborators can focus on what's relevant now.
Nothing is more frustrating for users than an issue remaining open for two years without any response. The Action communicates transparently: "Hey, nothing's happened here in a while. If this is still important, let us know, otherwise we'll clean this up soon." This comes across as more professional than dead silence.
As a maintainer, you have limited energy. A list of ~2500 open issues is mentally taxing (risk of burnout). When the Action automatically marks or closes the "junk," the list shrinks to the actual issues. This gives you the feeling of regaining control of the project.
Sometimes people simply forget they still have an open pull request. The "stale" warning serves as a gentle nudge. Users often write, "Oh sorry, totally forgot, I'll fix it tomorrow!" – and just like that, an old problem is solved.
It's unpleasant for a human to manually close hundreds of issues and tell people, "We're no longer interested in this." When a bot does this according to established rules, it's less personal and more readily accepted by the community as a necessary rule.
How does the process work?
issue/staleand comments on the issue.issue/stalestatus is removed.There are currently ~2500 open issues:
is:issue state:open updated:<@today-8yis:issue state:open updated:<@today-7yis:issue state:open updated:<@today-6yis:issue state:open updated:<@today-5yis:issue state:open updated:<@today-4yis:issue state:open updated:<@today-3yis:issue state:open updated:<@today-2yis:issue state:open updated:<@today-1yis:issue state:open updated:<@today-180dis:issue state:open updated:<@today-120dis:issue state:open updated:<@today-90dis:issue state:open updated:<@today-60dis:issue state:open updated:<@today-30dNote:
If the PR is accepted, the line
debug-only: trueshould be deleted after the first successful dry run.