Stop the stale bot from closing issues - #8076
Open
Scottcjn wants to merge 2 commits into
Open
Conversation
github-actions
Bot
added
BCOS-L1
The link-check job has not checked a link since 22 July. Every run on main fails the same way: error: unexpected argument '--exclude-mail' found Process completed with exit code 2 lychee removed --exclude-mail; mail links are now skipped by default and --include-mail opts back in. The action resolves to a lychee version that rejects the flag, so the binary exits before it reads a single file. This is the failure mode worth naming: the job did not report broken links, it reported that it could not start, and a red check for an unparseable argument looks much like a red check for a genuinely broken link. Three runs on main failed this way and the workflow kept being treated as if it were guarding something. Verified after removing it: a run with the same args minus this flag checks 2,261 links and reports 0 errors. Contributors have since opened several PRs against this bounty proposing new duplicate link-check workflows. None of them needed to. The existing workflow was one flag away from working. Signed-off-by: Scott <scottbphone12@gmail.com>
Contributor
|
| Metric | Value |
|---|---|
| Trust Score | 49/100 |
| Certificate ID | BCOS-da0834ff |
| Tier | L1 (not met) |
What does this mean?
The BCOS (Beacon Certified Open Source) engine scans for:
- SPDX license header compliance
- Known CVE vulnerabilities (OSV database)
- Static analysis findings (Semgrep)
- SBOM completeness
- Dependency freshness
- Test infrastructure evidence
- Review attestation tier
BCOS v2 Engine - Free & Open Source (MIT) - Elyan Labs
It closed #6749 today as not planned. That issue tracks a live cross-path double-credit race in epoch settlement, and the PRs fixing it were open and unreviewed at the time. It closed a security issue on beacon-skill the same morning and had armed five more. Adding labels to the exempt list, which is what I did first, only works if somebody remembers to apply them. Two days ago the same class of failure hit the bounty board: 29 real bounties had no bounty label, so they were invisible to the browse link and unprotected here, and ten were already marked stale while the spam around them was untouched. The wrong assumption is that silence means resolved. For a hard bug the opposite holds. Nobody comments on a consensus race for a month precisely because it is hard, and the reward for being hard was deletion. So issues are now marked and never auto-closed. Marking is information and a human can still close in one click. Closing is destruction and it falls hardest on the work that takes longest. Pull requests still auto-close, on a longer leash, because they genuinely rot against a moving main and reopening one is cheap. Also: stale window 30 to 45 days, milestoned and assigned issues exempt since a human already triaged them, and operations-per-run 50 to 300 because 50 could not cover the backlog so the queue never drained. Signed-off-by: Scott <scottbphone12@gmail.com>
Scottcjn
force-pushed
the
fix/stale-exempt-consensus-tracking
branch
from
July 27, 2026 17:31
b5e2ead to
28b1d46
Compare
Scottcjn
changed the title
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
2 participants
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
What happened
This workflow closed #6749 this morning as
not planned. That issue tracks a live cross-path double-credit race in epoch settlement, and the PRs fixing it (#8062, #8065) were open and unreviewed at the time. It closed a security issue on beacon-skill the same morning, and had armed five more including the production node drift issue and the federation bridge RFC.Why adding labels was not enough
My first attempt at this PR just added
consensus,trackingandoperationsto the exempt list. That only works if somebody remembers to apply the label.Two days ago the same class of failure hit the bounty board: 29 genuine bounty postings had no
bountylabel, which made them invisible to the README browse link and unprotected from that repo's stale bot. Ten were already marked stale. The spam around them was untouched, because spam gets commented on and real work goes quiet.The actual wrong assumption
Silence means resolved. For a hard bug the opposite is true. Nobody comments on a consensus race for a month precisely because it is hard, and the reward for being hard was deletion.
What changed
Issues are marked stale and never auto-closed.
days-before-issue-close: -1. Marking is information and a human can still close in one click. Closing is destruction, and it lands hardest on exactly the work that takes longest to fix.Pull requests still auto-close, on a longer leash (21 days to stale, 14 to close, up from 14/7). PRs genuinely rot against a moving main, the branch is not lost, and reopening is one click.
Also:
operations-per-run50 → 300. At 50 the queue never drained on a repo this size, so the same items were revisited while others were never reached at allNote
The six issues it hit were reopened and pinned by hand today. Two of them (#6624, #6627) turned out to be genuinely complete and were closed properly with pointers to the PRs that did the work — which is what closing should look like.