Repository navigation
CI: 38 open PR heads are queued at action_required, so their PRs show no checks #1631
Description
The workflow runs for the heads below were created but never started: GitHub holds pull_request runs from fork contributors until someone with write access approves them, so the run sits at action_required with no jobs, and the PR's rollup is empty.
gh run view 37790853088 --repo boykopovar/AnyPS5
X feat/libkernel-posix-rename Progress report #1192 · 37790853088
Triggered via pull_request about 50 minutes ago
JOBS
(empty)
gh pr checks 1192 --repo boykopovar/AnyPS5 → no checks reported on the 'feat/libkernel-posix-rename' branch
Why it is worth a line here. no checks reported reads as this PR has not been checked yet, and reviewers act on that: this morning two reviews cited "No CI ran on the current head" (#1192, #1446) and asked for local runs that a contributor cannot produce, while the runs were only waiting for approval. Merges that wait for green CI cannot proceed either, because the CI never starts.
38 open PR heads are in this state (oldest first; waiting since is the oldest unapproved run for that head):
| PR | author | head | waiting since (UTC) | run |
|---|---|---|---|---|
| #1580 | @Zheorgyan | cb7bb927 |
13:57 | 37788536215 |
| #1499 | @yabanana | 6e8058f0 |
13:58 | 37788614411 |
| #1596 | @neuralsin | 1504a831 |
14:03 | 37789321018 |
| #1490 | @cubersu | 98005984 |
14:05 | 37789558178 |
| #1599 | @andrey231 | adba58a8 |
14:06 | 37789744816 |
| #1604 | @Zheorgyan | ee96aec8 |
14:10 | 37790258217 |
| #1591 | @MarcZX100 | ae0bd4c9 |
14:10 | 37790353735 |
| #1592 | @MarcZX100 | 8e7a88d8 |
14:10 | 37790362442 |
| #1593 | @MarcZX100 | a9c88e29 |
14:11 | 37790373720 |
| #1594 | @MarcZX100 | e508a1ab |
14:11 | 37790378477 |
| #1595 | @MarcZX100 | a63e7fef |
14:11 | 37790391917 |
| #1559 | @ctj01 | aaf33add |
14:11 | 37790467942 |
| #1605 | @QRMOE | f5655a9c |
14:11 | 37790496535 |
| #1582 | @Deses | 0c6ce38b |
14:12 | 37790590398 |
| #1459 | @nauroth | b7096477 |
14:14 | 37790800432 |
| #1192 | @ps54all | 76db9fa0 |
14:14 | 37790853322 |
| #1583 | @Deses | 3fece9c6 |
14:17 | 37791310743 |
| #1608 | @neuralsin | 10af4ec2 |
14:23 | 37792064169 |
| #1282 | @Swastik-51 | 515df443 |
14:29 | 37792913325 |
| #1511 | @cubersu | 30d8c82e |
14:30 | 37793013676 |
| #1611 | @neuralsin | 39159bb4 |
14:31 | 37793158085 |
| #1607 | @saarthakpant | 186ebeda |
14:32 | 37793272907 |
| #1547 | @russhil | 54e5719b |
14:34 | 37793519107 |
| #1521 | @silentone12725 | da4df7ec |
14:35 | 37793750633 |
| #1523 | @silentone12725 | 8b94c136 |
14:35 | 37793753672 |
| #1522 | @silentone12725 | 065c7f0e |
14:35 | 37793759228 |
| #1610 | @silentone12725 | 1feb8793 |
14:35 | 37793766007 |
| #1446 | @sktharunbalaji13-cmd | 4da534fe |
14:45 | 37795115339 |
| #969 | @Saibamen | c0111b82 |
14:50 | 37795701848 |
| #1614 | @gdrosa | 6f4939f2 |
14:53 | 37796194793 |
| #1620 | @babyhuey | f63f4ff0 |
15:03 | 37797600461 |
| #1624 | @babyhuey | 4b9eb4ef |
15:11 | 37798671372 |
| #1625 | @shanu337 | 526930d0 |
15:12 | 37798849465 |
| #1508 | @MarcZX100 | 28e9713a |
15:13 | 37798915525 |
| #1626 | @mohdfaour03 | ce5a2136 |
15:13 | 37798924180 |
| #1606 | @mohdfaour03 | 26e99afc |
15:13 | 37799021947 |
| #1627 | @babyhuey | a2de24b9 |
15:15 | 37799246399 |
| #1628 | @babyhuey | ce3c5bf8 |
15:17 | 37799471053 |
37 of the 38 have no completed run for that head at all — nothing has ever executed on them. The exception is #969, which has an earlier run for the same head.
Reproduce / refresh:
gh pr list --state open --limit 400 --json number,author,headRefOid > prs.json
gh run list --limit 500 --json headSha,status,conclusion,event,createdAt,databaseId > runs.json
# a PR whose headRefOid has a run with conclusion == "action_required" is in this listCaveat: that samples the latest 400 PRs and 500 runs, so 38 is a floor rather than a total.
The ask, either way. If the approval gate is deliberate — GitHub's "require approval for first-time contributors", which is a normal choice — then a line in CONTRIBUTING.md or a pinned comment saying "your first workflow run needs a maintainer to approve it; no checks reported means queued, not skipped" would stop 21 contributors misreading a queue as a broken pipeline. If it is not deliberate, the runs above are the backlog, oldest first, and approving them converts "no checks" into an actual Linux and Windows result for PRs that currently have neither.
Raised first on #1192, where both review points were already resolved and the missing check was the only thing left.
Written with AI assistance.
Activity
Deses commentedon Oct 8, 2026
Workaround for fork contributors while a run waits:
build.ymlalso hasworkflow_dispatch, so the same Build can run in your fork on your branch.Actions must be enabled on the fork. If it answers
HTTP 404: workflow build.yml not found on the default branch, the fork lists no workflows yet;gh repo sync <user>/AnyPS5 -b mainfixed that here. The run does not show in the PR's checks, so link it in the PR.Differences from the upstream run: it builds the branch head, not the merge with current
main, so a head cut from amainthat was red at the time fails the same way; and the fork has no ccache (the cache is saved only onpush), so it builds cold: 16-24 min on Linux, 24-27 min on Windows here, against limits of 30 and 45.MarcZX10 commentedon Oct 9, 2026
The manual-build gate that caused
action_requiredis now implemented in merged #2113, and currentmaindocuments the collaborator-triggered run from upstreammainwith a PR number. The old fork command in this issue comment predates that workflow change and should not be treated as the current upstream procedure. This issue appears addressed by #2113; please close it if the new policy resolves the original report.oneandonlydean commentedon Oct 9, 2026
Closing: since #2113, Conventions runs on every PR head through
pull_request_target, and the full build is started by a maintainer, so heads no longer sit ataction_requiredwith no checks.