Repository navigation
ci(v1-gate): let gnome-e2e pin bumps through and flag hand-check holds - #1151
Conversation
Renovate's projectbluefin/actions digest bumps change one `uses:` line in .github/actions/gnome-e2e/action.yml, which the planner maps to the gnome-e2e-action lane the gate cannot boot, so every bump held v1 for a person (#1127, #1140). scripts/v1_gate_pins.py now drops a file under that action from the plan when its whole diff in the range only swaps one full commit SHA for another on the same action. Anything else keeps it in. A `hold` action can only be released by a hand check and a forced run, and it repeats on every push until then (#1149 sat unflagged). It now gets "Needs a maintainer" and needs-human straight away. priority/critical-urgent stays for the infrastructure and 24-hour cases. Assisted-by: Claude Opus 5.5 via Claude Code Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01WcWm7qQa538Zjcuvv8jBtn Signed-off-by: Danathar <Danathar@users.noreply.github.com>
|
Codecov Report❌ Patch coverage is
📢 Thoughts on this report? Let us know! |
15c9691
ℹ️ Sentinel notice — informational; author is trusted (owner), merge is not blockedHive flagged this PR (author @Danathar, head
Hive did not add the |
There was a problem hiding this comment.
Summary: This change lets routine dependency updates pass the release checks and immediately flags changes that need manual validation. One regression remains: a change waiting for manual validation no longer receives the urgent label after a day, making a stalled release easier to overlook.
What I need from you: Keep the age-based urgency check independent of the manual-validation check.
correctness
Medium — Preserve urgency for manual holds older than 24 hours. In .github/workflows/update-v1-tag.yml:302-308, ACTION=hold takes the new branch before the existing issue-age branch. Every subsequent manual hold therefore skips the age check, leaving URGENT empty indefinitely; the urgent label is only added when that variable is set at lines 327-329. Executing this shell step with stubbed GitHub calls reproduced the difference: a 48-hour manual hold gets only needs-human, while a 48-hour failed VM check gets both labels.
Run the age check independently, preserving the more specific manual-validation explanation:
--- a/.github/workflows/update-v1-tag.yml
+++ b/.github/workflows/update-v1-tag.yml
@@
elif [ "${ACTION}" = "hold" ]; then
ESCALATE="the change needs a lane the gate cannot boot, so only a maintainer's hand check and a forced run can release it. No contributor change clears this hold."
- elif [ -n "${HELD}" ]; then
+ fi
+ if [ -z "${URGENT}" ] && [ -n "${HELD}" ]; then
OPENED=$(gh issue view "${HELD}" --json createdAt --jq .createdAt)
if [ $(( $(date +%s) - $(date -d "${OPENED}" +%s) )) -gt 86400 ]; then
- ESCALATE="v1 has been held for more than 24 hours."
+ ESCALATE="${ESCALATE:-v1 has been held for more than 24 hours.}"
URGENT=yes
fiAdd a regression case for ACTION=hold with an issue older than 24 hours. The proposed patch preserved fresh holds, infrastructure escalation, failed checks, and successful moves in the same shell simulation; no live workflow or VM execution was performed.
No findings from: security, intent-alignment, style, docs-currency.
Confidence: 3/5 (needs attention) — 1 medium finding; 1 perspective requested changes
— hive: agent=reviewer backend=copilot model=gpt-6.1-sol copilot=1.0.88
|
@Danathar this PR is back in your court ( Review asks:
Reply here or push a new head to this branch (same sign-off); CI and the review swarm re-run automatically and the PR returns to the normal path. Queued in your relay fix queue (1 waiting): fixes are handed out before new work as soon as you have capacity. |
Review on #1151: the new hand-check branch came before the issue-age check, so a hold that needs a forced run never turned priority/critical-urgent, even after days. The age check now runs on its own and keeps the more specific explanation. tests/unit/test_update_v1_tag_escalation.py runs the real step with a stubbed gh and covers fresh and day-old hand-check holds, failed gates, infrastructure failures and a move. Assisted-by: Claude Opus 5.5 via Claude Code Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01WcWm7qQa538Zjcuvv8jBtn Signed-off-by: Danathar <Danathar@users.noreply.github.com>
|
Merged by hive contributor-lane auto-merge: CI green for ≥30m, reviewer verdict approve, @Danathar was auto-promoted after 5 clean merged PRs in this hive's repositories. |
Summary: two changes to the
v1gate, so it stops waiting on a person when it doesn't need one, and says so clearly when it does.v1. Renovate'sprojectbluefin/actionsdigest bumps change oneuses:line in.github/actions/gnome-e2e/action.yml. The gate can't boot that action, so every bump heldv1until someone forced it (chore(deps): update projectbluefin/actions digest to 4199208 #1127, v1 tag held by the post-merge VM gate #1140). Five of the last eight changes to that file were these bumps.needs-human, because the flag only covered infrastructure failures and holds older than a day.What I need from you:
update-v1-tag.ymlis a human-merge path, so a maintainer merges this.How the pin filter works
scripts/v1_gate_pins.pyruns betweengit diff --name-onlyand the planner..github/actions/gnome-e2e/is dropped only when its whole diff inv1..mainisuses:lines that swap one full 40-character commit SHA for another, on the same action. Comments after the pin may change.e2e.ymland other workflows still boot, as before.Replayed on real history, the last five Renovate bumps (#1127, #1104, #1102, #1095, #1088) drop the action file and boot the other workflow files they touched, instead of holding. #1139 and #1089, real action changes, still hold.
The hold flag
holdaction: the comment starts with Needs a maintainer, and the issue getsneeds-human(and losesgood first issue/help wanted).priority/critical-urgent. The hive's review caught that the first version dropped the 24-hour urgency for hand-check holds; fixed in the second commit.Files
scripts/v1_gate_pins.py: the filter (new)..github/workflows/update-v1-tag.yml: calls it in the plan step, and adds the hold flag.scripts/vm_validation_plan.py: the filter never reaches a VM.tests/unit/test_v1_gate_pins.py: new tests.tests/unit/test_update_v1_tag_escalation.py: runs the real "Move or hold v1" step with a stubbedghand checks the labels (new).docs/skills/ci-ops/ops/references/candidate-vm-validation.md: documents both changes.Tests
tests/unit/test_v1_gate_pins.py: diff cases, plus end-to-end through a scratch git repo. With the escalation tests,test_v1_gate.pyandtest_vm_validation_plan.py, 60 passed locally. PyYAML isn't on my machine, so the escalation test ran with a local stand-in foryaml.safe_load; CI installs the real one.ruff checkandactionlintare clean.test_update_v1_tag_escalation.pycovers fresh and day-old hand-check holds, fresh and day-old failed gates, an infrastructure failure, and a move. Against the first commit, only the day-old hand-check case fails, which is the bug the review found.validate_docs.py, which need PyYAML and behave. CI covers them.🤖 Generated with Claude Code
https://claude.ai/code/session_01WcWm7qQa538Zjcuvv8jBtn