Repository navigation
LAB: nightly train lane publish-dryrun (off the release verdict) #5153
Description
The nightly train's publish-dryrun lane moves from the release verdict to LAB (#5147 PR). It has no last green and no measured duration on main.
- Owner: aprender-w4941 (handed over by aprender-cd, 2026-10-12)
- While LAB, a red or not_measured night updates this ticket and stays silent. The lane never stops a merge or a release.
- After 7 days red or unmeasured, the lane stops running until its owner brings it back.
- It returns to the verdict only with an operator yes, after three green nights on main, with one gate retired for each one added.
- The tag-time step this lane rehearses stays on the release path unchanged.
keep-open: LAB ticket for the lane; closed by its owner when the lane is retired or returned to the verdict.
Activity
fix(release): models gates on crux; ladder and six night rows go to L…
noahgift commentedon Oct 12, 2026
First green night and first measured duration on
mainfor this lane. The body's "no last green and no measured duration" is out of date.measure: publish-dryrunpublish-dryrunmainwas ahead of crates.io:0.71.0against the newest published0.70.3. Before that, the lane skipped by design (release_lanes.shruns the gate only on a bumped tree), so those nights are not_measured, not green.rc_publish_gate.sh --verify, and all three of its parts ran. Log lines:package: 73/73 publishable members packaged (cargo package with verify --locked)andrc_publish_gate: graph=0 pathonly=0 package=0 -> green. That givesrelease_lanes publish-dryrun: green.cascade-publish.sh --checkis a receipt only (release_lanes.sh, the header comment at line 46), so it does not affect this verdict.So the lane has 1 green night and a measured duration of about 31 min. This is data for the owner. It does not change the lane's LAB state; returning it to the verdict still needs the operator's yes.
Agent: aprender-w4941
noahgift commentedon Oct 13, 2026
Nightly 2026-10-13, run 38342519178 @
7f9a15a52b:measure: publish-dryrun= failure (setup, not measured). The lane never ran, sopublish-dryrunwas skipped.The step
git worktree add --detach "$RUNNER_TEMP/lanes-c" "$REF"(.github/workflows/release-lanes-nightly.yml:209) exited 128:'.../_work/_temp/lanes-c' is a missing but already registered worktree; use 'add -f' to override, or 'prune' or 'remove' to clear.Cause: the runner deletes
$RUNNER_TEMPbetween jobs, but the checkout's.git/worktrees/lanes-cregistration from an earlier run survives. Every later run on that runner's checkout fails the same way until it is pruned. The previous night (run 38195895879,cc71bde029) was green.Proposed fix: run
git worktree prunebefore the add, in the same step. It is a workflow edit, so it is not made from this lane and goes to the cop. No other change.Owner: aprender-w4941.
noahgift commentedon Oct 14, 2026
Folded into #4715 as a checklist row (TRIAGE-100, 2026-10-14): the open-ticket cap is 100, so this ticket is now tracked on that master and closed as not planned. Nothing is lost: reopen it when its work resumes, and an open PR for it reopens it.
Agent: aprender-clerk