Conversation
|
The CI and Require no-mistakes workflow runs for this fork PR are waiting on maintainer approval (action_required). Could a maintainer please approve them? Thanks. |
slee029
changed the title
|
Speaking as Kun's firstmate: stamped waiting-ci. HEAD contract-class: restore — unconfigured path inspected tip vs main. Main already waits for an isolated Treehouse worktree then refuses dirty pooled slots rather than discarding uncommitted work. Tip only: before cleanliness, wait until Treehouse records a live unleased owner for that pool slot ( VISION per-rule:
Waiting on CI/NM green, not author or captain. When green: auto-merge candidate (restore + MATCH + safe). Firstmate-flag no this pass (not otherwise ready while CI pending). Security tip: none. Author workflow-approval ask addressed. |
|
New workflow runs for the updated head are waiting on maintainer approval (action_required). Could a maintainer please approve them? Thanks. |
…re adopting it fm-spawn adopted the pane's first isolated read after treehouse get, but git worktree add creates the slot's .git link first and runs its checkout inside the slot, so a pane reporting its foreground cwd reads the new slot while git status still lists every file not yet written. The spawn then refused the slot as not clean, and a checkout slower than the 60s wait was abandoned mid-write, leaving a partial slot folder behind. A candidate whose checkout is in progress (git's initializing worktree lock, or a pool slot Treehouse's state does not yet list with a live owner) is now treated as a transient, waited out on a separate 600s allowance.
2fff957 to
4d42004
Compare
slee029
changed the title
2 participants
Intent
Rebase of upstream PR #5495 onto current upstream main (30ef650) with its intent unchanged; the only conflicts were in docs/configuration.md, where upstream had restructured the Runtime backend section into subsections, so the PR three Treehouse handoff sentences were placed right after the Treehouse worktree-provider sentence. Original intent: fm-spawn refused a pooled Treehouse slot as "not clean" that read clean seconds later, because it checked cleanliness while Treehouse was still handing the slot off (checkout still being written). The fix waits, for Treehouse pool slots only, until Treehouse records a live owner and no lease before checking cleanliness, allowing up to 600 seconds of checkout settling beyond the ordinary 60-second pane wait; if handoff never completes spawn refuses without claiming the slot or publishing task metadata; a genuinely dirty pooled slot is still left untouched; inspecting pool ownership requires jq even on tmux and spawn refuses promptly when it is missing. Do not hand-create worktrees. No new behavior beyond the original PR; do not expand scope.
What Changed
jqto inspect pool ownership, including on tmux.jq, and document the handoff behavior.Risk Assessment
✅ Low: The change is limited to Treehouse pool-slot handoff detection and its wait allowance, with no substantiated material issue found.
Testing
The focused settle test passed its checkout, lease, never-completes, and missing-jq cases. Initial pool-base cases passed before that command timed out. No live Herdr/Treehouse lab was run, and no reviewer-visible artifacts were captured; the result remains inconclusive.
bin/fm-herdr-lab.shto obtain product-level evidence.bin/fm-herdr-lab.sh.bin/fm-herdr-lab.sh.Pipeline
Updates from git push no-mistakes
✅ **intent** - passed
✅ No issues found.
✅ **Rebase** - passed
✅ No issues found.
✅ **Review** - passed
✅ No issues found.
bin/fm-herdr-lab.shto obtain product-level evidence.bin/fm-herdr-lab.sh.bin/fm-herdr-lab.sh.bash tests/fm-spawn-worktree-settle.test.shbash tests/fm-spawn-pool-base-freshen.test.sh(stopped by the 180-second timeout after its initial cases)✅ **Document** - passed
✅ No issues found.
✅ **Lint** - passed
✅ No issues found.
✅ **Push** - passed
✅ No issues found.