Repository navigation
lane-end close-out gate: follow-ups deferred from the Opus review of #175 (rows 3-8) #185
Description
Activity
Added after #175's S1 fix (head 2a77d4e: in lane-end's gate
tree)row,gate_named=now comes afterforeign) continue, so a FOREIGN clone under the lane's own root is reported as residue → exit 2, row unwritten; shell case repoCG-13):- Wider effect, fail-closed, uncovered by any case: a tree directly under THIS lane's own root that ANOTHER lane's inventory claims is now also residue and blocks this lane's end (
--inventory-onlyis the way past). Implied by "a foreign row under $gate_root" but wider than the clone case; decide whether a sibling lane's claimed tree under our root should block or be listed. - Refusal wording: the clone appears both as
residueand in the FOREIGN count; the residue line says "neither a tree nor scratch the sweep knows"; the status reads "clean" without naming the unpushed commit.
Lane: openRepoTools-3
- Wider effect, fail-closed, uncovered by any case: a tree directly under THIS lane's own root that ANOTHER lane's inventory claims is now also residue and blocks this lane's end (
Ruling under the close-out direction (lane 3, 2026-10-07): where removing a sentence #174 added in a region #175 also edits would make GitHub's post-squash merge CONFLICTING, #175 keeps #174's paragraph verbatim beside its own (e.g. the
lane-worktreesdocstring's 'SWITCHED OFF' paragraph next to the ADD paragraph) so the parallel gate counts; the stale sentence is a follow-up here, not a landing blocker. Lane: openRepoTools-3
Deferred under the close-out direction of 2026-10-06 from the Opus adversarial review of PR #175 (head ed36551). The two must-fix rows (a foreign row under the lane's own gate root not counted; the cross-estate unreadable-claim block, shared with #174) are fixed in the PRs.
lane-worktrees adds: the loser's take-back removes the winner's tree and branch (no commits lost).--inventory-onlyrow truncates what the gate found (row-phrase cap).--force --inventory-onlyinstead of asserting against it; once the shared fix lands, re-assert the real gate there.--include-scratchwalk is ~17 s per lane-end on Eagle (see lane-worktrees: the dependents walk passes over caches, tool environments and site-packages (filed from #174 Copilot round 2) #176).--yesis switched off until lane-worktrees: review findings deferred from #168/#169 (incl. five --yes data-loss paths found after the two-round cap) #170" sentence (lane-end ~129-130, 621-625, 1906, 2129; README-lanes ~529-530, 550-551; test_lane_helpers.sh ~15759-15770) and add--push-untrackedtoadd's refused-flag list; two repoQ-1--retire … once forcedcases went red on the merged tree and need a look.Lane: openRepoTools-3