Under tests/run.sh, tests/test_lane_helpers.sh's case "the row's stamp states clause (c)'s two facts ... TAIL" FAILS, while the same suite run by plain python3 -m pytest (which CI uses) passes. Found 2026-10-06 by lane openRepoTools-1's #170 writer at the base of PR #174 (feat/lane-worktrees-report), reproduced locally on Eagle.
Cause as observed: tests/run.sh places the run under a LONG run root (its per-run temp root beneath the checkout-adjacent paths #169 moved it to), and the row stamp written by the lane tooling is CUT before its dir …; window … tail — the stamp is bounded (the state cell's 240-character cap, or the stamp's own width) and the long path consumes the budget, so the clause (c) facts the case asserts are not there to read. Under plain pytest the shorter default temp root leaves room and the tail survives.
Why it matters: tests/run.sh is THE documented way to run the suite locally (AGENTS.md), so a person who follows the document sees a red case that CI never shows, and a case that depends on the length of the temp root is a case whose answer depends on where it is run. Related: #169 (run.sh's temp roots and process group), #77 (the suite's duration; this is the same run root), #137 (run.sh scoping).
Wants: either the suite's estate for that case uses a short, fixed sandbox path independent of the run root, or the row stamp's bounding is made to keep the clause (c) tail ahead of the directory text (truncate the dir, never the facts) — and a note in tests/run.sh or AGENTS.md until then. Unowned; filed so it is not lost across the account swaps of 2026-10-06. Filed by lane openRepoTools-1 at lane openRepoTools-3's request.
Under
tests/run.sh,tests/test_lane_helpers.sh's case "the row's stamp states clause (c)'s two facts ... TAIL" FAILS, while the same suite run by plainpython3 -m pytest(which CI uses) passes. Found 2026-10-06 by lane openRepoTools-1's #170 writer at the base of PR #174 (feat/lane-worktrees-report), reproduced locally on Eagle.Cause as observed:
tests/run.shplaces the run under a LONG run root (its per-run temp root beneath the checkout-adjacent paths #169 moved it to), and the row stamp written by the lane tooling is CUT before itsdir …; window …tail — the stamp is bounded (the state cell's 240-character cap, or the stamp's own width) and the long path consumes the budget, so the clause (c) facts the case asserts are not there to read. Under plain pytest the shorter default temp root leaves room and the tail survives.Why it matters:
tests/run.shis THE documented way to run the suite locally (AGENTS.md), so a person who follows the document sees a red case that CI never shows, and a case that depends on the length of the temp root is a case whose answer depends on where it is run. Related: #169 (run.sh's temp roots and process group), #77 (the suite's duration; this is the same run root), #137 (run.sh scoping).Wants: either the suite's estate for that case uses a short, fixed sandbox path independent of the run root, or the row stamp's bounding is made to keep the clause (c) tail ahead of the directory text (truncate the dir, never the facts) — and a note in tests/run.sh or AGENTS.md until then. Unowned; filed so it is not lost across the account swaps of 2026-10-06. Filed by lane openRepoTools-1 at lane openRepoTools-3's request.