Repository navigation
Quota-bound wait recovery test fails on main #7435
Description
On main at aec2a1b, four cases of tests/test_quota_bound_wait_recovery.py::test_bound_wait_recovers_original_turn_before_selecting_successor fail (monitor_changed and todo_done, each with file and sqlite). The failure is an assertion on a result with 'ok': False.
Main's Python test workflow run 40371465209 also reports failures for this test (2 test IDs in its failed-job log).
The cause has not been investigated.
Activity
Aboysky commentedon Oct 27, 2026
Thanks for the clear repro. I'd like to take a shot at this one — I'll re-run the four failing parameterizations on current main to confirm they're still red, then dig into why the recovery path returns
ok: falseformonitor_changed/todo_done. I'll follow up here with a diagnosis and a PR.Aboysky commentedon Oct 28, 2026
Follow-up from my earlier comment: the root cause was 155dee1 retiring the unmigrated read fallback — this test still read the source Markdown before initializing canonical authority. #7502 (00f6be3) landed that migration on main: the test now seeds authority via
read_fixture_todos(state)before any CLI read. Verified on main 787f15e (Node 24):pytest tests/test_quota_bound_wait_recovery.py→ 4 passed. I'm closing my PR #7503 as superseded; this issue looks closable as fixed.