Scheduler: poll thread dies silently on transport error; timed-out one-off runs late #8

New Issue

2026-07-01T19:22:47-04:00

claude commented

2026-07-01 19:22:47 -04:00

From the review. obdcore/scheduler.py:

An uncaught transport exception in tick() kills the polling thread but leaves _running=True; the GUI keeps showing "Connected" with frozen data, and blocked run_oneoff callers hang the full timeout. Fix: wrap _loop in try/except -> set _running=False, fail pending one-offs with the real error, call an on_error callback; controller sets disconnected + closes link.
A run_oneoff that times out leaves the job queued, so it executes LATER on the shared link — a ghost/duplicate vehicle command (e.g. a clear-codes or action firing unexpectedly). Fix: cancel-on-timeout via a per-job flag test-and-set under lock in _drain_oneoffs/run_oneoff.

From the review. `obdcore/scheduler.py`: 1. An uncaught transport exception in `tick()` kills the polling thread but leaves `_running=True`; the GUI keeps showing "Connected" with frozen data, and blocked `run_oneoff` callers hang the full timeout. Fix: wrap `_loop` in try/except -> set `_running=False`, fail pending one-offs with the real error, call an `on_error` callback; controller sets disconnected + closes link. 2. A `run_oneoff` that times out leaves the job queued, so it executes LATER on the shared link — a **ghost/duplicate vehicle command** (e.g. a clear-codes or action firing unexpectedly). Fix: cancel-on-timeout via a per-job flag test-and-set under lock in `_drain_oneoffs`/`run_oneoff`.

claude added the bug P1 labels 2026-07-01 19:22:47 -04:00

justin referenced this issue from a commit

2026-07-01 19:33:35 -04:00

Fix #8: scheduler survives link death; timed-out one-offs cancelled

justin closed this issue

2026-07-01 19:33:35 -04:00

justin referenced this issue from a commit

2026-07-01 19:39:49 -04:00

Fix #10 + #11: transport hardening + controller resource leaks

Sign in to join this conversation.