Files
OrbitTv/VERIFICATION-0.40.0.md
pauljdandClaude Opus 5.5 a43d7ec70a
pauljd/davidson-ci: Davidson CI / quality (push) Failing after 1m46s
pauljd/davidson-ci: Davidson CI / ai-review (push) Failing after 1m22s
Record Phase B evidence and roadmap status
Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
2026-11-02 06:29:34 +00:00

2.1 KiB

Orbit TV 0.40.0 verification

Smart Channels Phase B: Terra and After Dark on the same engine as Paradox. Also a guide day jump and a clearer switch.

Changed

  • A film channel is a configuration with a rest period. Films past their rest are picked by a generator seeded from the start time, or the longest-waiting film if none has rested.
  • After Dark (102): Paul's chosen films plus 30 curated titles, exact title matches only, 48-hour rest. Terra (140): up to 8 of 10 curated nature series.
  • Holding left or right in the guide jumps a day. The switch says ON or OFF and ignores a second press within 5 seconds.

Checked on the box, 2 November 2026

Channel Pool Programmes in 7 days Requests Read Plan Stored Most airings Smallest gap
After Dark 33 films, all 30 picks found 101 30 22.6 s 107 ms 17.7 KB 4 48.3 h
Terra 63 episodes, 8 series 184 16 6.9 s 137 ms 31.5 KB 6 29.1 h
  • Film runtimes 84 to 133 minutes, Terra 49 to 59 minutes, all back to back.
  • Schedule files identical after two restarts, with 0 requests. catalogue.db fingerprint unchanged (b630828b...).
  • Film extension from 5.5 to 7 days: 425 ms, every committed film kept, smallest gap still 48.3 h.
  • Damaged After Dark schedule and backup: rebuilt in 232 ms with 0 requests. The aired place survived.
  • Paradox: every programme kept in the plan. Its file grew by 11 bytes at the upgrade, matching a new empty film field. The pre-upgrade copy wasn't kept for a byte comparison.
  • Tuning this morning, 9 changes: 10 to 20 s in total. Vortex, still on its loop, took 15 to 20 s. Last night's tunes took about 2 s. Measurements point to the provider or time of day, not the engine, but the cause isn't proven.

Unit tests

  • assembleDebug, testDebugUnitTest and lintDebug pass; 182 tests, 0 failures, 8 new.

Not checked

  • The held-key day jump on the real remote, film join-in-progress by eye, and a film becoming unavailable on the box (same code path as Paradox).
  • Several days of real running on these two channels.