Files
LlamaCasty/HANDOFF.md
T
gramps 72cb3ad55a feat(stream): retry a dropped push instead of ending the session (TASK 32·1)
A dead encoder used to run the whole stop path, so a one-second blip ended the broadcast. Now a dropped
push is relaunched against the same ingest while the broadcast is still alive.

- `Services/StreamRetryPolicy.cs` — immediate, +1 s, +2 s, +3 s. Minimum waits on purpose (creator
  ruling): YouTube's grace is short, so every second not pushing is grace spent.
- `Services/StreamRecovery.cs` — the loop, with the audio pipe recreated before every attempt (its
  server is single-instance and unusable after the old client dies, so a naive relaunch fails on a
  missing input). After a push comes up it asks the API whether the broadcast survived: "complete"
  ends recovery as `BroadcastComplete` instead of showing a live sign over a dead stream; an API
  hiccup counts as recovered rather than ending a healthy session.
- `FramePump.EncoderFailed` — a dead push is recoverable; a compositor fault still goes through
  `Failed` and ends the session, because retrying it would loop.
- `MainViewModel.Recovery.cs` — wires the steps, reports each episode into the left panel alert stack
  (one row per episode, updated in place, log range painted), and cancels on Stop.
- Alert stack polish: a rule and a gap now separate the stack from the scene layers (they read as one
  list before), and the licence indicator uses the creator's wording — `STEAM ID NOT PRESENT` /
  `Steam ID Limit: 30fps` / `Click for More Info`, becoming `STEAM ID PRESENT` / `60fps` with a close
  control once owned.

Tests: `StreamRecoveryTests` (schedule, recover on attempt 2 with the pipe restarted each time,
complete-broadcast honesty, unknown liveness, give-up, cancel, audio-restart fault),
`FramePumpTests` now pin that process death raises `EncoderFailed` and not `Failed`.
FAST gate: 538 passed (one documented pre-existing audio failure), 0 warnings.
2026-11-05 12:09:38 -08:00

2.0 KiB

HANDOFF — 2026-11-05 (alert stack + media/chroma fixes)

State

  • main. TASK 21 (video layer + soundtrack), TASK 64 (green/blue screen) and TASK 32·4 (fragmented MP4) are complete; TASK 49 is retired (YouTube owns chat moderation — we just don't paint tombstones).
  • Left panel alert stack shipped: Steam licence indicator pinned at row 1 (Entitlement 30 fps · Maximum 60 fps, pin → close once owned, permanent while owned), event rows LIFO beneath, click → About or the Logs tab with that event's lines painted. Detail: docs/ai/ui-surface.md.
  • Video layer fixes this session: the preview-pane trigger whitelist (a video layer was invisible), a quadratic frame copy that capped 1080p playback at 2 fps, and deadline pacing for video and audio.
  • Tests: 575 green, 1m50s; FAST=1 ./scripts/verify.sh = 527 in ~62s.

Next action

  • TASK 32·1 shipped (blip retry, immediate/+1/+2/+3 s, alert-stack episodes, complete-broadcast check). Next: 32·5 (ingest pre-flight), then 32·2/32·3 (countdown + Back on air) which need YOUR measurement take: pull the network on a private stream and time how long YouTube holds the broadcast.
  • Open loop: 32·4 is proven by construction only — kill ffmpeg mid-recording and play the file.
  • Do not push absent a checkpoint request.

Landmines

  • The preview pane draws frame-source layers from a whitelist of DataTriggers. A layer type with no trigger is invisible in the preview while its decode looks healthy (alert box 2026-09-26, scoreboard, video layer 2026-11-05). Add the trigger with the layer type; pin it with MediaPreviewDisplayTests.
  • Nothing in the suite drives the real cursor any more; the drag arithmetic is pure (DragReorderMathTests).
  • Integration tests constructing MainWindow must set a temporary LayoutPathOverride and clear it in finally.
  • Media sessions are keyed by path + look (MainViewModel.MediaSessionKey); never acquire by raw path.
  • Bundled FFmpeg stays outside the single-file exe (ExcludeFromSingleFile="true").