Lane D — Cast output (DEV-D1..D4): HLS CORS, cast driver, SPA flow, hardware validation

[aperim-agent]

CLAIMED — actively being implemented by the Claude devices/display program (started 2026-06-10). Do not pick up slices from this lane; coordination happens in comments here.

Spec: ADR-M011 + managed-devices brief §5. D1 is standalone (benefits all browser consumers) and ships first.

• DEV-D1 CORS on the HLS HTTP endpoints — branch ship/dev-d1-hls-cors
• DEV-D2 cast session actor + cast driver (rust_cast spike → pick) — branch ship/dev-d2-cast-driver
• DEV-D3 SPA cast flow (ad-hoc sheet, save-as-device, latency badge) — branch ship/dev-d3-cast-spa
• DEV-D4 hardware validation across device generations (incl. CASTV2-over-IPv6 probe) — gated on D2/D3

🤖 Generated with Claude Code

[aperim-agent]

Wave 1 started: DEV-D1 (ship/dev-d1-hls-cors) implementation underway in a dedicated worktree. Standalone slice; ships first.

[aperim-agent]

DEV-D1 implementation + security review complete; PR open. Review found 2 majors on the new unauthenticated surface — symlink confinement escape (working repro) and whole-segment RAM buffering — both fixed (canonicalize confinement + ReaderStream streaming, with red-then-green tests). Discovery: no Rust HLS endpoint existed before, so this slice ships the delivery surface + CORS together.

[aperim-agent]

✅ DEV-D1 merged (PR #70). HLS now serves with CORS + Range streaming; browser HLS consumers benefit immediately. DEV-D2 (cast driver) is unblocked once the Devices domain (DEV-A3) lands.

[aperim-agent]

DEV-D2 (cast driver) implementation + adversarial review + fixes complete; PR open. Review caught invisible LOAD failures, app_id-keyed preemption (could fight another sender), and the dead GET_STATUS reconnect — all fixed with hostile-receiver script tests. The rust_cast rejection is evidence-recorded; the hand-rolled CASTV2 channel has bounded framing + timeouts throughout. DEV-D3 (cast SPA) unblocks on merge.

[aperim-agent]

✅ DEV-D2 merged — PR #123, merge commit 92cac4f. Google Cast driver: CASTV2 session actor (TLS 8009, DMR CC1AD845), cast-session store + REST surface (/api/v1/cast-sessions), HLS handoff via cast_media_base, device-poller integration through the CompositePollerFactory, OpenAPI + schema.ts regenerated. All CI green.

▶️ DEV-D3 (cast SPA flow) unblocked — starting now.
Remaining in lane D: D3 (SPA), D4 (hardware validation vs the test Cast device).

[aperim-agent]

DEV-D3 PR opened: #127 — feat(web): cast SPA flow — ad-hoc sheet, session list, save-as-device, help (DEV-D3) (base feat/foundation-buildout, branch ship/dev-d3-cast-spa).

Shipped: ad-hoc cast sheet (target prefill from adopted cast devices + discovery hints; manual host[:port] cross-VLAN escape hatch, bracketed IPv6 first), ephemeral session list with lifecycle badge + honest Tier-D latency badge + stop, save-as-device via the D2 promotion endpoint, and the registered /help/casting guide. Gates: eslint 0 warnings, tsc clean, vitest 52 files/531 tests green, vite build green, Playwright real-browser guard 2/2 (which also caught + fixed the missing Lingui catalog entries that rendered as hash IDs in the built bundle).

Contract gaps reported in the PR (not worked around): CastSessionDoc has no started-at field; session list membership changes have no realtime topic (REST re-poll covers it).

Remaining in lane D: D4 hardware validation (real device, CASTV2-over-IPv6 item included).

[aperim-agent]

Status (verified vs main@e77bbd0): DEV-D1 CORS (#70), DEV-D2 cast driver (#123) merged. Remaining: DEV-D3 (SPA cast flow), DEV-D4 (hardware validation — gated on D3 + the cast fleet). NDI note: the operator supplied the SDK URL (held privately), unblocking the NDI-L lanes separately.

[aperim-agent]

✅ DEV-D3 merged — PR #127, merge commit 53e65c4. Cast SPA flow: ad-hoc cast sheet (device → rendition → cast), ephemeral session list with the honest Tier-D latency badge, save-as-device promotion, manual [IPv6]:8009 escape hatch, /help/casting. Review verdict APPROVE (0 major); 533/533 web tests.

▶️ DEV-D3.1 follow-up starting (review-confirmed D2 contract gaps): cast session started-at timestamp + realtime session-lifecycle events (membership currently only via 15 s re-poll).
Remaining in lane D after that: D4 hardware validation.

[aperim-agent]

DEV-D3.1 PR opened: #134 — #134

Closes the two refinements both DEV-D3 reviews flagged:

• Cast session started-at — stamped at the LOAD-accept moment (first MEDIA_STATUS attributing an active media session to the actor, not REST-accept), Unix-ns from the same AckClock/wall-epoch family the status registry ages against; surfaced on the session list/get REST + OpenAPI + SPA ("started N … ago").
• Realtime session-lifecycle events — cast.session.started/.removed on the lossless devices lane via the existing DeviceBroadcaster (mirrors device adopt/delete), so the SPA refreshes the session list immediately instead of waiting for the 15 s re-poll.

Gates green: cargo test --workspace (406 suites, 0 failed), control+events suites, clippy/fmt clean, web eslint/tsc/vitest (546)/Playwright/build all pass.

🤖 Generated with Claude Code