Lane B — Display out (DEV-B1..B6): display-kms sink, HAL affinity, render path, ALSA audio, multiview node

[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: display-out brief · ADR-0044/0045 · acceptance in work-schedule DEV-B items. All work in dedicated worktrees, PRs into feat/foundation-buildout.

• DEV-B1 display-kms sink (drm-rs+gbm, mailbox flip loop, Output::Display fully wired: 5 config matches + CLI build_outputs + SPA) — branch ship/dev-b1-display-kms (starts with the IN_FENCE_FD/OUT_FENCE_PTR spike)
• DEV-B2 HAL scanout affinity (probe/select/placement) — branch ship/dev-b2-scanout-affinity
• DEV-B3 render path (wgpu NV12→XRGB into GBM dmabuf; NV12 direct scanout) — branch ship/dev-b3-render-path
• DEV-B4 ALSA HDMI audio sink (ELD gating, buffer servo) — branch ship/dev-b4-alsa-hdmi
• DEV-B5 multiview node subcommand + systemd/container deployment — branch ship/dev-b5-node-mode
• DEV-B6 node enrollment/pairing + displaynode driver — branch ship/dev-b6-node-enroll

🤖 Generated with Claude Code

[aperim-agent]

Wave 1 started: DEV-B1 (ship/dev-b1-display-kms) implementation underway in a dedicated worktree — fence spike first, then the sink + full Output::Display wiring (config matches + CLI + SPA) in one slice per repo doctrine.

[aperim-agent]

DEV-B1 implementation + adversarial review (APPROVE) complete; PR #73 open. Reviewer independently proved invariant #1/#10 via the wedged-sink test (100k publishes complete with the sink permanently stuck), never-float-fps, NV12 confinement, zero unsafe. Flip-event-only v1 (fences deferred to B3 where they become load-bearing). Two non-blocking notes carried forward (FlipEvent timestamp semantics → B4). CI running.

[aperim-agent]

✅ DEV-B1 merged (PR #73). The display-kms sink + Output::Display are in. DEV-B2 (HAL scanout affinity), B3 (GPU render path), B4 (ALSA audio) are now unblocked.

[aperim-agent]

DEV-B2 (PR #81, scanout affinity) + DEV-B3 (PR #83, render path) implementation + adversarial review complete. B2 review caught a cross-probe DeviceId identity mismatch (would defeat the affinity once wired) — fixed by reusing the render-node DeviceId. B3 review caught a per-frame GEM-handle leak on the NV12-direct path — fixed with open/close parity + 3 new tests. Both judged sound (migration-refusal unreachability holds; wgpu deferral is honest). CI running.

[aperim-agent]

✅ DEV-B2 (PR #81) + DEV-B3 (PR #82) merged. Scanout affinity + the buffer-strategy/NV12-direct render path are in. DEV-B4 (ALSA HDMI audio) is now unblocked; B5 (multiview node) follows B4.

[aperim-agent]

DEV-B4 implementation + adversarial review + fixes complete; PR open. The review empirically proved the skew servo was open-loop (saturation after ~5 h of crystal drift) — fixed with content-frame accounting + a closed-loop drift suite over the real components; teardown bounded; PCM format negotiation + mono upmix added. 2664 workspace tests green. DEV-B5 (multiview node) unblocks on merge.

[aperim-agent]

✅ DEV-B4 merged (PR #110). Display audio is in (closed-loop three-clock servo, ELD gating, format negotiation, bounded teardown). DEV-B5 (multiview node) is now unblocked — the last big Lane B slice before node enrollment (B6).