m-19 TASK-321.02: persona registry seeded — 14 captured fixtures

[jvgomg]

Summary

Hardware-capture pass for TASK-321.02. 14 personas committed under packages/device-testing/src/personas/ with per-persona provenance.md linking raw Mac + Linux probe data back to capture sessions. Follows documents/persona-capture-playbook.md (also new in this PR — the canonical workflow doc the captures themselves were directed by).

Personas

Persona	Path	Linux	Notes
ipod-video-5g-iflash-1tb	SCSI	deferred	TERAPOD, 94 MiB MBR firmware gap
ipod-mini-2g-pink	SCSI	deferred	PID 0x1205 shared with mini 1G (documented)
ipod-nano-2g-green	SCSI	deferred	Pre-2006 SysInfo
ipod-nano-3g-black	USB	✓	Boundary device — USB-inquiry begins here
ipod-nano-4g-black	USB	✓	HFS+/APM topology sample
ipod-nano-7g-blue	USB	✓	HFS+/APM sibling, hashAB checksum
ipod-nano-7g-space-gray	USB	deferred	FAT32/MBR Windows-restored sibling
ipod-touch-5g-unsupported	rejection	N/A	iOS, no disk mode
echo-mini	mass-storage	✓	Dual-LUN, both lsblk dumps captured
sony-nw-hd5	rejection	deferred	Original Network Walkman, OpenMG v1.1 + MACLIST DRM
sony-nw-a1000 / a1200 / a3000	rejection	deferred	SonicStage-era HDD Walkmans
sony-nwz-e384	rejection	deferred	Newest Walkman; capability_00.xml authoritative

Linux strategy: 4 topologically-distinct devices fully reconciled; siblings have explicit "Linux capture deferred (shape expected to match X)" markers in their provenance. Re-plug is cheap if a downstream test surfaces a need.

Also in this PR

• documents/persona-capture-playbook.md — agent-directive doc covering full Mac → Linux capture pipeline, raw-probe file layout, expected-* derivation strategies, synthesised rejection cases.
• 3 device profiles in devices/ — sony-walkman-nw-a-series.md, sony-walkman-nw-hd-series.md, sony-walkman-nwz-e380.md. Family-level alongside existing devices/ipod.md / echo-mini.md.
• documents/test-devices.md updated with capture timestamps + persona IDs.

Backlog hygiene

• TASK-321.02 → Done. Full final summary documents over-delivery (14 personas vs originally-planned 3 starters). AC Music video sync support #4 (per-persona smoke tests) explicitly left unchecked — fixtures are informational + future-test inputs; smoke tests deferred to TASK-301..311 implementers.
• TASK-324 (Phase 5) description rewritten to acknowledge what 321.02 over-delivered; remaining scope narrowed to state variants (corrupt-db, populated), firmware variant (rockbox), and 3 synthesised rejection personas.
• TASK-331 created — extend ReadinessLevel with 'unsupported' variant; the current 'unknown' workaround in touch 5G + Sony personas is documented but pending normalisation.
• TASK-332 created — DevicePersona schema v2 (USB descriptor hierarchy, partition LUN field, nullable serial). Blocks TASK-322.05 (FunctionFS daemon).

Findings

• USB PID 0x1205 is shared between iPod mini 1G + 2G per linux-usb.org. packages/devices-ipod/src/tables/usb-ids.ts documents this and intentionally maps to mini_1g with a generic display name; SysInfo cascade resolves precise generation. Earlier provenance characterised this as a "known bug" — corrected. Codebase right; documentation now reflects it.
• Linux sysfs bNumConfigurations = 2 for nano 3G vs Mac ioreg 1. Not a host disagreement: Apple iPods advertise two USB configurations (Mass Storage + iAP) and macOS surfaces only the active one. Folded into TASK-332's schema v2 scope.

Test plan

• bun run typecheck --filter @podkit/device-testing — pass (verified pre-push)
• bun run test:unit --filter @podkit/device-testing — 83 pass / 2 skip / 0 fail (verified pre-push)
• All 14 personas load via the registry; runtime.test.ts personas.size > 0 assertion green
• CI on this PR confirms full workspace builds + tests still pass

Out of scope

• Per-persona smoke tests (deferred, see TASK-321.02 AC Music video sync support #4 + TASK-331 dep)
• Schema v2 migration (TASK-332)
• ReadinessLevel: 'unsupported' implementation (TASK-331)
• 9 deferred Linux captures (extrapolation-justified; re-plug if needed)

🤖 Generated with Claude Code