[brkosebay]

iPad 5 (J71t) port: three PRs, in release order

1. Add iPad 5 Wi-Fi (J71t) hardware support orchard#6: kernel series, device data, firmware recipe, recovery and runtime
2. Add the iPad 5 Wi-Fi (J71t) adapter and multi-adapter discovery #2: J71t adapter and multi-adapter discovery
3. Build Omarchy A for iPad 5 Wi-Fi (J71t) beside J71b omarchy-a#4: J71t image and release

Kernel source: brkosebay/linux a9/orchard-j71t · Project: brkosebay/omarchy-ipad5

Depends on the hardware PR (it needs a hardware wheel with devices/j71t).

Summary

• Discovery with several adapters.
  • Adapters raise UnsupportedDevice (a ValueError subclass in
    contracts.py) when the connected target is a different model.
  • choose() skips those refusals, reports shared USB errors once and refuses
    if two adapters claim a target.
  • When nothing matches, it says which devices are supported. The J71b adapter
    now raises the refusal for other chips/boards and for a foreign recovery
    product string; its own errors are unchanged.
  • docs/adapter-contract.md documents the rule.
• Commands.
  • orchard devices merges every adapter's observations.
  • orchard mixin picks the profile from the ADT model (matched against the
    hardware descriptors), or takes --profile.
  • resolve --profile still defaults to j71b but accepts only registered
    profiles.
• SmartIO extractor moved to apple/smartio.py. It takes the kext bundle
  as a parameter, and J71b's default is unchanged.
• Video-decoder firmware. apple/d5500.py copies the VXD (DEVA) image out
  of the AppleD5500 kext in the pinned kernelcache. The kext has no
  __firmware section, so the image is found through its pointer record in
  __DATA.__data (name, build, sizes, image pointer), not by offsets. The
  prelinked-kext lookup that SmartIO used moved to apple/prelink.py and is
  shared; SmartIO's behaviour and messages are unchanged. The kernelcache is
  extracted once for both.
• New orchard_cli.j71t (adapter, config, inputs, prepare, mixin):
  • touch.py: validates the N1 boot packet from J71t.mtprops (C1F15,1,
    header/length/checksums) and returns it unmodified. J71t needs no Z2FW
    wrapper.
  • calibration.py: the j71t-adt generator. It copies multi-touch-calibration
    (1024 B), alsCalibration, usb-input-limit-calibration and the Bluetooth
    local-mac-address from the fresh recovery ADT, and rejects unresolved
    syscfg/ placeholders, blank, wrong-size and non-unicast values.
  • inputs.py derives four inputs: the Bluetooth patchram, SmartIO, the N1
    packet and vxd-deva.bin.
  • capture_adt requires iPad6,11 and J71tAP. iPad6,11 also ships as J71s
    (S8000), which discovery refuses (chip 0x8000). The J71t unit we test
    reports CPID 0x8003 and BDID 0x10 in DFU.
• Packaging. pyproject, mypy and typecheck_adapter.py (J71T: Adapter)
  updated; README and NOTICE too.

Testing done

• Unit tests. python -m unittest discover -s tests with the hardware and
  CLI wheels installed: 153 tests OK on Python 3.12 and 3.9 (136 before). One
  skipped test runs only with ORCHARD_TEST_LOCAL_J71T. A synthetic
  kernelcache test covers the VXD record lookup, including no record, two
  records and a missing kext.
• Types. uv tool run --from 'mypy==1.18.2' mypy --config-file pyproject.toml
  reports no issues.
• Pinned hashes. With ORCHARD_TEST_LOCAL_J71T pointing at files from
  the same 20H350 build (its kernelcache, BlueTool, J71t.mtprops), the
  local test reproduces the recipe pins for s8000-smartio.bin,
  vxd-deva.bin, BCM4355C0-FearA-USI.hcd and j71t-n1.bin.
• ADT generator. Run on a J71t boot ADT captured from m1n1, it matched,
  byte for byte, the calibration files the same drivers use on that unit in
  the omarchy-ipad5 port. The
  outputs were discarded; no device data is in the repository.

Not tested

• inputs.ensure() against a real IPSW download (mocked in tests).
• Any device step through Orchard: discovery, recovery boot, ADT capture and
  mixin on a J71t. The kernel drivers themselves passed 13 of 13 device checks
  through the omarchy-ipad5 port's own boot path.

How to test

1. Install the hardware and CLI wheels, then run
   python -m unittest discover -s tests and the mypy command above.
2. With a J71t in DFU, orchard devices, then boot recovery and run
   orchard capture-adt, orchard mixin --adt … --firmware-dir … --output …
   and orchard prepare --recovery ….

Credits

• Touch format and VXD record. The N1 packet, calibration facts and the
  AppleD5500 firmware record layout come from the
  omarchy-ipad5 iPad 5 port,
  which builds on Paul Praschl's (Pauli1Go) HoolockLinux A10 work and
  HoolockLinux.
• Structure. The adapter structure follows the J71b reference adapter.

🤖 Generated with Claude Code

[brkosebay]

Update: J71t charging is validated; no change needed in this PR

The charging fix is kernel-only. It is a new patch in OrchardLinux/orchard#6 (commit 0dd7ae0 on a9/orchard-j71t).

The CLI is unaffected:

• no new firmware inputs;
• the charger still reads only its usb-input-limit-calibration from the device's ADT through the j71t-adt generator.

With the updated hardware wheel installed, the tests still pass: 153 OK and 1 skipped (local inputs only), with mypy clean.

Computers and USB-C chargers now charge at up to 2.4 A on the iPad 5 (validated 2026-10-27 through the omarchy-ipad5 port's boot path).

[henrikklee]

Thank you, Berkay. The shared parts of this PR are merged on main with your authorship: 91fea99 (your prelink.py, smartio.py and d5500.py) and dbf78a7 (your touch packet as a recipe step), plus the recipe steps the iPad 5 firmware uses. The J71t adapter and multi-adapter discovery aren't needed any more, because the CLI now has one adapter driven by device descriptors. Details are in docs/integration/j71t.md (topic 18) on the hardware branch feature/j71t-integration. Closing this one as merged that way.