[Haseenesss]

What and why

On an iPhone with a GameSir X5 Lite attached, the phone's gyro stopped driving the game a few seconds after the controller connected (LNPANDY/NX2Apple#971, Breath of the Wild 01007EF00011E000). The X5 Lite has no gyroscope, but GameController lists it with a rotation rate, so the host logs controller 'GameSir-X5 Lite_S' connected: player 1 (npad 0), with motion (gyro). motion_apple.h took any report that is not all zeroes as live motion. A fresh controller source wins over the phone's (DM_PHONE_MOTION=auto), so the phone's motion was no longer used.

The reporter's DM_MOTION_LOG=1 log shows the X5 Lite sends the same report every time: 26 controller lines, all acc +0.00 +0.00 -1.04 G, gyro +0.000 +0.003 -0.000 rev/s, mixed with the phone's lines while it is held and moved.

• dm/host/motion_apple.h: a controller becomes a source only after its rotation rate changes from its first report. A real gyroscope's noise changes it within a few reports. A missing one never does. The zero-only reports of Mac: controller motion (Switch Pro Controller gyro over HID) #296 are the same case, so the all-zero check is now part of this rule.
• dm/host/tests/t_motion_apple.m: a constant non-zero rotation gives no sample until it changes. This holds again after the controller is detached and attached again.
• docs/reference/services.md: the entry "A gyroscope advertised but not there (iOS)".

LNPANDY/NX2Apple carries this change as a patch until it lands here (LNPANDY/NX2Apple#975).

Checked

• Tests: make -C dm/host/tests GAME=01007EF00011E000 build/t_motion_apple and its run pass on this branch. python3 -m unittest dm.tools.tests.test_docs passes (11 tests). Without the change, the new constant-rotation cases fail.
• Games run, with title ID and version: none on this Mac. The reporter's iOS log for Breath of the Wild (01007EF00011E000) with the X5 Lite gives the values above. With DM_PHONE_MOTION=1 the phone's gyro works with the X5 Lite attached, which matches the cause. The iOS build with this change has not been run on a device yet.
• Devices and OS: Apple M-series Mac, macOS 27.2 (tests only). Reporter: iPhone with a GameSir X5 Lite over USB.
• regress.py: none. The change only decides whether a GameController controller's motion counts as a source; no game code, shaders or rewriter are involved.

Docs

• The finding is in its game note or its docs/reference/<subsystem>.md entry, and python3 dm/tools/docs_symptoms.py was run (the symptom index)
• A chain tool changed: dmsh_chain.py --header was run and dm/gfx/nvn/dmsh_chain.h is part of the change (no chain change)
• No game files, keys or data extracted from a game in the change or the attachments

🤖 Generated with Claude Code