kernel: retire 0040/0041 — upstream took the userspace fix for the Switch IMU and LED names

[mcfbytes]

Why

Upstream closed the two kernel PRs we sent for the Switch controller names and took the userspace alternatives instead:

Kernel PR (closed)	Maintainer's call	Userspace PR (merged)
Linux-Kernel_MiSTer #96 — keep the " IMU" suffix	"better to follow mainline to be compatible with upstream kernel source"	Main_MiSTer #1307 — also skip "(IMU)"
Linux-Kernel_MiSTer #97 — restore player1..4/home LED names	"we will use mainline version to be compatible with further updates"	Main_MiSTer #1308 — fall back to :green:player-N / :blue:player-5

Both Main_MiSTer PRs merged 2026-09-12 and are in Release 20260912, so our kernel no longer needs to carry the stock names.

What changes

• Deleted 0040-hid-nintendo-imu-name-suffix.patch and 0041-hid-nintendo-stock-led-classdev-names.patch from the shared series, plus the beta and DE25 symlinks to them, and their two series entries. Slots left empty (same convention as 0005).
• Docs/ledgers follow: provenance rows marked RETIRED with the upstream outcome; series header, rt-beta-kernel.md, buildroot-config.md, README and the DE25 series README/audit counts move 40 → 38 shared, 42 → 40 beta entries, 36 → 34 DE25; the two recon records drop the retired file from carried_patches and reduce.py's regenerated outputs are committed; upstream-candidates/README.md records the PR outcomes.
• Nothing else in the series moves. 0032's :combo node and 0035's non-fatal home-LED registration are unaffected (:combo has no mainline name to fall back to).

Verification

• Surviving hid-nintendo stack (0015 0032 0034 0035 0038 0039 0049) re-applied with patch -p1 -F0 on pristine 6.18.51 and, via the beta copies, 7.2.5: 7/7 each, zero fuzz; mainline's "%s (IMU)" and LED_FUNCTION_PLAYER5 names are back in both trees.
• scripts/lint-kernel-patches.sh: PASS, 39 patches in 2 series.
• python3 docs/kernel-recon/reduce.py: 136 records, 0 problems (no orphans).
• Beta series: 40 non-comment entries, every one resolves.
• Reverse-applied 0040/0041 in the existing output/build/linux-6.18.51 and linux-rt-7.2.5 trees: hid-nintendo.c came out byte-identical to the pristine replays above, and hid-nintendo.o cross-compiled W=1 with zero warnings on both (trees restored afterwards).
• Not done: no full kernel relink or board boot. The removed hunks only renamed strings.
• Code review (high) ran on the branch; its findings (stale references in 0049's header and three LED-class comments, README count drift, "not sent" wording in the fork-sync/export docs, missing build evidence, user-facing version floor) are folded in.

After checkout

Buildroot never re-patches an already-extracted tree, so a local make would keep the old names: run make linux-dirclean linux-rt-dirclean first. CI is unaffected.

Note for anyone flashing this

Requires Main_MiSTer 20260912 or newer on the card. On an older Main the Switch IMU node is pooled as a phantom controller and the player/home LEDs stay dark — exactly the gap the two patches closed.

Follow-ups (not in this PR)

1. Bump the sdcard payload pin to release_20260912. release.yml's STOCK_RELEASE_* (mirrored in scripts/fetch-sdcard-payload.sh and docs/ci.md) still stages release_20260907, whose files/MiSTer predates #1307/#1308 — a tag-built sdcard.img would pair this kernel with a Main that only matches the old names. Per docs/renovate.md that pin moves only by a deliberate, reviewed change with its own verification (docs/verification/stock-release-20260907.md §6.2 template), so it is a separate PR. Values already measured for it, from SD-Installer commit cd80db9c0ab64ba38be95071a090a80c367d63cf ("Release 20260912."):

   volumes	release_20260912.7z.001 (83,886,080 B) + .002 (42,660,398 B)
   joined size	126,546,478
   joined MD5	7cec2206e2a1133aa307c541219aa08f
   joined SHA-256	35fcbaca57cd2471b1d353f3dd4bae7c7e67256f8c5c8b2d6d5cbcc78269a7ea
   uboot.img / updateboot	unchanged (e2d46cf9… / 6ff2d50a… — the STOCK_UBOOT_*/STOCK_UPDATEBOOT_* pins hold)
   files/MiSTer	1,166,224 B, 2026-09-12 build; strings shows (IMU), :green:player-1, :blue:player-5

2. 0042 (lightbar :red/:green/:blue) and 0037's BTN_Z scoping are the same class of change and were never sent; given this outcome a Main_MiSTer fallback is the likelier route for 0042.

🤖 Generated with Claude Code

https://claude.ai/code/session_01Su9UEMbDZLx5pQRYgQfVuV

[mcfbytes]

Local build on this branch (WSL2 workstation, 2026-09-12), after make linux-dirclean linux-rt-dirclean:

• make all: exit 0. Buildroot applied 78 patches (38 shared on 6.18.51, 40 series entries on RT 7.2.5), zero FAILED hunks, zero fuzz; 0049 landed at the documented -23 offset on 6.18 and -2 on 7.2.5. Both built hid-nintendo.c files carry mainline's "%s (IMU)" / LED_FUNCTION_PLAYER5 and none of the retired strings. Fresh linux.img, zImage_dtb, zImage_dtb-rt; module trees 6.18.51 + 7.2.5; check-linux-img.sh all assertions passed.
• scripts/ci-tests.sh: PASS (176 ok, 3 skip for the missing stock-binary oracle, 0 fail).
• make sdcard on the new release_20260912 pin: archive verified (size/MD5/SHA-256/CRC), uboot.img/updateboot byte-identical, staged files/MiSTer is the 1,166,224 B 20260912 build with the (IMU) / :green:player-1 / :blue:player-5 fallbacks; check-sdcard.sh all assertions passed (payload inventory matches sdcard-payload.md, 30 entries).
• scripts/test-initramfs.sh (6.18.51, default series) and --kernel rt (7.2.5, beta series): 8/8 PASS each under QEMU.

Not done: no boot on the DE10-Nano rig.