motor_controller: migrate to the new USB device stack

[rosterloh]

Fixes west twister -T applications/motor_controller, which could not build this app at all.

Root cause

CONFIG_USB_DEVICE_STACK now carries select DEPRECATED, and the entire legacy stack — usb_device.c, usb_descriptor.c, usb_transfer.c, plus the SoC's bossa.c — is built from __deprecated APIs. Under twister's -Werror every one is fatal. Nothing in the app caused it; nothing in the app could suppress it.

src/usb.c already carried a complete port to the new stack, dead behind #if defined(CONFIG_USB_DEVICE_STACK_NEXT) with the Kconfig commented out. Enabling it needed one API fix: usbd_register_all_classes() has since gained a fourth blocklist argument.

BOSSA 1200-baud reset, reimplemented

soc/atmel/sam0/common/bossa.c hangs off CDC_ACM_DTE_RATE_CALLBACK_SUPPORT, a legacy-only symbol, so under the new stack BOOTLOADER_BOSSA_DEVICE_NAME is unsatisfiable and the file compiles to nothing. Without a replacement, west flash would need a manual double-tap of reset every time. It is reimplemented in usb.c off USBD_MSG_CDC_ACM_LINE_CODING, using the same magic value and SRAM location as bossa.c.

What was given up to make it fit

The new stack costs ~13 KB more flash and ~3.2 KB more RAM, on a part already at 98.89% flash / 99.51% RAM — 160 bytes of RAM free.

Cut	Saves
kernel/device/i2c shells + INIT_STACKS, THREAD_MONITOR, THREAD_STACK_INFO	~7.5 KB flash
SHELL_HELP, SHELL_HISTORY, SHELL_TAB	help strings + line editing; every command still works
DYNAMIXEL_LOG_LEVEL_DBG	debug log strings — set it locally when debugging the bus
LV_Z_VDB_SIZE 10% → 5%	~3.2 KB RAM; still ~13 display lines per flush

⚠️ Those shell symbols default to y once SHELL is on — deleting the lines is a no-op, they need an explicit =n. That cost me a build to discover and is called out in prj.conf.

	Before (main)	After
FLASH	234,936 B — 98.89%	236,780 B — 99.67% (788 B free)
RAM	32,608 B — 99.51%	31,712 B — 96.78%

RAM is healthier than before; flash is considerably tighter. This app has no room left — anything added from here needs something removed.

Also: USBD_CDC_ACM_LOG_LEVEL must be OFF, not ERR. cdc_acm_uart0 is zephyr,shell-uart, so with SHELL_LOG_BACKEND the driver would log over the transport it is logging about; usbd_cdc_acm.c makes that a hard #warning.

Two pre-existing problems surfaced and fixed

• tests.yaml had no platform_allow, so twister fanned out over ~60 platforms this board-specific app was never meant to build on, reporting their missing-devicetree-node failures as its own. 64 configs → 2.
• debug.conf set CONFIG_DEBUG_OPTIMIZATIONS=y (-Os → -Og), which adds ~56 KB to a 99%-full image. Verified against main: it overflows by 56,040 B there too — this config has never linked on any revision; it only looked green because twister aborted on the USB deprecation first. Dropped, with reasoning recorded: the normal build already compiles -gdwarf-4, so gdb has full symbols at -Os.

Verification

2 test scenarios (102 configurations) selected, 100 configurations filtered
0 failed, 0 errored, with no warnings in 45.60 seconds
2 test configurations were only built  (build_only: true)

Not runtime-verified — no hardware available to this session. Neither USB enumeration nor the reimplemented 1200-baud reset has been exercised on a board. If the reset path is wrong, west flash needs a manual double-tap until it's fixed; recoverable, but worth checking first thing.

🤖 Generated with Claude Code