[srcnert]

Realigns the RAK4631 support in this downstream with the latest state of the
official upstream RAK4631 pull request, so that future merges from
arduino/ArduinoCore-zephyr stay conflict-free.

MCUboot is now part of the normal build

rak4631.build.zephyr_args carries --sysbuild, so the loader always builds
through MCUboot and the --mcuboot flag disappears:

• extra/build.sh derives MCUboot mode from --sysbuild in the board's
  zephyr_args instead of a CLI flag, and the option parsing goes back to
  upstream's [ x$2 == x"--debug" ] form.
• The bootloader image is renamed firmwares/zephyr-<variant>.mcuboot.hex so
  extra/package_core.sh (which globs firmwares/zephyr-${variant}.*) ships it
  with the core. .vscode/flash_loader.jlink, .vscode/tasks.json,
  extra/rak/checkall.py (the loader build dir is build/<variant>/loader
  under sysbuild) and rak-build.yml follow.
• extra/bootstrap.sh installs MCUboot's Python requirements for image signing.

Loader and core sources back to upstream shape

• cores/arduino/main.cpp returns to upstream's unconditional
  SerialUSB.begin(115200); the shell-port exception moves into the loader.
• loader/main.c restores upstream's #if CONFIG_SHELL / #elif CONFIG_LOG
  blocks verbatim and adds the RAK case as a separate SHELL_ON_OWN_CDC_PORT
  branch, so the shell stays on its own CDC-ACM port instead of being moved
  onto the sketch's Serial.
• extra/debug.conf goes back to upstream plus the single
  CONFIG_SHELL_LOG_BACKEND=n line; the RAK-specific logging and LoRa shell
  tuning moves out (CONFIG_SIZE_OPTIMIZATIONS lands in the variant .conf).

USB identity

RAK4631 stops borrowing Nordic's 0x1915:0x5300 and uses 0x1209:0x0003
(0x0004 for MCUboot) in boards.txt, the variant .conf, the samples and the
MCUboot fragment.

Tooling and CI

• MCUmgr is built on a macOS runner (its macOS binaries need cgo, which cannot
  be cross-compiled from Linux) and extra/package_tool.sh uses
  shasum/wc -c so it runs on both hosts; dependabot now watches
  tools/mcumgr.
• package_core.yml picks the loader's memory table out of the sysbuild
  output, which contains one table per image.
• .editorconfig-checker.json and license_check.py exclusions follow the
  files touched above.

[claude]

Claude finished @srcnert's task in 10m 7s —— View job

---

Review

• Read CLAUDE.md and gather PR context
• Review build/CI tooling changes (extra/build.sh, package_*.yml, checkall.py, package_tool.sh)
• Review loader + core sources (loader/main.c, cores/arduino/main.cpp, debug.conf)
• Review variant consistency (overlay, .conf, boards.txt, samples, docs)
• Post inline findings + summary

---

This PR realigns the RAK4631 port with the latest state of the upstream RAK4631 PR: MCUboot becomes part of the normal build (rak4631.build.zephyr_args=--sysbuild, --mcuboot gone, bootloader hex renamed to firmwares/zephyr-<variant>.mcuboot.hex), loader/main.c and cores/arduino/main.cpp go back to upstream shape with the RAK shell-on-its-own-CDC-port case added as a separate SHELL_ON_OWN_CDC_PORT branch, the USB identity moves off Nordic's 0x1915:0x5300, and MCUmgr is built on a macOS runner.

The direction is good and the downstream delta genuinely shrinks. I diffed loader/main.c against the upstream tip merged in here (38dbedf) and the remaining delta really is additive and isolated — the 1200-bps touch in loader_usb_msg_cb, the SHELL_ON_OWN_CDC_PORT branch, and the sketch_valid && guard. cores/arduino/main.cpp reverting to the unconditional SerialUSB.begin() is a no-op for RAK, since zephyr,shell-uart = &board_cdc_acm_shell already makes the old guard false. The variant overlay/.conf/boards.txt/samples/MCUboot fragment all agree on 0x1209:0x0003 + 0x0004, endpoint counts still fit the three CDC-ACM ports, .github/dependabot.yml matches what extra/ci_dependabot_config.sh generates (so the package_tool.yml verify step passes), there are no stale --mcuboot or firmwares/mcuboot-* references left, and the double-tap recovery text in variants/rak4631_nrf52840/doc/index.txt now matches modules/rak_double_tap/double_tap.c — that one is resolved.

New findings this round:

• extra/debug.conf — dropping the # Logging block loses CONFIG_LOG_DEFAULT_LEVEL=3, so the variant's =2 (WRN) now applies to --debug builds and every LOG_INF/LOG_DBG in the loader is compiled out, including the new shell: port open (DTR set) message. Fix this →
• extra/rak/checkall.py — the sysbuild /loader path was fixed here and in rak-build.yml, but README.rak.md:142-150 still runs west build -d build/rak4631_nrf52840 -t footprint|ram_report|rom_report, which now fails with an unknown-target error; README.rak.md:276 / extra/rak/lint.py:8-9 and .vscode/launch.json:8 are stale too. Fix this →

Still open from earlier rounds (no code change since, only the rak-main merge):

• extra/debug.conf:32 — CONFIG_LOG_BACKEND_UART_AUTOSTART=y vs the variant's =n; both fragments reach the loader image through EXTRA_CONF_FILE, so the merge order silently decides. Fixing it the same way as the new finding above resolves both at once.
• extra/build.sh:65 — MCUboot mode is only derived inside the boards.txt branch, so ./extra/build.sh rak4631/nrf52840 --sysbuild leaves mcuboot unset, LOADER_DIR points at the sysbuild top level, and -t llext-edk fails after a full two-image build.
• .github/workflows/package_core.yml:229 — the new awk has no END rule (the loader table is dropped if stdout ends inside it) and discriminates MCUboot on the line following the table.
• variants/rak4631_nrf52840/rak4631_nrf52840.conf:26 — 0x1209:0x0003/0x0004 are pid.codes test IDs, and boards.txt keys board detection plus the 1200-bps touch on them.
• .vscode/launch.json:8 — covered in the second inline comment above.

Nothing here blocks the sync in principle; the debug.conf items are the ones I would settle before merge, since ./extra/build.sh rak4631 --debug is not covered by rak-build.yml or rak-checkall and both findings only show up in that configuration.

I did not run a build — no Zephyr workspace in this checkout — so the --debug claims are read off the Kconfig fragments and loader/main.c, not observed.
• branch dev/variant-rak4631-sync