feat: add UF2 uploader for CDC-touch bootloaders

[cumin777]

Summary

Add uf2upload, a self-contained C++17 host uploader for STM32 boards using a USB CDC 1200-bps touch and UF2 mass-storage bootloader.

This PR implements the following features:

• Convert a firmware .bin image to UF2 with board-supplied address, family ID and maximum-size validation
• Detect an already-mounted UF2 volume, or perform a CDC 1200-bps touch and wait for the bootloader volume before copying the UF2 image
• Add Windows, Linux x86_64 and Linux ARM64/AArch64 tool artifacts
• Breaking changes

Arduino_Core_STM32 can describe a UF2 upload recipe, but STM32Tools currently has no uploader that can perform the complete CDC-touch, UF2 conversion and mass-storage-copy flow. This tool provides that missing user-facing upload path without requiring Python or shell dependencies at upload time.

On Windows, uf2upload tolerates the transient serial error caused by the expected reset after the 1200-bps touch and continues waiting for the UF2 volume.

The Linux ARM64 artifact is cross-compiled from the same source with the AArch64 GNU toolchain and is placed at linux/aarch64/uf2upload. The build scripts and Dockerfile also include this artifact in the ARM64 tool build.

Validation

Validated on a Seeed XIAO STM32C5 running the TinyUF2 bootloader:

• Upload when the XIAOC5BOOT volume is already mounted
• Full automatic flow from a running application: CDC 1200-bps touch, bootloader entry, UF2 conversion and copy
• Windows serial-reset status reporting during the device reset transition

The Linux x86_64 host build was compiled and functionally exercised under Ubuntu 24.04 on WSL2. The Linux ARM64 binary was cross-compiled on x86_64 and executed under QEMU user-mode emulation for command-line, UF2 generation, simulated volume discovery and file-copy checks. Physical ARM64 USB CDC and TinyUF2 hardware upload have not been validated.

Packaging notes

The Windows and Linux x86_64 artifacts are included, together with the Linux ARM64/AArch64 artifact. macOS build integration is included, but a macOS artifact and hardware validation are not part of this PR.

Arduino_Core_STM32 follow-up PR #3061 consumes this executable from STM32Tools. After this PR merges, a corresponding STM32Tools release/package-index update is still required before end users can install the tool through Board Manager.

CI and AStyle checks are pending GitHub Actions.

[cumin777]

Arduino_Core_STM32 follow-up: generic UF2 upload recipe is now split into core PR #3061, and XIAO STM32C5 board support is #3048. The core recipe invokes the uf2upload executable from this PR; after this tools PR merges, a STM32Tools release/package-index update will still be needed before end users can install it through Board Manager. End-to-end C5 CDC 1200-bps upload also depends on core USB HAL v2 PR #3059.

[fpistm]

Thanks for the PR.
Anyway it seems some host target are missing ? arm ?

[cumin777]

I have added a Linux ARM64/AArch64 uf2upload artifact at linux/aarch64/uf2upload, built from the same source with the AArch64 GNU toolchain.

The ARM64 artifact was cross-compiled on Ubuntu 24.04 under WSL2 (x86_64), checked as an AArch64 ELF, and executed under QEMU user-mode emulation with an ARM64 sysroot. But physical ARM64 USB CDC and TinyUF2 hardware upload have not been validated yet.

I also completed Linux x86_64 validation on Ubuntu 24.04.3 under WSL2. The following all passed:

• Linux x86_64 compilation with make -C src/uf2upload, without warnings or errors
• ELF and dynamic-library checks (file and ldd)
• --help and invalid-argument handling
• UF2 conversion from a test binary, including 512-byte block size, UF2 magic values, application address 0x08008000, and family ID 0x00C5C5C5
• Simulated XIAOC5BOOT volume discovery and UF2 copy
• Missing input, empty input, size-limit, invalid-parameter, missing-port, and timeout error paths
• Linux termios 1200-bps path using a pseudo-terminal

And the Windows build and upload flow have also been exercised in my regular Windows development environment.

Arduino Core is not required for these host-tool tests. It is only needed when reproducing the complete Arduino compile-and-upload workflow with a real xiao-STM32C5 board.