Attempt 1: circular DMA half-transfer smoke compile
Failure identity: imported target stm32f103, system
configs/systems/onboarding/stm32f103.yaml, firmware crate
firmware-f103-i2c-demo (thumbv7m-none-eabi). The attempt ran against the
working tree based on da17e8a0d28c683a94dd21feddb392ec85bbaa38; source
crates/firmware-f103-i2c-demo/src/main.rs had SHA-256
993b42602b542cddb4868af3bfe0b063028747023dc9015e8ef6360c7a260d2d.
The exact command was:
PATH=/usr/bin:/home/andrii/.cargo/bin:/usr/local/bin:/bin CARGO_BUILD_JOBS=1 CARGO_INCREMENTAL=0 /usr/bin/python3 scripts/run_renode_onboarding_inventory.py --only stm32f103
At 2026-11-18 18:57:50–18:57:59 UTC, the runner returned fail at
build_firmware. No firmware execution or boot evidence was produced. Rust
reported E0428: NVIC_ICPR was defined both at line 59 and the newly added
DMA constants at line 162. The existing module-level pointer already serves
this use; the second definition was accidental.
The metrics, summary, and complete build/run stage logs are retained beside this note. The cheapest discriminating test is to remove the duplicate constant and rerun the same one-target command: a successful firmware build eliminates this compile cause, after which the smoke assertions distinguish HTIF/HTIE behavior.
No simulation or hardware conclusion is drawn from this build failure.