Fix ATmega32U4 _writeDutyCycle3PWM signature mismatch with hardware_api.h

[94xhn]

Description

_writeDutyCycle3PWM in src/drivers/hardware_specific/atmega/atmega32u4_mcu.cpp had an extra, unused int pinA parameter, giving it a 5-parameter signature:

void _writeDutyCycle3PWM(float dc_a,  float dc_b, float dc_c, int pinA, void* params){

This mismatches the prototype declared in src/drivers/hardware_api.h:

void _writeDutyCycle3PWM(float dc_a,  float dc_b, float dc_c, void* params);

...and the 4-parameter signature used by every other MCU implementation (atmega2560, atmega328, esp32 x2, renesas, rp2040, samd, silabs/efr32, stm32, teensy), the weak fallback in generic_mcu.cpp, and the sole call site in BLDCDriver3PWM.cpp.

Because differing parameter lists make these distinct overloads/symbols in C++, the ATmega32U4-specific _writeDutyCycle3PWM could never actually be called by anything in the codebase — it was permanently dead code, and any call from BLDCDriver3PWM on an ATmega32U4 board silently resolved to the generic weak fallback instead of the MCU-specific override.

pinA is never referenced anywhere in the 5-line function body (just three analogWrite calls), so removing it is a pure signature fix with no behavior change to the body itself — it just makes the function reachable again as the intended per-MCU override.

Fix

Remove the unused int pinA parameter so the function matches hardware_api.h and the pattern used by every other MCU:

void _writeDutyCycle3PWM(float dc_a,  float dc_b, float dc_c, void* params){

Note on impact

Today the ATmega32U4-specific body and the generic weak-fallback body are functionally identical (both just do three analogWrite calls with the same pin mapping), so there's no currently-observable runtime symptom from this bug on existing boards. This fix restores the correct per-MCU override/interface intent and removes unreachable/dead code — it also prevents a silent no-op trap for any future edit to the ATmega32U4-specific body (e.g. to actually use a pinA-style parameter), which today would never take effect due to this signature mismatch.

Testing

Verified by inspecting current dev branch source directly (this repo has no CI build matrix I could run locally for AVR). Confirmed via grep that:

• hardware_api.h, the weak fallback in generic_mcu.cpp, all 10 other MCU implementations, and the single call site in BLDCDriver3PWM.cpp all use the 4-parameter signature.
• No caller anywhere in the codebase passes 5 arguments to _writeDutyCycle3PWM.

Also built an isolated minimal repro (separate from this repo) transplanting the exact buggy 5-param body and the exact 4-param call site, compiled with g++, and confirmed via nm that the 5-param version's mangled symbol never matches what the call site needs — while the fixed 4-param version's symbol matches exactly and links/runs correctly.