digitalio, pwmio, keypad, bitbangio: subclasses and arguments taken o… - #11381

Code written by Claude Code, guided and corrected by @peterbay.

The problem

Five defects in the pin and bit-banged bus bindings: three arguments narrowed or compared before being checked, one accepted without being checked at all, and a bus error read as an acknowledgement.

The changes

• drive_mode accepted anything. Anything that was not OPEN_DRAIN silently became PUSH_PULL, so switch_to_output(drive_mode=42) and pin.drive_mode = 42 configured a push-pull output rather than raising.

• PWMOut narrowed duty_cycle before validating it. The constructor put it into a uint16_t and checked afterwards, so duty_cycle=65536 silently became 0 and -1 became 65535. The setter next door already validated the value first.

• keypad's range checks admitted one past the last index. key_count, row_count and column_count were passed to mp_arg_validate_int_range as the maximum, and it is inclusive at both ends. On a two by three matrix, key_number_to_row_column(6) returned (2, 0) and row_column_to_key_number(2, 0) returned 6.

• bitbangio did not range-check its rate. Neither the I2C frequency nor the SPI baudrate went through any validation, and both are divisors when the module works out its bit delay. The first version of this also added a maximum of 500000; that was wrong and only the lower bound remains.

• bitbangio.I2C read a bus error as an acknowledgement. write_byte returns -1 for a bus error, 0 for a NAK and 1 for an ACK, and both the data loop in i2c_write and the address byte in i2c_read tested !write_byte(...), which is true only for the NAK.

Testing

Seeed XIAO nRF52840 Sense, on two builds differing only by these changes. Nothing is wired to the pins.

The bus-error row is not covered: it needs a bus fault to produce, which nothing here can arrange.

No new translatable strings.