TwoWire::read_from and TwoWire::write_to millis rollover bug

[delta-G]

The code in bot TwoWire::read_from and TwoWire::write_to includes these lines:

 timeout_ms = millis() + timeout_ms;
 while(millis() < timeout_ms && bus_status == WIRE_STATUS_UNSET && err == FSP_SUCCESS) {

 }

which is not safe for millis rollover. In the rare occurrence that read_from or write_to is called within timeout_ms of rollover, the code will immediately timeout. The AVR code for this is written using subtraction as is preached daily on the forum.

  startMicros = micros();
  while(TWI_MRX == twi_state){
    if((twi_timeout_us > 0ul) && ((micros() - startMicros) > twi_timeout_us)) {
      twi_handleTimeout(twi_do_reset_on_timeout);
      return 0;
    }
  }

I propose fixing the code to use subtraction like the AVR code.

[JM-FRANCE]

twi_timeout_us is a uint32_tso it's always positive - you only want to check if it's not null

  startMicros = micros();
  while(TWI_MRX == twi_state){
    if((twi_timeout_us != 0ul) && ((micros() - startMicros) > twi_timeout_us)) {
      twi_handleTimeout(twi_do_reset_on_timeout);
      return 0;
    }
  }

I guess this is for efficiency reasons, avoid the call to micros() and the subtraction and compare if it's null...

maybe the compiler is smart enough to only check for 0 rather than the sign of an unsigned variable.