Redefine F() as no-op

[MCUdude]

See #11 and #53 for details

[facchinm]

Manually merged as d316aa6 , thanks!

[MCUdude]

So how about the digital_pin_to arrays where pgm_read_byte? We can save some memory (a couple of hundred bytes actually!) and clock cycles by removing all this.

[facchinm]

If you want to target it I'll be glad to merge the PR 😉

[MCUdude]

Sure! But I'll have to do some proper testing first, so give me a few days.

[facchinm]

Agree, thank you!

[matthijskooijman]

I wonder if this is really the cleanest way to fix this. I'm actually a bit surprised that the F macro is defined by ArduinoCore-API, since it is rather target-specific. I would have expected each target to define it as a no-op or functional depending on what the target needs. Instead, it seems there is now AVR-specific code (guarded by an #ifdef __AVR__) in the generic API repo (https://github.com/arduino/ArduinoCore-API/blob/master/api/String.h#L29-L42), which is a bit smelly. In addition to the definition of F(), it seems the API repo also defines AVR-specific handling of FlashStringHelper (e.g. how to print it: https://github.com/arduino/ArduinoCore-API/blob/master/api/Print.cpp#L41-L53

I wonder if this should somehow be done differently? Maybe the Print::print(__FlashStringHelper*) should be declared by API but defined by the core? Or should the core and API communicate using macros somehow (so the API can define default nop versions for cores that do not need this, and leave things undefined for cores that do need it)? Hm.