Fix for off-by-one in get_max_duty()

[codepainters]

Summary

This PR fixes off-by-one issue in TC/TCC get_max_duty() function, for details see #911.

Change verified with Adafruit ItsiBitsy M4 and Sparkfun ATSAMD21G Mini boards.

[jbeaurivage]

Thanks @codepainters. However I would suggest, if we want the functions to keep returning u16s, to use saturating_add() instead of the regular + operator to cover the case where CC == u16::MAX. What do you think?

[codepainters]

Hmm, fair point. Actually it's quite interesting what the chip actually do if 0 is used as a duty cycle value, let me check the datasheet (and perhaps the hardware).

[codepainters]

Finally found a moment to take a look, I've done some experiments with ItsyBits M4.

I've tried setting maximum period (2^16 1) and try various duty cycle values (CC1 register):

• 0 - output constantly low
• 2^16 - 1 - gives 100% - 1 tick (as expected)

So, with maximum period value it is not possible to get 100% PWM.

All that said, I guess that saturating add is a good approach here (the configuration with maximum period is somewhat pathological anyway).

[jbeaurivage]

Sounds good, if you just want to make that change, I can get this merged.

[codepainters]

Just added saturating_add() (and rebased on master branch).

Note: I only added saturating_add() for TC timer, where CC0 register is 16-bit and return type is u16. For TCC, PER is 24-bit and return type is u32, so there's no overflow possibility.