Pluggable delay method?

[mikepurvis]

I raised this briefly here, but I'd like to file a separate issue— when integrating with an RTOS, it's critical that stm32plus's delay method be a yield rather than just a busy-wait. A few thoughts for how to implement this:

• Supply a function that allows registering an alternative delay implementation.
• Supply a function that allows supplying an alternative MillisecondTimer implementation.
• Have delay call a weakref'd function that the user can supply an alternative implementation of.

Thoughts?

[hellow554]

Maybe use the weak-attribute or weak pragma of gcc?

[mikepurvis]

I'm interested in this. Would you entertain a PR for a weak-attribute based solution?

It's unclear to me if weak linking can work for a static class method like MillisecondTimer, or if it would need to pass through to a separate free function. Either way, the inline is going to be gone, whether effectively or explicitly.

[andysworkshop]

weak could work. The only problem with weak is that when multiple implementations are weak then it's undefined who wins. That shouldn't be an issue here though.

[andysworkshop]

closed by #169