LowPower.idle() doesn't work as expected, and the argument to LowPower.idle(millis) has absolutely no effect. about arduinolowpower HOT 5 OPEN

Possible solution:

void ArduinoLowPowerClass::idle(uint32_t delay)
{
  uint32_t ticks = delay;
  while (ticks > 0)
  {
    uint32_t lastMillis = millis();
    idle();
    uint32_t delta = millis() - lastMillis;
    if (delta > ticks) {
      ticks = 0;
    } else {
      ticks -= delta;
    }
  }
}

from arduinolowpower.

Note that it is possible to turn off systick which prevents idle from returning every millisecond:

SysTick->CTRL &= ~SysTick_CTRL_TICKINT_Msk;
LowPower.idle();
SysTick->CTRL |= SysTick_CTRL_TICKINT_Msk;

from arduinolowpower.

Sure, I know.

But what's the interest ? Without SysTick, delay(), millis(), etc. don't work anymore.

If the device is internally driven, say take a measure every so often and record it somehow, putting the controller to sleep is fine, and 50 ms of full speed CPU is an acceptable price to pay, even once per sec.

If you have to handle interrupts and debounce button presses, it's not an option.

from arduinolowpower.

Agree, use case is limited, but sometimes it might be useful. It relates to this issue because it solves the problem of "idle() doesn't work as expected" by preventing it from returning every millisecond.

However, I did some testing testing and looping on idle() every millisecond (ex: 100 loops for 100 milliseconds) takes only a few percent more power than waiting for an interrupt from the RTC with idle(100), so not a lot to be gained.

from arduinolowpower.

The way LowPower.idle() (the version without parameter) works is perfect for me, just as I said in my initial post.
It just needs to be documented.

And yes, said documentation could explain how to disable SysTick, or even how to configure the controller such that it returns to idle state after having handled an interrupt.

One the other hand, LowPower.idle(uint32_t millis) doesn't at all do what it's supposed to, and needs to be fixed.

from arduinolowpower.