Different V1/V2 behaviour of printChar in two fibers

[martinwork]

With two fibers displaying 0,1,2 and 3,4,5, V1 shows 0,1,2,3,4,5 V2 shows 0,1,2,5

Related: https://github.com/microsoft/pxt-microbit/issues/3870

Hex: test-forever.zip

#include "MicroBit.h"

MicroBit uBit;

int interval = 750;

void loopA()
{
  while (true)
  {
    uBit.display.printChar('0', interval);
    uBit.display.printChar('1', interval);
    uBit.display.printChar('2', interval);
    uBit.sleep(20);
  }
}

void loopB()
{
  while (true)
  {
    uBit.display.printChar('3', interval);
    uBit.display.printChar('4', interval);
    uBit.display.printChar('5', interval);
    uBit.sleep(20);
  }
}

int  main()
{
    uBit.init();

    create_fiber(loopA);
    create_fiber(loopB);

    release_fiber();
}

[martinwork]

Here's the difference...

https://github.com/lancaster-university/microbit-dal/blob/77d679c87f3550ab01f7ee3f5de3c295d8161ab1/source/drivers/MicroBitDisplay.cpp#L447

void MicroBitDisplay::waitForFreeDisplay()
{
    // If there's an ongoing animation, wait for our turn to display.
    while (animationMode != ANIMATION_MODE_NONE && animationMode != ANIMATION_MODE_STOPPED)
        fiber_wait_for_event(MICROBIT_ID_NOTIFY, MICROBIT_DISPLAY_EVT_FREE);
}

https://github.com/lancaster-university/codal-core/blob/master/source/drivers/AnimatedDisplay.cpp#L274

void AnimatedDisplay::waitForFreeDisplay()
{
    // If there's an ongoing animation, wait for our turn to display.
    if (animationMode != ANIMATION_MODE_NONE && animationMode != ANIMATION_MODE_STOPPED)
        fiber_wait_for_event(DEVICE_ID_NOTIFY, DISPLAY_EVT_FREE);
}

This fix was made in DAL: https://github.com/lancaster-university/microbit-dal/commit/77d679c87f3550ab01f7ee3f5de3c295d8161ab1

[martinwork]

Created PR https://github.com/lancaster-university/codal-core/pull/140

[martinwork]

@JohnVidler As noted in the PR, this behaves differently on V1, V2 and the MakeCode simulator.

V2 is definitely wrong, and the PR changes V2 to match V1.

The simulator allows a fiber switch after each display call, but V1 only switches in the implicit sleep at the end of the forever loop. V1 is right. because a program can insert sleeps to get the extra fiber switches, but could not stop them if the extra switches were the default. Consider this example https://makecode.microbit.org/_87wWRrYLM7bg.

[microbit-carlos]

@martinwork once the latest CODAL tag is out let's retest and list here the behaviours of the sim vs v1 vs v2.

[microbit-carlos]

The https://github.com/lancaster-university/codal-core/pull/140 patch didn't fix the issue as expected. @finneyj will look into this before the next CODAL tag is release.

[microbit-carlos]

This issue, to match V1 and V2 behaviour, has been resolved with the https://github.com/lancaster-university/codal-core/pull/140 patch.

However, we have an open question about changing this behaviour to match the MakeCode simulator.

@finneyj will create a new issue to describe the two behaviours.

[microbit-carlos]

Issue https://github.com/lancaster-university/codal-microbit-v2/issues/253 opened summarising the current behaviour and possible options, so we can close this issue.

• https://github.com/lancaster-university/codal-microbit-v2/issues/253