Emul backend API

[tristan-google]

For initial feedback

[yperess]

Good start, we need to have generic test that will replace this one and the ICM tests

[yperess]

This doesn't always work, if shift is negative some architectures will do strange things, you need to check the sign and shift.

CLAMP(
    shift < 0 ? (value >> -shift) : (value << shift),
    AKM09918C_MAGN_MIN_MICRO_GAUSS,
    AKM09918C_MAGN_MAX_MICRO_GAUSS
);

[yperess]

I didn't notice this last time, but you're using uint8_t for the shift value, it's not unsigned. I think that's where the confusion is.

[yperess]

shift < 0 ? ((int64_t)q >> -shift) : ((int64_t)q << shift)

[yperess]

This doesn't seem right, if the backend takes a q31_t and a shift, we should expect the result to be the same, so I would expect this assert to be:

zassert_equal(0, shift); // Since the call emul_sensor_backend_set_channel used a 0 shift
zassert_equal(expected_x, q);

[yperess]

I don't think we guarantee the order so another driver might not return the order you asked it in. Same here, if you change

iodev_akm09918c_channels[0] = SENSOR_CHAN_MAGN_Z;
iodev_akm09918c_channels[1] = SENSOR_CHAN_MAGN_Y;
iodev_akm09918c_channels[2] = SENSOR_CHAN_MAGN_X;

You'll still get X, Y, Z order. Its the natural register order that dictates the result order.

[yperess]

I didn't notice this last time, but you're using uint8_t for the shift value, it's not unsigned. I think that's where the confusion is.

[yperess]

arithmetic_shift_right doesn't work here, you'll need to check if shift is negative and basically shift left if positive, shift right if negative. Since this is a test, you're welcome to use the zDSP library for this (the scale function will handle this for you) or you can manually do the check shift < 0 ? (value >> -shift) : (value << shift)

[yperess]

The final step once this is finished is to figure out how to move this test out of the akm09918c specific test and run it on all sensors in an agnostic way. You can enable sensor info with Kconfig and that will allow you to iterate over the sensors, the one bit that's missing is that we also need a way to map a const struct device* to its corresponding const struct emul*. It's a bit costly, but I think it's fine since it's a test, each struct emul contains a pointer to the device it's emulating, so you can just go through the emul iterable section and find the emulator with emul->dev == dev.