Re: [RFC] Issues with Zephyr Sensors API and ways to resolve them

Paul Sokolovsky

Hello Anas,

On Thu, 25 Jan 2018 14:16:20 +0000
"Nashif, Anas" <> wrote:

If I remember correctly, the PCM units for humidity are coming from
Linux sysfs.
Yes, the original author of Sensors API, Vlad Dogaru, replied on Github
and pointed that is the case .
However, Linux uses milli-degrees for temperature, but we degrees.

When this interface was originally proposed, it tried to
follow a few things from the Linux world.


In general and for other units, you really need to think about high
accuracy sensors.
We don't have problems with that, allowing for precision of one
millionth of a unit. Of all physical (well, let's say "environmental")
quantities, relative humidity is the least to worry about re:
precision - it's a challenge to find a sensor offering precision
of 0.1 unit (unit being a percent).

I do not think it is terribly bad how the units are
defined as long as they are defined and are consistent. So this is
basically not a bug, but if there is a general agreement that it
should be changed and for a good reason other than avoiding
additional calculation, I am fine with that.
They aren't consistent, that's the problem. It's not a bug indeed, but
an inconsistency, which can only become more glaring going forward.
Even now, few have few confused samples and even one driver which mix
up milli-percents and percents.

So far, nobody seems to object to the change, and it seems that's last
reasonable chance to do it, to allow one release of "stability" before
the expected 1.12 LTS.

Thanks for the comments. Any thoughts on the 2 other points?

2. By extension, we may want to re-review unit of
measurements used for other values too.
3. Definition of struct sensor_value is vague, leading to
Best Regards,
Paul | Open source software for ARM SoCs
Follow Linaro:!/linaroorg -

Join to automatically receive all group messages.