Acer Switch 3 SW312-31 Incorrect Screen Rotation

[RoadrunnerWMC]

systemd version the issue has been seen with
240 (Ubuntu), 243-1.fc31 (Fedora)

Used distribution
I tested with two distros:

• Ubuntu 19.04
• Fedora 31 Beta 1.1

Expected behaviour you didn't see
The screen graphics (in my case, the entire GNOME desktop environment) are oriented correctly, and when you rotate the device, the screen graphics correctly rotate to match.

Unexpected behaviour you saw
The screen graphics are always rotated incorrectly. The graphics are 90 degrees counterclockwise if the device is right-side up or upside down, and 90 degrees clockwise if the device is sideways.

Steps to reproduce the problem

• Have an Acer Switch 3 SW312-31
• Boot it into Ubuntu or Fedora from a live USB
• When the desktop appears, the screen orientation is incorrectly detected as described above.

More information
I believe this is a hwdb issue, since I was able to make the graphics rotate completely upside-down by messing with 60-sensor.hwdb. Copying the latest Acer rules from GitHub does not help in my case, though.

Using the instructions from https://unix.stackexchange.com/a/411241, here are the identification parameters for my device:

Driver name: BOSC0200
Vendor: Acer
Product name: Switch SW312-31

Output of cat /sys/`udevadm info -q path -n /dev/iio:device0`/../modalias:

acpi:BOSC0200:BOSC0200:

Output of cat /sys/class/dmi/id/modalias:

dmi:bvnInsydeCorp.:bvrV1.14:bd11/08/2018:svnAcer:pnSwitchSW312-31:pvrV1.14:rvnAPL:rnTahiti_S:rvrV1.14:cvnAcer:ct32:cvrChassisVersion:

Based on all of the above, I think the rule name for this device would be sensor:modalias:acpi:BOSC0200*:dmi:*:svnAcer*:pnSwitchSW312-31:*? I'm not sure exactly where to place the glob asterisks, though (they seem to be pretty inconsistent among the existing rules for Acer products).

I'm happy to open a PR for this once I know what the correct rotation matrix is, but despite some searching, I couldn't figure out how you're supposed to determine it. I could probably swap and negate rows of the identity matrix more or less at random until it seems to work, but is that actually how it's done, or is there a more systematic way?

Thank you for any help you can provide!

[poettering]

/cc @hadess

[hadess]

I could probably swap and negate rows of the identity matrix more or less at random until it seems to work, but is that actually how it's done, or is there a more systematic way?

That's how it's done, as there's only one correct way.

We use the same orientation as Windows:
https://docs.microsoft.com/en-gb/previous-versions/windows/apps/dn440593(v=win.10)

I'll try to get this documented and added to iio-sensor-proxy at some point.

[hadess]

Documentation issues will be tracked in: hadess/iio-sensor-proxy#285

[RoadrunnerWMC]

I figured it would make the most sense to try existing rotation matrices for other Acer products instead of guessing randomly, and the first one I tried worked perfectly. The pull request above uses that.