Axis Swap and Inversion are deprecated by Canonical's multi-pointer patches, and those X versions only support the Coordinate Transformation Matrix. Is this supported by xinput_calibrator? We're seeing some calibration "drift" with transformed input devices.
E.g.: We (mis-)calibrate the screen repeatedly by clicking 5mm below the crosshair. With non-rotated devices, the calibration keeps constant (± finger size) – after the initial calibration, taps on the crosshair are registered on them. As it should.
If we rotate the device, the calibration starts to drift off – with each iteration, the pointer is more and more off, rapidly going off screen.
Some example calibration results:
Intial Calibration: 0,4096,0,496
- Iteration: 4100,80,4070,80
Second: 4200,200,4200,170
Third: 4400,400,4250,350
Fourth: 4800,800,4600,600
...and so on.
Is there a way to fix/circumvent this (apart from manually resetting the calibration)?
Axis Swap and Inversion are deprecated by Canonical's multi-pointer patches, and those X versions only support the Coordinate Transformation Matrix. Is this supported by xinput_calibrator? We're seeing some calibration "drift" with transformed input devices.
E.g.: We (mis-)calibrate the screen repeatedly by clicking 5mm below the crosshair. With non-rotated devices, the calibration keeps constant (± finger size) – after the initial calibration, taps on the crosshair are registered on them. As it should.
If we rotate the device, the calibration starts to drift off – with each iteration, the pointer is more and more off, rapidly going off screen.
Some example calibration results:
Intial Calibration: 0,4096,0,496
Second: 4200,200,4200,170
Third: 4400,400,4250,350
Fourth: 4800,800,4600,600
...and so on.
Is there a way to fix/circumvent this (apart from manually resetting the calibration)?