[PATCH libinput] evdev: fix device_transform_ functions
benjamin.tissoires at gmail.com
Mon Feb 17 10:42:52 PST 2014
X and Y are li_fixed_t, which is 24.8 fixed point real number.
li_fixed_t max is thus ~8388607.
On a touchscreen with a range of 32767 values (like a 3M sensor), and
mapped on monitor with a resolution of 1920x1080, we currently have:
(x - li_fixed_from_int(device->abs.min_x)) * width == 62912640
which is 7 times bigger than li_fixed_t max.
To keep the precision of the sensor, first compute the uniformized
coordinate (in range 0 .. 1.0) of the touch point, then multiply it
by the screen dimension, and revert it to a li_fixed_t.
Signed-off-by: Benjamin Tissoires <benjamin.tissoires at gmail.com>
I have hit this problem by playing with a touchscreen reporting 4096 values, on
xf86-input-libinput. xf86-input-libinput does not use the real screen size, but
0xffff instead. This allows to report a touchscreen with a range of 128 values
to work properly :(
I went through the multitouch device database, and took one example of a more
real use-case (xf86-input-libinput is still in an early shape).
src/evdev.c | 6 ++++--
1 file changed, 4 insertions(+), 2 deletions(-)
diff --git a/src/evdev.c b/src/evdev.c
index d8dff65..0d033b8 100644
@@ -91,8 +91,9 @@ evdev_device_transform_x(struct evdev_device *device,
- return (x - li_fixed_from_int(device->abs.min_x)) * width /
+ double x_scaled = (li_fixed_to_double(x) - device->abs.min_x) /
(device->abs.max_x - device->abs.min_x + 1);
+ return li_fixed_from_double(x_scaled * width);
@@ -100,8 +101,9 @@ evdev_device_transform_y(struct evdev_device *device,
- return (y - li_fixed_from_int(device->abs.min_y)) * height /
+ double y_scaled = (li_fixed_to_double(y) - device->abs.min_y) /
(device->abs.max_y - device->abs.min_y + 1);
+ return li_fixed_from_double(y_scaled * height);
More information about the wayland-devel