Consider ID_INPUT_POINTINGSTICK as touchpad (BugFix) #1447
Conversation
khfeng
force-pushed
the
resistive-touchpad
branch
from
28c86a9
to
eda0824
Compare
khfeng
changed the title
Most resistive touchpads don't support ABS coordination nor multitouch. When both capabilities are absent, the touchpad gets tagged as ID_INPUT_MOUSE. Since there's no such device as I2C mouse, if the touchpad is on I2C bus, it will also get tagged as ID_INPUT_POINTINGSTICK. To really distinguish a true pointing stick and a resistive touchpad, the parser needs to find the slibling node created by HID Report Descriptor and ensure it's a "touchpad" (i.e. ABS or MT capability). It's not easy to traverse to a sibling node in the current parser, so simply treat ID_INPUT_POINTINGSTICK as resistive touchpad to resolve the issue.
khfeng
force-pushed
the
resistive-touchpad
branch
from
eda0824
to
7877a9c
Compare
There was a problem hiding this comment.
See below inline comment: it seems introducing this fix would create a regression on devices with a trackpoint, which is currently recognizes as a mouse, but would start being recognized as a touchpad if we landed your change.
Is there a better way to identify the model you are focusing on, to make sure it's detected as a touchpad nonetheless?
| self.assertEqual(self.count(devices, "TOUCHPAD"), 2) | ||
| self.assertEqual(self.count(devices, "CARDREADER"), 1) | ||
| self.assertEqual(self.count(devices, "MOUSE"), 1) | ||
| self.assertEqual(self.count(devices, "MOUSE"), 0) |
There was a problem hiding this comment.
This doesn't seem right.
I went and check what device this udev DB came from, and it is a Dell Latitude 5580 which comes with a trackpoint, similar to the ones on the Lenovo laptops.
I asked one of our colleagues and they provided a sample of a Lenovo unit with such a trackpoint:
4649 P: /devices/pci0000:00/0000:00:1f.4/i2c-0/0-0015/input/input8
4650 L: 0
4651 E: DEVPATH=/devices/pci0000:00/0000:00:1f.4/i2c-0/0-0015/input/input8
4652 E: SUBSYSTEM=input
4653 E: PRODUCT=18/4f3/67/0
4654 E: NAME="Elan TrackPoint"
4655 E: PROP=21
4656 E: EV=7
4657 E: KEY=70000 0 0 0 0
4658 E: REL=3
4659 E: MODALIAS=input:b0018v04F3p0067e0000-e0,1,2,k110,111,112,r0,1,amlsfw
4660 E: USEC_INITIALIZED=3293319
4661 E: ID_INPUT=1
4662 E: ID_INPUT_POINTINGSTICK=1
4663 E: ID_INPUT_MOUSE=1
4664 E: ID_SERIAL=noserial
4665 E: ID_PATH=pci-0000:00:1f.4
4666 E: ID_PATH_TAG=pci-0000_00_1f_4
4667 E: ID_FOR_SEAT=input-pci-0000_00_1f_4
4668 E: TAGS=:seat:
And, as confirmed with them, in this case, when ID_INPUT_POINTINGSTICK=1 and ID_INPUT_MOUSE=1, the TrackPoint should be a mouse, not a touchpad. (otherwise the touchpad test cases would be run twice when running Checkbox on this device, which doesn't really make sense)
There was a problem hiding this comment.
The question is, why should a trackpoint be a mouse?
There was a problem hiding this comment.
Is it possible to add a separate test case for pointingsitck? If that's present then it's okay to have no touchpad?
What makes a pointingstick a mouse? |
Most resistive touchpads don't support ABS coordination nor multitouch. When both capabilities are absent, the touchpad gets tagged as ID_INPUT_MOUSE.
Since there's no such device as I2C mouse, if the touchpad is on I2C bus, it will also get tagged as ID_INPUT_POINTINGSTICK.
To really distinguish a true pointing stick and a resistive touchpad, the parser needs to find the slibling node created by HID Report Descriptor and ensure it's a "touchpad" (i.e. ABS or MT capability). It's not easy to traverse to a sibling node in the current parser, so simply treat ID_INPUT_POINTINGSTICK as resistive touchpad to resolve the issue.
Description
Resolved issues
Documentation
Tests