Created attachment 1037467 [details] relatively slow finger rolling movement Description of problem: at the default setting of 50% the pointer movement as controlled by the trackpad is much too sensitive. This makes it very difficult to control the trackpad with any granularity. Decreasing "pointer speed" to lower helps somewhat, but then it takes too many swipes to get the pointer to the opposite end of the screen (see #1230459). Version-Release number of selected component (if applicable): 0.17.0-1.fc22 How reproducible: Very reproducible Steps to Reproduce: 1. Place finger on the trackpad 2. "roll" your finger around in wide circles without actually sliding your finger on the trackpad Actual results: Pointer moves about 1 sq centimeter. Expected results: I would expect that at a setting with this acceleration that the cursor would only move the circle of a few pixels. Additional info: Note, the "pointer speed" setting doesn't seem to affect the actual results.
as I thought, the hysteresis calculation was off. this touchpad has a low axis range (870 across vs the ca 4000 of a synaptics touchpad) so the margin of error was only one device-unit - much too low. give this one a try please: http://koji.fedoraproject.org/koji/taskinfo?taskID=10014362
In this particular test case, I don't notice much change. It still seems too fast in a localized area.
hmm, I wonder if I fixed the wrong thing. I expected that before, leaving the finger (mostly) still on the touchpad would still wobble the pointer around a bit. With this change, the pointer wobbling should stop now. If I replay your recording, there is no pointer movement anymore now. Acceleration otherwise is unaffected, this really applies only to the filtering of small movements. Is that what you tested?
Yes, I tested the same test case as listed in the bug description. Could it be that the above koji build didn't have the fix you intended? I noticed that the rpm changelog referenced another bug: - Cap the minimum acceleration slowdown at 0.3 (#1227796) Also, in the above test case and associated recording, I was doing quite a "vigorous" fingertip roll. I would expect the result would be that it still moves a little, just not as much as it currently does.
install this one please, then check the log for message with "::::", that should print what the hysteresis is set to. http://koji.fedoraproject.org/koji/taskinfo?taskID=10027812
From the xorg log: (EE) libinput bug: :::: Hysteresis is 5/5 Hrm, when I replay it, I still get lots of movement. See the attached screencast. ?
Created attachment 1037894 [details] screencast
you did restart X after installing?
yes :) I also took your patch from the mailing list: http://lists.freedesktop.org/archives/wayland-devel/2015-June/022623.html and patched the 0.17-1 rpm and got the same results as the screencast.
libinput-0.17.0-5.fc22 has been submitted as an update for Fedora 22. https://admin.fedoraproject.org/updates/libinput-0.17.0-5.fc22
This does indeed help the "fidgety"ness tremendously. But the ev recording playback still looks the same as the screencast... ? Why would that recording not match the screencast for you?
Package libinput-0.17.0-5.fc22: * should fix your issue, * was pushed to the Fedora 22 testing repository, * should be available at your local mirror within two days. Update it with: # su -c 'yum update --enablerepo=updates-testing libinput-0.17.0-5.fc22' as soon as you are able to. Please go to the following url: https://admin.fedoraproject.org/updates/FEDORA-2015-10372/libinput-0.17.0-5.fc22 then log in and leave karma (feedback).
are you replaying it through your device or the one created with evemu? if the latter you'll need to do some extra tweaks to get the resolution on the uinput device.
libinput-0.17.0-5.fc22 has been pushed to the Fedora 22 stable repository. If problems still persist, please make note of it in this bug report.
Created attachment 1042421 [details] evemu-play shell output This is shell output log of how I play back the recording that produces the screencast. Maybe I'm not doing it correctly?
weird, I replayed that recording again this time and I can see the pointer still move. I must've done something wrong the first time testing. I wonder if I mixed up test devices. Quick check of your recording: the finger moves around by just over 4mm on the x axis and about 5mm in the y axis. Pressure information ranges from 0 to 63, which essentially only tells us "finger". What this means is that we can't rely on hysteresis or general touchpad data to fix this as an erroneous event, we need to handle this through touchpad accel. There's an upstream bug for touchpad accel, but I'm tackling two other acceleration bugs atm that need to be fixed first. https://bugs.freedesktop.org/show_bug.cgi?id=90735
actually, let's call this one closed, the fidgety part is fixed and the remainder is an acceleration issue tracked in bug 1230459
Sounds good, thanks again.
Wade, can I get you to evemu-record a sequence where you just hold the finger still on the touchpad and attach it here please? The hysteresis is too high for some other device, causing some issues with slow pointer motion so I'm trying to figure out how to reduce it without breaking your device. Thanks.
Created attachment 1119981 [details] holding finger on touchpad Attached is a recording of holding my finger on the touchpad. Thanks again.
Oh, here's the package versions (latest in fc23): libinput-1.1.5-1.fc23.x86_64 xorg-x11-drv-libinput-0.16.0-1.fc23.x86_64
Thanks heaps. If you can, please keep an eye on https://bugs.freedesktop.org/show_bug.cgi?id=93503, the hysteresis will be removed by default (but I'll leave it enabled on your touchpad to avoid jitter).