Touchpad on an XPS 13 9360

[Peter Hutterer]

On Wed, May 03, 2017 at 05:41:34PM -0500, Michael Wisniewski wrote:
> On Tue, May 2, 2017 at 6:08 PM, Peter Hutterer <peter.hutterer at who-t.net>
> wrote:
> 
> > fwiw, this stuff would be better on a public mailing list, so others can
> > learn too.
> 
> 
> OK.  :)  Adding the public list back on here.
> 
> 
> >
> > On Tue, May 02, 2017 at 07:47:31AM -0500, Michael Wisniewski wrote:
> > > Peter,
> > >
> > > I hope you don't mind me asking a couple more questions; you seem to be
> > an
> > > expert on this topic.  :) On my XPS13, for whatever reason, there are two
> > > touchpads; a Synaptic one and a " DLL075B:01 06CB:76AF Touchpad" one.
> > Many
> > > of the articles I've read said that you should disable the Synaptic
> > > touchpad or else bad things happen.  Just for kicks, I tried to disable
> > the
> > >  DLL075B:01 06CB:76AF Touchpad one and the touchpad didn't work at all.
> > > Disable the synaptic one and the touchpad works.
> >
> > I'd like to see those articles, because, well, that's just plain BS. The
> > devices are the same touchpad, the only difference is whether the kernel
> > talks to the device over ps2 ("Synaptics...") or over i2c ("DLL0...").
> > Because of the way modules are loaded in the kernel, you always get the ps2
> > device, then the kernel tries i2c and if it works you get the DLL0 device.
> > But once the touchpad is initialized over i2c, it won't send events over
> > the
> > ps2 event node, that device is simply mute. Disabling it or not has no
> > effects on the events.
> >
> 
> You're right with this, but just having the synaptics (ps2) device and
> disabling the DLL0 device makes the touchpad not work.  Now mind you, I did
> this though xorg configs....I'm wondering what would happen if I
> blacklisted the i2c_hid that drives the i2c touchpad.  Here's the bug
> report that kind of talks about it.
> 
> https://bugs.launchpad.net/ubuntu/+source/linux/+bug/1591669

yeah, that's a synclient/syndaemon bug because they can't handle multple
devices. Same would apply if you ever attached an external touchpad.
some info here:
https://who-t.blogspot.com.au/2017/01/the-definitive-guide-to-synclient.html

i2c is the preferred bus, ps/2 is emulated in the firmware and we had some
fun with that not getting sufficient testing by hardware vendors.

> > > In any case, when I execute 'xinput list-props' while using the Synaptic
> > > driver, there are a ton of properties that can be set.  If I use the
> > > libinput driver, there are nearly not as many options available.  I'm
> > > wondering if maybe part of my problem is that it's not reading all the
> > > available options for my touchpad.  Just as an example, I have "Device
> > > Accel Constant Deceleration (266): 2.500000" when using the Synaptic
> > driver
> > > but not the libinput one.
> >
> > libinput doesn't have the same number of config options, that's expected.
> > a bunch of things that synaptics has options for are set within libinput.
> >
> > https://who-t.blogspot.com.au/2016/04/why-libinput-doesnt-ha
> > ve-lot-of-config.html
> >
> > so yeah, afacit everything works fine here.
> >
> 
> I saw your blog late yesterday and it helped explain a number of things,
> why it doesn't have a lot of config options is one of them.
> 
> I have another laptop that has an ALPS touchpad using libinput and it feels
> wonderful.  Everything about moving the cursor/pointer around on the screen
> seems right.  On this XPS13, it seems like I've been struggling with it for
> a few days now.  I would say that the issue I'm having is with very small
> movements....it's like you move your finger quickly to the area you want to
> click, start to slow down, then move your finger slowly the rest of the way
> to that point.  I have the problem where when you slowly move it around to
> that point, it doesn't react smoothly.
> 
> I read another message by you somewhere else giving instructions on how one
> might be able to submit data that was collected throughout the day to help
> improve libinput.  Would you be open to taking a look at the data if I
> collected it throughout the day?


there's at least one bug in the fdo bugzilla for this
https://bugs.freedesktop.org/show_bug.cgi?id=99383

simply said, more data right now doesn't help because it's one of the things 
that are hard to analyse without moving the finger. There's (maybe) a
mismatch between finger movement and generated data, or somewhere else. Who
knows. I don't have access to an XPS, so I can't replicate it here with the
lenovos that I have around. A lot of people are affected by this, no-one has
so far stepped up to analyse it and find the cause (aside from the usual
call for more configuration options)

Cheers,
   Peter