Two-finger width setting on Dell XPS 15 9650

[Peter Hutterer]

On Tue, Jun 20, 2017 at 05:07:44AM +0000, Vahe Oughourlian wrote:
> On Mon, Jun 19, 2017 at 5:58 PM Peter Hutterer <peter.hutterer at who-t.net>
> wrote:
> 
> > On Fri, Jun 16, 2017 at 05:48:31PM +0000, Vahe Oughourlian wrote:
> > > On Thu, Jun 15, 2017 at 3:23 PM Peter Hutterer <peter.hutterer at who-t.net
> > >
> > > wrote:
> > >
> > > > On Tue, Jun 13, 2017 at 01:46:12AM +0000, Vahe Oughourlian wrote:
> > > > > I'm working with a Dell XPS 15 9650 using libinput. Using the
> > Synaptics
> > > > > driver, a property is available to set the maximum distance two
> > fingers
> > > > may
> > > > > be apart before a two-finger scroll can be triggered, called
> > "Synaptics
> > > > > Two-Finger Width." At the moment, two fingers anywhere on the
> > touchpad
> > > > will
> > > > > switch it over to a two-finger scroll.
> > > > >
> > > > > I ran across a post from Peter Hutterer (
> > > > >
> > > >
> > https://who-t.blogspot.de/2016/04/why-libinput-doesnt-have-lot-of-config.html
> > > > )
> > > > > discussing the intention to minimize options in libinput, and there
> > > > doesn't
> > > > > seem to be a setting that affects this behavior that I could find.
> > Given
> > > > > that, is the most appropriate thing to file a bug for this particular
> > > > > trackpad to have a reasonable default added for this device, or is
> > this a
> > > > > setting libinput eventually intends to support? Or is this setting
> > > > already
> > > > > available?
> > > >
> > > > it's not an available configuration, no. But the main question is what
> > > > you're actually trying to achive? I'm assuming that any finger distance
> > > > limits are not the issue per se but that there's something else that
> > > > affects
> > > > how you can interact.
> > > >
> > > > Cheers,
> > > >    Peter
> > > >
> > > >
> > > There are a couple common cases:
> > >
> > > 1. My left thumb may drift onto the upper left corner of the touchpad (in
> > > the usual home-row typing position) while my right hand is controlling
> > the
> > > touchpad. Despite the 70mm-ish distance between the fingers, the next
> > > cursor motion by the right hand is treated as a two-finger scroll.
> >
> > https://bugs.freedesktop.org/show_bug.cgi?id=99703
> >
> > > 2. I often use the physical click as part of a drag-and-highlight. While
> > > there are dead zones at the bottom of the mouse for the virtual buttons,
> > if
> > > my thumb isn't quite low enough on the right hand, the moment before
> > > actually clicking the physical button underneath the trackpad will turn
> > > into a two-finger scroll. This also can occur at the end of the click as
> > > I'm lifting the thumb, and the next motion is a scroll.
> >
> > not quite the same but similar to the bug above. I think that's something
> > we'd have to revisit once we manage to get a hand on 99703
> >
> 
> Sure, I'll keep an eye on 99703 and give it a shot when it comes out.

I've just attached a branch for testing to that bug, please give it a shot,
we can revisit the other questions once we get that one sorted

Cheers,
   Peter

> > > My intention was to set the distance between two fingers such that they'd
> > > have to be much closer together (within 20mm, for example) to avoid these
> > > accidental scroll events.
> >
> > tbh, I don't think that's the right solution and that's also why there's no
> > config option for it. Any distance-based guessing of intentions should
> > always be internal to libinput because once we expose the option we'll have
> > to keep it around forever. And exposing an option promises behaviour.
> >
> 
> Completely agree. Mostly this is me working backwards from an existing
> option in the old version to find a possible solution.
> 
> 
> >
> > I think fixing 99703 will remove most of your issues.
> >
> 
> I'm not sure in this case. While it looks like 99703 will solve the case
> mid-movement, it doesn't really address the cases at the start of the
> movement when two fingers are on the pad.
> 
> 
> >
> > > One thing I have also noticed on other trackpads (on other OSes,
> > > admittedly) is that both fingers seem to have to be in motion to trigger
> > a
> > > two finger scroll, whereas I seem to be able to trigger a two-finger
> > scroll
> > > on the Dell's trackpad by moving one finger while two fingers are on the
> > > trackpad at a time. Perhaps that's the solution here?
> >
> > yeah, that's been a long-standing feature of the 2fg scroll on linux. not
> > sure we can/should get rid of that.
> >
> 
> I'm entirely sympathetic to your perspective here about changing existing
> and expected behavior, and no doubt you'd hear from a cadre of unhappy
> users if the behavior were to change. However, I'd like to push back here.
> I think two fingers moving gives significantly more intention to the
> gesture, and as I'm seeing in my experience, it's happening enough
> unintentionally to me to raise the question. I don't think it makes sense
> from a user perspective to have any two touch points anywhere on the pad
> trigger a scroll.
> 
> Adding a configuration seems like the wrong solution per the direction the
> project is taking, but changing the existing behavior is also undesirable.
> What kind of consideration might move you to change this scroll behavior?
> 
> 
> >
> > Cheers,
> >    Peter
> >