[RFC libinput 0/9] touchpad: force resolutions for all touchpads

[Peter Hutterer]

On Wed, Jul 01, 2015 at 01:22:45PM -0700, Bill Spitzak wrote:
> On Tue, Jun 30, 2015 at 11:08 PM, Peter Hutterer <peter.hutterer at who-t.net>
> wrote:
> 
> >
> > Not all touchpads provide x/y axis resolution. some of them (apple
> > touchpads) we fix up in the systemd hwdb but others are harder to detect or
> > fix in a generic manner. so we set a fake resolution (1 unit/mm), but
> > that's
> > just to avoid divide-by-zero.
> >
> > the touchpad code handles physical distances where possible but always has
> > an escape path to handle the resolution-less touchpads, most of that by
> > some
> > magic numbers or percentages but those may or may not be correct.
> > especially
> > on the touchpads we don't have available for testing we have no idea how
> > close the behaviour is to the intended one.
> >
> > this patchset does away with the fake resolution handling. a udev callout
> > checks for the firmware version and assigns a tag based on that, the hwdb
> > we
> > ship then assigns size hints for the device. anything that doesn't get a
> > size hint is simply set to a default size.
> > this may make some touchpads worse in the short term, but long term we
> > should be able to narrow down the various sizes and provide hwdb entries
> > for
> > them - and get the same behaviour across devices.
> >
> > Note that the hwdb entries are just approximates so far, the ALPS fw
> > version
> > 8 is a measurement of such a device, everything else is just assumption or
> > blind guesses. I'll try to collect some real-world sizes before pushing
> > this if we're happy with this approach.
> >
> > 1/9 and 2/9 are independent and can go in regardless.
> >
> 
> Looks correct to me.
> 
> I hope my attempt to do this was helpful, but this approach where it can
> get the size from the database is much better.

yeah, thanks. it did help, if only to make it clear that the magic numbers
approach won't work :)

Cheers,
   Peter