platform/upstream/libinput.git
10 years agoRemove old touchpad code
Peter Hutterer [Thu, 22 May 2014 06:13:12 +0000 (16:13 +1000)]
Remove old touchpad code

With the addition of software buttons for clickpads, the new touchpad code has
overtaken this driver in terms of features. The older driver was disabled in
6a6103262530d8fca66ee3847d5664824158c12b.

Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
Reviewed-by: Hans de Goede <hdegoede@redhat.com>
10 years agoevdev: INPUT_PROP_DIRECT devices can't be touchpads
Peter Hutterer [Thu, 29 May 2014 05:43:38 +0000 (15:43 +1000)]
evdev: INPUT_PROP_DIRECT devices can't be touchpads

Devices that are direct input devices are marked by the kernel with the
INPUT_PROP_DIRECT property. Touchpads are always indirect input devices, so
let's do the easiest check first before we try device-specific capabilities.

Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
10 years agofilter: Ignore non-suitable trackers when calculating initial velocity
Jonas Ådahl [Sat, 24 May 2014 19:49:13 +0000 (21:49 +0200)]
filter: Ignore non-suitable trackers when calculating initial velocity

calculate_velocity() didn't skip pointer trackers far away in time when
calculating the initial velocity. This check was done later when
iterating the rest, so while at it, simplify the function by doing both
iterations in one single loop.

Signed-off-by: Jonas Ådahl <jadahl@gmail.com>
Reviewed-by: Hans de Goede <hdegoede@redhat.com>
10 years agofilter: Fix get_direction for short vectors
Jonas Ådahl [Sat, 24 May 2014 19:38:08 +0000 (21:38 +0200)]
filter: Fix get_direction for short vectors

Short vectors alongside the x axis produced inverted directions.

Signed-off-by: Jonas Ådahl <jadahl@gmail.com>
Reviewed-by: Hans de Goede <hdegoede@redhat.com>
10 years agotouchpad: Make anisotropic motion deltas isotropic
Jonas Ådahl [Sat, 24 May 2014 14:53:45 +0000 (16:53 +0200)]
touchpad: Make anisotropic motion deltas isotropic

The x and y absolute axis may have different resolutions, meaning 1 unit
long motion delta on one axis is not physically as long as 1 unit motion
delta on the other axis.

In order to make these anisotropic input motion deltas output as isotropic
motion deltas, apply scaling to one of the axis making it have the same
dimension as the other, before passing it through the motion filter
which assumes all deltas are isotropic.

https://bugs.freedesktop.org/show_bug.cgi?id=79056

Signed-off-by: Jonas Ådahl <jadahl@gmail.com>
Reviewed-by: Hans de Goede <hdegoede@redhat.com>
10 years agoAdd basic mouse pointer acceleration
Jonas Ådahl [Sun, 18 May 2014 17:20:39 +0000 (19:20 +0200)]
Add basic mouse pointer acceleration

This patch reimplements the simple smooth pointer acceleration profile
from X.org xserver. The algorithm is identical to the classic profile
with a non-zero pointer acceleration threshold.

When support for changable parameters is in place, to get a pointer
acceleration the same as the default classic profile of X.org a
polynomial acceleration profile should be used for when the threshold
parameter is zero.

Signed-off-by: Jonas Ådahl <jadahl@gmail.com>
Reviewed-by: Hans de Goede <hdegoede@redhat.com>
10 years agotest: Queue two motion events in pointer event conversion test
Jonas Ådahl [Mon, 26 May 2014 21:18:28 +0000 (23:18 +0200)]
test: Queue two motion events in pointer event conversion test

Pointer acceleration filters may absorb the first event, so queue two,
just in case.

Signed-off-by: Jonas Ådahl <jadahl@gmail.com>
Reviewed-by: Hans de Goede <hdegoede@redhat.com>
10 years agotest: Make relative pointer event test more accepting
Jonas Ådahl [Mon, 26 May 2014 20:12:25 +0000 (22:12 +0200)]
test: Make relative pointer event test more accepting

A test cannot exactly predict the resulting motion event from a given
evdev event series without having to reimplement the acceleration
algorithm. To still be able to test that sane relative motion events are
produced, check that the length and direction of the resulting motion
event vectors are close to the same as the expected vectors.

Signed-off-by: Jonas Ådahl <jadahl@gmail.com>
Reviewed-by: Hans de Goede <hdegoede@redhat.com>
10 years agofilter: Add motion filter destruction helper
Jonas Ådahl [Mon, 26 May 2014 21:20:42 +0000 (23:20 +0200)]
filter: Add motion filter destruction helper

Signed-off-by: Jonas Ådahl <jadahl@gmail.com>
Reviewed-by: Hans de Goede <hdegoede@redhat.com>
10 years agotouchpad: Stop scrolling on a button click / tap
Hans de Goede [Fri, 23 May 2014 14:06:25 +0000 (16:06 +0200)]
touchpad: Stop scrolling on a button click / tap

On a button click / tap the scrolling event handler no longer gets called,
ensure that any in progress scrolling is stopped.

Signed-off-by: Hans de Goede <hdegoede@redhat.com>
Reviewed-by: Peter Hutterer <peter.hutterer@who-t.net>
Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
10 years agotouchpad: Refactor tp_post_scroll_events()
Hans de Goede [Fri, 23 May 2014 14:06:24 +0000 (16:06 +0200)]
touchpad: Refactor tp_post_scroll_events()

Put the actual scroll event posting in the straight path.

Signed-off-by: Hans de Goede <hdegoede@redhat.com>
Reviewed-by: Peter Hutterer <peter.hutterer@who-t.net>
Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
10 years agotouchpad: Fix sending of scroll stop events
Hans de Goede [Fri, 23 May 2014 14:06:23 +0000 (16:06 +0200)]
touchpad: Fix sending of scroll stop events

Setting tp->scroll.direction = 0 before checking tp->scroll.direction
to see if we need to send stop scroll events for vert / horz scrolling does
not really work well.

Also we need to check direction with an axis mask, not the axis number.

While at it also add a tp_stop_scroll_events() helper function for this.

Signed-off-by: Hans de Goede <hdegoede@redhat.com>
Reviewed-by: Peter Hutterer <peter.hutterer@who-t.net>
Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
10 years agotouchpad: Clear touches being the pointer when doing 2 finger scrolling
Hans de Goede [Fri, 23 May 2014 14:06:22 +0000 (16:06 +0200)]
touchpad: Clear touches being the pointer when doing 2 finger scrolling

When doing 2 finger scrolling we don't want any spurious movement events after
scrolling. touchpad_2fg_no_motion tests for this, but it lifts touch 0
(which is the pointer as it came down first) first, so it only catches the
case where touch 1 suddenly gets promoted to being the pointer.

However if touch 1 is lifted first, then touch 0 is still the pointer and
will cause spurious movement events. Swap the 2 litest_touch_up calls to
catch this (and make the test fail), and add code to clear the is_pointer
flag on all touched when doing 2 finger scrolling to fix it again.

Signed-off-by: Hans de Goede <hdegoede@redhat.com>
Reviewed-by: Peter Hutterer <peter.hutterer@who-t.net>
Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
10 years agotools: plug leak in event-debug
Peter Hutterer [Tue, 27 May 2014 06:14:41 +0000 (16:14 +1000)]
tools: plug leak in event-debug

Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
10 years agofilter: Fix coding style
Jonas Ådahl [Sat, 24 May 2014 20:00:46 +0000 (22:00 +0200)]
filter: Fix coding style

Signed-off-by: Jonas Ådahl <jadahl@gmail.com>
10 years agoMerge branch 'clickpad-improvements-v2' of git://people.freedesktop.org/~jwrdegoede...
Peter Hutterer [Thu, 22 May 2014 22:51:29 +0000 (08:51 +1000)]
Merge branch 'clickpad-improvements-v2' of git://people.freedesktop.org/~jwrdegoede/libinput

10 years agotouchpad: Make pointer movements a bit faster by default
Jonas Ådahl [Tue, 20 May 2014 21:12:28 +0000 (23:12 +0200)]
touchpad: Make pointer movements a bit faster by default

Alter the so far hard coded parameters a bit to make the pointer
acceleration profile accelerate the pointer movement a bit more than
before.

Signed-off-by: Jonas Ådahl <jadahl@gmail.com>
Reviewed-by: Peter Hutterer <peter.hutterer@who-t.net>
10 years agoChange internal timestamps to uint64_t to properly deal with wrapping
Hans de Goede [Tue, 8 Apr 2014 10:29:45 +0000 (12:29 +0200)]
Change internal timestamps to uint64_t to properly deal with wrapping

We store timestamps in ms since system boot (CLOCK_MONOTONIC). This will wrap
after circa 50 days.

I've considered making our code wrapping safe, but that won't work. We also
use our internal timestamps to program timer-fds for timeouts. And we store
ms in a single integer but the kernel uses 2 integers, one for seconds and
one for usec/nanosec. So at 32 bits our ms containing integer will wrap
in 50 days, while the kernels seconds storing integer lasts a lot longer.
So when we wrap our ms timestamps, we will be programming the timer-fds
with a seconds value in the past.

So change all our internal timestamps to uint64_t to avoid the wrapping
when programming the timer-fds. Note that we move from 64-bit timestamps to
32-bit timestamps when calling the foo_notify_bar functions from
libinput-private.h. Having 64 bit timestamps has no use past this point,
since the wayland input protocol uses 32 bit timestamps (and clients will
have to deal with wrapping).

Signed-off-by: Hans de Goede <hdegoede@redhat.com>
Reviewed-by: Peter Hutterer <peter.hutterer@who-t.net>
10 years agotouchpad: handle_timeouts: Remove unused return value
Hans de Goede [Tue, 8 Apr 2014 10:39:24 +0000 (12:39 +0200)]
touchpad: handle_timeouts: Remove unused return value

Signed-off-by: Hans de Goede <hdegoede@redhat.com>
Reviewed-by: Peter Hutterer <peter.hutterer@who-t.net>
10 years agotouchpad: Remove clickpad clicked test from 2 finger scrolling handling
Hans de Goede [Tue, 8 Apr 2014 06:39:11 +0000 (08:39 +0200)]
touchpad: Remove clickpad clicked test from 2 finger scrolling handling

This is no longer needed now that we take the button area and pinned fingers
into account.

Signed-off-by: Hans de Goede <hdegoede@redhat.com>
Reviewed-by: Peter Hutterer <peter.hutterer@who-t.net>
10 years agotouchpad: Ignore fingers in button area for 2 finger scroll
Hans de Goede [Tue, 8 Apr 2014 06:34:20 +0000 (08:34 +0200)]
touchpad: Ignore fingers in button area for 2 finger scroll

Signed-off-by: Hans de Goede <hdegoede@redhat.com>
Reviewed-by: Peter Hutterer <peter.hutterer@who-t.net>
10 years agotouchpad: softbuttons: Deal with a click arriving before any touches
Hans de Goede [Mon, 7 Apr 2014 14:49:36 +0000 (16:49 +0200)]
touchpad: softbuttons: Deal with a click arriving before any touches

It is possible for a click to get reported before any related touch events
get reported, here is the relevant part of an evemu-record session on a T440s:

E: 3.985585 0000 0000 0000 # ------------ SYN_REPORT (0) ----------
E: 3.997419 0003 0039 -001 # EV_ABS / ABS_MT_TRACKING_ID   -1
E: 3.997419 0001 014a 0000 # EV_KEY / BTN_TOUCH            0
E: 3.997419 0003 0018 0000 # EV_ABS / ABS_PRESSURE         0
E: 3.997419 0001 0145 0000 # EV_KEY / BTN_TOOL_FINGER      0
E: 3.997419 0000 0000 0000 # ------------ SYN_REPORT (0) ----------
E: 5.117881 0001 0110 0001 # EV_KEY / BTN_LEFT             1
E: 5.117881 0000 0000 0000 # ------------ SYN_REPORT (0) ----------
E: 5.133422 0003 0039 0187 # EV_ABS / ABS_MT_TRACKING_ID   187
E: 5.133422 0003 0035 3098 # EV_ABS / ABS_MT_POSITION_X    3098
E: 5.133422 0003 0036 3282 # EV_ABS / ABS_MT_POSITION_Y    3282
E: 5.133422 0003 003a 0046 # EV_ABS / ABS_MT_PRESSURE      46
E: 5.133422 0001 014a 0001 # EV_KEY / BTN_TOUCH            1
E: 5.133422 0003 0000 3102 # EV_ABS / ABS_X                3102
E: 5.133422 0003 0001 3282 # EV_ABS / ABS_Y                3282
E: 5.133422 0003 0018 0046 # EV_ABS / ABS_PRESSURE         46
E: 5.133422 0001 0145 0001 # EV_KEY / BTN_TOOL_FINGER      1
E: 5.133422 0000 0000 0000 # ------------ SYN_REPORT (0) ----------

Notice the BTN_LEFT event all by itself!

To deal with this if a physical click registers before we get any touches,
wait for the first touch to resolve the click.

Also see the new activity diagram for the tp_post_softbutton_buttons
method which has been added to doc/touchpad-softbutton-state-machine.svg

Signed-off-by: Hans de Goede <hdegoede@redhat.com>
Reviewed-by: Peter Hutterer <peter.hutterer@who-t.net>
10 years agotouchpad: post_button_events: Remove TOUCHPAD_EVENT_BUTTON_PRESS/RELEASE test
Hans de Goede [Mon, 7 Apr 2014 14:09:03 +0000 (16:09 +0200)]
touchpad: post_button_events: Remove TOUCHPAD_EVENT_BUTTON_PRESS/RELEASE test

We already check for old != current everywhere, so this is not needed;
And in tp_post_softbutton_buttons we want to do delay button down reporting if
we don't have touch info yet in which case this check actually gets in the way.

Signed-off-by: Hans de Goede <hdegoede@redhat.com>
Reviewed-by: Peter Hutterer <peter.hutterer@who-t.net>
10 years agotouchpad: Ignore non left clicks on clickpads
Hans de Goede [Mon, 7 Apr 2014 11:50:11 +0000 (13:50 +0200)]
touchpad: Ignore non left clicks on clickpads

We should never get any non left button events on clickpads, but if we
do these might confuse our state, so complain about it and ignore these.

Signed-off-by: Hans de Goede <hdegoede@redhat.com>
Reviewed-by: Peter Hutterer <peter.hutterer@who-t.net>
10 years agotouchpad: Use INPUT_PROP_BUTTONPAD instead of checking for buttons
Hans de Goede [Mon, 7 Apr 2014 11:44:07 +0000 (13:44 +0200)]
touchpad: Use INPUT_PROP_BUTTONPAD instead of checking for buttons

And warn if INPUT_PROP_BUTTONPAD mismatches right/middle buttons presence.

Also fix the bcm5974 to properly advertise INPUT_PROP_BUTTONPAD.

Signed-off-by: Hans de Goede <hdegoede@redhat.com>
Reviewed-by: Peter Hutterer <peter.hutterer@who-t.net>
10 years agotouchpad: Rework is_pointer handling
Hans de Goede [Fri, 4 Apr 2014 15:22:02 +0000 (17:22 +0200)]
touchpad: Rework is_pointer handling

We don't want touches in the button area to cause the pointer to move. So
instead of making a touch the pointer when it moves to TOUCH_BEGIN, wait
with making it the pointer until its buttons state moves to BUTTON_STATE_AREA.

Note that a touch in the main area of the touchpad will move to
BUTTON_STATE_AREA immediately. If software-buttons are not enabled, any finger
is in the BUTTON_STATE_AREA.

While at it also refactor the is_pointer setting in general, removing
code duplicition wrt checking that another touch is not already
the pointer on unpinning a finger, and add safeguards that unpinning
does not make a finger which is not in button state BUTTON_STATE_AREA the
pointer, nor that the button code makes a pinned finger the pointer.

All these sanity checks are combined into a new tp_button_active function,
since they should be taken into account for 2 finger scrolling, etc. too.

Signed-off-by: Hans de Goede <hdegoede@redhat.com>
Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
Reviewed-by: Peter Hutterer <peter.hutterer@who-t.net>
10 years agotouchpad: Add clickpad-style software buttons
Peter Hutterer [Thu, 27 Mar 2014 23:44:11 +0000 (09:44 +1000)]
touchpad: Add clickpad-style software buttons

Almost all non Apple touchpads have visible markings for software button areas,
so limit clickfinger behavior to Apple clickpads, and implement software button
areas for others.

This is a slightly fancier implementation than the simplest model and ported
over from libtouchpad. It implements a state machine for the software buttons
with left and right buttons currently implemented. Buttons are oriented
left-to-right, in a horizontal bar. No random button placement allowed.

In general, the procedure is:
- if a finger sets down in the left button area, a click is a left click
- if a finger sets down in the right button area, a click is a right click
- if a finger leaves the button area, a click is a left click
- if a finger starts outside the button area, a click is a left click

Two timeouts are used to handle buttons more smoothly:
- if a finger sets down in a button area but "immediately" moves over
  to a different area, that area takes effect on a click.
- if a finger leaves a button area and "immediately" clicks or moves back into
  the area, the button still takes effect on a click.
- if a finger changes between areas and stays there for a timeout, that area
  takes effect on a click.

Note the button area states are named BOTTOM_foo to make it easier to later
add support for a top button area such as can be found on the Thinkpad [2-5]40
series.

Co-authored-by: Hans de Goede <hdegoede@redhat.com>
Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
Signed-off-by: Hans de Goede <hdegoede@redhat.com>
Reviewed-by: Jonas Ådahl <jadahl@gmail.com>
Reviewed-by: Hans de Goede <hdegoede@redhat.com>
Reviewed-by: Peter Hutterer <peter.hutterer@who-t.net>
10 years agotouchpad: save the active clickfinger button
Peter Hutterer [Thu, 27 Mar 2014 05:51:12 +0000 (15:51 +1000)]
touchpad: save the active clickfinger button

To avoid having a button left press and a button right release if the number
of fingers changes.

Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
Signed-off-by: Hans de Goede <hdegoede@redhat.com>
Reviewed-by: Jonas Ådahl <jadahl@gmail.com>
Reviewed-by: Hans de Goede <hdegoede@redhat.com>
10 years agodoc: add state machine SVG to EXTRA_DIST
Peter Hutterer [Fri, 28 Mar 2014 05:51:08 +0000 (15:51 +1000)]
doc: add state machine SVG to EXTRA_DIST

Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
Signed-off-by: Hans de Goede <hdegoede@redhat.com>
Reviewed-by: Hans de Goede <hdegoede@redhat.com>
10 years agotouchpad: move button-related code into a separate file
Peter Hutterer [Thu, 27 Mar 2014 03:09:06 +0000 (13:09 +1000)]
touchpad: move button-related code into a separate file

This is about to become more complicated with the support for software button
areas. Move it to a separate file to have it logically grouped together.
No functional changes.

Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
Signed-off-by: Hans de Goede <hdegoede@redhat.com>
Reviewed-by: Hans de Goede <hdegoede@redhat.com>
10 years agotouchpad: reset the tap timer_fd to -1 on destroy
Peter Hutterer [Tue, 15 Apr 2014 12:28:00 +0000 (14:28 +0200)]
touchpad: reset the tap timer_fd to -1 on destroy

No real effect, just for safety

Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
Signed-off-by: Hans de Goede <hdegoede@redhat.com>
Reviewed-by: Jonas Ådahl <jadahl@gmail.com>
Reviewed-by: Hans de Goede <hdegoede@redhat.com>
10 years agotouchpad: after a click, lock the finger to its current position
Peter Hutterer [Tue, 15 Apr 2014 12:27:59 +0000 (14:27 +0200)]
touchpad: after a click, lock the finger to its current position

On clickpads, clicking the pad usually causes some motion events. To avoid
erroneous movements, lock all fingers into position on the click and don't
allow for motion events until a finger moves past a given threshold
(currently 2% of the touchpad diagonal).

Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
Signed-off-by: Hans de Goede <hdegoede@redhat.com>
Reviewed-by: Jonas Ådahl <jadahl@gmail.com>
Reviewed-by: Hans de Goede <hdegoede@redhat.com>
Reviewed-by: Peter Hutterer <peter.hutterer@who-t.net>
10 years agotouchpad: set ntouches for single-touch pads depending on key bits
Peter Hutterer [Tue, 15 Apr 2014 12:27:58 +0000 (14:27 +0200)]
touchpad: set ntouches for single-touch pads depending on key bits

A single-touch touchpad that provides BTN_TOOL_TRIPLETAP has 3 touches, etc.
There aren't a lot of these out there, but some touchpads don't have slots but
do provide two- or three-finger detection.

Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
Signed-off-by: Hans de Goede <hdegoede@redhat.com>
Reviewed-by: Jonas Ådahl <jadahl@gmail.com>
Reviewed-by: Hans de Goede <hdegoede@redhat.com>
10 years agoconfigure.ac: libinput 0.2 0.2.0
Jonas Ådahl [Thu, 22 May 2014 06:10:01 +0000 (08:10 +0200)]
configure.ac: libinput 0.2

Bump the libinput version and the libtool version.

Signed-off-by: Jonas Ådahl <jadahl@gmail.com>
10 years agoconfigure.ac: Add libtool versioning setting to configure.ac
Jonas Ådahl [Thu, 22 May 2014 06:06:25 +0000 (08:06 +0200)]
configure.ac: Add libtool versioning setting to configure.ac

Add the explanatory description of the version components from libevdev
as well.

Signed-off-by: Jonas Ådahl <jadahl@gmail.com>
10 years agotouchpad: check calloc result
Carlos Olmedo Escobar [Mon, 28 Apr 2014 23:26:28 +0000 (01:26 +0200)]
touchpad: check calloc result

Check the value returned by calloc.

Signed-off-by: Carlos Olmedo Escobar <carlos.olmedo.e@gmail.com>
Reviewed-by: Peter Hutterer <peter.hutterer@who-t.net>
Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
10 years agotest: disable parallel build in test directory
Peter Hutterer [Wed, 23 Apr 2014 21:25:19 +0000 (07:25 +1000)]
test: disable parallel build in test directory

We depend on device creation on the host system, having the tests run in
parallel runs a risk of random failure.

Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
Reviewed-by: Jonas Ådahl <jadahl@gmail.com>
10 years agoevdev: Define KEY_LIGHTS_TOGGLE macro if missing
Jonas Ådahl [Wed, 23 Apr 2014 07:54:49 +0000 (09:54 +0200)]
evdev: Define KEY_LIGHTS_TOGGLE macro if missing

When building on a system with an older kernel, some KEY_ macros might
be missing. To be able to build on such system, define them if they are
missing.

It is probably better to keep our own copy of input.h somewhere in our
tree, and include that one instead of the system one, but that can be
added later.

Signed-off-by: Jonas Ådahl <jadahl@opera.com>
10 years agotest: fix a couple of memleaks in the tests
Peter Hutterer [Wed, 23 Apr 2014 01:04:48 +0000 (11:04 +1000)]
test: fix a couple of memleaks in the tests

Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
10 years agotools: add newline before function name
Peter Hutterer [Wed, 23 Apr 2014 00:47:09 +0000 (10:47 +1000)]
tools: add newline before function name

Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
10 years agoevdev: indentation fix
Carlos Garnacho [Mon, 21 Apr 2014 17:20:44 +0000 (19:20 +0200)]
evdev: indentation fix

Signed-off-by: Carlos Garnacho <carlosg@gnome.org>
Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
10 years agoFix doc typo
Carlos Garnacho [Mon, 21 Apr 2014 17:20:42 +0000 (19:20 +0200)]
Fix doc typo

The infinitive reads a bit odd there.

Signed-off-by: Carlos Garnacho <carlosg@gnome.org>
Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
10 years agoFix doc typo in function name
Carlos Garnacho [Mon, 21 Apr 2014 17:20:41 +0000 (19:20 +0200)]
Fix doc typo in function name

Signed-off-by: Carlos Garnacho <carlosg@gnome.org>
Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
10 years agoRemove doc references to non-existing function
Carlos Garnacho [Mon, 21 Apr 2014 17:20:40 +0000 (19:20 +0200)]
Remove doc references to non-existing function

Signed-off-by: Carlos Garnacho <carlosg@gnome.org>
Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
10 years agoFix doc typo
Carlos Garnacho [Mon, 21 Apr 2014 17:20:39 +0000 (19:20 +0200)]
Fix doc typo

Signed-off-by: Carlos Garnacho <carlosg@gnome.org>
Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
10 years agotouchpad: Silence -Wmaybe-uninitialized compiler warning
Jonas Ådahl [Fri, 18 Apr 2014 06:43:14 +0000 (08:43 +0200)]
touchpad: Silence -Wmaybe-uninitialized compiler warning

Signed-off-by: Jonas Ådahl <jadahl@gmail.com>
10 years agotest: Test seat wide button and key count helpers
Jonas Ådahl [Tue, 1 Apr 2014 20:24:10 +0000 (22:24 +0200)]
test: Test seat wide button and key count helpers

Signed-off-by: Jonas Ådahl <jadahl@gmail.com>
Reviewed-by: Peter Hutterer <peter.hutterer@who-t.net>
10 years agotest: Add ability to add test devices to existing libinput context
Jonas Ådahl [Sat, 22 Feb 2014 14:06:34 +0000 (15:06 +0100)]
test: Add ability to add test devices to existing libinput context

Signed-off-by: Jonas Ådahl <jadahl@gmail.com>
10 years agoIntroduce seat wide button and key count API
Jonas Ådahl [Tue, 1 Apr 2014 19:57:45 +0000 (21:57 +0200)]
Introduce seat wide button and key count API

Compositors will need to keep provide virtual devices of supported
generic device types (pointer, keyboard, touch etc). Events from each
device capable of a certain device type abstraction should be combined
as if it was only one device.

For key and button events this means counting presses of every key or
button. With this patch, libinput provides two new API for doing just
this; libinput_event_pointer_get_seat_button_count() and
libinput_event_keyboard_get_seat_key_count().

With these functions, a compositor can sort out what key or button events
that should be ignored for a virtual device. This could for example
look like:

event = libinput_get_event(libinput);
switch (libinput_event_get_type(event)) {
...
case LIBINPUT_EVENT_POINTER_BUTTON:
device = libinput_event_get_device(event);
seat = libinput_event_get_seat(device);
pevent = libinput_event_get_pointer_event(event);

if (libinput_event_pointer_get_button_state(pevent) &&
    libinput_event_pointer_get_seat_button_count(pevent) == 1)
notify_pointer_button_press(seat);
else if (libinput_event_pointer_get_button_state(pevent) &&
 libinput_event_pointer_get_seat_button_count(pevent) == 0)
notify_pointer_button_release(seat);
break;
...
}

Signed-off-by: Jonas Ådahl <jadahl@gmail.com>
Reviewed-by: Peter Hutterer <peter.hutterer@who-t.net>
10 years agotest: Fix test device type validity check
Jonas Ådahl [Tue, 25 Mar 2014 21:56:01 +0000 (22:56 +0100)]
test: Fix test device type validity check

Signed-off-by: Jonas Ådahl <jadahl@gmail.com>
Reviewed-by: Peter Hutterer <peter.hutterer@who-t.net>
10 years agotest: Check that libinput doesn't send double touch down/up events
Jonas Ådahl [Mon, 24 Mar 2014 22:05:21 +0000 (23:05 +0100)]
test: Check that libinput doesn't send double touch down/up events

Signed-off-by: Jonas Ådahl <jadahl@gmail.com>
10 years agotest: Test handling of many touch points
Jonas Ådahl [Mon, 24 Mar 2014 21:37:00 +0000 (22:37 +0100)]
test: Test handling of many touch points

libinput currently handles 16 per device touch points. Test that we
behave as expected when a device has an even higher number of active
touch points.

Signed-off-by: Jonas Ådahl <jadahl@gmail.com>
Reviewed-by: Peter Hutterer <peter.hutterer@who-t.net>
10 years agoevdev: Avoid double touch down/up events
Jonas Ådahl [Mon, 24 Mar 2014 22:37:09 +0000 (23:37 +0100)]
evdev: Avoid double touch down/up events

When the kernel sends multiple touch down or touch up for the same slot
in a row, ignore any such subsequent event ensuring libinput always
produces 1 x touch down -> [n x touch motion] -> 1 x touch up event
series.

Signed-off-by: Jonas Ådahl <jadahl@gmail.com>
Reviewed-by: Peter Hutterer <peter.hutterer@who-t.net>
10 years agoevdev: Dynamically allocate slot array
Jonas Ådahl [Tue, 22 Apr 2014 21:02:14 +0000 (23:02 +0200)]
evdev: Dynamically allocate slot array

Don't have a hard coded slot array size; instead allocate the array
needed according to the abs info reported by either libmtdev or libevdev.

Signed-off-by: Jonas Ådahl <jadahl@gmail.com>
10 years agoevdev: Use temporary variable when passing libevdev pointer
Jonas Ådahl [Mon, 24 Mar 2014 22:12:36 +0000 (23:12 +0100)]
evdev: Use temporary variable when passing libevdev pointer

Signed-off-by: Jonas Ådahl <jadahl@gmail.com>
10 years agoAdd a log_bug macro
Hans de Goede [Tue, 15 Apr 2014 12:28:09 +0000 (14:28 +0200)]
Add a log_bug macro

For logging when things happen which should not happen. We may want to do
something more fancy in the future but for now this suffices.

Modelled after log_bug in libevdev.

Signed-off-by: Hans de Goede <hdegoede@redhat.com>
Acked-by: Peter Hutterer <peter.hutterer@who-t.net>
10 years agoevdev: log configured device info
U. Artie Eoff [Fri, 11 Apr 2014 23:27:01 +0000 (16:27 -0700)]
evdev: log configured device info

Bring back the device configure logging that was originally part of
Weston's evdev.

Signed-off-by: U. Artie Eoff <ullysses.a.eoff@intel.com>
10 years agotest: add tests for event conversion and back
Peter Hutterer [Wed, 26 Mar 2014 05:08:02 +0000 (15:08 +1000)]
test: add tests for event conversion and back

Looks a bit excessive given how simple the base is but hey, we don't want to
ever break that bit. That'd be embarrassing.

And while we're at it make sure that the 'wrong' event getters return NULL for
each event.

Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
Reviewed-by: Jonas Ådahl <jadahl@gmail.com>
10 years agoAdd functions to convert back to the base event
Peter Hutterer [Tue, 25 Mar 2014 03:55:47 +0000 (13:55 +1000)]
Add functions to convert back to the base event

A few functions only work on the base event but once we've converted to the
target event we can't go back. Casting works for now but that would expose
internal ABI.

Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
Reviewed-by: Jonas Ådahl <jadahl@gmail.com>
10 years agotest: automatically run the tests against valgrind for leaks
Peter Hutterer [Thu, 27 Mar 2014 01:04:24 +0000 (11:04 +1000)]
test: automatically run the tests against valgrind for leaks

Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
Reviewed-by: Jonas Ådahl <jadahl@gmail.com>
10 years agotest: mark the synaptics clickpad as buttonpad
Peter Hutterer [Thu, 27 Mar 2014 02:43:21 +0000 (12:43 +1000)]
test: mark the synaptics clickpad as buttonpad

Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
Reviewed-by: Jonas Ådahl <jadahl@gmail.com>
10 years agotools: add --verbose to event-debug
Peter Hutterer [Wed, 26 Mar 2014 05:49:34 +0000 (15:49 +1000)]
tools: add --verbose to event-debug

And redirect the log to stdout. libinput logs to stderr by default, but if
we're running with --verbose we want all msgs on the same stream.

Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
Reviewed-by: Jonas Ådahl <jadahl@gmail.com>
10 years agoFix up ARRAY_FOR_EACH macro
Peter Hutterer [Tue, 25 Mar 2014 01:19:52 +0000 (11:19 +1000)]
Fix up ARRAY_FOR_EACH macro

Remove compiler warning about signed/unsigned comparison. And while we're at
it, rename i to _i in the macro to avoid name clashes.

Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
Reviewed-by: Jonas Ådahl <jadahl@gmail.com>
10 years agoUse log_error instead of fprintf in old touchpad code
Peter Hutterer [Thu, 27 Mar 2014 05:05:36 +0000 (15:05 +1000)]
Use log_error instead of fprintf in old touchpad code

Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
Reviewed-by: Jonas Ådahl <jadahl@gmail.com>
10 years agotest: fix memleak in touch test
Peter Hutterer [Thu, 10 Apr 2014 01:11:43 +0000 (11:11 +1000)]
test: fix memleak in touch test

Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
10 years agoStructure libinput_seat struct
Jonas Ådahl [Wed, 19 Mar 2014 22:38:45 +0000 (23:38 +0100)]
Structure libinput_seat struct

Signed-off-by: Jonas Ådahl <jadahl@gmail.com>
10 years agotest: Fix name of litest_button_click() definition
Jonas Ådahl [Tue, 25 Mar 2014 21:56:58 +0000 (22:56 +0100)]
test: Fix name of litest_button_click() definition

In litest.h it was called litest_button_click() while in litest.c
litest_click(); update the definition to be the same as the declaration.

Signed-off-by: Jonas Ådahl <jadahl@gmail.com>
10 years agotest: Use generic test device API in touch_abs_transform test
Jonas Ådahl [Thu, 3 Apr 2014 20:35:47 +0000 (22:35 +0200)]
test: Use generic test device API in touch_abs_transform test

Instead of having a test device which only purpose is to test absolute
coordinate transformation, use the litest_create_device_with_overrides()
API to create a specially crafted wacom touch device with high
resolution.

Signed-off-by: Jonas Ådahl <jadahl@gmail.com>
Reviewed-by: Peter Hutterer <peter.hutterer@who-t.net>
10 years agotest: switch the remaining devices to a description-based device
Peter Hutterer [Mon, 31 Mar 2014 05:23:25 +0000 (15:23 +1000)]
test: switch the remaining devices to a description-based device

Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
Reviewed-by: Hans de Goede <hdegoede@redhat.com>
10 years agotest: allow partial overriding the test devices
Peter Hutterer [Mon, 31 Mar 2014 04:54:46 +0000 (14:54 +1000)]
test: allow partial overriding the test devices

For specific tests we need something that e.g. looks like a touchpad, but has
a different name, a different number of slots, etc. In this case, the
following code will do exactly that:

struct input_absinfo overrides[] = {
 { .value = ABS_MT_SLOT, .minimum = 0, .maximum = 100 },
 { .value = -1 },
};

litest_create_device_with_overrides(LITEST_SYNAPTICS_CLICKPAD,
    NULL, NULL, &overrides, NULL);

For general event codes, overrides can only add to the set of events, they
can't remove.

Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
Reviewed-by: Hans de Goede <hdegoede@redhat.com>
10 years agotest: allow for description-based test devices
Peter Hutterer [Mon, 31 Mar 2014 00:00:16 +0000 (10:00 +1000)]
test: allow for description-based test devices

Most of the test devices now are static descriptions anyway, make them fully
static now, including for touch events.

Switch the synaptics device now as example, the rest comes later for easier
patch review.

Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
Reviewed-by: Hans de Goede <hdegoede@redhat.com>
10 years agotest: if no teardown func is set, use the default
Peter Hutterer [Sun, 30 Mar 2014 23:38:49 +0000 (09:38 +1000)]
test: if no teardown func is set, use the default

Reduces the amount of boilerplate code.

Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
Reviewed-by: Hans de Goede <hdegoede@redhat.com>
10 years agotest: add litest helper functions for creating uinput devices
Peter Hutterer [Wed, 26 Mar 2014 10:09:42 +0000 (20:09 +1000)]
test: add litest helper functions for creating uinput devices

Both functions accept a series of event types/codes tuples, terminated by -1.
For the even type INPUT_PROP_MAX (an invalid type otherwise) the code is used
as a property to enable.

The _abs function als takes an array of absinfo, with absinfo.value
determining the axis to change. If none are given, abs axes are initialized
with default settings.

Both functions abort on failure, so the caller does not need to check the
return value.

Example code for creating a rel device:

struct libevdev_uinput *uinput;
struct input_id id = { ... };
uinput = litest_create_uinput_device("foo", &id,
                                     EV_REL, REL_X,
                                     EV_REL, REL_Y,
                                     EV_KEY, BTN_LEFT,
                                     INPUT_PROP_MAX, INPUT_PROP_BUTTONPAD,
                                     -1);
libevdev_uinput_write_event(uinput, EV_REL, REL_X, -1);
libevdev_uinput_write_event(uinput, EV_SYN, SYN_REPORT, 0);
...
libevdev_uinput_destroy(uinput);

Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
Reviewed-by: Hans de Goede <hdegoede@redhat.com>
10 years agoconfigure.ac: Always check for C++ compiler
Jonas Ådahl [Sat, 29 Mar 2014 13:39:42 +0000 (14:39 +0100)]
configure.ac: Always check for C++ compiler

Even though libinput uses no C++, it should be supported to include
libinput.h from C++. Therefore a build test ensuring this possibility
exist. However, since we can not conditionally invoke AC_PROG_CXX
in configure.ac just do it always.

Signed-off-by: Jonas Ådahl <jadahl@gmail.com>
10 years agotest: Add include from C++ build test
Jonas Ådahl [Mon, 24 Mar 2014 22:45:05 +0000 (23:45 +0100)]
test: Add include from C++ build test

10 years agoMake libinput.h safe to be included from C++
Jonas Ådahl [Mon, 24 Mar 2014 22:40:39 +0000 (23:40 +0100)]
Make libinput.h safe to be included from C++

Signed-off-by: Jonas Ådahl <jadahl@gmail.com>
10 years agotest/Makefile.am: Use $(GCC_CFLAGS) from configure.ac
Jonas Ådahl [Mon, 24 Mar 2014 22:09:09 +0000 (23:09 +0100)]
test/Makefile.am: Use $(GCC_CFLAGS) from configure.ac

This commit also passes AM_CFLAGS instead of AM_CPPFLAGS to the test
cases CFLAGS.

Signed-off-by: Jonas Ådahl <jadahl@gmail.com>
10 years agotest: plug a couple of memory leaks
Peter Hutterer [Wed, 26 Mar 2014 22:43:04 +0000 (08:43 +1000)]
test: plug a couple of memory leaks

Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
10 years agotouchpad: clean up tap bits on destroy
Peter Hutterer [Tue, 25 Mar 2014 02:14:39 +0000 (12:14 +1000)]
touchpad: clean up tap bits on destroy

Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
10 years agotouchpad: use log_error instead of fprintf
Peter Hutterer [Tue, 25 Mar 2014 01:43:52 +0000 (11:43 +1000)]
touchpad: use log_error instead of fprintf

Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
10 years agotouchpad: add missing flag for ABS_X events
Peter Hutterer [Tue, 25 Mar 2014 01:30:42 +0000 (11:30 +1000)]
touchpad: add missing flag for ABS_X events

Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
10 years agotouchpad: fix copy/paste error
Peter Hutterer [Tue, 25 Mar 2014 01:10:23 +0000 (11:10 +1000)]
touchpad: fix copy/paste error

Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
10 years agotest: add framework for a single-touch synaptics device
Peter Hutterer [Mon, 24 Mar 2014 05:25:32 +0000 (15:25 +1000)]
test: add framework for a single-touch synaptics device

Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
10 years agoMerge branch 'wip/mt-touchpad'
Peter Hutterer [Tue, 25 Mar 2014 00:47:59 +0000 (10:47 +1000)]
Merge branch 'wip/mt-touchpad'

10 years agotest: add a couple of touchpad tests
Peter Hutterer [Fri, 21 Mar 2014 04:41:32 +0000 (14:41 +1000)]
test: add a couple of touchpad tests

Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
10 years agotest: make sure BTN_TOOL_FINGER and BTN_TOUCH are down
Peter Hutterer [Fri, 21 Mar 2014 04:37:27 +0000 (14:37 +1000)]
test: make sure BTN_TOOL_FINGER and BTN_TOUCH are down

Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
10 years agotest: fix keyboard capabilities
Peter Hutterer [Mon, 24 Mar 2014 04:37:16 +0000 (14:37 +1000)]
test: fix keyboard capabilities

LITEST_KEYBOARD is the device type, not a feature.

Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
10 years agoevdev: drop hook to init old touchpad driver
Peter Hutterer [Mon, 17 Feb 2014 05:40:09 +0000 (15:40 +1000)]
evdev: drop hook to init old touchpad driver

Still leaving the driver itself in place for removal later, but only
initialize the new driver now.

Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
10 years agotouchpad: Support finger-pinnnig during physical button presses
Peter Hutterer [Mon, 17 Feb 2014 04:24:20 +0000 (14:24 +1000)]
touchpad: Support finger-pinnnig during physical button presses

On a clickpad, one finger has be on the trackpad to trigger a physical button
press. For drag and drop, we still want motion events though when a second
finger is down.

This patch adds finger-pinning. If the touchpad is pressed, the pressing
finger is "pinned" and ignored for further motion events. A second finger may
then be used to drag.

Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
10 years agotouchpad: mark the first finger as pointer-controlling finger
Peter Hutterer [Fri, 14 Feb 2014 03:59:41 +0000 (13:59 +1000)]
touchpad: mark the first finger as pointer-controlling finger

Unused at the moment, but will be used later to determine if a finger should
trigger motion events.

Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
10 years agotouchpad: add support for clickfingers
Peter Hutterer [Mon, 17 Feb 2014 01:14:29 +0000 (11:14 +1000)]
touchpad: add support for clickfingers

On touchpads without physical buttons, the number of fingers on the touchpad
at the time the physical click happens decides the button type. 1/2/3 fingers
is handled left/right/middle.

We also swallow the motion event on the actual click event, this reduces
erroneous motion events by a bit. More processing is needed here though.

Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
10 years agotouchpad: add fake-touch support for BTN_TOOL_DOUBLETAP and friends
Peter Hutterer [Fri, 14 Feb 2014 05:12:22 +0000 (15:12 +1000)]
touchpad: add fake-touch support for BTN_TOOL_DOUBLETAP and friends

This enables two-finger scrolling and two- and three-finger tapping on a
single-touch touchpad if BTN_TOOL_DOUBLETAP and BTN_TOOL_TRIPLETAP is set.

These require a bit of special processing:
BTN_TOUCH is set with the first finger down, but somewhat randomly unset and
re-set when switching between the various BTN_TOOL_*TAP values.
BTN_TOOL_<N>TAP is only set for N fingers down, thus a double->triple move
will see a release for DOUBLETAP and a press for TRIPLETAP. This may happen in
the same event, or across two consecutive events.

This patch adds a fake_touches mask to the touchpad struct. The mask is set
for each matching BTN_* event and used to count the number of expected
fake touchpoints. From that we begin/end the number of actual touchpoints
required. Fake touchpoints take their x/y coordinates from the first
touchpoint, which reads ABS_X/ABS_Y.

Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
10 years agotouchpad: support single-touch touchpads
Peter Hutterer [Fri, 14 Feb 2014 04:18:27 +0000 (14:18 +1000)]
touchpad: support single-touch touchpads

Touchpads without ABS_MT_SLOT create 5 slots by default (for up to QUINTTAP)
and ABS_X/Y is mapped to the 0-slot touchpoint. This commit adds handling for
a single finger, no BTN_TOOL_DOUBLETAP or similar is being processed yet.

Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
10 years agotouchpad: Only move the pointer when there's a single finger down
Peter Hutterer [Fri, 14 Feb 2014 05:48:49 +0000 (15:48 +1000)]
touchpad: Only move the pointer when there's a single finger down

Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
10 years agotouchpad: require minimum scroll distance and lock scroll direction
Peter Hutterer [Fri, 14 Feb 2014 02:31:26 +0000 (12:31 +1000)]
touchpad: require minimum scroll distance and lock scroll direction

This is a fairly rough approach, but can be handled more fine-grained later.
Require a minimum of 1 unit to start scrolling and lock the scrolling in the
initial direction, so further scroll events are limited to that direction
only.

Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
10 years agotouchpad: Filter motion in a certain number of tap states
Peter Hutterer [Sun, 9 Feb 2014 21:44:59 +0000 (07:44 +1000)]
touchpad: Filter motion in a certain number of tap states

Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
10 years agotouchpad: add support for multi-finger tapping
Peter Hutterer [Fri, 7 Feb 2014 05:18:17 +0000 (15:18 +1000)]
touchpad: add support for multi-finger tapping

Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
10 years agotouchpad: Make touchpad_get_delta() available from other files
Peter Hutterer [Fri, 7 Feb 2014 04:39:21 +0000 (14:39 +1000)]
touchpad: Make touchpad_get_delta() available from other files

No functional changes.

Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
10 years agotouchpad: add a struct for handling physical button event state changes
Peter Hutterer [Fri, 7 Feb 2014 03:59:38 +0000 (13:59 +1000)]
touchpad: add a struct for handling physical button event state changes

On ClickPads (touchpads without phys. middle/right buttons) it is important to
know whether a physical click is queued up. The finger position or number of
fingers decide which button event to send.

This isn't currently used, we still just send the button number at the moment.

Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
10 years agotouchpad: mark which events are currently pending processing
Peter Hutterer [Fri, 7 Feb 2014 03:48:06 +0000 (13:48 +1000)]
touchpad: mark which events are currently pending processing

Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>