| Commit message (Collapse) | Author | Age | Files | Lines |
|---|
| |
|
|
|
|
|
|
|
| |
To support semi-mt devices, buttons need to be scanned
for touch count information. This patch adds EV_KEY to
the set of event types sent through utouch-frame.
Signed-off-by: Chase Douglas <chase.douglas@canonical.com>
Signed-off-by: Henrik Rydberg <rydberg@euromail.se>
|
| |
|
|
|
|
|
|
|
| |
Buffer all events as planned, but strip the stream of all events
involved with pointer and MT, rather than bluntly removing all
events. In practise, this means button and switch events are let
through, undistorted although possibly slightly delayed.
Signed-off-by: Henrik Rydberg <rydberg@euromail.se>
|
| |
|
|
|
|
|
|
|
| |
A client listening only for multi-finger gestures will not get
the chance to recognize unless there is a gesture timeout. This
patch waits until the drag hold time has passed before releasing
the touch.
Signed-off-by: Henrik Rydberg <rydberg@euromail.se>
|
| |
|
|
|
|
|
|
|
| |
The timeout of a gesture is integral to the gesture type.
This patch introduces timeout per gesture type, potentially
enabling more control over gesture timing aspects. The timeout
logic replaces the recently added recognition state.
Signed-off-by: Henrik Rydberg <rydberg@euromail.se>
|
| |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
50 ms: Minimum time waiting for touch configuration to settle before
flushing touch events to X. We don't want a two finger gesture
triggering X events for the first finger before the second finger
lands.
200 ms: Maximum time for touches to begin a gesture. If no gesture is
found within this time period, the events are flushed to X.
If after 50 ms there are no possible gestures for the number of active
touches, the events are immediately flushed to X.
If events are flushed to X, no more gesture recognition is possible until
all touches are lifted. Likewise, if any gesture is recognized, no events
may be flushed to X until all touches are lifted.
[rydberg@euromail.se: this patch is severely restrictive and not final]
Signed-off-by: Chase Douglas <chase.douglas@canonical.com>
Signed-off-by: Henrik Rydberg <rydberg@euromail.se>
|
| |
|
|
|
|
|
|
| |
With the assumption that X will be handle to handle the gesture
cases correctly, remove the pointer emulation
Signed-off-by: Chase Douglas <chase.douglas@canonical.com>
Signed-off-by: Henrik Rydberg <rydberg@euromail.se>
|
| |
|
|
|
|
|
| |
Found by using -Wall -Werror
Signed-off-by: Chase Douglas <chase.douglas@canonical.com>
Signed-off-by: Henrik Rydberg <rydberg@euromail.se>
|
| |
|
|
|
|
|
|
|
|
|
| |
The original grail uses an internal touch framework and exports some
touch information as extended attributes to gesture events. This was
never quite the intention, but rather to expose gestures and touches
on a similar footing, such that grail can be input agnostic. This
patch starts off by replacing the internal touch framework with the
utouch-frame engine.
Signed-off-by: Henrik Rydberg <rydberg@euromail.se>
|
| |
|
|
|
|
|
|
|
| |
The current grail operates via callbacks. In order to make grail work
as a pipe, where input and output can be controlled externally, add
the missing event queue for gestures, and align the code to simplify a
switch.
Signed-off-by: Henrik Rydberg <rydberg@euromail.se>
|
| |
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
The existing kernel MT devices together use a handful of different
pointer emulation strategies, each with its own subtlety. This creates
a problem for grail, which needs to rebuild pointer emulation during
gesture recognition; relying on the correct pointer information to be
available at the right time becomes almost impossible.
This patch implements a common pointer emulation model for grail,
following a strategy recently discussed on LKML. This resolves the
problems stemming from different emulation strategies, and results
in a common pointer emulation experience for all devices.
Signed-off-by: Henrik Rydberg <rydberg@euromail.se>
|
| |
|
|
|
|
|
|
|
|
|
|
|
| |
The current code mistakenly set the pointer logic based on
the single variable, which is false at the entry to one-finger
mode, but true immediately afterwards. This leads to grail
sometimes failing to emit the proper emulation sequence, resulting
in a missed touch in X.
This patch corrects the logic such that the pointer is based directly
on one-finger mode, thereby solving the problem.
Signed-off-by: Henrik Rydberg <rydberg@euromail.se>
|
| |
|
|
|
|
|
|
|
|
|
| |
Touchscreens always emit a button event when touching the
screen. Trackpads, however, are considered hovering while
touching the pad, and instead uses taps to emits button events,
which is traditionally implemented elsewhere, and thus leaves
a functionality discrepancy. This patch adds tap emulation to
grail, utilizing gesture engine.
Signed-off-by: Henrik Rydberg <rydberg@euromail.se>
|
| |
|
|
|
|
|
| |
Separate out reporting of the touch state, such that reporting
can be deferred. Does not introduce any logic changes.
Signed-off-by: Henrik Rydberg <rydberg@euromail.se>
|
| |
|
|
|
|
|
| |
This bug was introduced in v1.0.11. Only report events when
the pointer position is actually changing.
Signed-off-by: Henrik Rydberg <rydberg@euromail.se>
|
| |
|
|
|
|
|
|
|
| |
Inject the emulated events into the event buffer instead,
and move the pointer logic into the emulation handler.
Introduces no logical changes, but reveals the bug in the
position event frequency.
Signed-off-by: Henrik Rydberg <rydberg@euromail.se>
|
| |
|
|
|
|
|
|
| |
There is no need to buffer events that will be ignored later on,
so move the event filter to the upstream side. This also allows
the event buffer to be used for injected events.
Signed-off-by: Henrik Rydberg <rydberg@euromail.se>
|
| |
|
|
|
|
|
|
|
| |
The current frame computation returns the event types to filter,
which is less useful than knowing what gesture types are active
in the frame. Add the active gesture types to the frame struct
and simplify usage.
Signed-off-by: Henrik Rydberg <rydberg@euromail.se>
|
| |
|
|
|
|
|
|
|
|
|
|
|
| |
This bug was hidden behind the motion inhibit during tapping,
but resurfaced when tapping from resting finger was introduced.
With this patch, pointer movement is inhibited also when lifting
fingers from the surface.
Ideally, tapping should be treated as a transient event, which would
allow zero pointer event also _before_ the tap, but presently unity
seems to depend on the initial pointer movement as a touch-down event.
Signed-off-by: Henrik Rydberg <rydberg@euromail.se>
|
| |
|
|
|
|
|
|
|
|
| |
The current code lets abs events through during one-finger drags,
and tapping does not work properly due to abs events not always
being present during taps. Track the pointer position and emit abs
events only if no gestures are in effect, and only if not explicitly
filtered.
Signed-off-by: Henrik Rydberg <rydberg@euromail.se>
|
| |
|
|
| |
Signed-off-by: Henrik Rydberg <rydberg@euromail.se>
|
| |
|
|
|
|
|
|
|
|
| |
To be able to model touch information uniformly across all supported
devices, the semantics of the touch attributes has to be treated
explicitly. Drop the generic touch property array and replace it
by an actual structure. In addition, the touch-engine abstraction
turned out to not be needed, so drop it as well.
Signed-off-by: Henrik Rydberg <rydberg@euromail.se>
|
| |
|
|
| |
Signed-off-by: Henrik Rydberg <rydberg@euromail.se>
|
| | |
|
| |
|
|
| |
Signed-off-by: Henrik Rydberg <rydberg@euromail.se>
|
| |
|
|
|
|
|
| |
Not counting the BTN_TOUCH event implied not emitting the sync
event, which confused evdev.
Signed-off-by: Henrik Rydberg <rydberg@euromail.se>
|
| |
|
|
|
|
| |
Reinsert BTN_TOUCH depending on the drag gesture
Signed-off-by: Henrik Rydberg <rydberg@euromail.se>
|
| |
|
|
|
|
|
| |
Always send drag events for one-finger touches, but only filter
the ABS_X/Y events if grail_filter_abs_events() is set to true.
Signed-off-by: Henrik Rydberg <rydberg@euromail.se>
|
| |
|
|
|
|
| |
Add the author to the copyright, courtesy of Canonical Ltd.
Signed-off-by: Henrik Rydberg <rydberg@bitmath.org>
|
| |
|
