Tablets may have different ABS_PRESSURE ranges with the oldest tablets
having 1k pressure range, then 2k, and the newer ones 8k.
If the same tool is used across two tablets with different ABS_PRESSURE
ranges, the first tablet in proximity calculated the range on where to
normalize to. As a result the other tablet either couldn't reach the
full pressure (2k pressure first, then 8k) or the full pressure range
was reached at a fraction of the full range (8k pressure first, then
2k).
Fix this by moving the threshold handling into a separate struct and
hardcoding up to 4 of those per tool. That is 2 more than the more
complicated setups I've heard of (and this only applies to tracking the
same stylus across those tablets anyway).
This duplicates the pressure offset heuristics but that's easier than
figuring out how to handle heuristics across potentially two tablets.
The range configuration is left as-is on the assumption that this one is
per tool, not per tablet.
Closes#1089
Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
Part-of: <https://gitlab.freedesktop.org/libinput/libinput/-/merge_requests/1143>
Similar to the condition just north of here, if we have 3 fingers in a
roughly linear line and 3fg dragging is enabled, assume we're actually
trying to drag. This reduces the minimum movement otherwise required
to detect the type of gesture.
Part-of: <https://gitlab.freedesktop.org/libinput/libinput/-/merge_requests/1042>
Exposed via new configuration option this enables 3 and 4 finger
dragging on touchpads. When enabled a 3/4 finger swipe
gesture is actually a button down + motion + button up sequence.
If tapping is disabled the drag starts immediately, if tapping is
enabled the drag starts after the tap timeout/motion so we can distinguish
between a tap and a drag.
When fingers are released:
- if two fingers remain -> keep dragging
- if one finger remains -> release drag, switch to pointer motion
When 3/4 fingers are set down immediately after releasing all fingers
the drag continues, similar to the tap drag lock feature. This drag lock
is not currently configurable.
This matches the macos behavior for the same feature.
Part-of: <https://gitlab.freedesktop.org/libinput/libinput/-/merge_requests/1042>
Where the tapping code calls into the timeout function make sure we have
a separate event for this. This way we know whether the current gesture
event is caused by the hold timeout or something else.
Part-of: <https://gitlab.freedesktop.org/libinput/libinput/-/merge_requests/1042>
This shouldn't have any effect in the current setup as there is
"coincidentally" no overlap between the two state machines so this is
effectively a noop.
Nonetheless, if we're about to send a tap button event we cannot be in
any other gesture than tapping so all swipe/pinch/holds need to be
cancelled.
Part-of: <https://gitlab.freedesktop.org/libinput/libinput/-/merge_requests/1042>
This is a workaround for a kernel bug with the Wacom Mobile Studio Pro
13: this device sends erroneous ABS_WHEEL events when pressing
buttons:
- evdev:
- [ 0, 0, 1, 256, 1] # EV_KEY / BTN_0 1
- [ 0, 0, 3, 8, 17] # EV_ABS / ABS_WHEEL 17 (+17)
- [ 0, 0, 3, 8, 0] # EV_ABS / ABS_WHEEL 0 (-17)
- [ 0, 0, 3, 40, 15] # EV_ABS / ABS_MISC 15 (+15)
- [ 0, 0, 0, 0, 0] # ------------ SYN_REPORT (0) ---------- +0ms
ABS_WHEEL to 17 then to 0 in the same frame. We should (and do) treat this
as a zero event and drop the 17 to the floor. Alas, we then generate a
ring event for the zero value despite our current value being zero
anyway. This again causes confusing behavior.
This is simple enough to work around, at least until the kernel is fixed
but also to prevent this happening in the future: warn if we get
multiple events and if so and the later event is zero, undo the axis
change state.
Closes: #1081
Part-of: <https://gitlab.freedesktop.org/libinput/libinput/-/merge_requests/1136>
Slightly less efficient but easier to read and it's not possible to
accidentally provide the wrong length. Plus it handles null pointers
correctly so get to skip the checks (which weren't needed for strneq()
either, but still).
Part-of: <https://gitlab.freedesktop.org/libinput/libinput/-/merge_requests/1121>
The previous invocation was gated behind a TABLET_OUT_OF_AREA check
resulting in a nonresponsive tool when the tablet was moved out of
proximity outside the tablet area and the area was changed.
Move the actual status bit changes up into tablet_flush()
so we unconditionally set those.
Part-of: <https://gitlab.freedesktop.org/libinput/libinput/-/merge_requests/1118>
litest supports ranged tests but they are not enough, doubly so with
tests where we want to parametrize across multiple options.
This patch adds support for just that, in clunky C style.
The typical invocation for a test is by giving the test parameter
a name, a number of values and then the values themselves:
struct litest_parameters *params = litest_parameters_new("axis", 's', 2, "ABS_X", "ABS_Y",
"enabled", 'b', '2', true, false,
"number", 'u', '2', 10, 11,
NULL);
litest_add_parametrized(sometest, LITEST_ANY, LITEST_ANY, params);
litest_parameters_unref(params);
Currently supported are u (uint32), i (int32), d (double), b (bool),
c (char) and s (string).
In the test itself, the `test_env->params` variable is available and
retrieval of the parameters works like this:
const char *axis;
uint32_t number;
bool enabled;
litest_test_param_fetch(test_env->params,
"axis", &axis,
"enabled", &enabled,
"number", &number,
NULL);
Note that since this is an effectively internal test-suite only
functionality we don't do type-checking here, it's assumed that if you
write the code to pass parameters into a test you remember the type
of said params when you write the test code.
Because we don't have hashmaps or anything useful other than lists the
implementation is a bit clunky: we copy the parameter into the test
during litest_add_*, permutate it for our test list which gives us yet
another linked list C struct, and finally copy the actual value into
the test and test environment as it's executed. Not pretty, but it
works.
A few tests are switched as simple demonstration. The name of the
test has the parameters with their names and values appended now, e.g.:
"pointer:pointer_scroll_wheel_hires_send_only_lores:ms-surface-cover:axis:ABS_X"
"pointer:pointer_motion_relative_min_decel:mouse-roccat:direction:NW"
Filtering by parameters can be done via globs of their string
representation:
libinput-test-suite --filter-params="axis:ABS_*,enabled:true,number:10*"
Part-of: <https://gitlab.freedesktop.org/libinput/libinput/-/merge_requests/1109>
../../../src/util-files.h:61:3: warning: The 1st argument to 'close' is <= -2 but should be >= -1 [unix.StdCLibraryFunctions]
61 | close(*fd);
../../../test/test-quirks.c:66:8: warning: Null pointer passed to 2nd parameter expecting 'nonnull' [core.NonNullParamChecker]
66 | rc = fputs(file_content, fp);
The latter is bogus because we have a litest_assert for this but
somehow this is ignored, so... shrug.
Part-of: <https://gitlab.freedesktop.org/libinput/libinput/-/merge_requests/1111>
If we get to SWIPE_START we send out the BEGIN event and transition to
state SWIPE. We must not process that state immediately to avoid sending
out a spurious UPDATE event when nothing has changed.
Same for the PINCH_START/PINCH and SCROLL_START/SCROLL states.
This also fixes a crasher where we end up with NaN in the custom
acceleration function because passing the same timestamp in twice causes
a division by zero (delta time is zero).
Closes#1053
Part-of: <https://gitlab.freedesktop.org/libinput/libinput/-/merge_requests/1088>
There appears to be a race condition where an ABS_MT_TRACKING_ID -1
event is on the wire but libevdev_fetch_slot_value() for that slot
already gives us -1 as well.
If we just (re)opened our device, synching our slots would thus set zero
active slots and then trigger the assert when that event is being
processed.
It's unclear how to reliably reproduce this issue but removing the
assert and simply ignoring this event if we don't have active slots
is correct anyway.
Closes#1050
Part-of: <https://gitlab.freedesktop.org/libinput/libinput/-/merge_requests/1081>
If a tablet has an area configured and the pen goes into proximity
outside this area, ignore all events from this sequence. This truly
deactivates that area so it can even be used for e.g. placing a pen
there.
For simplicity, a sequence that starts outside the configured area will
be completely ignored, i.e. moving into the tablet area will not trigger
any fake proximity events as we cross into the allowed area. This
requires quite a bit of effort and it's unclear if it's really needed by
users - we can reconsider when we get complaints.
We do however accept a proximity event within within 3% of the
configured area. This gives us 6mm on a 200mm tablet where we can move
in from the area and still have events work, i.e. some error margin for
where a user needs both an area and work closes to the edge of that
area.
Part-of: <https://gitlab.freedesktop.org/libinput/libinput/-/merge_requests/1013>
For external tablets like the Intuos series we now expose the area
rectangle configuration and the (minimum) implementation required to
make this work.
Because an area configuration may apply late and tablet events usually
get scaled by the compositor we need to store the current axis extents
in each event. This is to behave correctly in this events sequence:
1. tool proximity in
2. caller changes config, config is pending
3. tool moves, generates events
4. tool goes out of prox, new config applies
5. caller processes motion events from step 3
If the caller in step five uses any of the get_x_transformed calls these
need to be scaled relative to the original area, not the one set in
step 2.
The current implementation merely clips into the area so moving a stylus
outside the area will be equivalent to moving it along the respective
edge of the area. It's not a true dead zone yet.
Part-of: <https://gitlab.freedesktop.org/libinput/libinput/-/merge_requests/1013>
This adds the configuration option to define a rectangle that serves as
an input area on external tablets such as an Intuos.
The intention behind this is to make this input area behave as if it was
the only physical input area on this tablet with libinput emulating
proximity events as required for where the tools moves in and out
of this area.
This could also be achieved with the existing calibration setting but
area configuration is not calibration and we don't want to expose other
side-effects of the matrix (e.g. scaling and rotation) for these
devices.
Part-of: <https://gitlab.freedesktop.org/libinput/libinput/-/merge_requests/1013>
device->abs.absinfo_x/y points to the x/y axis we want to use and
that axis is used in all the evdev helper function. For consistency
use that one here too.
This is for consistency only and has no effect on tablet devices, the
only time this isn't ABS_X/Y is on multitouch devices where that points to
ABS_MT_POSITION_X/Y.
Part-of: <https://gitlab.freedesktop.org/libinput/libinput/-/merge_requests/1013>
The vast majority of devices that libwacom doesn't know about are the
various built-in ones. Since the only effect in our code here is that we
enable the calibration matrix, let's default to built-in if we don't
know any better - better to have the matrix and not use it than to not
be able to calibrate a tablet.
Note that libwacom 2.11 and later also now default to a built-in tablet.
Part-of: <https://gitlab.freedesktop.org/libinput/libinput/-/merge_requests/1074>
Instead of re-creating the the libwacom device from the database every
time we need it let's create it once during tablet|pad_init and pass it
down to the functions.
This allows us to have one point per tablet/pad where we can log an
error if the device is not supported by libwacom - previously this was
printed during the left-handed setup.
Part-of: <https://gitlab.freedesktop.org/libinput/libinput/-/merge_requests/1073>
This was historically based on the LEDs on the device: we'd loop through
the LEDs and assign them to pad mode groups, then figure out which
button is the mode toggle for that group and finally which buttons
are in the same position as that toggle button.
Devices like the XP Pen ACK05 Remote don't have LEDs though but they do
have a mode toggle button [1] inside the dial. Let's support those by
switching the initialization on its head: search for mode toggle
buttons, create a mode group per toggle button, then associate
LEDs with that group.
The outcome should be functionally the same for devices with LEDs but
allows us to create mode groups where no LEDs exist.
Closes#1045
[1] As per Windows default button behavior
Part-of: <https://gitlab.freedesktop.org/libinput/libinput/-/merge_requests/1070>
This prevents accidentally leaving the fd set after closing.
And it includes the -1 check so we don't need this everywhere ourselves
(not that we use it right now but valgrind likes to complain about
this).
Part-of: <https://gitlab.freedesktop.org/libinput/libinput/-/merge_requests/1067>
Since our API doesn't accomodate for "dunno" we need to pick a mode that
we are actually in. This happens on the Intuos Pro 2 (PTH-660) which has
all LEDs on brightness zero, resulting in a failure to set up the modes
and we're left without a mode button.
Fix it by just picking zero as the default mode until specified
otherwise.
Part-of: <https://gitlab.freedesktop.org/libinput/libinput/-/merge_requests/1069>
