Keep things simple by handling server managed fds in the common parts
of the open and close paths.
Signed-off-by: Hans de Goede <hdegoede@redhat.com>
Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
This is a preparation patch for adding support for server managed fds.
Signed-off-by: Hans de Goede <hdegoede@redhat.com>
Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
This is a preparation patch for adding support for server managed fds, this
also fixes a missing free() in an error handling path in the evdev back-end.
Signed-off-by: Hans de Goede <hdegoede@redhat.com>
Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
see discussion: https://bugs.gentoo.org/show_bug.cgi?id=403003
distributions should not ship the module enabled by default, because a lot
of users have the joystick module installed without knowing, resulting in
confusion and bad user experience when trying to play games, the more
likely usecase when plugging in a joystick.
the default configuration only fits for a limited number of users.
it does not make much sense to automatically enable hotplugging of the
module without explicit user consent and reviewing of the configuration.
Signed-off-by: Sascha Hlusiak <saschahlusiak@arcor.de>
reference: https://bugs.freedesktop.org/show_bug.cgi?id=42399
joysticks with a rectangular field have a
corner position of (32768,32768), joysticks with a
circular field have (23170,23170).
make sure that diagonal movement feels fast. either:
1) linear
f(32768) ~= f(23170) + f(23170)
f(32768) ~= a * f(23170)
a = 2.0
on circular joysticks, the time needed for xy movement is
exactly the time needed for x + the time for y separately.
absolute diagonal travel speed (in cm/s) is 0.707 times as fast,
which feels pretty slow.
on square joysticks, diagonal travel speed is always 1.41 times
faster than orthogonal travel speed. time needed for diagonal
movement is always 0.5 times as long as for orthogonal movement.
the value of a = 2.0 results in a nice, non-linear acceleration.
or
2) trigonometric
f(32768) ~= sqrt(f(23170)^2 + f(23170)^2))
f(32768) ~= a * f(23170)
a = 1.414
on circular joysticks, the absolute pointer travel speed
(in cm/s) is now the same for both linear and diagonal movement,
which feels natural. moving diagonally takes 0.707 times the time
of moving orthogonally.
on square joysticks, values are as in 1)
the value of a = 1.414 results in linear acceleration, which feels
too slow.
to maintain non-linear acceleration, make sure that:
a >>= 1.414
the following formula achieves results inbetween,
so it should feel natural on both devices while maintaining a
nice acceleration:
f(32768) ~= 1.620 * f(23170)
The module scales all axis values from the kernel to the range of
-32768 .. 32768, for compatibility with the old joystick kernel module.
The current implementation had an integer overflow, if the axis had a high
resolution of > 16384, like the popular XBox 360 controller.
This commitmakes the scaling use float instead to fix erratic behaviour
on high resolution joysticks. The joystick backend was not affected.
Fixes bug: https://bugs.freedesktop.org/show_bug.cgi?id=42399
Excerpt https://lists.gnu.org/archive/html/automake/2012-12/msg00038.html
- Support for the long-deprecated INCLUDES variable will be removed
altogether in Automake 1.14. The AM_CPPFLAGS variable should be
used instead.
This variable was deprecated in Automake releases prior to 1.10, which is the
current minimum level required to build X.
Signed-off-by: Gaetan Nadon <memsize@videotron.ca>
Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
there are rare cases when disabling debug code gives any benefit but the lack of debugging output,
which needs to be enabled anyway with the DebugLevel option
udev does copy the Driver line to the list of options, but when manually
specifying the driver in xorg.conf, the option "Driver" is unset. Because we
do hotplug a sub-device from within the core device, we need the "Driver"
option to be present in the list.
This should fix archlinux bug #23577:
https://bugs.archlinux.org/task/23577
Thanks to Malek for coming up with a fix.
Signed-off-by: Sascha Hlusiak <contact@saschahlusiak.de>
ABI 14 made the InputOption type opaque, move the existing code to ifdefs
and use the new function calls otherwise.
Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
Reviewed-by: Jeremy Huddleston <jeremyhu@apple.com>
Reviewed-by: Chase Douglas <chase.douglas@canonical.com>
xf86OptionListDuplicate() duplicates an XF86Option list, not an InputOption
list.
Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
Reviewed-by: Jeremy Huddleston <jeremyhu@apple.com>
Buttons are labeled "Button %d", starting with 0 and representing the button in X _after_ mapping.
Mapping can be changed while running so the labels will be constant.
Axes are labeled "Axis %d", starting with 1, representing the _physical_ axis that reports the valuator
data. The raw valuators can't be dynamically mapped, the first two valuators always are labeled "Rel X"
and "Rel Y", representing the aggregated post-calculation data from all axes.
Signed-off-by: Sascha Hlusiak <saschahlusiak@arcor.de>
the axis keys high/low properties were inconsistently capitalized,
leading to potential confusion as to why one works but not the other.
Signed-off-by: Devin J. Pohly <djpohly+xorg@gmail.com>
Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
The fdi file was never meant to be installed by default but exists merely as an
example and documentation for distributors who may decide to ship it by default or
install it in share/doc/...
This reverts commit 1d67a2da26.
The hotplugged keyboard device and the main device share the same pInfo->private data
and the same jstkCoreUnInit, so the data is freed twice. Furthermore, since the keyboard
device will delete itself, we must not delete it from within the deletion of the main device.
Freeing pInfo->private is done by the main device, hopefully done independently of the order
in which the two devices will be removed by the server.
Should fix bug #35391
This commit adds support for input ABI 12, consisting of a number of
changes:
- requires an X server with an ABI of 12
- valuators have a per-mode setting
- new PreInit prototype.
Because of the new PreInit prototype, the hotplug system has been switched
around too (should have probably been done in a separate commit before,
but...).
The old hotplug mechanism added a separate ModuleInfoRec for the keyboard
part of the driver. This isn't feasable for InputClass configurations, the
driver part may get overwritten.
On entering the driver, after checking a few default values, hotplug the
keyboard device (wacom-style) and let it initialize. Because NIDR calls
DEVICE_INIT and DEVICE_ON the keyboard must initialise the private pointer
and pass it back to the original device.
Call order is:
NewInputDeviceRequest
- jstkCorePreInit
- jstkKeyboardHotplug
- NewInputDeviceRequest
- jstkCorePreInit
immediately return jstkKeyboardPreInit()
- keyboard DEVICE_INIT
- keyboard DEVICE_ON
return keyboard device
- copy keyboard->priv to joystick->priv
- finish jstkCorePreInit as normal
- joystick DEVICE_INIT
- joystick DEVICE_ON
Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
Reviewed-by: Trevor Woerner <twoerner@gmail.com>
Doesn't matter which one we take, but ABI 12 dropped the IDevRec.
Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
Reviewed-by: Trevor Woerner <twoerner@gmail.com>
All options used by the driver will show up in the log anyway.
And new xserver versions will call this for debugging purposes for us.
Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
Reviewed-by: Trevor Woerner <twoerner@gmail.com>
This has been handled in the server for quite a while now, just initialize
with the defaults and ignore it otherwise.
Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
Reviewed-by: Trevor Woerner <twoerner@gmail.com>
Flag was write-only for a number of server generations already.
Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
Reviewed-by: Trevor Woerner <twoerner@gmail.com>
Both typedefs describe the same struct, LocalDevicePtr has been removed with
input ABI 12.
Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
Reviewed-by: Trevor Woerner <twoerner@gmail.com>
The InputDriverRec data structure in xserver/hw/xfree86/common/xf86Xinput.h
no longer contains an integer refCount member; as per commit
d568221710959cf7d783e6ff0fb80fb43a231124.
Signed-off-by: Trevor Woerner <twoerner@gmail.com>
Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
