Native input dispatch rewrite work in progress.
The old dispatch mechanism has been left in place and continues to
be used by default for now. To enable native input dispatch,
edit the ENABLE_NATIVE_DISPATCH constant in WindowManagerPolicy.
Includes part of the new input event NDK API. Some details TBD.
To wire up input dispatch, as the ViewRoot adds a window to the
window session it receives an InputChannel object as an output
argument. The InputChannel encapsulates the file descriptors for a
shared memory region and two pipe end-points. The ViewRoot then
provides the InputChannel to the InputQueue. Behind the
scenes, InputQueue simply attaches handlers to the native PollLoop object
that underlies the MessageQueue. This way MessageQueue doesn't need
to know anything about input dispatch per-se, it just exposes (in native
code) a PollLoop that other components can use to monitor file descriptor
state changes.
There can be zero or more targets for any given input event. Each
input target is specified by its input channel and some parameters
including flags, an X/Y coordinate offset, and the dispatch timeout.
An input target can request either synchronous dispatch (for foreground apps)
or asynchronous dispatch (fire-and-forget for wallpapers and "outside"
targets). Currently, finding the appropriate input targets for an event
requires a call back into the WindowManagerServer from native code.
In the future this will be refactored to avoid most of these callbacks
except as required to handle pending focus transitions.
End-to-end event dispatch mostly works!
To do: event injection, rate limiting, ANRs, testing, optimization, etc.
Change-Id: I8c36b2b9e0a2d27392040ecda0f51b636456de25
2010-04-23 01:58:52 +00:00
|
|
|
//
|
|
|
|
// Copyright 2010 The Android Open Source Project
|
|
|
|
//
|
|
|
|
// The input reader.
|
|
|
|
//
|
|
|
|
#define LOG_TAG "InputReader"
|
|
|
|
|
|
|
|
//#define LOG_NDEBUG 0
|
|
|
|
|
|
|
|
// Log debug messages for each raw event received from the EventHub.
|
|
|
|
#define DEBUG_RAW_EVENTS 0
|
|
|
|
|
|
|
|
// Log debug messages about touch screen filtering hacks.
|
2010-06-22 08:27:15 +00:00
|
|
|
#define DEBUG_HACKS 0
|
Native input dispatch rewrite work in progress.
The old dispatch mechanism has been left in place and continues to
be used by default for now. To enable native input dispatch,
edit the ENABLE_NATIVE_DISPATCH constant in WindowManagerPolicy.
Includes part of the new input event NDK API. Some details TBD.
To wire up input dispatch, as the ViewRoot adds a window to the
window session it receives an InputChannel object as an output
argument. The InputChannel encapsulates the file descriptors for a
shared memory region and two pipe end-points. The ViewRoot then
provides the InputChannel to the InputQueue. Behind the
scenes, InputQueue simply attaches handlers to the native PollLoop object
that underlies the MessageQueue. This way MessageQueue doesn't need
to know anything about input dispatch per-se, it just exposes (in native
code) a PollLoop that other components can use to monitor file descriptor
state changes.
There can be zero or more targets for any given input event. Each
input target is specified by its input channel and some parameters
including flags, an X/Y coordinate offset, and the dispatch timeout.
An input target can request either synchronous dispatch (for foreground apps)
or asynchronous dispatch (fire-and-forget for wallpapers and "outside"
targets). Currently, finding the appropriate input targets for an event
requires a call back into the WindowManagerServer from native code.
In the future this will be refactored to avoid most of these callbacks
except as required to handle pending focus transitions.
End-to-end event dispatch mostly works!
To do: event injection, rate limiting, ANRs, testing, optimization, etc.
Change-Id: I8c36b2b9e0a2d27392040ecda0f51b636456de25
2010-04-23 01:58:52 +00:00
|
|
|
|
|
|
|
// Log debug messages about virtual key processing.
|
2010-06-22 08:27:15 +00:00
|
|
|
#define DEBUG_VIRTUAL_KEYS 0
|
Native input dispatch rewrite work in progress.
The old dispatch mechanism has been left in place and continues to
be used by default for now. To enable native input dispatch,
edit the ENABLE_NATIVE_DISPATCH constant in WindowManagerPolicy.
Includes part of the new input event NDK API. Some details TBD.
To wire up input dispatch, as the ViewRoot adds a window to the
window session it receives an InputChannel object as an output
argument. The InputChannel encapsulates the file descriptors for a
shared memory region and two pipe end-points. The ViewRoot then
provides the InputChannel to the InputQueue. Behind the
scenes, InputQueue simply attaches handlers to the native PollLoop object
that underlies the MessageQueue. This way MessageQueue doesn't need
to know anything about input dispatch per-se, it just exposes (in native
code) a PollLoop that other components can use to monitor file descriptor
state changes.
There can be zero or more targets for any given input event. Each
input target is specified by its input channel and some parameters
including flags, an X/Y coordinate offset, and the dispatch timeout.
An input target can request either synchronous dispatch (for foreground apps)
or asynchronous dispatch (fire-and-forget for wallpapers and "outside"
targets). Currently, finding the appropriate input targets for an event
requires a call back into the WindowManagerServer from native code.
In the future this will be refactored to avoid most of these callbacks
except as required to handle pending focus transitions.
End-to-end event dispatch mostly works!
To do: event injection, rate limiting, ANRs, testing, optimization, etc.
Change-Id: I8c36b2b9e0a2d27392040ecda0f51b636456de25
2010-04-23 01:58:52 +00:00
|
|
|
|
|
|
|
// Log debug messages about pointers.
|
2010-06-22 08:27:15 +00:00
|
|
|
#define DEBUG_POINTERS 0
|
Native input dispatch rewrite work in progress.
The old dispatch mechanism has been left in place and continues to
be used by default for now. To enable native input dispatch,
edit the ENABLE_NATIVE_DISPATCH constant in WindowManagerPolicy.
Includes part of the new input event NDK API. Some details TBD.
To wire up input dispatch, as the ViewRoot adds a window to the
window session it receives an InputChannel object as an output
argument. The InputChannel encapsulates the file descriptors for a
shared memory region and two pipe end-points. The ViewRoot then
provides the InputChannel to the InputQueue. Behind the
scenes, InputQueue simply attaches handlers to the native PollLoop object
that underlies the MessageQueue. This way MessageQueue doesn't need
to know anything about input dispatch per-se, it just exposes (in native
code) a PollLoop that other components can use to monitor file descriptor
state changes.
There can be zero or more targets for any given input event. Each
input target is specified by its input channel and some parameters
including flags, an X/Y coordinate offset, and the dispatch timeout.
An input target can request either synchronous dispatch (for foreground apps)
or asynchronous dispatch (fire-and-forget for wallpapers and "outside"
targets). Currently, finding the appropriate input targets for an event
requires a call back into the WindowManagerServer from native code.
In the future this will be refactored to avoid most of these callbacks
except as required to handle pending focus transitions.
End-to-end event dispatch mostly works!
To do: event injection, rate limiting, ANRs, testing, optimization, etc.
Change-Id: I8c36b2b9e0a2d27392040ecda0f51b636456de25
2010-04-23 01:58:52 +00:00
|
|
|
|
2010-06-16 08:53:36 +00:00
|
|
|
// Log debug messages about pointer assignment calculations.
|
|
|
|
#define DEBUG_POINTER_ASSIGNMENT 0
|
|
|
|
|
Native input dispatch rewrite work in progress.
The old dispatch mechanism has been left in place and continues to
be used by default for now. To enable native input dispatch,
edit the ENABLE_NATIVE_DISPATCH constant in WindowManagerPolicy.
Includes part of the new input event NDK API. Some details TBD.
To wire up input dispatch, as the ViewRoot adds a window to the
window session it receives an InputChannel object as an output
argument. The InputChannel encapsulates the file descriptors for a
shared memory region and two pipe end-points. The ViewRoot then
provides the InputChannel to the InputQueue. Behind the
scenes, InputQueue simply attaches handlers to the native PollLoop object
that underlies the MessageQueue. This way MessageQueue doesn't need
to know anything about input dispatch per-se, it just exposes (in native
code) a PollLoop that other components can use to monitor file descriptor
state changes.
There can be zero or more targets for any given input event. Each
input target is specified by its input channel and some parameters
including flags, an X/Y coordinate offset, and the dispatch timeout.
An input target can request either synchronous dispatch (for foreground apps)
or asynchronous dispatch (fire-and-forget for wallpapers and "outside"
targets). Currently, finding the appropriate input targets for an event
requires a call back into the WindowManagerServer from native code.
In the future this will be refactored to avoid most of these callbacks
except as required to handle pending focus transitions.
End-to-end event dispatch mostly works!
To do: event injection, rate limiting, ANRs, testing, optimization, etc.
Change-Id: I8c36b2b9e0a2d27392040ecda0f51b636456de25
2010-04-23 01:58:52 +00:00
|
|
|
#include <cutils/log.h>
|
|
|
|
#include <ui/InputReader.h>
|
|
|
|
|
|
|
|
#include <stddef.h>
|
|
|
|
#include <unistd.h>
|
|
|
|
#include <errno.h>
|
|
|
|
#include <limits.h>
|
2010-07-15 01:48:53 +00:00
|
|
|
#include <math.h>
|
Native input dispatch rewrite work in progress.
The old dispatch mechanism has been left in place and continues to
be used by default for now. To enable native input dispatch,
edit the ENABLE_NATIVE_DISPATCH constant in WindowManagerPolicy.
Includes part of the new input event NDK API. Some details TBD.
To wire up input dispatch, as the ViewRoot adds a window to the
window session it receives an InputChannel object as an output
argument. The InputChannel encapsulates the file descriptors for a
shared memory region and two pipe end-points. The ViewRoot then
provides the InputChannel to the InputQueue. Behind the
scenes, InputQueue simply attaches handlers to the native PollLoop object
that underlies the MessageQueue. This way MessageQueue doesn't need
to know anything about input dispatch per-se, it just exposes (in native
code) a PollLoop that other components can use to monitor file descriptor
state changes.
There can be zero or more targets for any given input event. Each
input target is specified by its input channel and some parameters
including flags, an X/Y coordinate offset, and the dispatch timeout.
An input target can request either synchronous dispatch (for foreground apps)
or asynchronous dispatch (fire-and-forget for wallpapers and "outside"
targets). Currently, finding the appropriate input targets for an event
requires a call back into the WindowManagerServer from native code.
In the future this will be refactored to avoid most of these callbacks
except as required to handle pending focus transitions.
End-to-end event dispatch mostly works!
To do: event injection, rate limiting, ANRs, testing, optimization, etc.
Change-Id: I8c36b2b9e0a2d27392040ecda0f51b636456de25
2010-04-23 01:58:52 +00:00
|
|
|
|
2010-06-15 08:31:58 +00:00
|
|
|
/** Amount that trackball needs to move in order to generate a key event. */
|
|
|
|
#define TRACKBALL_MOVEMENT_THRESHOLD 6
|
|
|
|
|
|
|
|
|
Native input dispatch rewrite work in progress.
The old dispatch mechanism has been left in place and continues to
be used by default for now. To enable native input dispatch,
edit the ENABLE_NATIVE_DISPATCH constant in WindowManagerPolicy.
Includes part of the new input event NDK API. Some details TBD.
To wire up input dispatch, as the ViewRoot adds a window to the
window session it receives an InputChannel object as an output
argument. The InputChannel encapsulates the file descriptors for a
shared memory region and two pipe end-points. The ViewRoot then
provides the InputChannel to the InputQueue. Behind the
scenes, InputQueue simply attaches handlers to the native PollLoop object
that underlies the MessageQueue. This way MessageQueue doesn't need
to know anything about input dispatch per-se, it just exposes (in native
code) a PollLoop that other components can use to monitor file descriptor
state changes.
There can be zero or more targets for any given input event. Each
input target is specified by its input channel and some parameters
including flags, an X/Y coordinate offset, and the dispatch timeout.
An input target can request either synchronous dispatch (for foreground apps)
or asynchronous dispatch (fire-and-forget for wallpapers and "outside"
targets). Currently, finding the appropriate input targets for an event
requires a call back into the WindowManagerServer from native code.
In the future this will be refactored to avoid most of these callbacks
except as required to handle pending focus transitions.
End-to-end event dispatch mostly works!
To do: event injection, rate limiting, ANRs, testing, optimization, etc.
Change-Id: I8c36b2b9e0a2d27392040ecda0f51b636456de25
2010-04-23 01:58:52 +00:00
|
|
|
namespace android {
|
|
|
|
|
|
|
|
// --- Static Functions ---
|
|
|
|
|
|
|
|
template<typename T>
|
|
|
|
inline static T abs(const T& value) {
|
|
|
|
return value < 0 ? - value : value;
|
|
|
|
}
|
|
|
|
|
|
|
|
template<typename T>
|
|
|
|
inline static T min(const T& a, const T& b) {
|
|
|
|
return a < b ? a : b;
|
|
|
|
}
|
|
|
|
|
2010-06-16 08:53:36 +00:00
|
|
|
template<typename T>
|
|
|
|
inline static void swap(T& a, T& b) {
|
|
|
|
T temp = a;
|
|
|
|
a = b;
|
|
|
|
b = temp;
|
|
|
|
}
|
|
|
|
|
|
|
|
|
Native input dispatch rewrite work in progress.
The old dispatch mechanism has been left in place and continues to
be used by default for now. To enable native input dispatch,
edit the ENABLE_NATIVE_DISPATCH constant in WindowManagerPolicy.
Includes part of the new input event NDK API. Some details TBD.
To wire up input dispatch, as the ViewRoot adds a window to the
window session it receives an InputChannel object as an output
argument. The InputChannel encapsulates the file descriptors for a
shared memory region and two pipe end-points. The ViewRoot then
provides the InputChannel to the InputQueue. Behind the
scenes, InputQueue simply attaches handlers to the native PollLoop object
that underlies the MessageQueue. This way MessageQueue doesn't need
to know anything about input dispatch per-se, it just exposes (in native
code) a PollLoop that other components can use to monitor file descriptor
state changes.
There can be zero or more targets for any given input event. Each
input target is specified by its input channel and some parameters
including flags, an X/Y coordinate offset, and the dispatch timeout.
An input target can request either synchronous dispatch (for foreground apps)
or asynchronous dispatch (fire-and-forget for wallpapers and "outside"
targets). Currently, finding the appropriate input targets for an event
requires a call back into the WindowManagerServer from native code.
In the future this will be refactored to avoid most of these callbacks
except as required to handle pending focus transitions.
End-to-end event dispatch mostly works!
To do: event injection, rate limiting, ANRs, testing, optimization, etc.
Change-Id: I8c36b2b9e0a2d27392040ecda0f51b636456de25
2010-04-23 01:58:52 +00:00
|
|
|
int32_t updateMetaState(int32_t keyCode, bool down, int32_t oldMetaState) {
|
|
|
|
int32_t mask;
|
|
|
|
switch (keyCode) {
|
2010-06-30 23:10:35 +00:00
|
|
|
case AKEYCODE_ALT_LEFT:
|
2010-07-15 01:48:53 +00:00
|
|
|
mask = AMETA_ALT_LEFT_ON;
|
Native input dispatch rewrite work in progress.
The old dispatch mechanism has been left in place and continues to
be used by default for now. To enable native input dispatch,
edit the ENABLE_NATIVE_DISPATCH constant in WindowManagerPolicy.
Includes part of the new input event NDK API. Some details TBD.
To wire up input dispatch, as the ViewRoot adds a window to the
window session it receives an InputChannel object as an output
argument. The InputChannel encapsulates the file descriptors for a
shared memory region and two pipe end-points. The ViewRoot then
provides the InputChannel to the InputQueue. Behind the
scenes, InputQueue simply attaches handlers to the native PollLoop object
that underlies the MessageQueue. This way MessageQueue doesn't need
to know anything about input dispatch per-se, it just exposes (in native
code) a PollLoop that other components can use to monitor file descriptor
state changes.
There can be zero or more targets for any given input event. Each
input target is specified by its input channel and some parameters
including flags, an X/Y coordinate offset, and the dispatch timeout.
An input target can request either synchronous dispatch (for foreground apps)
or asynchronous dispatch (fire-and-forget for wallpapers and "outside"
targets). Currently, finding the appropriate input targets for an event
requires a call back into the WindowManagerServer from native code.
In the future this will be refactored to avoid most of these callbacks
except as required to handle pending focus transitions.
End-to-end event dispatch mostly works!
To do: event injection, rate limiting, ANRs, testing, optimization, etc.
Change-Id: I8c36b2b9e0a2d27392040ecda0f51b636456de25
2010-04-23 01:58:52 +00:00
|
|
|
break;
|
2010-06-30 23:10:35 +00:00
|
|
|
case AKEYCODE_ALT_RIGHT:
|
2010-07-15 01:48:53 +00:00
|
|
|
mask = AMETA_ALT_RIGHT_ON;
|
Native input dispatch rewrite work in progress.
The old dispatch mechanism has been left in place and continues to
be used by default for now. To enable native input dispatch,
edit the ENABLE_NATIVE_DISPATCH constant in WindowManagerPolicy.
Includes part of the new input event NDK API. Some details TBD.
To wire up input dispatch, as the ViewRoot adds a window to the
window session it receives an InputChannel object as an output
argument. The InputChannel encapsulates the file descriptors for a
shared memory region and two pipe end-points. The ViewRoot then
provides the InputChannel to the InputQueue. Behind the
scenes, InputQueue simply attaches handlers to the native PollLoop object
that underlies the MessageQueue. This way MessageQueue doesn't need
to know anything about input dispatch per-se, it just exposes (in native
code) a PollLoop that other components can use to monitor file descriptor
state changes.
There can be zero or more targets for any given input event. Each
input target is specified by its input channel and some parameters
including flags, an X/Y coordinate offset, and the dispatch timeout.
An input target can request either synchronous dispatch (for foreground apps)
or asynchronous dispatch (fire-and-forget for wallpapers and "outside"
targets). Currently, finding the appropriate input targets for an event
requires a call back into the WindowManagerServer from native code.
In the future this will be refactored to avoid most of these callbacks
except as required to handle pending focus transitions.
End-to-end event dispatch mostly works!
To do: event injection, rate limiting, ANRs, testing, optimization, etc.
Change-Id: I8c36b2b9e0a2d27392040ecda0f51b636456de25
2010-04-23 01:58:52 +00:00
|
|
|
break;
|
2010-06-30 23:10:35 +00:00
|
|
|
case AKEYCODE_SHIFT_LEFT:
|
2010-07-15 01:48:53 +00:00
|
|
|
mask = AMETA_SHIFT_LEFT_ON;
|
Native input dispatch rewrite work in progress.
The old dispatch mechanism has been left in place and continues to
be used by default for now. To enable native input dispatch,
edit the ENABLE_NATIVE_DISPATCH constant in WindowManagerPolicy.
Includes part of the new input event NDK API. Some details TBD.
To wire up input dispatch, as the ViewRoot adds a window to the
window session it receives an InputChannel object as an output
argument. The InputChannel encapsulates the file descriptors for a
shared memory region and two pipe end-points. The ViewRoot then
provides the InputChannel to the InputQueue. Behind the
scenes, InputQueue simply attaches handlers to the native PollLoop object
that underlies the MessageQueue. This way MessageQueue doesn't need
to know anything about input dispatch per-se, it just exposes (in native
code) a PollLoop that other components can use to monitor file descriptor
state changes.
There can be zero or more targets for any given input event. Each
input target is specified by its input channel and some parameters
including flags, an X/Y coordinate offset, and the dispatch timeout.
An input target can request either synchronous dispatch (for foreground apps)
or asynchronous dispatch (fire-and-forget for wallpapers and "outside"
targets). Currently, finding the appropriate input targets for an event
requires a call back into the WindowManagerServer from native code.
In the future this will be refactored to avoid most of these callbacks
except as required to handle pending focus transitions.
End-to-end event dispatch mostly works!
To do: event injection, rate limiting, ANRs, testing, optimization, etc.
Change-Id: I8c36b2b9e0a2d27392040ecda0f51b636456de25
2010-04-23 01:58:52 +00:00
|
|
|
break;
|
2010-06-30 23:10:35 +00:00
|
|
|
case AKEYCODE_SHIFT_RIGHT:
|
2010-07-15 01:48:53 +00:00
|
|
|
mask = AMETA_SHIFT_RIGHT_ON;
|
Native input dispatch rewrite work in progress.
The old dispatch mechanism has been left in place and continues to
be used by default for now. To enable native input dispatch,
edit the ENABLE_NATIVE_DISPATCH constant in WindowManagerPolicy.
Includes part of the new input event NDK API. Some details TBD.
To wire up input dispatch, as the ViewRoot adds a window to the
window session it receives an InputChannel object as an output
argument. The InputChannel encapsulates the file descriptors for a
shared memory region and two pipe end-points. The ViewRoot then
provides the InputChannel to the InputQueue. Behind the
scenes, InputQueue simply attaches handlers to the native PollLoop object
that underlies the MessageQueue. This way MessageQueue doesn't need
to know anything about input dispatch per-se, it just exposes (in native
code) a PollLoop that other components can use to monitor file descriptor
state changes.
There can be zero or more targets for any given input event. Each
input target is specified by its input channel and some parameters
including flags, an X/Y coordinate offset, and the dispatch timeout.
An input target can request either synchronous dispatch (for foreground apps)
or asynchronous dispatch (fire-and-forget for wallpapers and "outside"
targets). Currently, finding the appropriate input targets for an event
requires a call back into the WindowManagerServer from native code.
In the future this will be refactored to avoid most of these callbacks
except as required to handle pending focus transitions.
End-to-end event dispatch mostly works!
To do: event injection, rate limiting, ANRs, testing, optimization, etc.
Change-Id: I8c36b2b9e0a2d27392040ecda0f51b636456de25
2010-04-23 01:58:52 +00:00
|
|
|
break;
|
2010-06-30 23:10:35 +00:00
|
|
|
case AKEYCODE_SYM:
|
2010-07-15 01:48:53 +00:00
|
|
|
mask = AMETA_SYM_ON;
|
Native input dispatch rewrite work in progress.
The old dispatch mechanism has been left in place and continues to
be used by default for now. To enable native input dispatch,
edit the ENABLE_NATIVE_DISPATCH constant in WindowManagerPolicy.
Includes part of the new input event NDK API. Some details TBD.
To wire up input dispatch, as the ViewRoot adds a window to the
window session it receives an InputChannel object as an output
argument. The InputChannel encapsulates the file descriptors for a
shared memory region and two pipe end-points. The ViewRoot then
provides the InputChannel to the InputQueue. Behind the
scenes, InputQueue simply attaches handlers to the native PollLoop object
that underlies the MessageQueue. This way MessageQueue doesn't need
to know anything about input dispatch per-se, it just exposes (in native
code) a PollLoop that other components can use to monitor file descriptor
state changes.
There can be zero or more targets for any given input event. Each
input target is specified by its input channel and some parameters
including flags, an X/Y coordinate offset, and the dispatch timeout.
An input target can request either synchronous dispatch (for foreground apps)
or asynchronous dispatch (fire-and-forget for wallpapers and "outside"
targets). Currently, finding the appropriate input targets for an event
requires a call back into the WindowManagerServer from native code.
In the future this will be refactored to avoid most of these callbacks
except as required to handle pending focus transitions.
End-to-end event dispatch mostly works!
To do: event injection, rate limiting, ANRs, testing, optimization, etc.
Change-Id: I8c36b2b9e0a2d27392040ecda0f51b636456de25
2010-04-23 01:58:52 +00:00
|
|
|
break;
|
|
|
|
default:
|
|
|
|
return oldMetaState;
|
|
|
|
}
|
|
|
|
|
|
|
|
int32_t newMetaState = down ? oldMetaState | mask : oldMetaState & ~ mask
|
2010-07-15 01:48:53 +00:00
|
|
|
& ~ (AMETA_ALT_ON | AMETA_SHIFT_ON);
|
Native input dispatch rewrite work in progress.
The old dispatch mechanism has been left in place and continues to
be used by default for now. To enable native input dispatch,
edit the ENABLE_NATIVE_DISPATCH constant in WindowManagerPolicy.
Includes part of the new input event NDK API. Some details TBD.
To wire up input dispatch, as the ViewRoot adds a window to the
window session it receives an InputChannel object as an output
argument. The InputChannel encapsulates the file descriptors for a
shared memory region and two pipe end-points. The ViewRoot then
provides the InputChannel to the InputQueue. Behind the
scenes, InputQueue simply attaches handlers to the native PollLoop object
that underlies the MessageQueue. This way MessageQueue doesn't need
to know anything about input dispatch per-se, it just exposes (in native
code) a PollLoop that other components can use to monitor file descriptor
state changes.
There can be zero or more targets for any given input event. Each
input target is specified by its input channel and some parameters
including flags, an X/Y coordinate offset, and the dispatch timeout.
An input target can request either synchronous dispatch (for foreground apps)
or asynchronous dispatch (fire-and-forget for wallpapers and "outside"
targets). Currently, finding the appropriate input targets for an event
requires a call back into the WindowManagerServer from native code.
In the future this will be refactored to avoid most of these callbacks
except as required to handle pending focus transitions.
End-to-end event dispatch mostly works!
To do: event injection, rate limiting, ANRs, testing, optimization, etc.
Change-Id: I8c36b2b9e0a2d27392040ecda0f51b636456de25
2010-04-23 01:58:52 +00:00
|
|
|
|
2010-07-15 01:48:53 +00:00
|
|
|
if (newMetaState & (AMETA_ALT_LEFT_ON | AMETA_ALT_RIGHT_ON)) {
|
|
|
|
newMetaState |= AMETA_ALT_ON;
|
Native input dispatch rewrite work in progress.
The old dispatch mechanism has been left in place and continues to
be used by default for now. To enable native input dispatch,
edit the ENABLE_NATIVE_DISPATCH constant in WindowManagerPolicy.
Includes part of the new input event NDK API. Some details TBD.
To wire up input dispatch, as the ViewRoot adds a window to the
window session it receives an InputChannel object as an output
argument. The InputChannel encapsulates the file descriptors for a
shared memory region and two pipe end-points. The ViewRoot then
provides the InputChannel to the InputQueue. Behind the
scenes, InputQueue simply attaches handlers to the native PollLoop object
that underlies the MessageQueue. This way MessageQueue doesn't need
to know anything about input dispatch per-se, it just exposes (in native
code) a PollLoop that other components can use to monitor file descriptor
state changes.
There can be zero or more targets for any given input event. Each
input target is specified by its input channel and some parameters
including flags, an X/Y coordinate offset, and the dispatch timeout.
An input target can request either synchronous dispatch (for foreground apps)
or asynchronous dispatch (fire-and-forget for wallpapers and "outside"
targets). Currently, finding the appropriate input targets for an event
requires a call back into the WindowManagerServer from native code.
In the future this will be refactored to avoid most of these callbacks
except as required to handle pending focus transitions.
End-to-end event dispatch mostly works!
To do: event injection, rate limiting, ANRs, testing, optimization, etc.
Change-Id: I8c36b2b9e0a2d27392040ecda0f51b636456de25
2010-04-23 01:58:52 +00:00
|
|
|
}
|
|
|
|
|
2010-07-15 01:48:53 +00:00
|
|
|
if (newMetaState & (AMETA_SHIFT_LEFT_ON | AMETA_SHIFT_RIGHT_ON)) {
|
|
|
|
newMetaState |= AMETA_SHIFT_ON;
|
Native input dispatch rewrite work in progress.
The old dispatch mechanism has been left in place and continues to
be used by default for now. To enable native input dispatch,
edit the ENABLE_NATIVE_DISPATCH constant in WindowManagerPolicy.
Includes part of the new input event NDK API. Some details TBD.
To wire up input dispatch, as the ViewRoot adds a window to the
window session it receives an InputChannel object as an output
argument. The InputChannel encapsulates the file descriptors for a
shared memory region and two pipe end-points. The ViewRoot then
provides the InputChannel to the InputQueue. Behind the
scenes, InputQueue simply attaches handlers to the native PollLoop object
that underlies the MessageQueue. This way MessageQueue doesn't need
to know anything about input dispatch per-se, it just exposes (in native
code) a PollLoop that other components can use to monitor file descriptor
state changes.
There can be zero or more targets for any given input event. Each
input target is specified by its input channel and some parameters
including flags, an X/Y coordinate offset, and the dispatch timeout.
An input target can request either synchronous dispatch (for foreground apps)
or asynchronous dispatch (fire-and-forget for wallpapers and "outside"
targets). Currently, finding the appropriate input targets for an event
requires a call back into the WindowManagerServer from native code.
In the future this will be refactored to avoid most of these callbacks
except as required to handle pending focus transitions.
End-to-end event dispatch mostly works!
To do: event injection, rate limiting, ANRs, testing, optimization, etc.
Change-Id: I8c36b2b9e0a2d27392040ecda0f51b636456de25
2010-04-23 01:58:52 +00:00
|
|
|
}
|
|
|
|
|
|
|
|
return newMetaState;
|
|
|
|
}
|
|
|
|
|
|
|
|
static const int32_t keyCodeRotationMap[][4] = {
|
|
|
|
// key codes enumerated counter-clockwise with the original (unrotated) key first
|
|
|
|
// no rotation, 90 degree rotation, 180 degree rotation, 270 degree rotation
|
2010-06-30 23:10:35 +00:00
|
|
|
{ AKEYCODE_DPAD_DOWN, AKEYCODE_DPAD_RIGHT, AKEYCODE_DPAD_UP, AKEYCODE_DPAD_LEFT },
|
|
|
|
{ AKEYCODE_DPAD_RIGHT, AKEYCODE_DPAD_UP, AKEYCODE_DPAD_LEFT, AKEYCODE_DPAD_DOWN },
|
|
|
|
{ AKEYCODE_DPAD_UP, AKEYCODE_DPAD_LEFT, AKEYCODE_DPAD_DOWN, AKEYCODE_DPAD_RIGHT },
|
|
|
|
{ AKEYCODE_DPAD_LEFT, AKEYCODE_DPAD_DOWN, AKEYCODE_DPAD_RIGHT, AKEYCODE_DPAD_UP },
|
Native input dispatch rewrite work in progress.
The old dispatch mechanism has been left in place and continues to
be used by default for now. To enable native input dispatch,
edit the ENABLE_NATIVE_DISPATCH constant in WindowManagerPolicy.
Includes part of the new input event NDK API. Some details TBD.
To wire up input dispatch, as the ViewRoot adds a window to the
window session it receives an InputChannel object as an output
argument. The InputChannel encapsulates the file descriptors for a
shared memory region and two pipe end-points. The ViewRoot then
provides the InputChannel to the InputQueue. Behind the
scenes, InputQueue simply attaches handlers to the native PollLoop object
that underlies the MessageQueue. This way MessageQueue doesn't need
to know anything about input dispatch per-se, it just exposes (in native
code) a PollLoop that other components can use to monitor file descriptor
state changes.
There can be zero or more targets for any given input event. Each
input target is specified by its input channel and some parameters
including flags, an X/Y coordinate offset, and the dispatch timeout.
An input target can request either synchronous dispatch (for foreground apps)
or asynchronous dispatch (fire-and-forget for wallpapers and "outside"
targets). Currently, finding the appropriate input targets for an event
requires a call back into the WindowManagerServer from native code.
In the future this will be refactored to avoid most of these callbacks
except as required to handle pending focus transitions.
End-to-end event dispatch mostly works!
To do: event injection, rate limiting, ANRs, testing, optimization, etc.
Change-Id: I8c36b2b9e0a2d27392040ecda0f51b636456de25
2010-04-23 01:58:52 +00:00
|
|
|
};
|
|
|
|
static const int keyCodeRotationMapSize =
|
|
|
|
sizeof(keyCodeRotationMap) / sizeof(keyCodeRotationMap[0]);
|
|
|
|
|
|
|
|
int32_t rotateKeyCode(int32_t keyCode, int32_t orientation) {
|
2010-06-15 08:31:58 +00:00
|
|
|
if (orientation != InputReaderPolicyInterface::ROTATION_0) {
|
Native input dispatch rewrite work in progress.
The old dispatch mechanism has been left in place and continues to
be used by default for now. To enable native input dispatch,
edit the ENABLE_NATIVE_DISPATCH constant in WindowManagerPolicy.
Includes part of the new input event NDK API. Some details TBD.
To wire up input dispatch, as the ViewRoot adds a window to the
window session it receives an InputChannel object as an output
argument. The InputChannel encapsulates the file descriptors for a
shared memory region and two pipe end-points. The ViewRoot then
provides the InputChannel to the InputQueue. Behind the
scenes, InputQueue simply attaches handlers to the native PollLoop object
that underlies the MessageQueue. This way MessageQueue doesn't need
to know anything about input dispatch per-se, it just exposes (in native
code) a PollLoop that other components can use to monitor file descriptor
state changes.
There can be zero or more targets for any given input event. Each
input target is specified by its input channel and some parameters
including flags, an X/Y coordinate offset, and the dispatch timeout.
An input target can request either synchronous dispatch (for foreground apps)
or asynchronous dispatch (fire-and-forget for wallpapers and "outside"
targets). Currently, finding the appropriate input targets for an event
requires a call back into the WindowManagerServer from native code.
In the future this will be refactored to avoid most of these callbacks
except as required to handle pending focus transitions.
End-to-end event dispatch mostly works!
To do: event injection, rate limiting, ANRs, testing, optimization, etc.
Change-Id: I8c36b2b9e0a2d27392040ecda0f51b636456de25
2010-04-23 01:58:52 +00:00
|
|
|
for (int i = 0; i < keyCodeRotationMapSize; i++) {
|
|
|
|
if (keyCode == keyCodeRotationMap[i][0]) {
|
|
|
|
return keyCodeRotationMap[i][orientation];
|
|
|
|
}
|
|
|
|
}
|
|
|
|
}
|
|
|
|
return keyCode;
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
|
|
// --- InputReader ---
|
|
|
|
|
|
|
|
InputReader::InputReader(const sp<EventHubInterface>& eventHub,
|
2010-06-15 08:31:58 +00:00
|
|
|
const sp<InputReaderPolicyInterface>& policy,
|
Native input dispatch rewrite work in progress.
The old dispatch mechanism has been left in place and continues to
be used by default for now. To enable native input dispatch,
edit the ENABLE_NATIVE_DISPATCH constant in WindowManagerPolicy.
Includes part of the new input event NDK API. Some details TBD.
To wire up input dispatch, as the ViewRoot adds a window to the
window session it receives an InputChannel object as an output
argument. The InputChannel encapsulates the file descriptors for a
shared memory region and two pipe end-points. The ViewRoot then
provides the InputChannel to the InputQueue. Behind the
scenes, InputQueue simply attaches handlers to the native PollLoop object
that underlies the MessageQueue. This way MessageQueue doesn't need
to know anything about input dispatch per-se, it just exposes (in native
code) a PollLoop that other components can use to monitor file descriptor
state changes.
There can be zero or more targets for any given input event. Each
input target is specified by its input channel and some parameters
including flags, an X/Y coordinate offset, and the dispatch timeout.
An input target can request either synchronous dispatch (for foreground apps)
or asynchronous dispatch (fire-and-forget for wallpapers and "outside"
targets). Currently, finding the appropriate input targets for an event
requires a call back into the WindowManagerServer from native code.
In the future this will be refactored to avoid most of these callbacks
except as required to handle pending focus transitions.
End-to-end event dispatch mostly works!
To do: event injection, rate limiting, ANRs, testing, optimization, etc.
Change-Id: I8c36b2b9e0a2d27392040ecda0f51b636456de25
2010-04-23 01:58:52 +00:00
|
|
|
const sp<InputDispatcherInterface>& dispatcher) :
|
|
|
|
mEventHub(eventHub), mPolicy(policy), mDispatcher(dispatcher) {
|
2010-06-15 08:31:58 +00:00
|
|
|
configureExcludedDevices();
|
Native input dispatch rewrite work in progress.
The old dispatch mechanism has been left in place and continues to
be used by default for now. To enable native input dispatch,
edit the ENABLE_NATIVE_DISPATCH constant in WindowManagerPolicy.
Includes part of the new input event NDK API. Some details TBD.
To wire up input dispatch, as the ViewRoot adds a window to the
window session it receives an InputChannel object as an output
argument. The InputChannel encapsulates the file descriptors for a
shared memory region and two pipe end-points. The ViewRoot then
provides the InputChannel to the InputQueue. Behind the
scenes, InputQueue simply attaches handlers to the native PollLoop object
that underlies the MessageQueue. This way MessageQueue doesn't need
to know anything about input dispatch per-se, it just exposes (in native
code) a PollLoop that other components can use to monitor file descriptor
state changes.
There can be zero or more targets for any given input event. Each
input target is specified by its input channel and some parameters
including flags, an X/Y coordinate offset, and the dispatch timeout.
An input target can request either synchronous dispatch (for foreground apps)
or asynchronous dispatch (fire-and-forget for wallpapers and "outside"
targets). Currently, finding the appropriate input targets for an event
requires a call back into the WindowManagerServer from native code.
In the future this will be refactored to avoid most of these callbacks
except as required to handle pending focus transitions.
End-to-end event dispatch mostly works!
To do: event injection, rate limiting, ANRs, testing, optimization, etc.
Change-Id: I8c36b2b9e0a2d27392040ecda0f51b636456de25
2010-04-23 01:58:52 +00:00
|
|
|
resetGlobalMetaState();
|
|
|
|
resetDisplayProperties();
|
2010-06-15 08:31:58 +00:00
|
|
|
updateExportedVirtualKeyState();
|
Native input dispatch rewrite work in progress.
The old dispatch mechanism has been left in place and continues to
be used by default for now. To enable native input dispatch,
edit the ENABLE_NATIVE_DISPATCH constant in WindowManagerPolicy.
Includes part of the new input event NDK API. Some details TBD.
To wire up input dispatch, as the ViewRoot adds a window to the
window session it receives an InputChannel object as an output
argument. The InputChannel encapsulates the file descriptors for a
shared memory region and two pipe end-points. The ViewRoot then
provides the InputChannel to the InputQueue. Behind the
scenes, InputQueue simply attaches handlers to the native PollLoop object
that underlies the MessageQueue. This way MessageQueue doesn't need
to know anything about input dispatch per-se, it just exposes (in native
code) a PollLoop that other components can use to monitor file descriptor
state changes.
There can be zero or more targets for any given input event. Each
input target is specified by its input channel and some parameters
including flags, an X/Y coordinate offset, and the dispatch timeout.
An input target can request either synchronous dispatch (for foreground apps)
or asynchronous dispatch (fire-and-forget for wallpapers and "outside"
targets). Currently, finding the appropriate input targets for an event
requires a call back into the WindowManagerServer from native code.
In the future this will be refactored to avoid most of these callbacks
except as required to handle pending focus transitions.
End-to-end event dispatch mostly works!
To do: event injection, rate limiting, ANRs, testing, optimization, etc.
Change-Id: I8c36b2b9e0a2d27392040ecda0f51b636456de25
2010-04-23 01:58:52 +00:00
|
|
|
}
|
|
|
|
|
|
|
|
InputReader::~InputReader() {
|
|
|
|
for (size_t i = 0; i < mDevices.size(); i++) {
|
|
|
|
delete mDevices.valueAt(i);
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
|
|
|
void InputReader::loopOnce() {
|
|
|
|
RawEvent rawEvent;
|
|
|
|
mEventHub->getEvent(& rawEvent.deviceId, & rawEvent.type, & rawEvent.scanCode,
|
|
|
|
& rawEvent.keyCode, & rawEvent.flags, & rawEvent.value, & rawEvent.when);
|
|
|
|
|
|
|
|
// Replace the event timestamp so it is in same timebase as java.lang.System.nanoTime()
|
|
|
|
// and android.os.SystemClock.uptimeMillis() as expected by the rest of the system.
|
|
|
|
rawEvent.when = systemTime(SYSTEM_TIME_MONOTONIC);
|
|
|
|
|
|
|
|
#if DEBUG_RAW_EVENTS
|
|
|
|
LOGD("Input event: device=0x%x type=0x%x scancode=%d keycode=%d value=%d",
|
|
|
|
rawEvent.deviceId, rawEvent.type, rawEvent.scanCode, rawEvent.keyCode,
|
|
|
|
rawEvent.value);
|
|
|
|
#endif
|
|
|
|
|
|
|
|
process(& rawEvent);
|
|
|
|
}
|
|
|
|
|
|
|
|
void InputReader::process(const RawEvent* rawEvent) {
|
|
|
|
switch (rawEvent->type) {
|
|
|
|
case EventHubInterface::DEVICE_ADDED:
|
|
|
|
handleDeviceAdded(rawEvent);
|
|
|
|
break;
|
|
|
|
|
|
|
|
case EventHubInterface::DEVICE_REMOVED:
|
|
|
|
handleDeviceRemoved(rawEvent);
|
|
|
|
break;
|
|
|
|
|
|
|
|
case EV_SYN:
|
|
|
|
handleSync(rawEvent);
|
|
|
|
break;
|
|
|
|
|
|
|
|
case EV_KEY:
|
|
|
|
handleKey(rawEvent);
|
|
|
|
break;
|
|
|
|
|
|
|
|
case EV_REL:
|
|
|
|
handleRelativeMotion(rawEvent);
|
|
|
|
break;
|
|
|
|
|
|
|
|
case EV_ABS:
|
|
|
|
handleAbsoluteMotion(rawEvent);
|
|
|
|
break;
|
|
|
|
|
|
|
|
case EV_SW:
|
|
|
|
handleSwitch(rawEvent);
|
|
|
|
break;
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
|
|
|
void InputReader::handleDeviceAdded(const RawEvent* rawEvent) {
|
|
|
|
InputDevice* device = getDevice(rawEvent->deviceId);
|
|
|
|
if (device) {
|
|
|
|
LOGW("Ignoring spurious device added event for deviceId %d.", rawEvent->deviceId);
|
|
|
|
return;
|
|
|
|
}
|
|
|
|
|
|
|
|
addDevice(rawEvent->when, rawEvent->deviceId);
|
|
|
|
}
|
|
|
|
|
|
|
|
void InputReader::handleDeviceRemoved(const RawEvent* rawEvent) {
|
|
|
|
InputDevice* device = getDevice(rawEvent->deviceId);
|
|
|
|
if (! device) {
|
|
|
|
LOGW("Ignoring spurious device removed event for deviceId %d.", rawEvent->deviceId);
|
|
|
|
return;
|
|
|
|
}
|
|
|
|
|
|
|
|
removeDevice(rawEvent->when, device);
|
|
|
|
}
|
|
|
|
|
|
|
|
void InputReader::handleSync(const RawEvent* rawEvent) {
|
|
|
|
InputDevice* device = getNonIgnoredDevice(rawEvent->deviceId);
|
|
|
|
if (! device) return;
|
|
|
|
|
|
|
|
if (rawEvent->scanCode == SYN_MT_REPORT) {
|
|
|
|
// MultiTouch Sync: The driver has returned all data for *one* of the pointers.
|
|
|
|
// We drop pointers with pressure <= 0 since that indicates they are not down.
|
|
|
|
if (device->isMultiTouchScreen()) {
|
|
|
|
uint32_t pointerIndex = device->multiTouchScreen.accumulator.pointerCount;
|
|
|
|
|
|
|
|
if (device->multiTouchScreen.accumulator.pointers[pointerIndex].fields) {
|
|
|
|
if (pointerIndex == MAX_POINTERS) {
|
|
|
|
LOGW("MultiTouch device driver returned more than maximum of %d pointers.",
|
|
|
|
MAX_POINTERS);
|
|
|
|
} else {
|
|
|
|
pointerIndex += 1;
|
|
|
|
device->multiTouchScreen.accumulator.pointerCount = pointerIndex;
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
|
|
|
device->multiTouchScreen.accumulator.pointers[pointerIndex].clear();
|
|
|
|
}
|
|
|
|
} else if (rawEvent->scanCode == SYN_REPORT) {
|
|
|
|
// General Sync: The driver has returned all data for the current event update.
|
|
|
|
if (device->isMultiTouchScreen()) {
|
|
|
|
if (device->multiTouchScreen.accumulator.isDirty()) {
|
|
|
|
onMultiTouchScreenStateChanged(rawEvent->when, device);
|
|
|
|
device->multiTouchScreen.accumulator.clear();
|
|
|
|
}
|
|
|
|
} else if (device->isSingleTouchScreen()) {
|
|
|
|
if (device->singleTouchScreen.accumulator.isDirty()) {
|
|
|
|
onSingleTouchScreenStateChanged(rawEvent->when, device);
|
|
|
|
device->singleTouchScreen.accumulator.clear();
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
|
|
|
if (device->trackball.accumulator.isDirty()) {
|
|
|
|
onTrackballStateChanged(rawEvent->when, device);
|
|
|
|
device->trackball.accumulator.clear();
|
|
|
|
}
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
|
|
|
void InputReader::handleKey(const RawEvent* rawEvent) {
|
|
|
|
InputDevice* device = getNonIgnoredDevice(rawEvent->deviceId);
|
|
|
|
if (! device) return;
|
|
|
|
|
|
|
|
bool down = rawEvent->value != 0;
|
|
|
|
int32_t scanCode = rawEvent->scanCode;
|
|
|
|
|
2010-06-30 23:10:35 +00:00
|
|
|
if (device->isSingleTouchScreen()) {
|
Native input dispatch rewrite work in progress.
The old dispatch mechanism has been left in place and continues to
be used by default for now. To enable native input dispatch,
edit the ENABLE_NATIVE_DISPATCH constant in WindowManagerPolicy.
Includes part of the new input event NDK API. Some details TBD.
To wire up input dispatch, as the ViewRoot adds a window to the
window session it receives an InputChannel object as an output
argument. The InputChannel encapsulates the file descriptors for a
shared memory region and two pipe end-points. The ViewRoot then
provides the InputChannel to the InputQueue. Behind the
scenes, InputQueue simply attaches handlers to the native PollLoop object
that underlies the MessageQueue. This way MessageQueue doesn't need
to know anything about input dispatch per-se, it just exposes (in native
code) a PollLoop that other components can use to monitor file descriptor
state changes.
There can be zero or more targets for any given input event. Each
input target is specified by its input channel and some parameters
including flags, an X/Y coordinate offset, and the dispatch timeout.
An input target can request either synchronous dispatch (for foreground apps)
or asynchronous dispatch (fire-and-forget for wallpapers and "outside"
targets). Currently, finding the appropriate input targets for an event
requires a call back into the WindowManagerServer from native code.
In the future this will be refactored to avoid most of these callbacks
except as required to handle pending focus transitions.
End-to-end event dispatch mostly works!
To do: event injection, rate limiting, ANRs, testing, optimization, etc.
Change-Id: I8c36b2b9e0a2d27392040ecda0f51b636456de25
2010-04-23 01:58:52 +00:00
|
|
|
switch (rawEvent->scanCode) {
|
|
|
|
case BTN_TOUCH:
|
|
|
|
device->singleTouchScreen.accumulator.fields |=
|
|
|
|
InputDevice::SingleTouchScreenState::Accumulator::FIELD_BTN_TOUCH;
|
|
|
|
device->singleTouchScreen.accumulator.btnTouch = down;
|
2010-06-30 23:10:35 +00:00
|
|
|
return;
|
Native input dispatch rewrite work in progress.
The old dispatch mechanism has been left in place and continues to
be used by default for now. To enable native input dispatch,
edit the ENABLE_NATIVE_DISPATCH constant in WindowManagerPolicy.
Includes part of the new input event NDK API. Some details TBD.
To wire up input dispatch, as the ViewRoot adds a window to the
window session it receives an InputChannel object as an output
argument. The InputChannel encapsulates the file descriptors for a
shared memory region and two pipe end-points. The ViewRoot then
provides the InputChannel to the InputQueue. Behind the
scenes, InputQueue simply attaches handlers to the native PollLoop object
that underlies the MessageQueue. This way MessageQueue doesn't need
to know anything about input dispatch per-se, it just exposes (in native
code) a PollLoop that other components can use to monitor file descriptor
state changes.
There can be zero or more targets for any given input event. Each
input target is specified by its input channel and some parameters
including flags, an X/Y coordinate offset, and the dispatch timeout.
An input target can request either synchronous dispatch (for foreground apps)
or asynchronous dispatch (fire-and-forget for wallpapers and "outside"
targets). Currently, finding the appropriate input targets for an event
requires a call back into the WindowManagerServer from native code.
In the future this will be refactored to avoid most of these callbacks
except as required to handle pending focus transitions.
End-to-end event dispatch mostly works!
To do: event injection, rate limiting, ANRs, testing, optimization, etc.
Change-Id: I8c36b2b9e0a2d27392040ecda0f51b636456de25
2010-04-23 01:58:52 +00:00
|
|
|
}
|
2010-06-30 23:10:35 +00:00
|
|
|
}
|
|
|
|
|
|
|
|
if (device->isTrackball()) {
|
Native input dispatch rewrite work in progress.
The old dispatch mechanism has been left in place and continues to
be used by default for now. To enable native input dispatch,
edit the ENABLE_NATIVE_DISPATCH constant in WindowManagerPolicy.
Includes part of the new input event NDK API. Some details TBD.
To wire up input dispatch, as the ViewRoot adds a window to the
window session it receives an InputChannel object as an output
argument. The InputChannel encapsulates the file descriptors for a
shared memory region and two pipe end-points. The ViewRoot then
provides the InputChannel to the InputQueue. Behind the
scenes, InputQueue simply attaches handlers to the native PollLoop object
that underlies the MessageQueue. This way MessageQueue doesn't need
to know anything about input dispatch per-se, it just exposes (in native
code) a PollLoop that other components can use to monitor file descriptor
state changes.
There can be zero or more targets for any given input event. Each
input target is specified by its input channel and some parameters
including flags, an X/Y coordinate offset, and the dispatch timeout.
An input target can request either synchronous dispatch (for foreground apps)
or asynchronous dispatch (fire-and-forget for wallpapers and "outside"
targets). Currently, finding the appropriate input targets for an event
requires a call back into the WindowManagerServer from native code.
In the future this will be refactored to avoid most of these callbacks
except as required to handle pending focus transitions.
End-to-end event dispatch mostly works!
To do: event injection, rate limiting, ANRs, testing, optimization, etc.
Change-Id: I8c36b2b9e0a2d27392040ecda0f51b636456de25
2010-04-23 01:58:52 +00:00
|
|
|
switch (rawEvent->scanCode) {
|
|
|
|
case BTN_MOUSE:
|
|
|
|
device->trackball.accumulator.fields |=
|
|
|
|
InputDevice::TrackballState::Accumulator::FIELD_BTN_MOUSE;
|
|
|
|
device->trackball.accumulator.btnMouse = down;
|
2010-07-17 00:21:06 +00:00
|
|
|
|
|
|
|
// Process the trackball change now since we may not receive a sync immediately.
|
|
|
|
onTrackballStateChanged(rawEvent->when, device);
|
|
|
|
device->trackball.accumulator.clear();
|
2010-06-30 23:10:35 +00:00
|
|
|
return;
|
Native input dispatch rewrite work in progress.
The old dispatch mechanism has been left in place and continues to
be used by default for now. To enable native input dispatch,
edit the ENABLE_NATIVE_DISPATCH constant in WindowManagerPolicy.
Includes part of the new input event NDK API. Some details TBD.
To wire up input dispatch, as the ViewRoot adds a window to the
window session it receives an InputChannel object as an output
argument. The InputChannel encapsulates the file descriptors for a
shared memory region and two pipe end-points. The ViewRoot then
provides the InputChannel to the InputQueue. Behind the
scenes, InputQueue simply attaches handlers to the native PollLoop object
that underlies the MessageQueue. This way MessageQueue doesn't need
to know anything about input dispatch per-se, it just exposes (in native
code) a PollLoop that other components can use to monitor file descriptor
state changes.
There can be zero or more targets for any given input event. Each
input target is specified by its input channel and some parameters
including flags, an X/Y coordinate offset, and the dispatch timeout.
An input target can request either synchronous dispatch (for foreground apps)
or asynchronous dispatch (fire-and-forget for wallpapers and "outside"
targets). Currently, finding the appropriate input targets for an event
requires a call back into the WindowManagerServer from native code.
In the future this will be refactored to avoid most of these callbacks
except as required to handle pending focus transitions.
End-to-end event dispatch mostly works!
To do: event injection, rate limiting, ANRs, testing, optimization, etc.
Change-Id: I8c36b2b9e0a2d27392040ecda0f51b636456de25
2010-04-23 01:58:52 +00:00
|
|
|
}
|
|
|
|
}
|
2010-06-30 23:10:35 +00:00
|
|
|
|
|
|
|
if (device->isKeyboard()) {
|
|
|
|
int32_t keyCode = rawEvent->keyCode;
|
|
|
|
onKey(rawEvent->when, device, down, keyCode, scanCode, rawEvent->flags);
|
|
|
|
}
|
Native input dispatch rewrite work in progress.
The old dispatch mechanism has been left in place and continues to
be used by default for now. To enable native input dispatch,
edit the ENABLE_NATIVE_DISPATCH constant in WindowManagerPolicy.
Includes part of the new input event NDK API. Some details TBD.
To wire up input dispatch, as the ViewRoot adds a window to the
window session it receives an InputChannel object as an output
argument. The InputChannel encapsulates the file descriptors for a
shared memory region and two pipe end-points. The ViewRoot then
provides the InputChannel to the InputQueue. Behind the
scenes, InputQueue simply attaches handlers to the native PollLoop object
that underlies the MessageQueue. This way MessageQueue doesn't need
to know anything about input dispatch per-se, it just exposes (in native
code) a PollLoop that other components can use to monitor file descriptor
state changes.
There can be zero or more targets for any given input event. Each
input target is specified by its input channel and some parameters
including flags, an X/Y coordinate offset, and the dispatch timeout.
An input target can request either synchronous dispatch (for foreground apps)
or asynchronous dispatch (fire-and-forget for wallpapers and "outside"
targets). Currently, finding the appropriate input targets for an event
requires a call back into the WindowManagerServer from native code.
In the future this will be refactored to avoid most of these callbacks
except as required to handle pending focus transitions.
End-to-end event dispatch mostly works!
To do: event injection, rate limiting, ANRs, testing, optimization, etc.
Change-Id: I8c36b2b9e0a2d27392040ecda0f51b636456de25
2010-04-23 01:58:52 +00:00
|
|
|
}
|
|
|
|
|
|
|
|
void InputReader::handleRelativeMotion(const RawEvent* rawEvent) {
|
|
|
|
InputDevice* device = getNonIgnoredDevice(rawEvent->deviceId);
|
|
|
|
if (! device) return;
|
|
|
|
|
|
|
|
if (device->isTrackball()) {
|
|
|
|
switch (rawEvent->scanCode) {
|
|
|
|
case REL_X:
|
|
|
|
device->trackball.accumulator.fields |=
|
|
|
|
InputDevice::TrackballState::Accumulator::FIELD_REL_X;
|
|
|
|
device->trackball.accumulator.relX = rawEvent->value;
|
|
|
|
break;
|
|
|
|
case REL_Y:
|
|
|
|
device->trackball.accumulator.fields |=
|
|
|
|
InputDevice::TrackballState::Accumulator::FIELD_REL_Y;
|
|
|
|
device->trackball.accumulator.relY = rawEvent->value;
|
|
|
|
break;
|
|
|
|
}
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
|
|
|
void InputReader::handleAbsoluteMotion(const RawEvent* rawEvent) {
|
|
|
|
InputDevice* device = getNonIgnoredDevice(rawEvent->deviceId);
|
|
|
|
if (! device) return;
|
|
|
|
|
|
|
|
if (device->isMultiTouchScreen()) {
|
|
|
|
uint32_t pointerIndex = device->multiTouchScreen.accumulator.pointerCount;
|
|
|
|
InputDevice::MultiTouchScreenState::Accumulator::Pointer* pointer =
|
|
|
|
& device->multiTouchScreen.accumulator.pointers[pointerIndex];
|
|
|
|
|
|
|
|
switch (rawEvent->scanCode) {
|
|
|
|
case ABS_MT_POSITION_X:
|
|
|
|
pointer->fields |=
|
|
|
|
InputDevice::MultiTouchScreenState::Accumulator::FIELD_ABS_MT_POSITION_X;
|
|
|
|
pointer->absMTPositionX = rawEvent->value;
|
|
|
|
break;
|
|
|
|
case ABS_MT_POSITION_Y:
|
|
|
|
pointer->fields |=
|
|
|
|
InputDevice::MultiTouchScreenState::Accumulator::FIELD_ABS_MT_POSITION_Y;
|
|
|
|
pointer->absMTPositionY = rawEvent->value;
|
|
|
|
break;
|
|
|
|
case ABS_MT_TOUCH_MAJOR:
|
|
|
|
pointer->fields |=
|
|
|
|
InputDevice::MultiTouchScreenState::Accumulator::FIELD_ABS_MT_TOUCH_MAJOR;
|
|
|
|
pointer->absMTTouchMajor = rawEvent->value;
|
|
|
|
break;
|
2010-07-15 01:48:53 +00:00
|
|
|
case ABS_MT_TOUCH_MINOR:
|
|
|
|
pointer->fields |=
|
|
|
|
InputDevice::MultiTouchScreenState::Accumulator::FIELD_ABS_MT_TOUCH_MINOR;
|
|
|
|
pointer->absMTTouchMinor = rawEvent->value;
|
|
|
|
break;
|
Native input dispatch rewrite work in progress.
The old dispatch mechanism has been left in place and continues to
be used by default for now. To enable native input dispatch,
edit the ENABLE_NATIVE_DISPATCH constant in WindowManagerPolicy.
Includes part of the new input event NDK API. Some details TBD.
To wire up input dispatch, as the ViewRoot adds a window to the
window session it receives an InputChannel object as an output
argument. The InputChannel encapsulates the file descriptors for a
shared memory region and two pipe end-points. The ViewRoot then
provides the InputChannel to the InputQueue. Behind the
scenes, InputQueue simply attaches handlers to the native PollLoop object
that underlies the MessageQueue. This way MessageQueue doesn't need
to know anything about input dispatch per-se, it just exposes (in native
code) a PollLoop that other components can use to monitor file descriptor
state changes.
There can be zero or more targets for any given input event. Each
input target is specified by its input channel and some parameters
including flags, an X/Y coordinate offset, and the dispatch timeout.
An input target can request either synchronous dispatch (for foreground apps)
or asynchronous dispatch (fire-and-forget for wallpapers and "outside"
targets). Currently, finding the appropriate input targets for an event
requires a call back into the WindowManagerServer from native code.
In the future this will be refactored to avoid most of these callbacks
except as required to handle pending focus transitions.
End-to-end event dispatch mostly works!
To do: event injection, rate limiting, ANRs, testing, optimization, etc.
Change-Id: I8c36b2b9e0a2d27392040ecda0f51b636456de25
2010-04-23 01:58:52 +00:00
|
|
|
case ABS_MT_WIDTH_MAJOR:
|
|
|
|
pointer->fields |=
|
|
|
|
InputDevice::MultiTouchScreenState::Accumulator::FIELD_ABS_MT_WIDTH_MAJOR;
|
|
|
|
pointer->absMTWidthMajor = rawEvent->value;
|
|
|
|
break;
|
2010-07-15 01:48:53 +00:00
|
|
|
case ABS_MT_WIDTH_MINOR:
|
|
|
|
pointer->fields |=
|
|
|
|
InputDevice::MultiTouchScreenState::Accumulator::FIELD_ABS_MT_WIDTH_MINOR;
|
|
|
|
pointer->absMTWidthMinor = rawEvent->value;
|
|
|
|
break;
|
|
|
|
case ABS_MT_ORIENTATION:
|
|
|
|
pointer->fields |=
|
|
|
|
InputDevice::MultiTouchScreenState::Accumulator::FIELD_ABS_MT_ORIENTATION;
|
|
|
|
pointer->absMTOrientation = rawEvent->value;
|
|
|
|
break;
|
Native input dispatch rewrite work in progress.
The old dispatch mechanism has been left in place and continues to
be used by default for now. To enable native input dispatch,
edit the ENABLE_NATIVE_DISPATCH constant in WindowManagerPolicy.
Includes part of the new input event NDK API. Some details TBD.
To wire up input dispatch, as the ViewRoot adds a window to the
window session it receives an InputChannel object as an output
argument. The InputChannel encapsulates the file descriptors for a
shared memory region and two pipe end-points. The ViewRoot then
provides the InputChannel to the InputQueue. Behind the
scenes, InputQueue simply attaches handlers to the native PollLoop object
that underlies the MessageQueue. This way MessageQueue doesn't need
to know anything about input dispatch per-se, it just exposes (in native
code) a PollLoop that other components can use to monitor file descriptor
state changes.
There can be zero or more targets for any given input event. Each
input target is specified by its input channel and some parameters
including flags, an X/Y coordinate offset, and the dispatch timeout.
An input target can request either synchronous dispatch (for foreground apps)
or asynchronous dispatch (fire-and-forget for wallpapers and "outside"
targets). Currently, finding the appropriate input targets for an event
requires a call back into the WindowManagerServer from native code.
In the future this will be refactored to avoid most of these callbacks
except as required to handle pending focus transitions.
End-to-end event dispatch mostly works!
To do: event injection, rate limiting, ANRs, testing, optimization, etc.
Change-Id: I8c36b2b9e0a2d27392040ecda0f51b636456de25
2010-04-23 01:58:52 +00:00
|
|
|
case ABS_MT_TRACKING_ID:
|
|
|
|
pointer->fields |=
|
|
|
|
InputDevice::MultiTouchScreenState::Accumulator::FIELD_ABS_MT_TRACKING_ID;
|
|
|
|
pointer->absMTTrackingId = rawEvent->value;
|
|
|
|
break;
|
|
|
|
}
|
|
|
|
} else if (device->isSingleTouchScreen()) {
|
|
|
|
switch (rawEvent->scanCode) {
|
|
|
|
case ABS_X:
|
|
|
|
device->singleTouchScreen.accumulator.fields |=
|
|
|
|
InputDevice::SingleTouchScreenState::Accumulator::FIELD_ABS_X;
|
|
|
|
device->singleTouchScreen.accumulator.absX = rawEvent->value;
|
|
|
|
break;
|
|
|
|
case ABS_Y:
|
|
|
|
device->singleTouchScreen.accumulator.fields |=
|
|
|
|
InputDevice::SingleTouchScreenState::Accumulator::FIELD_ABS_Y;
|
|
|
|
device->singleTouchScreen.accumulator.absY = rawEvent->value;
|
|
|
|
break;
|
|
|
|
case ABS_PRESSURE:
|
|
|
|
device->singleTouchScreen.accumulator.fields |=
|
|
|
|
InputDevice::SingleTouchScreenState::Accumulator::FIELD_ABS_PRESSURE;
|
|
|
|
device->singleTouchScreen.accumulator.absPressure = rawEvent->value;
|
|
|
|
break;
|
|
|
|
case ABS_TOOL_WIDTH:
|
|
|
|
device->singleTouchScreen.accumulator.fields |=
|
|
|
|
InputDevice::SingleTouchScreenState::Accumulator::FIELD_ABS_TOOL_WIDTH;
|
|
|
|
device->singleTouchScreen.accumulator.absToolWidth = rawEvent->value;
|
|
|
|
break;
|
|
|
|
}
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
|
|
|
void InputReader::handleSwitch(const RawEvent* rawEvent) {
|
|
|
|
InputDevice* device = getNonIgnoredDevice(rawEvent->deviceId);
|
|
|
|
if (! device) return;
|
|
|
|
|
2010-06-15 08:31:58 +00:00
|
|
|
onSwitch(rawEvent->when, device, rawEvent->scanCode, rawEvent->value);
|
Native input dispatch rewrite work in progress.
The old dispatch mechanism has been left in place and continues to
be used by default for now. To enable native input dispatch,
edit the ENABLE_NATIVE_DISPATCH constant in WindowManagerPolicy.
Includes part of the new input event NDK API. Some details TBD.
To wire up input dispatch, as the ViewRoot adds a window to the
window session it receives an InputChannel object as an output
argument. The InputChannel encapsulates the file descriptors for a
shared memory region and two pipe end-points. The ViewRoot then
provides the InputChannel to the InputQueue. Behind the
scenes, InputQueue simply attaches handlers to the native PollLoop object
that underlies the MessageQueue. This way MessageQueue doesn't need
to know anything about input dispatch per-se, it just exposes (in native
code) a PollLoop that other components can use to monitor file descriptor
state changes.
There can be zero or more targets for any given input event. Each
input target is specified by its input channel and some parameters
including flags, an X/Y coordinate offset, and the dispatch timeout.
An input target can request either synchronous dispatch (for foreground apps)
or asynchronous dispatch (fire-and-forget for wallpapers and "outside"
targets). Currently, finding the appropriate input targets for an event
requires a call back into the WindowManagerServer from native code.
In the future this will be refactored to avoid most of these callbacks
except as required to handle pending focus transitions.
End-to-end event dispatch mostly works!
To do: event injection, rate limiting, ANRs, testing, optimization, etc.
Change-Id: I8c36b2b9e0a2d27392040ecda0f51b636456de25
2010-04-23 01:58:52 +00:00
|
|
|
}
|
|
|
|
|
|
|
|
void InputReader::onKey(nsecs_t when, InputDevice* device,
|
|
|
|
bool down, int32_t keyCode, int32_t scanCode, uint32_t policyFlags) {
|
|
|
|
/* Refresh display properties so we can rotate key codes according to display orientation */
|
|
|
|
|
|
|
|
if (! refreshDisplayProperties()) {
|
|
|
|
return;
|
|
|
|
}
|
|
|
|
|
|
|
|
/* Update device state */
|
|
|
|
|
|
|
|
int32_t oldMetaState = device->keyboard.current.metaState;
|
|
|
|
int32_t newMetaState = updateMetaState(keyCode, down, oldMetaState);
|
|
|
|
if (oldMetaState != newMetaState) {
|
|
|
|
device->keyboard.current.metaState = newMetaState;
|
|
|
|
resetGlobalMetaState();
|
|
|
|
}
|
|
|
|
|
|
|
|
// FIXME if we send a down event about a rotated key press we should ensure that we send
|
|
|
|
// a corresponding up event about the rotated key press even if the orientation
|
|
|
|
// has changed in the meantime
|
|
|
|
keyCode = rotateKeyCode(keyCode, mDisplayOrientation);
|
|
|
|
|
|
|
|
if (down) {
|
|
|
|
device->keyboard.current.downTime = when;
|
|
|
|
}
|
|
|
|
|
|
|
|
/* Apply policy */
|
|
|
|
|
|
|
|
int32_t policyActions = mPolicy->interceptKey(when, device->id,
|
|
|
|
down, keyCode, scanCode, policyFlags);
|
|
|
|
|
|
|
|
if (! applyStandardInputDispatchPolicyActions(when, policyActions, & policyFlags)) {
|
|
|
|
return; // event dropped
|
|
|
|
}
|
|
|
|
|
|
|
|
/* Enqueue key event for dispatch */
|
|
|
|
|
|
|
|
int32_t keyEventAction;
|
|
|
|
if (down) {
|
|
|
|
device->keyboard.current.downTime = when;
|
2010-07-15 01:48:53 +00:00
|
|
|
keyEventAction = AKEY_EVENT_ACTION_DOWN;
|
Native input dispatch rewrite work in progress.
The old dispatch mechanism has been left in place and continues to
be used by default for now. To enable native input dispatch,
edit the ENABLE_NATIVE_DISPATCH constant in WindowManagerPolicy.
Includes part of the new input event NDK API. Some details TBD.
To wire up input dispatch, as the ViewRoot adds a window to the
window session it receives an InputChannel object as an output
argument. The InputChannel encapsulates the file descriptors for a
shared memory region and two pipe end-points. The ViewRoot then
provides the InputChannel to the InputQueue. Behind the
scenes, InputQueue simply attaches handlers to the native PollLoop object
that underlies the MessageQueue. This way MessageQueue doesn't need
to know anything about input dispatch per-se, it just exposes (in native
code) a PollLoop that other components can use to monitor file descriptor
state changes.
There can be zero or more targets for any given input event. Each
input target is specified by its input channel and some parameters
including flags, an X/Y coordinate offset, and the dispatch timeout.
An input target can request either synchronous dispatch (for foreground apps)
or asynchronous dispatch (fire-and-forget for wallpapers and "outside"
targets). Currently, finding the appropriate input targets for an event
requires a call back into the WindowManagerServer from native code.
In the future this will be refactored to avoid most of these callbacks
except as required to handle pending focus transitions.
End-to-end event dispatch mostly works!
To do: event injection, rate limiting, ANRs, testing, optimization, etc.
Change-Id: I8c36b2b9e0a2d27392040ecda0f51b636456de25
2010-04-23 01:58:52 +00:00
|
|
|
} else {
|
2010-07-15 01:48:53 +00:00
|
|
|
keyEventAction = AKEY_EVENT_ACTION_UP;
|
Native input dispatch rewrite work in progress.
The old dispatch mechanism has been left in place and continues to
be used by default for now. To enable native input dispatch,
edit the ENABLE_NATIVE_DISPATCH constant in WindowManagerPolicy.
Includes part of the new input event NDK API. Some details TBD.
To wire up input dispatch, as the ViewRoot adds a window to the
window session it receives an InputChannel object as an output
argument. The InputChannel encapsulates the file descriptors for a
shared memory region and two pipe end-points. The ViewRoot then
provides the InputChannel to the InputQueue. Behind the
scenes, InputQueue simply attaches handlers to the native PollLoop object
that underlies the MessageQueue. This way MessageQueue doesn't need
to know anything about input dispatch per-se, it just exposes (in native
code) a PollLoop that other components can use to monitor file descriptor
state changes.
There can be zero or more targets for any given input event. Each
input target is specified by its input channel and some parameters
including flags, an X/Y coordinate offset, and the dispatch timeout.
An input target can request either synchronous dispatch (for foreground apps)
or asynchronous dispatch (fire-and-forget for wallpapers and "outside"
targets). Currently, finding the appropriate input targets for an event
requires a call back into the WindowManagerServer from native code.
In the future this will be refactored to avoid most of these callbacks
except as required to handle pending focus transitions.
End-to-end event dispatch mostly works!
To do: event injection, rate limiting, ANRs, testing, optimization, etc.
Change-Id: I8c36b2b9e0a2d27392040ecda0f51b636456de25
2010-04-23 01:58:52 +00:00
|
|
|
}
|
|
|
|
|
2010-07-15 01:48:53 +00:00
|
|
|
int32_t keyEventFlags = AKEY_EVENT_FLAG_FROM_SYSTEM;
|
2010-06-15 08:31:58 +00:00
|
|
|
if (policyActions & InputReaderPolicyInterface::ACTION_WOKE_HERE) {
|
2010-07-15 01:48:53 +00:00
|
|
|
keyEventFlags = keyEventFlags | AKEY_EVENT_FLAG_WOKE_HERE;
|
Native input dispatch rewrite work in progress.
The old dispatch mechanism has been left in place and continues to
be used by default for now. To enable native input dispatch,
edit the ENABLE_NATIVE_DISPATCH constant in WindowManagerPolicy.
Includes part of the new input event NDK API. Some details TBD.
To wire up input dispatch, as the ViewRoot adds a window to the
window session it receives an InputChannel object as an output
argument. The InputChannel encapsulates the file descriptors for a
shared memory region and two pipe end-points. The ViewRoot then
provides the InputChannel to the InputQueue. Behind the
scenes, InputQueue simply attaches handlers to the native PollLoop object
that underlies the MessageQueue. This way MessageQueue doesn't need
to know anything about input dispatch per-se, it just exposes (in native
code) a PollLoop that other components can use to monitor file descriptor
state changes.
There can be zero or more targets for any given input event. Each
input target is specified by its input channel and some parameters
including flags, an X/Y coordinate offset, and the dispatch timeout.
An input target can request either synchronous dispatch (for foreground apps)
or asynchronous dispatch (fire-and-forget for wallpapers and "outside"
targets). Currently, finding the appropriate input targets for an event
requires a call back into the WindowManagerServer from native code.
In the future this will be refactored to avoid most of these callbacks
except as required to handle pending focus transitions.
End-to-end event dispatch mostly works!
To do: event injection, rate limiting, ANRs, testing, optimization, etc.
Change-Id: I8c36b2b9e0a2d27392040ecda0f51b636456de25
2010-04-23 01:58:52 +00:00
|
|
|
}
|
|
|
|
|
2010-07-15 01:48:53 +00:00
|
|
|
mDispatcher->notifyKey(when, device->id, AINPUT_SOURCE_KEYBOARD, policyFlags,
|
Native input dispatch rewrite work in progress.
The old dispatch mechanism has been left in place and continues to
be used by default for now. To enable native input dispatch,
edit the ENABLE_NATIVE_DISPATCH constant in WindowManagerPolicy.
Includes part of the new input event NDK API. Some details TBD.
To wire up input dispatch, as the ViewRoot adds a window to the
window session it receives an InputChannel object as an output
argument. The InputChannel encapsulates the file descriptors for a
shared memory region and two pipe end-points. The ViewRoot then
provides the InputChannel to the InputQueue. Behind the
scenes, InputQueue simply attaches handlers to the native PollLoop object
that underlies the MessageQueue. This way MessageQueue doesn't need
to know anything about input dispatch per-se, it just exposes (in native
code) a PollLoop that other components can use to monitor file descriptor
state changes.
There can be zero or more targets for any given input event. Each
input target is specified by its input channel and some parameters
including flags, an X/Y coordinate offset, and the dispatch timeout.
An input target can request either synchronous dispatch (for foreground apps)
or asynchronous dispatch (fire-and-forget for wallpapers and "outside"
targets). Currently, finding the appropriate input targets for an event
requires a call back into the WindowManagerServer from native code.
In the future this will be refactored to avoid most of these callbacks
except as required to handle pending focus transitions.
End-to-end event dispatch mostly works!
To do: event injection, rate limiting, ANRs, testing, optimization, etc.
Change-Id: I8c36b2b9e0a2d27392040ecda0f51b636456de25
2010-04-23 01:58:52 +00:00
|
|
|
keyEventAction, keyEventFlags, keyCode, scanCode,
|
|
|
|
device->keyboard.current.metaState,
|
|
|
|
device->keyboard.current.downTime);
|
|
|
|
}
|
|
|
|
|
2010-06-15 08:31:58 +00:00
|
|
|
void InputReader::onSwitch(nsecs_t when, InputDevice* device, int32_t switchCode,
|
|
|
|
int32_t switchValue) {
|
|
|
|
int32_t policyActions = mPolicy->interceptSwitch(when, switchCode, switchValue);
|
|
|
|
|
|
|
|
uint32_t policyFlags = 0;
|
|
|
|
applyStandardInputDispatchPolicyActions(when, policyActions, & policyFlags);
|
Native input dispatch rewrite work in progress.
The old dispatch mechanism has been left in place and continues to
be used by default for now. To enable native input dispatch,
edit the ENABLE_NATIVE_DISPATCH constant in WindowManagerPolicy.
Includes part of the new input event NDK API. Some details TBD.
To wire up input dispatch, as the ViewRoot adds a window to the
window session it receives an InputChannel object as an output
argument. The InputChannel encapsulates the file descriptors for a
shared memory region and two pipe end-points. The ViewRoot then
provides the InputChannel to the InputQueue. Behind the
scenes, InputQueue simply attaches handlers to the native PollLoop object
that underlies the MessageQueue. This way MessageQueue doesn't need
to know anything about input dispatch per-se, it just exposes (in native
code) a PollLoop that other components can use to monitor file descriptor
state changes.
There can be zero or more targets for any given input event. Each
input target is specified by its input channel and some parameters
including flags, an X/Y coordinate offset, and the dispatch timeout.
An input target can request either synchronous dispatch (for foreground apps)
or asynchronous dispatch (fire-and-forget for wallpapers and "outside"
targets). Currently, finding the appropriate input targets for an event
requires a call back into the WindowManagerServer from native code.
In the future this will be refactored to avoid most of these callbacks
except as required to handle pending focus transitions.
End-to-end event dispatch mostly works!
To do: event injection, rate limiting, ANRs, testing, optimization, etc.
Change-Id: I8c36b2b9e0a2d27392040ecda0f51b636456de25
2010-04-23 01:58:52 +00:00
|
|
|
}
|
|
|
|
|
|
|
|
void InputReader::onMultiTouchScreenStateChanged(nsecs_t when,
|
|
|
|
InputDevice* device) {
|
|
|
|
static const uint32_t REQUIRED_FIELDS =
|
|
|
|
InputDevice::MultiTouchScreenState::Accumulator::FIELD_ABS_MT_POSITION_X
|
|
|
|
| InputDevice::MultiTouchScreenState::Accumulator::FIELD_ABS_MT_POSITION_Y
|
|
|
|
| InputDevice::MultiTouchScreenState::Accumulator::FIELD_ABS_MT_TOUCH_MAJOR
|
|
|
|
| InputDevice::MultiTouchScreenState::Accumulator::FIELD_ABS_MT_WIDTH_MAJOR;
|
|
|
|
|
|
|
|
/* Refresh display properties so we can map touch screen coords into display coords */
|
|
|
|
|
|
|
|
if (! refreshDisplayProperties()) {
|
|
|
|
return;
|
|
|
|
}
|
|
|
|
|
|
|
|
/* Update device state */
|
|
|
|
|
|
|
|
InputDevice::MultiTouchScreenState* in = & device->multiTouchScreen;
|
|
|
|
InputDevice::TouchData* out = & device->touchScreen.currentTouch;
|
|
|
|
|
|
|
|
uint32_t inCount = in->accumulator.pointerCount;
|
|
|
|
uint32_t outCount = 0;
|
|
|
|
bool havePointerIds = true;
|
|
|
|
|
|
|
|
out->clear();
|
|
|
|
|
|
|
|
for (uint32_t inIndex = 0; inIndex < inCount; inIndex++) {
|
|
|
|
uint32_t fields = in->accumulator.pointers[inIndex].fields;
|
|
|
|
|
|
|
|
if ((fields & REQUIRED_FIELDS) != REQUIRED_FIELDS) {
|
|
|
|
#if DEBUG_POINTERS
|
|
|
|
LOGD("Pointers: Missing required multitouch pointer fields: index=%d, fields=%d",
|
|
|
|
inIndex, fields);
|
|
|
|
continue;
|
|
|
|
#endif
|
|
|
|
}
|
|
|
|
|
|
|
|
if (in->accumulator.pointers[inIndex].absMTTouchMajor <= 0) {
|
|
|
|
// Pointer is not down. Drop it.
|
|
|
|
continue;
|
|
|
|
}
|
|
|
|
|
2010-07-15 01:48:53 +00:00
|
|
|
out->pointers[outCount].x = in->accumulator.pointers[inIndex].absMTPositionX;
|
|
|
|
out->pointers[outCount].y = in->accumulator.pointers[inIndex].absMTPositionY;
|
|
|
|
|
|
|
|
out->pointers[outCount].touchMajor = in->accumulator.pointers[inIndex].absMTTouchMajor;
|
|
|
|
out->pointers[outCount].touchMinor = (fields
|
|
|
|
& InputDevice::MultiTouchScreenState::Accumulator::FIELD_ABS_MT_TOUCH_MINOR) != 0
|
|
|
|
? in->accumulator.pointers[inIndex].absMTTouchMinor
|
|
|
|
: in->accumulator.pointers[inIndex].absMTTouchMajor;
|
|
|
|
|
|
|
|
out->pointers[outCount].toolMajor = in->accumulator.pointers[inIndex].absMTWidthMajor;
|
|
|
|
out->pointers[outCount].toolMinor = (fields
|
|
|
|
& InputDevice::MultiTouchScreenState::Accumulator::FIELD_ABS_MT_WIDTH_MINOR) != 0
|
|
|
|
? in->accumulator.pointers[inIndex].absMTWidthMinor
|
|
|
|
: in->accumulator.pointers[inIndex].absMTWidthMajor;
|
|
|
|
|
|
|
|
out->pointers[outCount].orientation = (fields
|
|
|
|
& InputDevice::MultiTouchScreenState::Accumulator::FIELD_ABS_MT_ORIENTATION) != 0
|
|
|
|
? in->accumulator.pointers[inIndex].absMTOrientation : 0;
|
|
|
|
|
|
|
|
// Derive an approximation of pressure and size.
|
Native input dispatch rewrite work in progress.
The old dispatch mechanism has been left in place and continues to
be used by default for now. To enable native input dispatch,
edit the ENABLE_NATIVE_DISPATCH constant in WindowManagerPolicy.
Includes part of the new input event NDK API. Some details TBD.
To wire up input dispatch, as the ViewRoot adds a window to the
window session it receives an InputChannel object as an output
argument. The InputChannel encapsulates the file descriptors for a
shared memory region and two pipe end-points. The ViewRoot then
provides the InputChannel to the InputQueue. Behind the
scenes, InputQueue simply attaches handlers to the native PollLoop object
that underlies the MessageQueue. This way MessageQueue doesn't need
to know anything about input dispatch per-se, it just exposes (in native
code) a PollLoop that other components can use to monitor file descriptor
state changes.
There can be zero or more targets for any given input event. Each
input target is specified by its input channel and some parameters
including flags, an X/Y coordinate offset, and the dispatch timeout.
An input target can request either synchronous dispatch (for foreground apps)
or asynchronous dispatch (fire-and-forget for wallpapers and "outside"
targets). Currently, finding the appropriate input targets for an event
requires a call back into the WindowManagerServer from native code.
In the future this will be refactored to avoid most of these callbacks
except as required to handle pending focus transitions.
End-to-end event dispatch mostly works!
To do: event injection, rate limiting, ANRs, testing, optimization, etc.
Change-Id: I8c36b2b9e0a2d27392040ecda0f51b636456de25
2010-04-23 01:58:52 +00:00
|
|
|
// FIXME assignment of pressure may be incorrect, probably better to let
|
|
|
|
// pressure = touch / width. Later on we pass width to MotionEvent as a size, which
|
|
|
|
// isn't quite right either. Should be using touch for that.
|
|
|
|
out->pointers[outCount].pressure = in->accumulator.pointers[inIndex].absMTTouchMajor;
|
|
|
|
out->pointers[outCount].size = in->accumulator.pointers[inIndex].absMTWidthMajor;
|
|
|
|
|
|
|
|
if (havePointerIds) {
|
|
|
|
if (fields & InputDevice::MultiTouchScreenState::Accumulator::
|
|
|
|
FIELD_ABS_MT_TRACKING_ID) {
|
|
|
|
uint32_t id = uint32_t(in->accumulator.pointers[inIndex].absMTTrackingId);
|
|
|
|
|
|
|
|
if (id > MAX_POINTER_ID) {
|
|
|
|
#if DEBUG_POINTERS
|
|
|
|
LOGD("Pointers: Ignoring driver provided pointer id %d because "
|
|
|
|
"it is larger than max supported id %d for optimizations",
|
|
|
|
id, MAX_POINTER_ID);
|
|
|
|
#endif
|
|
|
|
havePointerIds = false;
|
|
|
|
}
|
|
|
|
else {
|
|
|
|
out->pointers[outCount].id = id;
|
|
|
|
out->idToIndex[id] = outCount;
|
|
|
|
out->idBits.markBit(id);
|
|
|
|
}
|
|
|
|
} else {
|
|
|
|
havePointerIds = false;
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
|
|
|
outCount += 1;
|
|
|
|
}
|
|
|
|
|
|
|
|
out->pointerCount = outCount;
|
|
|
|
|
|
|
|
onTouchScreenChanged(when, device, havePointerIds);
|
|
|
|
}
|
|
|
|
|
|
|
|
void InputReader::onSingleTouchScreenStateChanged(nsecs_t when,
|
|
|
|
InputDevice* device) {
|
|
|
|
/* Refresh display properties so we can map touch screen coords into display coords */
|
|
|
|
|
|
|
|
if (! refreshDisplayProperties()) {
|
|
|
|
return;
|
|
|
|
}
|
|
|
|
|
|
|
|
/* Update device state */
|
|
|
|
|
|
|
|
InputDevice::SingleTouchScreenState* in = & device->singleTouchScreen;
|
|
|
|
InputDevice::TouchData* out = & device->touchScreen.currentTouch;
|
|
|
|
|
|
|
|
uint32_t fields = in->accumulator.fields;
|
|
|
|
|
|
|
|
if (fields & InputDevice::SingleTouchScreenState::Accumulator::FIELD_BTN_TOUCH) {
|
|
|
|
in->current.down = in->accumulator.btnTouch;
|
|
|
|
}
|
|
|
|
|
2010-06-22 08:27:15 +00:00
|
|
|
if (fields & InputDevice::SingleTouchScreenState::Accumulator::FIELD_ABS_X) {
|
Native input dispatch rewrite work in progress.
The old dispatch mechanism has been left in place and continues to
be used by default for now. To enable native input dispatch,
edit the ENABLE_NATIVE_DISPATCH constant in WindowManagerPolicy.
Includes part of the new input event NDK API. Some details TBD.
To wire up input dispatch, as the ViewRoot adds a window to the
window session it receives an InputChannel object as an output
argument. The InputChannel encapsulates the file descriptors for a
shared memory region and two pipe end-points. The ViewRoot then
provides the InputChannel to the InputQueue. Behind the
scenes, InputQueue simply attaches handlers to the native PollLoop object
that underlies the MessageQueue. This way MessageQueue doesn't need
to know anything about input dispatch per-se, it just exposes (in native
code) a PollLoop that other components can use to monitor file descriptor
state changes.
There can be zero or more targets for any given input event. Each
input target is specified by its input channel and some parameters
including flags, an X/Y coordinate offset, and the dispatch timeout.
An input target can request either synchronous dispatch (for foreground apps)
or asynchronous dispatch (fire-and-forget for wallpapers and "outside"
targets). Currently, finding the appropriate input targets for an event
requires a call back into the WindowManagerServer from native code.
In the future this will be refactored to avoid most of these callbacks
except as required to handle pending focus transitions.
End-to-end event dispatch mostly works!
To do: event injection, rate limiting, ANRs, testing, optimization, etc.
Change-Id: I8c36b2b9e0a2d27392040ecda0f51b636456de25
2010-04-23 01:58:52 +00:00
|
|
|
in->current.x = in->accumulator.absX;
|
2010-06-22 08:27:15 +00:00
|
|
|
}
|
|
|
|
|
|
|
|
if (fields & InputDevice::SingleTouchScreenState::Accumulator::FIELD_ABS_Y) {
|
Native input dispatch rewrite work in progress.
The old dispatch mechanism has been left in place and continues to
be used by default for now. To enable native input dispatch,
edit the ENABLE_NATIVE_DISPATCH constant in WindowManagerPolicy.
Includes part of the new input event NDK API. Some details TBD.
To wire up input dispatch, as the ViewRoot adds a window to the
window session it receives an InputChannel object as an output
argument. The InputChannel encapsulates the file descriptors for a
shared memory region and two pipe end-points. The ViewRoot then
provides the InputChannel to the InputQueue. Behind the
scenes, InputQueue simply attaches handlers to the native PollLoop object
that underlies the MessageQueue. This way MessageQueue doesn't need
to know anything about input dispatch per-se, it just exposes (in native
code) a PollLoop that other components can use to monitor file descriptor
state changes.
There can be zero or more targets for any given input event. Each
input target is specified by its input channel and some parameters
including flags, an X/Y coordinate offset, and the dispatch timeout.
An input target can request either synchronous dispatch (for foreground apps)
or asynchronous dispatch (fire-and-forget for wallpapers and "outside"
targets). Currently, finding the appropriate input targets for an event
requires a call back into the WindowManagerServer from native code.
In the future this will be refactored to avoid most of these callbacks
except as required to handle pending focus transitions.
End-to-end event dispatch mostly works!
To do: event injection, rate limiting, ANRs, testing, optimization, etc.
Change-Id: I8c36b2b9e0a2d27392040ecda0f51b636456de25
2010-04-23 01:58:52 +00:00
|
|
|
in->current.y = in->accumulator.absY;
|
2010-06-22 08:27:15 +00:00
|
|
|
}
|
|
|
|
|
|
|
|
if (fields & InputDevice::SingleTouchScreenState::Accumulator::FIELD_ABS_PRESSURE) {
|
Native input dispatch rewrite work in progress.
The old dispatch mechanism has been left in place and continues to
be used by default for now. To enable native input dispatch,
edit the ENABLE_NATIVE_DISPATCH constant in WindowManagerPolicy.
Includes part of the new input event NDK API. Some details TBD.
To wire up input dispatch, as the ViewRoot adds a window to the
window session it receives an InputChannel object as an output
argument. The InputChannel encapsulates the file descriptors for a
shared memory region and two pipe end-points. The ViewRoot then
provides the InputChannel to the InputQueue. Behind the
scenes, InputQueue simply attaches handlers to the native PollLoop object
that underlies the MessageQueue. This way MessageQueue doesn't need
to know anything about input dispatch per-se, it just exposes (in native
code) a PollLoop that other components can use to monitor file descriptor
state changes.
There can be zero or more targets for any given input event. Each
input target is specified by its input channel and some parameters
including flags, an X/Y coordinate offset, and the dispatch timeout.
An input target can request either synchronous dispatch (for foreground apps)
or asynchronous dispatch (fire-and-forget for wallpapers and "outside"
targets). Currently, finding the appropriate input targets for an event
requires a call back into the WindowManagerServer from native code.
In the future this will be refactored to avoid most of these callbacks
except as required to handle pending focus transitions.
End-to-end event dispatch mostly works!
To do: event injection, rate limiting, ANRs, testing, optimization, etc.
Change-Id: I8c36b2b9e0a2d27392040ecda0f51b636456de25
2010-04-23 01:58:52 +00:00
|
|
|
in->current.pressure = in->accumulator.absPressure;
|
2010-06-22 08:27:15 +00:00
|
|
|
}
|
|
|
|
|
|
|
|
if (fields & InputDevice::SingleTouchScreenState::Accumulator::FIELD_ABS_TOOL_WIDTH) {
|
Native input dispatch rewrite work in progress.
The old dispatch mechanism has been left in place and continues to
be used by default for now. To enable native input dispatch,
edit the ENABLE_NATIVE_DISPATCH constant in WindowManagerPolicy.
Includes part of the new input event NDK API. Some details TBD.
To wire up input dispatch, as the ViewRoot adds a window to the
window session it receives an InputChannel object as an output
argument. The InputChannel encapsulates the file descriptors for a
shared memory region and two pipe end-points. The ViewRoot then
provides the InputChannel to the InputQueue. Behind the
scenes, InputQueue simply attaches handlers to the native PollLoop object
that underlies the MessageQueue. This way MessageQueue doesn't need
to know anything about input dispatch per-se, it just exposes (in native
code) a PollLoop that other components can use to monitor file descriptor
state changes.
There can be zero or more targets for any given input event. Each
input target is specified by its input channel and some parameters
including flags, an X/Y coordinate offset, and the dispatch timeout.
An input target can request either synchronous dispatch (for foreground apps)
or asynchronous dispatch (fire-and-forget for wallpapers and "outside"
targets). Currently, finding the appropriate input targets for an event
requires a call back into the WindowManagerServer from native code.
In the future this will be refactored to avoid most of these callbacks
except as required to handle pending focus transitions.
End-to-end event dispatch mostly works!
To do: event injection, rate limiting, ANRs, testing, optimization, etc.
Change-Id: I8c36b2b9e0a2d27392040ecda0f51b636456de25
2010-04-23 01:58:52 +00:00
|
|
|
in->current.size = in->accumulator.absToolWidth;
|
|
|
|
}
|
|
|
|
|
|
|
|
out->clear();
|
|
|
|
|
|
|
|
if (in->current.down) {
|
|
|
|
out->pointerCount = 1;
|
|
|
|
out->pointers[0].id = 0;
|
|
|
|
out->pointers[0].x = in->current.x;
|
|
|
|
out->pointers[0].y = in->current.y;
|
|
|
|
out->pointers[0].pressure = in->current.pressure;
|
|
|
|
out->pointers[0].size = in->current.size;
|
2010-07-15 01:48:53 +00:00
|
|
|
out->pointers[0].touchMajor = in->current.pressure;
|
|
|
|
out->pointers[0].touchMinor = in->current.pressure;
|
|
|
|
out->pointers[0].toolMajor = in->current.size;
|
|
|
|
out->pointers[0].toolMinor = in->current.size;
|
|
|
|
out->pointers[0].orientation = 0;
|
Native input dispatch rewrite work in progress.
The old dispatch mechanism has been left in place and continues to
be used by default for now. To enable native input dispatch,
edit the ENABLE_NATIVE_DISPATCH constant in WindowManagerPolicy.
Includes part of the new input event NDK API. Some details TBD.
To wire up input dispatch, as the ViewRoot adds a window to the
window session it receives an InputChannel object as an output
argument. The InputChannel encapsulates the file descriptors for a
shared memory region and two pipe end-points. The ViewRoot then
provides the InputChannel to the InputQueue. Behind the
scenes, InputQueue simply attaches handlers to the native PollLoop object
that underlies the MessageQueue. This way MessageQueue doesn't need
to know anything about input dispatch per-se, it just exposes (in native
code) a PollLoop that other components can use to monitor file descriptor
state changes.
There can be zero or more targets for any given input event. Each
input target is specified by its input channel and some parameters
including flags, an X/Y coordinate offset, and the dispatch timeout.
An input target can request either synchronous dispatch (for foreground apps)
or asynchronous dispatch (fire-and-forget for wallpapers and "outside"
targets). Currently, finding the appropriate input targets for an event
requires a call back into the WindowManagerServer from native code.
In the future this will be refactored to avoid most of these callbacks
except as required to handle pending focus transitions.
End-to-end event dispatch mostly works!
To do: event injection, rate limiting, ANRs, testing, optimization, etc.
Change-Id: I8c36b2b9e0a2d27392040ecda0f51b636456de25
2010-04-23 01:58:52 +00:00
|
|
|
out->idToIndex[0] = 0;
|
|
|
|
out->idBits.markBit(0);
|
|
|
|
}
|
|
|
|
|
|
|
|
onTouchScreenChanged(when, device, true);
|
|
|
|
}
|
|
|
|
|
|
|
|
void InputReader::onTouchScreenChanged(nsecs_t when,
|
|
|
|
InputDevice* device, bool havePointerIds) {
|
|
|
|
/* Apply policy */
|
|
|
|
|
|
|
|
int32_t policyActions = mPolicy->interceptTouch(when);
|
|
|
|
|
|
|
|
uint32_t policyFlags = 0;
|
|
|
|
if (! applyStandardInputDispatchPolicyActions(when, policyActions, & policyFlags)) {
|
|
|
|
device->touchScreen.lastTouch.clear();
|
|
|
|
return; // event dropped
|
|
|
|
}
|
|
|
|
|
|
|
|
/* Preprocess pointer data */
|
|
|
|
|
|
|
|
if (device->touchScreen.parameters.useBadTouchFilter) {
|
|
|
|
if (device->touchScreen.applyBadTouchFilter()) {
|
|
|
|
havePointerIds = false;
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
|
|
|
if (device->touchScreen.parameters.useJumpyTouchFilter) {
|
|
|
|
if (device->touchScreen.applyJumpyTouchFilter()) {
|
|
|
|
havePointerIds = false;
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
|
|
|
if (! havePointerIds) {
|
|
|
|
device->touchScreen.calculatePointerIds();
|
|
|
|
}
|
|
|
|
|
|
|
|
InputDevice::TouchData temp;
|
|
|
|
InputDevice::TouchData* savedTouch;
|
|
|
|
if (device->touchScreen.parameters.useAveragingTouchFilter) {
|
|
|
|
temp.copyFrom(device->touchScreen.currentTouch);
|
|
|
|
savedTouch = & temp;
|
|
|
|
|
|
|
|
device->touchScreen.applyAveragingTouchFilter();
|
|
|
|
} else {
|
|
|
|
savedTouch = & device->touchScreen.currentTouch;
|
|
|
|
}
|
|
|
|
|
|
|
|
/* Process virtual keys or touches */
|
|
|
|
|
|
|
|
if (! consumeVirtualKeyTouches(when, device, policyFlags)) {
|
|
|
|
dispatchTouches(when, device, policyFlags);
|
|
|
|
}
|
|
|
|
|
|
|
|
// Copy current touch to last touch in preparation for the next cycle.
|
|
|
|
device->touchScreen.lastTouch.copyFrom(*savedTouch);
|
|
|
|
}
|
|
|
|
|
|
|
|
bool InputReader::consumeVirtualKeyTouches(nsecs_t when,
|
|
|
|
InputDevice* device, uint32_t policyFlags) {
|
2010-07-02 22:37:36 +00:00
|
|
|
switch (device->touchScreen.currentVirtualKey.status) {
|
|
|
|
case InputDevice::TouchScreenState::CurrentVirtualKeyState::STATUS_CANCELED:
|
Native input dispatch rewrite work in progress.
The old dispatch mechanism has been left in place and continues to
be used by default for now. To enable native input dispatch,
edit the ENABLE_NATIVE_DISPATCH constant in WindowManagerPolicy.
Includes part of the new input event NDK API. Some details TBD.
To wire up input dispatch, as the ViewRoot adds a window to the
window session it receives an InputChannel object as an output
argument. The InputChannel encapsulates the file descriptors for a
shared memory region and two pipe end-points. The ViewRoot then
provides the InputChannel to the InputQueue. Behind the
scenes, InputQueue simply attaches handlers to the native PollLoop object
that underlies the MessageQueue. This way MessageQueue doesn't need
to know anything about input dispatch per-se, it just exposes (in native
code) a PollLoop that other components can use to monitor file descriptor
state changes.
There can be zero or more targets for any given input event. Each
input target is specified by its input channel and some parameters
including flags, an X/Y coordinate offset, and the dispatch timeout.
An input target can request either synchronous dispatch (for foreground apps)
or asynchronous dispatch (fire-and-forget for wallpapers and "outside"
targets). Currently, finding the appropriate input targets for an event
requires a call back into the WindowManagerServer from native code.
In the future this will be refactored to avoid most of these callbacks
except as required to handle pending focus transitions.
End-to-end event dispatch mostly works!
To do: event injection, rate limiting, ANRs, testing, optimization, etc.
Change-Id: I8c36b2b9e0a2d27392040ecda0f51b636456de25
2010-04-23 01:58:52 +00:00
|
|
|
if (device->touchScreen.currentTouch.pointerCount == 0) {
|
2010-07-02 22:37:36 +00:00
|
|
|
// Pointer went up after virtual key canceled.
|
|
|
|
device->touchScreen.currentVirtualKey.status =
|
|
|
|
InputDevice::TouchScreenState::CurrentVirtualKeyState::STATUS_UP;
|
|
|
|
}
|
|
|
|
return true; // consumed
|
Native input dispatch rewrite work in progress.
The old dispatch mechanism has been left in place and continues to
be used by default for now. To enable native input dispatch,
edit the ENABLE_NATIVE_DISPATCH constant in WindowManagerPolicy.
Includes part of the new input event NDK API. Some details TBD.
To wire up input dispatch, as the ViewRoot adds a window to the
window session it receives an InputChannel object as an output
argument. The InputChannel encapsulates the file descriptors for a
shared memory region and two pipe end-points. The ViewRoot then
provides the InputChannel to the InputQueue. Behind the
scenes, InputQueue simply attaches handlers to the native PollLoop object
that underlies the MessageQueue. This way MessageQueue doesn't need
to know anything about input dispatch per-se, it just exposes (in native
code) a PollLoop that other components can use to monitor file descriptor
state changes.
There can be zero or more targets for any given input event. Each
input target is specified by its input channel and some parameters
including flags, an X/Y coordinate offset, and the dispatch timeout.
An input target can request either synchronous dispatch (for foreground apps)
or asynchronous dispatch (fire-and-forget for wallpapers and "outside"
targets). Currently, finding the appropriate input targets for an event
requires a call back into the WindowManagerServer from native code.
In the future this will be refactored to avoid most of these callbacks
except as required to handle pending focus transitions.
End-to-end event dispatch mostly works!
To do: event injection, rate limiting, ANRs, testing, optimization, etc.
Change-Id: I8c36b2b9e0a2d27392040ecda0f51b636456de25
2010-04-23 01:58:52 +00:00
|
|
|
|
2010-07-02 22:37:36 +00:00
|
|
|
case InputDevice::TouchScreenState::CurrentVirtualKeyState::STATUS_DOWN:
|
|
|
|
if (device->touchScreen.currentTouch.pointerCount == 0) {
|
|
|
|
// Pointer went up while virtual key was down.
|
|
|
|
device->touchScreen.currentVirtualKey.status =
|
|
|
|
InputDevice::TouchScreenState::CurrentVirtualKeyState::STATUS_UP;
|
Native input dispatch rewrite work in progress.
The old dispatch mechanism has been left in place and continues to
be used by default for now. To enable native input dispatch,
edit the ENABLE_NATIVE_DISPATCH constant in WindowManagerPolicy.
Includes part of the new input event NDK API. Some details TBD.
To wire up input dispatch, as the ViewRoot adds a window to the
window session it receives an InputChannel object as an output
argument. The InputChannel encapsulates the file descriptors for a
shared memory region and two pipe end-points. The ViewRoot then
provides the InputChannel to the InputQueue. Behind the
scenes, InputQueue simply attaches handlers to the native PollLoop object
that underlies the MessageQueue. This way MessageQueue doesn't need
to know anything about input dispatch per-se, it just exposes (in native
code) a PollLoop that other components can use to monitor file descriptor
state changes.
There can be zero or more targets for any given input event. Each
input target is specified by its input channel and some parameters
including flags, an X/Y coordinate offset, and the dispatch timeout.
An input target can request either synchronous dispatch (for foreground apps)
or asynchronous dispatch (fire-and-forget for wallpapers and "outside"
targets). Currently, finding the appropriate input targets for an event
requires a call back into the WindowManagerServer from native code.
In the future this will be refactored to avoid most of these callbacks
except as required to handle pending focus transitions.
End-to-end event dispatch mostly works!
To do: event injection, rate limiting, ANRs, testing, optimization, etc.
Change-Id: I8c36b2b9e0a2d27392040ecda0f51b636456de25
2010-04-23 01:58:52 +00:00
|
|
|
#if DEBUG_VIRTUAL_KEYS
|
|
|
|
LOGD("VirtualKeys: Generating key up: keyCode=%d, scanCode=%d",
|
|
|
|
device->touchScreen.currentVirtualKey.keyCode,
|
|
|
|
device->touchScreen.currentVirtualKey.scanCode);
|
|
|
|
#endif
|
2010-07-15 01:48:53 +00:00
|
|
|
dispatchVirtualKey(when, device, policyFlags, AKEY_EVENT_ACTION_UP,
|
|
|
|
AKEY_EVENT_FLAG_FROM_SYSTEM | AKEY_EVENT_FLAG_VIRTUAL_HARD_KEY);
|
Native input dispatch rewrite work in progress.
The old dispatch mechanism has been left in place and continues to
be used by default for now. To enable native input dispatch,
edit the ENABLE_NATIVE_DISPATCH constant in WindowManagerPolicy.
Includes part of the new input event NDK API. Some details TBD.
To wire up input dispatch, as the ViewRoot adds a window to the
window session it receives an InputChannel object as an output
argument. The InputChannel encapsulates the file descriptors for a
shared memory region and two pipe end-points. The ViewRoot then
provides the InputChannel to the InputQueue. Behind the
scenes, InputQueue simply attaches handlers to the native PollLoop object
that underlies the MessageQueue. This way MessageQueue doesn't need
to know anything about input dispatch per-se, it just exposes (in native
code) a PollLoop that other components can use to monitor file descriptor
state changes.
There can be zero or more targets for any given input event. Each
input target is specified by its input channel and some parameters
including flags, an X/Y coordinate offset, and the dispatch timeout.
An input target can request either synchronous dispatch (for foreground apps)
or asynchronous dispatch (fire-and-forget for wallpapers and "outside"
targets). Currently, finding the appropriate input targets for an event
requires a call back into the WindowManagerServer from native code.
In the future this will be refactored to avoid most of these callbacks
except as required to handle pending focus transitions.
End-to-end event dispatch mostly works!
To do: event injection, rate limiting, ANRs, testing, optimization, etc.
Change-Id: I8c36b2b9e0a2d27392040ecda0f51b636456de25
2010-04-23 01:58:52 +00:00
|
|
|
return true; // consumed
|
|
|
|
}
|
|
|
|
|
2010-07-02 22:37:36 +00:00
|
|
|
if (device->touchScreen.currentTouch.pointerCount == 1) {
|
|
|
|
const InputDevice::VirtualKey* virtualKey = device->touchScreen.findVirtualKeyHit();
|
|
|
|
if (virtualKey
|
|
|
|
&& virtualKey->keyCode == device->touchScreen.currentVirtualKey.keyCode) {
|
|
|
|
// Pointer is still within the space of the virtual key.
|
|
|
|
return true; // consumed
|
|
|
|
}
|
Native input dispatch rewrite work in progress.
The old dispatch mechanism has been left in place and continues to
be used by default for now. To enable native input dispatch,
edit the ENABLE_NATIVE_DISPATCH constant in WindowManagerPolicy.
Includes part of the new input event NDK API. Some details TBD.
To wire up input dispatch, as the ViewRoot adds a window to the
window session it receives an InputChannel object as an output
argument. The InputChannel encapsulates the file descriptors for a
shared memory region and two pipe end-points. The ViewRoot then
provides the InputChannel to the InputQueue. Behind the
scenes, InputQueue simply attaches handlers to the native PollLoop object
that underlies the MessageQueue. This way MessageQueue doesn't need
to know anything about input dispatch per-se, it just exposes (in native
code) a PollLoop that other components can use to monitor file descriptor
state changes.
There can be zero or more targets for any given input event. Each
input target is specified by its input channel and some parameters
including flags, an X/Y coordinate offset, and the dispatch timeout.
An input target can request either synchronous dispatch (for foreground apps)
or asynchronous dispatch (fire-and-forget for wallpapers and "outside"
targets). Currently, finding the appropriate input targets for an event
requires a call back into the WindowManagerServer from native code.
In the future this will be refactored to avoid most of these callbacks
except as required to handle pending focus transitions.
End-to-end event dispatch mostly works!
To do: event injection, rate limiting, ANRs, testing, optimization, etc.
Change-Id: I8c36b2b9e0a2d27392040ecda0f51b636456de25
2010-04-23 01:58:52 +00:00
|
|
|
}
|
|
|
|
|
2010-07-02 22:37:36 +00:00
|
|
|
// Pointer left virtual key area or another pointer also went down.
|
|
|
|
// Send key cancellation.
|
|
|
|
device->touchScreen.currentVirtualKey.status =
|
|
|
|
InputDevice::TouchScreenState::CurrentVirtualKeyState::STATUS_CANCELED;
|
Native input dispatch rewrite work in progress.
The old dispatch mechanism has been left in place and continues to
be used by default for now. To enable native input dispatch,
edit the ENABLE_NATIVE_DISPATCH constant in WindowManagerPolicy.
Includes part of the new input event NDK API. Some details TBD.
To wire up input dispatch, as the ViewRoot adds a window to the
window session it receives an InputChannel object as an output
argument. The InputChannel encapsulates the file descriptors for a
shared memory region and two pipe end-points. The ViewRoot then
provides the InputChannel to the InputQueue. Behind the
scenes, InputQueue simply attaches handlers to the native PollLoop object
that underlies the MessageQueue. This way MessageQueue doesn't need
to know anything about input dispatch per-se, it just exposes (in native
code) a PollLoop that other components can use to monitor file descriptor
state changes.
There can be zero or more targets for any given input event. Each
input target is specified by its input channel and some parameters
including flags, an X/Y coordinate offset, and the dispatch timeout.
An input target can request either synchronous dispatch (for foreground apps)
or asynchronous dispatch (fire-and-forget for wallpapers and "outside"
targets). Currently, finding the appropriate input targets for an event
requires a call back into the WindowManagerServer from native code.
In the future this will be refactored to avoid most of these callbacks
except as required to handle pending focus transitions.
End-to-end event dispatch mostly works!
To do: event injection, rate limiting, ANRs, testing, optimization, etc.
Change-Id: I8c36b2b9e0a2d27392040ecda0f51b636456de25
2010-04-23 01:58:52 +00:00
|
|
|
#if DEBUG_VIRTUAL_KEYS
|
2010-07-02 22:37:36 +00:00
|
|
|
LOGD("VirtualKeys: Canceling key: keyCode=%d, scanCode=%d",
|
|
|
|
device->touchScreen.currentVirtualKey.keyCode,
|
|
|
|
device->touchScreen.currentVirtualKey.scanCode);
|
Native input dispatch rewrite work in progress.
The old dispatch mechanism has been left in place and continues to
be used by default for now. To enable native input dispatch,
edit the ENABLE_NATIVE_DISPATCH constant in WindowManagerPolicy.
Includes part of the new input event NDK API. Some details TBD.
To wire up input dispatch, as the ViewRoot adds a window to the
window session it receives an InputChannel object as an output
argument. The InputChannel encapsulates the file descriptors for a
shared memory region and two pipe end-points. The ViewRoot then
provides the InputChannel to the InputQueue. Behind the
scenes, InputQueue simply attaches handlers to the native PollLoop object
that underlies the MessageQueue. This way MessageQueue doesn't need
to know anything about input dispatch per-se, it just exposes (in native
code) a PollLoop that other components can use to monitor file descriptor
state changes.
There can be zero or more targets for any given input event. Each
input target is specified by its input channel and some parameters
including flags, an X/Y coordinate offset, and the dispatch timeout.
An input target can request either synchronous dispatch (for foreground apps)
or asynchronous dispatch (fire-and-forget for wallpapers and "outside"
targets). Currently, finding the appropriate input targets for an event
requires a call back into the WindowManagerServer from native code.
In the future this will be refactored to avoid most of these callbacks
except as required to handle pending focus transitions.
End-to-end event dispatch mostly works!
To do: event injection, rate limiting, ANRs, testing, optimization, etc.
Change-Id: I8c36b2b9e0a2d27392040ecda0f51b636456de25
2010-04-23 01:58:52 +00:00
|
|
|
#endif
|
2010-07-15 01:48:53 +00:00
|
|
|
dispatchVirtualKey(when, device, policyFlags, AKEY_EVENT_ACTION_UP,
|
|
|
|
AKEY_EVENT_FLAG_FROM_SYSTEM | AKEY_EVENT_FLAG_VIRTUAL_HARD_KEY
|
|
|
|
| AKEY_EVENT_FLAG_CANCELED);
|
2010-07-02 22:37:36 +00:00
|
|
|
return true; // consumed
|
Native input dispatch rewrite work in progress.
The old dispatch mechanism has been left in place and continues to
be used by default for now. To enable native input dispatch,
edit the ENABLE_NATIVE_DISPATCH constant in WindowManagerPolicy.
Includes part of the new input event NDK API. Some details TBD.
To wire up input dispatch, as the ViewRoot adds a window to the
window session it receives an InputChannel object as an output
argument. The InputChannel encapsulates the file descriptors for a
shared memory region and two pipe end-points. The ViewRoot then
provides the InputChannel to the InputQueue. Behind the
scenes, InputQueue simply attaches handlers to the native PollLoop object
that underlies the MessageQueue. This way MessageQueue doesn't need
to know anything about input dispatch per-se, it just exposes (in native
code) a PollLoop that other components can use to monitor file descriptor
state changes.
There can be zero or more targets for any given input event. Each
input target is specified by its input channel and some parameters
including flags, an X/Y coordinate offset, and the dispatch timeout.
An input target can request either synchronous dispatch (for foreground apps)
or asynchronous dispatch (fire-and-forget for wallpapers and "outside"
targets). Currently, finding the appropriate input targets for an event
requires a call back into the WindowManagerServer from native code.
In the future this will be refactored to avoid most of these callbacks
except as required to handle pending focus transitions.
End-to-end event dispatch mostly works!
To do: event injection, rate limiting, ANRs, testing, optimization, etc.
Change-Id: I8c36b2b9e0a2d27392040ecda0f51b636456de25
2010-04-23 01:58:52 +00:00
|
|
|
|
2010-07-02 22:37:36 +00:00
|
|
|
default:
|
|
|
|
if (device->touchScreen.currentTouch.pointerCount == 1
|
|
|
|
&& device->touchScreen.lastTouch.pointerCount == 0) {
|
|
|
|
// Pointer just went down. Check for virtual key hit.
|
|
|
|
const InputDevice::VirtualKey* virtualKey = device->touchScreen.findVirtualKeyHit();
|
|
|
|
if (virtualKey) {
|
|
|
|
device->touchScreen.currentVirtualKey.status =
|
|
|
|
InputDevice::TouchScreenState::CurrentVirtualKeyState::STATUS_DOWN;
|
Native input dispatch rewrite work in progress.
The old dispatch mechanism has been left in place and continues to
be used by default for now. To enable native input dispatch,
edit the ENABLE_NATIVE_DISPATCH constant in WindowManagerPolicy.
Includes part of the new input event NDK API. Some details TBD.
To wire up input dispatch, as the ViewRoot adds a window to the
window session it receives an InputChannel object as an output
argument. The InputChannel encapsulates the file descriptors for a
shared memory region and two pipe end-points. The ViewRoot then
provides the InputChannel to the InputQueue. Behind the
scenes, InputQueue simply attaches handlers to the native PollLoop object
that underlies the MessageQueue. This way MessageQueue doesn't need
to know anything about input dispatch per-se, it just exposes (in native
code) a PollLoop that other components can use to monitor file descriptor
state changes.
There can be zero or more targets for any given input event. Each
input target is specified by its input channel and some parameters
including flags, an X/Y coordinate offset, and the dispatch timeout.
An input target can request either synchronous dispatch (for foreground apps)
or asynchronous dispatch (fire-and-forget for wallpapers and "outside"
targets). Currently, finding the appropriate input targets for an event
requires a call back into the WindowManagerServer from native code.
In the future this will be refactored to avoid most of these callbacks
except as required to handle pending focus transitions.
End-to-end event dispatch mostly works!
To do: event injection, rate limiting, ANRs, testing, optimization, etc.
Change-Id: I8c36b2b9e0a2d27392040ecda0f51b636456de25
2010-04-23 01:58:52 +00:00
|
|
|
device->touchScreen.currentVirtualKey.downTime = when;
|
2010-07-02 22:37:36 +00:00
|
|
|
device->touchScreen.currentVirtualKey.keyCode = virtualKey->keyCode;
|
|
|
|
device->touchScreen.currentVirtualKey.scanCode = virtualKey->scanCode;
|
Native input dispatch rewrite work in progress.
The old dispatch mechanism has been left in place and continues to
be used by default for now. To enable native input dispatch,
edit the ENABLE_NATIVE_DISPATCH constant in WindowManagerPolicy.
Includes part of the new input event NDK API. Some details TBD.
To wire up input dispatch, as the ViewRoot adds a window to the
window session it receives an InputChannel object as an output
argument. The InputChannel encapsulates the file descriptors for a
shared memory region and two pipe end-points. The ViewRoot then
provides the InputChannel to the InputQueue. Behind the
scenes, InputQueue simply attaches handlers to the native PollLoop object
that underlies the MessageQueue. This way MessageQueue doesn't need
to know anything about input dispatch per-se, it just exposes (in native
code) a PollLoop that other components can use to monitor file descriptor
state changes.
There can be zero or more targets for any given input event. Each
input target is specified by its input channel and some parameters
including flags, an X/Y coordinate offset, and the dispatch timeout.
An input target can request either synchronous dispatch (for foreground apps)
or asynchronous dispatch (fire-and-forget for wallpapers and "outside"
targets). Currently, finding the appropriate input targets for an event
requires a call back into the WindowManagerServer from native code.
In the future this will be refactored to avoid most of these callbacks
except as required to handle pending focus transitions.
End-to-end event dispatch mostly works!
To do: event injection, rate limiting, ANRs, testing, optimization, etc.
Change-Id: I8c36b2b9e0a2d27392040ecda0f51b636456de25
2010-04-23 01:58:52 +00:00
|
|
|
#if DEBUG_VIRTUAL_KEYS
|
2010-07-02 22:37:36 +00:00
|
|
|
LOGD("VirtualKeys: Generating key down: keyCode=%d, scanCode=%d",
|
|
|
|
device->touchScreen.currentVirtualKey.keyCode,
|
|
|
|
device->touchScreen.currentVirtualKey.scanCode);
|
Native input dispatch rewrite work in progress.
The old dispatch mechanism has been left in place and continues to
be used by default for now. To enable native input dispatch,
edit the ENABLE_NATIVE_DISPATCH constant in WindowManagerPolicy.
Includes part of the new input event NDK API. Some details TBD.
To wire up input dispatch, as the ViewRoot adds a window to the
window session it receives an InputChannel object as an output
argument. The InputChannel encapsulates the file descriptors for a
shared memory region and two pipe end-points. The ViewRoot then
provides the InputChannel to the InputQueue. Behind the
scenes, InputQueue simply attaches handlers to the native PollLoop object
that underlies the MessageQueue. This way MessageQueue doesn't need
to know anything about input dispatch per-se, it just exposes (in native
code) a PollLoop that other components can use to monitor file descriptor
state changes.
There can be zero or more targets for any given input event. Each
input target is specified by its input channel and some parameters
including flags, an X/Y coordinate offset, and the dispatch timeout.
An input target can request either synchronous dispatch (for foreground apps)
or asynchronous dispatch (fire-and-forget for wallpapers and "outside"
targets). Currently, finding the appropriate input targets for an event
requires a call back into the WindowManagerServer from native code.
In the future this will be refactored to avoid most of these callbacks
except as required to handle pending focus transitions.
End-to-end event dispatch mostly works!
To do: event injection, rate limiting, ANRs, testing, optimization, etc.
Change-Id: I8c36b2b9e0a2d27392040ecda0f51b636456de25
2010-04-23 01:58:52 +00:00
|
|
|
#endif
|
2010-07-15 01:48:53 +00:00
|
|
|
dispatchVirtualKey(when, device, policyFlags, AKEY_EVENT_ACTION_DOWN,
|
|
|
|
AKEY_EVENT_FLAG_FROM_SYSTEM | AKEY_EVENT_FLAG_VIRTUAL_HARD_KEY);
|
Native input dispatch rewrite work in progress.
The old dispatch mechanism has been left in place and continues to
be used by default for now. To enable native input dispatch,
edit the ENABLE_NATIVE_DISPATCH constant in WindowManagerPolicy.
Includes part of the new input event NDK API. Some details TBD.
To wire up input dispatch, as the ViewRoot adds a window to the
window session it receives an InputChannel object as an output
argument. The InputChannel encapsulates the file descriptors for a
shared memory region and two pipe end-points. The ViewRoot then
provides the InputChannel to the InputQueue. Behind the
scenes, InputQueue simply attaches handlers to the native PollLoop object
that underlies the MessageQueue. This way MessageQueue doesn't need
to know anything about input dispatch per-se, it just exposes (in native
code) a PollLoop that other components can use to monitor file descriptor
state changes.
There can be zero or more targets for any given input event. Each
input target is specified by its input channel and some parameters
including flags, an X/Y coordinate offset, and the dispatch timeout.
An input target can request either synchronous dispatch (for foreground apps)
or asynchronous dispatch (fire-and-forget for wallpapers and "outside"
targets). Currently, finding the appropriate input targets for an event
requires a call back into the WindowManagerServer from native code.
In the future this will be refactored to avoid most of these callbacks
except as required to handle pending focus transitions.
End-to-end event dispatch mostly works!
To do: event injection, rate limiting, ANRs, testing, optimization, etc.
Change-Id: I8c36b2b9e0a2d27392040ecda0f51b636456de25
2010-04-23 01:58:52 +00:00
|
|
|
return true; // consumed
|
|
|
|
}
|
|
|
|
}
|
2010-07-02 22:37:36 +00:00
|
|
|
return false; // not consumed
|
Native input dispatch rewrite work in progress.
The old dispatch mechanism has been left in place and continues to
be used by default for now. To enable native input dispatch,
edit the ENABLE_NATIVE_DISPATCH constant in WindowManagerPolicy.
Includes part of the new input event NDK API. Some details TBD.
To wire up input dispatch, as the ViewRoot adds a window to the
window session it receives an InputChannel object as an output
argument. The InputChannel encapsulates the file descriptors for a
shared memory region and two pipe end-points. The ViewRoot then
provides the InputChannel to the InputQueue. Behind the
scenes, InputQueue simply attaches handlers to the native PollLoop object
that underlies the MessageQueue. This way MessageQueue doesn't need
to know anything about input dispatch per-se, it just exposes (in native
code) a PollLoop that other components can use to monitor file descriptor
state changes.
There can be zero or more targets for any given input event. Each
input target is specified by its input channel and some parameters
including flags, an X/Y coordinate offset, and the dispatch timeout.
An input target can request either synchronous dispatch (for foreground apps)
or asynchronous dispatch (fire-and-forget for wallpapers and "outside"
targets). Currently, finding the appropriate input targets for an event
requires a call back into the WindowManagerServer from native code.
In the future this will be refactored to avoid most of these callbacks
except as required to handle pending focus transitions.
End-to-end event dispatch mostly works!
To do: event injection, rate limiting, ANRs, testing, optimization, etc.
Change-Id: I8c36b2b9e0a2d27392040ecda0f51b636456de25
2010-04-23 01:58:52 +00:00
|
|
|
}
|
|
|
|
}
|
|
|
|
|
|
|
|
void InputReader::dispatchVirtualKey(nsecs_t when,
|
|
|
|
InputDevice* device, uint32_t policyFlags,
|
|
|
|
int32_t keyEventAction, int32_t keyEventFlags) {
|
2010-06-22 08:27:15 +00:00
|
|
|
updateExportedVirtualKeyState();
|
|
|
|
|
Native input dispatch rewrite work in progress.
The old dispatch mechanism has been left in place and continues to
be used by default for now. To enable native input dispatch,
edit the ENABLE_NATIVE_DISPATCH constant in WindowManagerPolicy.
Includes part of the new input event NDK API. Some details TBD.
To wire up input dispatch, as the ViewRoot adds a window to the
window session it receives an InputChannel object as an output
argument. The InputChannel encapsulates the file descriptors for a
shared memory region and two pipe end-points. The ViewRoot then
provides the InputChannel to the InputQueue. Behind the
scenes, InputQueue simply attaches handlers to the native PollLoop object
that underlies the MessageQueue. This way MessageQueue doesn't need
to know anything about input dispatch per-se, it just exposes (in native
code) a PollLoop that other components can use to monitor file descriptor
state changes.
There can be zero or more targets for any given input event. Each
input target is specified by its input channel and some parameters
including flags, an X/Y coordinate offset, and the dispatch timeout.
An input target can request either synchronous dispatch (for foreground apps)
or asynchronous dispatch (fire-and-forget for wallpapers and "outside"
targets). Currently, finding the appropriate input targets for an event
requires a call back into the WindowManagerServer from native code.
In the future this will be refactored to avoid most of these callbacks
except as required to handle pending focus transitions.
End-to-end event dispatch mostly works!
To do: event injection, rate limiting, ANRs, testing, optimization, etc.
Change-Id: I8c36b2b9e0a2d27392040ecda0f51b636456de25
2010-04-23 01:58:52 +00:00
|
|
|
int32_t keyCode = device->touchScreen.currentVirtualKey.keyCode;
|
|
|
|
int32_t scanCode = device->touchScreen.currentVirtualKey.scanCode;
|
|
|
|
nsecs_t downTime = device->touchScreen.currentVirtualKey.downTime;
|
|
|
|
int32_t metaState = globalMetaState();
|
|
|
|
|
2010-07-15 01:48:53 +00:00
|
|
|
if (keyEventAction == AKEY_EVENT_ACTION_DOWN) {
|
2010-07-02 22:37:36 +00:00
|
|
|
mPolicy->virtualKeyDownFeedback();
|
|
|
|
}
|
Native input dispatch rewrite work in progress.
The old dispatch mechanism has been left in place and continues to
be used by default for now. To enable native input dispatch,
edit the ENABLE_NATIVE_DISPATCH constant in WindowManagerPolicy.
Includes part of the new input event NDK API. Some details TBD.
To wire up input dispatch, as the ViewRoot adds a window to the
window session it receives an InputChannel object as an output
argument. The InputChannel encapsulates the file descriptors for a
shared memory region and two pipe end-points. The ViewRoot then
provides the InputChannel to the InputQueue. Behind the
scenes, InputQueue simply attaches handlers to the native PollLoop object
that underlies the MessageQueue. This way MessageQueue doesn't need
to know anything about input dispatch per-se, it just exposes (in native
code) a PollLoop that other components can use to monitor file descriptor
state changes.
There can be zero or more targets for any given input event. Each
input target is specified by its input channel and some parameters
including flags, an X/Y coordinate offset, and the dispatch timeout.
An input target can request either synchronous dispatch (for foreground apps)
or asynchronous dispatch (fire-and-forget for wallpapers and "outside"
targets). Currently, finding the appropriate input targets for an event
requires a call back into the WindowManagerServer from native code.
In the future this will be refactored to avoid most of these callbacks
except as required to handle pending focus transitions.
End-to-end event dispatch mostly works!
To do: event injection, rate limiting, ANRs, testing, optimization, etc.
Change-Id: I8c36b2b9e0a2d27392040ecda0f51b636456de25
2010-04-23 01:58:52 +00:00
|
|
|
|
2010-06-22 08:27:15 +00:00
|
|
|
int32_t policyActions = mPolicy->interceptKey(when, device->id,
|
2010-07-15 01:48:53 +00:00
|
|
|
keyEventAction == AKEY_EVENT_ACTION_DOWN, keyCode, scanCode, policyFlags);
|
2010-06-22 08:27:15 +00:00
|
|
|
|
|
|
|
if (applyStandardInputDispatchPolicyActions(when, policyActions, & policyFlags)) {
|
2010-07-15 01:48:53 +00:00
|
|
|
mDispatcher->notifyKey(when, device->id, AINPUT_SOURCE_KEYBOARD, policyFlags,
|
2010-06-22 08:27:15 +00:00
|
|
|
keyEventAction, keyEventFlags, keyCode, scanCode, metaState, downTime);
|
|
|
|
}
|
Native input dispatch rewrite work in progress.
The old dispatch mechanism has been left in place and continues to
be used by default for now. To enable native input dispatch,
edit the ENABLE_NATIVE_DISPATCH constant in WindowManagerPolicy.
Includes part of the new input event NDK API. Some details TBD.
To wire up input dispatch, as the ViewRoot adds a window to the
window session it receives an InputChannel object as an output
argument. The InputChannel encapsulates the file descriptors for a
shared memory region and two pipe end-points. The ViewRoot then
provides the InputChannel to the InputQueue. Behind the
scenes, InputQueue simply attaches handlers to the native PollLoop object
that underlies the MessageQueue. This way MessageQueue doesn't need
to know anything about input dispatch per-se, it just exposes (in native
code) a PollLoop that other components can use to monitor file descriptor
state changes.
There can be zero or more targets for any given input event. Each
input target is specified by its input channel and some parameters
including flags, an X/Y coordinate offset, and the dispatch timeout.
An input target can request either synchronous dispatch (for foreground apps)
or asynchronous dispatch (fire-and-forget for wallpapers and "outside"
targets). Currently, finding the appropriate input targets for an event
requires a call back into the WindowManagerServer from native code.
In the future this will be refactored to avoid most of these callbacks
except as required to handle pending focus transitions.
End-to-end event dispatch mostly works!
To do: event injection, rate limiting, ANRs, testing, optimization, etc.
Change-Id: I8c36b2b9e0a2d27392040ecda0f51b636456de25
2010-04-23 01:58:52 +00:00
|
|
|
}
|
|
|
|
|
|
|
|
void InputReader::dispatchTouches(nsecs_t when,
|
|
|
|
InputDevice* device, uint32_t policyFlags) {
|
|
|
|
uint32_t currentPointerCount = device->touchScreen.currentTouch.pointerCount;
|
|
|
|
uint32_t lastPointerCount = device->touchScreen.lastTouch.pointerCount;
|
|
|
|
if (currentPointerCount == 0 && lastPointerCount == 0) {
|
|
|
|
return; // nothing to do!
|
|
|
|
}
|
|
|
|
|
|
|
|
BitSet32 currentIdBits = device->touchScreen.currentTouch.idBits;
|
|
|
|
BitSet32 lastIdBits = device->touchScreen.lastTouch.idBits;
|
|
|
|
|
|
|
|
if (currentIdBits == lastIdBits) {
|
|
|
|
// No pointer id changes so this is a move event.
|
|
|
|
// The dispatcher takes care of batching moves so we don't have to deal with that here.
|
2010-07-15 01:48:53 +00:00
|
|
|
int32_t motionEventAction = AMOTION_EVENT_ACTION_MOVE;
|
Native input dispatch rewrite work in progress.
The old dispatch mechanism has been left in place and continues to
be used by default for now. To enable native input dispatch,
edit the ENABLE_NATIVE_DISPATCH constant in WindowManagerPolicy.
Includes part of the new input event NDK API. Some details TBD.
To wire up input dispatch, as the ViewRoot adds a window to the
window session it receives an InputChannel object as an output
argument. The InputChannel encapsulates the file descriptors for a
shared memory region and two pipe end-points. The ViewRoot then
provides the InputChannel to the InputQueue. Behind the
scenes, InputQueue simply attaches handlers to the native PollLoop object
that underlies the MessageQueue. This way MessageQueue doesn't need
to know anything about input dispatch per-se, it just exposes (in native
code) a PollLoop that other components can use to monitor file descriptor
state changes.
There can be zero or more targets for any given input event. Each
input target is specified by its input channel and some parameters
including flags, an X/Y coordinate offset, and the dispatch timeout.
An input target can request either synchronous dispatch (for foreground apps)
or asynchronous dispatch (fire-and-forget for wallpapers and "outside"
targets). Currently, finding the appropriate input targets for an event
requires a call back into the WindowManagerServer from native code.
In the future this will be refactored to avoid most of these callbacks
except as required to handle pending focus transitions.
End-to-end event dispatch mostly works!
To do: event injection, rate limiting, ANRs, testing, optimization, etc.
Change-Id: I8c36b2b9e0a2d27392040ecda0f51b636456de25
2010-04-23 01:58:52 +00:00
|
|
|
dispatchTouch(when, device, policyFlags, & device->touchScreen.currentTouch,
|
2010-07-16 22:01:56 +00:00
|
|
|
currentIdBits, -1, motionEventAction);
|
Native input dispatch rewrite work in progress.
The old dispatch mechanism has been left in place and continues to
be used by default for now. To enable native input dispatch,
edit the ENABLE_NATIVE_DISPATCH constant in WindowManagerPolicy.
Includes part of the new input event NDK API. Some details TBD.
To wire up input dispatch, as the ViewRoot adds a window to the
window session it receives an InputChannel object as an output
argument. The InputChannel encapsulates the file descriptors for a
shared memory region and two pipe end-points. The ViewRoot then
provides the InputChannel to the InputQueue. Behind the
scenes, InputQueue simply attaches handlers to the native PollLoop object
that underlies the MessageQueue. This way MessageQueue doesn't need
to know anything about input dispatch per-se, it just exposes (in native
code) a PollLoop that other components can use to monitor file descriptor
state changes.
There can be zero or more targets for any given input event. Each
input target is specified by its input channel and some parameters
including flags, an X/Y coordinate offset, and the dispatch timeout.
An input target can request either synchronous dispatch (for foreground apps)
or asynchronous dispatch (fire-and-forget for wallpapers and "outside"
targets). Currently, finding the appropriate input targets for an event
requires a call back into the WindowManagerServer from native code.
In the future this will be refactored to avoid most of these callbacks
except as required to handle pending focus transitions.
End-to-end event dispatch mostly works!
To do: event injection, rate limiting, ANRs, testing, optimization, etc.
Change-Id: I8c36b2b9e0a2d27392040ecda0f51b636456de25
2010-04-23 01:58:52 +00:00
|
|
|
} else {
|
|
|
|
// There may be pointers going up and pointers going down at the same time when pointer
|
|
|
|
// ids are reported by the device driver.
|
|
|
|
BitSet32 upIdBits(lastIdBits.value & ~ currentIdBits.value);
|
|
|
|
BitSet32 downIdBits(currentIdBits.value & ~ lastIdBits.value);
|
|
|
|
BitSet32 activeIdBits(lastIdBits.value);
|
|
|
|
|
|
|
|
while (! upIdBits.isEmpty()) {
|
|
|
|
uint32_t upId = upIdBits.firstMarkedBit();
|
|
|
|
upIdBits.clearBit(upId);
|
|
|
|
BitSet32 oldActiveIdBits = activeIdBits;
|
|
|
|
activeIdBits.clearBit(upId);
|
|
|
|
|
|
|
|
int32_t motionEventAction;
|
|
|
|
if (activeIdBits.isEmpty()) {
|
2010-07-15 01:48:53 +00:00
|
|
|
motionEventAction = AMOTION_EVENT_ACTION_UP;
|
Native input dispatch rewrite work in progress.
The old dispatch mechanism has been left in place and continues to
be used by default for now. To enable native input dispatch,
edit the ENABLE_NATIVE_DISPATCH constant in WindowManagerPolicy.
Includes part of the new input event NDK API. Some details TBD.
To wire up input dispatch, as the ViewRoot adds a window to the
window session it receives an InputChannel object as an output
argument. The InputChannel encapsulates the file descriptors for a
shared memory region and two pipe end-points. The ViewRoot then
provides the InputChannel to the InputQueue. Behind the
scenes, InputQueue simply attaches handlers to the native PollLoop object
that underlies the MessageQueue. This way MessageQueue doesn't need
to know anything about input dispatch per-se, it just exposes (in native
code) a PollLoop that other components can use to monitor file descriptor
state changes.
There can be zero or more targets for any given input event. Each
input target is specified by its input channel and some parameters
including flags, an X/Y coordinate offset, and the dispatch timeout.
An input target can request either synchronous dispatch (for foreground apps)
or asynchronous dispatch (fire-and-forget for wallpapers and "outside"
targets). Currently, finding the appropriate input targets for an event
requires a call back into the WindowManagerServer from native code.
In the future this will be refactored to avoid most of these callbacks
except as required to handle pending focus transitions.
End-to-end event dispatch mostly works!
To do: event injection, rate limiting, ANRs, testing, optimization, etc.
Change-Id: I8c36b2b9e0a2d27392040ecda0f51b636456de25
2010-04-23 01:58:52 +00:00
|
|
|
} else {
|
2010-07-16 22:01:56 +00:00
|
|
|
motionEventAction = AMOTION_EVENT_ACTION_POINTER_UP;
|
Native input dispatch rewrite work in progress.
The old dispatch mechanism has been left in place and continues to
be used by default for now. To enable native input dispatch,
edit the ENABLE_NATIVE_DISPATCH constant in WindowManagerPolicy.
Includes part of the new input event NDK API. Some details TBD.
To wire up input dispatch, as the ViewRoot adds a window to the
window session it receives an InputChannel object as an output
argument. The InputChannel encapsulates the file descriptors for a
shared memory region and two pipe end-points. The ViewRoot then
provides the InputChannel to the InputQueue. Behind the
scenes, InputQueue simply attaches handlers to the native PollLoop object
that underlies the MessageQueue. This way MessageQueue doesn't need
to know anything about input dispatch per-se, it just exposes (in native
code) a PollLoop that other components can use to monitor file descriptor
state changes.
There can be zero or more targets for any given input event. Each
input target is specified by its input channel and some parameters
including flags, an X/Y coordinate offset, and the dispatch timeout.
An input target can request either synchronous dispatch (for foreground apps)
or asynchronous dispatch (fire-and-forget for wallpapers and "outside"
targets). Currently, finding the appropriate input targets for an event
requires a call back into the WindowManagerServer from native code.
In the future this will be refactored to avoid most of these callbacks
except as required to handle pending focus transitions.
End-to-end event dispatch mostly works!
To do: event injection, rate limiting, ANRs, testing, optimization, etc.
Change-Id: I8c36b2b9e0a2d27392040ecda0f51b636456de25
2010-04-23 01:58:52 +00:00
|
|
|
}
|
|
|
|
|
|
|
|
dispatchTouch(when, device, policyFlags, & device->touchScreen.lastTouch,
|
2010-07-16 22:01:56 +00:00
|
|
|
oldActiveIdBits, upId, motionEventAction);
|
Native input dispatch rewrite work in progress.
The old dispatch mechanism has been left in place and continues to
be used by default for now. To enable native input dispatch,
edit the ENABLE_NATIVE_DISPATCH constant in WindowManagerPolicy.
Includes part of the new input event NDK API. Some details TBD.
To wire up input dispatch, as the ViewRoot adds a window to the
window session it receives an InputChannel object as an output
argument. The InputChannel encapsulates the file descriptors for a
shared memory region and two pipe end-points. The ViewRoot then
provides the InputChannel to the InputQueue. Behind the
scenes, InputQueue simply attaches handlers to the native PollLoop object
that underlies the MessageQueue. This way MessageQueue doesn't need
to know anything about input dispatch per-se, it just exposes (in native
code) a PollLoop that other components can use to monitor file descriptor
state changes.
There can be zero or more targets for any given input event. Each
input target is specified by its input channel and some parameters
including flags, an X/Y coordinate offset, and the dispatch timeout.
An input target can request either synchronous dispatch (for foreground apps)
or asynchronous dispatch (fire-and-forget for wallpapers and "outside"
targets). Currently, finding the appropriate input targets for an event
requires a call back into the WindowManagerServer from native code.
In the future this will be refactored to avoid most of these callbacks
except as required to handle pending focus transitions.
End-to-end event dispatch mostly works!
To do: event injection, rate limiting, ANRs, testing, optimization, etc.
Change-Id: I8c36b2b9e0a2d27392040ecda0f51b636456de25
2010-04-23 01:58:52 +00:00
|
|
|
}
|
|
|
|
|
|
|
|
while (! downIdBits.isEmpty()) {
|
|
|
|
uint32_t downId = downIdBits.firstMarkedBit();
|
|
|
|
downIdBits.clearBit(downId);
|
|
|
|
BitSet32 oldActiveIdBits = activeIdBits;
|
|
|
|
activeIdBits.markBit(downId);
|
|
|
|
|
|
|
|
int32_t motionEventAction;
|
|
|
|
if (oldActiveIdBits.isEmpty()) {
|
2010-07-15 01:48:53 +00:00
|
|
|
motionEventAction = AMOTION_EVENT_ACTION_DOWN;
|
Native input dispatch rewrite work in progress.
The old dispatch mechanism has been left in place and continues to
be used by default for now. To enable native input dispatch,
edit the ENABLE_NATIVE_DISPATCH constant in WindowManagerPolicy.
Includes part of the new input event NDK API. Some details TBD.
To wire up input dispatch, as the ViewRoot adds a window to the
window session it receives an InputChannel object as an output
argument. The InputChannel encapsulates the file descriptors for a
shared memory region and two pipe end-points. The ViewRoot then
provides the InputChannel to the InputQueue. Behind the
scenes, InputQueue simply attaches handlers to the native PollLoop object
that underlies the MessageQueue. This way MessageQueue doesn't need
to know anything about input dispatch per-se, it just exposes (in native
code) a PollLoop that other components can use to monitor file descriptor
state changes.
There can be zero or more targets for any given input event. Each
input target is specified by its input channel and some parameters
including flags, an X/Y coordinate offset, and the dispatch timeout.
An input target can request either synchronous dispatch (for foreground apps)
or asynchronous dispatch (fire-and-forget for wallpapers and "outside"
targets). Currently, finding the appropriate input targets for an event
requires a call back into the WindowManagerServer from native code.
In the future this will be refactored to avoid most of these callbacks
except as required to handle pending focus transitions.
End-to-end event dispatch mostly works!
To do: event injection, rate limiting, ANRs, testing, optimization, etc.
Change-Id: I8c36b2b9e0a2d27392040ecda0f51b636456de25
2010-04-23 01:58:52 +00:00
|
|
|
device->touchScreen.downTime = when;
|
|
|
|
} else {
|
2010-07-16 22:01:56 +00:00
|
|
|
motionEventAction = AMOTION_EVENT_ACTION_POINTER_DOWN;
|
Native input dispatch rewrite work in progress.
The old dispatch mechanism has been left in place and continues to
be used by default for now. To enable native input dispatch,
edit the ENABLE_NATIVE_DISPATCH constant in WindowManagerPolicy.
Includes part of the new input event NDK API. Some details TBD.
To wire up input dispatch, as the ViewRoot adds a window to the
window session it receives an InputChannel object as an output
argument. The InputChannel encapsulates the file descriptors for a
shared memory region and two pipe end-points. The ViewRoot then
provides the InputChannel to the InputQueue. Behind the
scenes, InputQueue simply attaches handlers to the native PollLoop object
that underlies the MessageQueue. This way MessageQueue doesn't need
to know anything about input dispatch per-se, it just exposes (in native
code) a PollLoop that other components can use to monitor file descriptor
state changes.
There can be zero or more targets for any given input event. Each
input target is specified by its input channel and some parameters
including flags, an X/Y coordinate offset, and the dispatch timeout.
An input target can request either synchronous dispatch (for foreground apps)
or asynchronous dispatch (fire-and-forget for wallpapers and "outside"
targets). Currently, finding the appropriate input targets for an event
requires a call back into the WindowManagerServer from native code.
In the future this will be refactored to avoid most of these callbacks
except as required to handle pending focus transitions.
End-to-end event dispatch mostly works!
To do: event injection, rate limiting, ANRs, testing, optimization, etc.
Change-Id: I8c36b2b9e0a2d27392040ecda0f51b636456de25
2010-04-23 01:58:52 +00:00
|
|
|
}
|
|
|
|
|
|
|
|
dispatchTouch(when, device, policyFlags, & device->touchScreen.currentTouch,
|
2010-07-16 22:01:56 +00:00
|
|
|
activeIdBits, downId, motionEventAction);
|
Native input dispatch rewrite work in progress.
The old dispatch mechanism has been left in place and continues to
be used by default for now. To enable native input dispatch,
edit the ENABLE_NATIVE_DISPATCH constant in WindowManagerPolicy.
Includes part of the new input event NDK API. Some details TBD.
To wire up input dispatch, as the ViewRoot adds a window to the
window session it receives an InputChannel object as an output
argument. The InputChannel encapsulates the file descriptors for a
shared memory region and two pipe end-points. The ViewRoot then
provides the InputChannel to the InputQueue. Behind the
scenes, InputQueue simply attaches handlers to the native PollLoop object
that underlies the MessageQueue. This way MessageQueue doesn't need
to know anything about input dispatch per-se, it just exposes (in native
code) a PollLoop that other components can use to monitor file descriptor
state changes.
There can be zero or more targets for any given input event. Each
input target is specified by its input channel and some parameters
including flags, an X/Y coordinate offset, and the dispatch timeout.
An input target can request either synchronous dispatch (for foreground apps)
or asynchronous dispatch (fire-and-forget for wallpapers and "outside"
targets). Currently, finding the appropriate input targets for an event
requires a call back into the WindowManagerServer from native code.
In the future this will be refactored to avoid most of these callbacks
except as required to handle pending focus transitions.
End-to-end event dispatch mostly works!
To do: event injection, rate limiting, ANRs, testing, optimization, etc.
Change-Id: I8c36b2b9e0a2d27392040ecda0f51b636456de25
2010-04-23 01:58:52 +00:00
|
|
|
}
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
|
|
|
void InputReader::dispatchTouch(nsecs_t when, InputDevice* device, uint32_t policyFlags,
|
2010-07-16 22:01:56 +00:00
|
|
|
InputDevice::TouchData* touch, BitSet32 idBits, uint32_t changedId,
|
Native input dispatch rewrite work in progress.
The old dispatch mechanism has been left in place and continues to
be used by default for now. To enable native input dispatch,
edit the ENABLE_NATIVE_DISPATCH constant in WindowManagerPolicy.
Includes part of the new input event NDK API. Some details TBD.
To wire up input dispatch, as the ViewRoot adds a window to the
window session it receives an InputChannel object as an output
argument. The InputChannel encapsulates the file descriptors for a
shared memory region and two pipe end-points. The ViewRoot then
provides the InputChannel to the InputQueue. Behind the
scenes, InputQueue simply attaches handlers to the native PollLoop object
that underlies the MessageQueue. This way MessageQueue doesn't need
to know anything about input dispatch per-se, it just exposes (in native
code) a PollLoop that other components can use to monitor file descriptor
state changes.
There can be zero or more targets for any given input event. Each
input target is specified by its input channel and some parameters
including flags, an X/Y coordinate offset, and the dispatch timeout.
An input target can request either synchronous dispatch (for foreground apps)
or asynchronous dispatch (fire-and-forget for wallpapers and "outside"
targets). Currently, finding the appropriate input targets for an event
requires a call back into the WindowManagerServer from native code.
In the future this will be refactored to avoid most of these callbacks
except as required to handle pending focus transitions.
End-to-end event dispatch mostly works!
To do: event injection, rate limiting, ANRs, testing, optimization, etc.
Change-Id: I8c36b2b9e0a2d27392040ecda0f51b636456de25
2010-04-23 01:58:52 +00:00
|
|
|
int32_t motionEventAction) {
|
|
|
|
int32_t orientedWidth, orientedHeight;
|
|
|
|
switch (mDisplayOrientation) {
|
2010-06-15 08:31:58 +00:00
|
|
|
case InputReaderPolicyInterface::ROTATION_90:
|
|
|
|
case InputReaderPolicyInterface::ROTATION_270:
|
Native input dispatch rewrite work in progress.
The old dispatch mechanism has been left in place and continues to
be used by default for now. To enable native input dispatch,
edit the ENABLE_NATIVE_DISPATCH constant in WindowManagerPolicy.
Includes part of the new input event NDK API. Some details TBD.
To wire up input dispatch, as the ViewRoot adds a window to the
window session it receives an InputChannel object as an output
argument. The InputChannel encapsulates the file descriptors for a
shared memory region and two pipe end-points. The ViewRoot then
provides the InputChannel to the InputQueue. Behind the
scenes, InputQueue simply attaches handlers to the native PollLoop object
that underlies the MessageQueue. This way MessageQueue doesn't need
to know anything about input dispatch per-se, it just exposes (in native
code) a PollLoop that other components can use to monitor file descriptor
state changes.
There can be zero or more targets for any given input event. Each
input target is specified by its input channel and some parameters
including flags, an X/Y coordinate offset, and the dispatch timeout.
An input target can request either synchronous dispatch (for foreground apps)
or asynchronous dispatch (fire-and-forget for wallpapers and "outside"
targets). Currently, finding the appropriate input targets for an event
requires a call back into the WindowManagerServer from native code.
In the future this will be refactored to avoid most of these callbacks
except as required to handle pending focus transitions.
End-to-end event dispatch mostly works!
To do: event injection, rate limiting, ANRs, testing, optimization, etc.
Change-Id: I8c36b2b9e0a2d27392040ecda0f51b636456de25
2010-04-23 01:58:52 +00:00
|
|
|
orientedWidth = mDisplayHeight;
|
|
|
|
orientedHeight = mDisplayWidth;
|
|
|
|
break;
|
|
|
|
default:
|
|
|
|
orientedWidth = mDisplayWidth;
|
|
|
|
orientedHeight = mDisplayHeight;
|
|
|
|
break;
|
|
|
|
}
|
|
|
|
|
|
|
|
uint32_t pointerCount = 0;
|
|
|
|
int32_t pointerIds[MAX_POINTERS];
|
|
|
|
PointerCoords pointerCoords[MAX_POINTERS];
|
|
|
|
|
2010-06-29 23:52:21 +00:00
|
|
|
const InputDevice::TouchScreenState::Precalculated& precalculated =
|
|
|
|
device->touchScreen.precalculated;
|
|
|
|
|
Native input dispatch rewrite work in progress.
The old dispatch mechanism has been left in place and continues to
be used by default for now. To enable native input dispatch,
edit the ENABLE_NATIVE_DISPATCH constant in WindowManagerPolicy.
Includes part of the new input event NDK API. Some details TBD.
To wire up input dispatch, as the ViewRoot adds a window to the
window session it receives an InputChannel object as an output
argument. The InputChannel encapsulates the file descriptors for a
shared memory region and two pipe end-points. The ViewRoot then
provides the InputChannel to the InputQueue. Behind the
scenes, InputQueue simply attaches handlers to the native PollLoop object
that underlies the MessageQueue. This way MessageQueue doesn't need
to know anything about input dispatch per-se, it just exposes (in native
code) a PollLoop that other components can use to monitor file descriptor
state changes.
There can be zero or more targets for any given input event. Each
input target is specified by its input channel and some parameters
including flags, an X/Y coordinate offset, and the dispatch timeout.
An input target can request either synchronous dispatch (for foreground apps)
or asynchronous dispatch (fire-and-forget for wallpapers and "outside"
targets). Currently, finding the appropriate input targets for an event
requires a call back into the WindowManagerServer from native code.
In the future this will be refactored to avoid most of these callbacks
except as required to handle pending focus transitions.
End-to-end event dispatch mostly works!
To do: event injection, rate limiting, ANRs, testing, optimization, etc.
Change-Id: I8c36b2b9e0a2d27392040ecda0f51b636456de25
2010-04-23 01:58:52 +00:00
|
|
|
// Walk through the the active pointers and map touch screen coordinates (TouchData) into
|
|
|
|
// display coordinates (PointerCoords) and adjust for display orientation.
|
|
|
|
while (! idBits.isEmpty()) {
|
|
|
|
uint32_t id = idBits.firstMarkedBit();
|
|
|
|
idBits.clearBit(id);
|
|
|
|
uint32_t index = touch->idToIndex[id];
|
|
|
|
|
2010-06-29 23:52:21 +00:00
|
|
|
float x = float(touch->pointers[index].x
|
|
|
|
- precalculated.xOrigin) * precalculated.xScale;
|
|
|
|
float y = float(touch->pointers[index].y
|
|
|
|
- precalculated.yOrigin) * precalculated.yScale;
|
|
|
|
float pressure = float(touch->pointers[index].pressure
|
|
|
|
- precalculated.pressureOrigin) * precalculated.pressureScale;
|
|
|
|
float size = float(touch->pointers[index].size
|
|
|
|
- precalculated.sizeOrigin) * precalculated.sizeScale;
|
Native input dispatch rewrite work in progress.
The old dispatch mechanism has been left in place and continues to
be used by default for now. To enable native input dispatch,
edit the ENABLE_NATIVE_DISPATCH constant in WindowManagerPolicy.
Includes part of the new input event NDK API. Some details TBD.
To wire up input dispatch, as the ViewRoot adds a window to the
window session it receives an InputChannel object as an output
argument. The InputChannel encapsulates the file descriptors for a
shared memory region and two pipe end-points. The ViewRoot then
provides the InputChannel to the InputQueue. Behind the
scenes, InputQueue simply attaches handlers to the native PollLoop object
that underlies the MessageQueue. This way MessageQueue doesn't need
to know anything about input dispatch per-se, it just exposes (in native
code) a PollLoop that other components can use to monitor file descriptor
state changes.
There can be zero or more targets for any given input event. Each
input target is specified by its input channel and some parameters
including flags, an X/Y coordinate offset, and the dispatch timeout.
An input target can request either synchronous dispatch (for foreground apps)
or asynchronous dispatch (fire-and-forget for wallpapers and "outside"
targets). Currently, finding the appropriate input targets for an event
requires a call back into the WindowManagerServer from native code.
In the future this will be refactored to avoid most of these callbacks
except as required to handle pending focus transitions.
End-to-end event dispatch mostly works!
To do: event injection, rate limiting, ANRs, testing, optimization, etc.
Change-Id: I8c36b2b9e0a2d27392040ecda0f51b636456de25
2010-04-23 01:58:52 +00:00
|
|
|
|
2010-07-15 01:48:53 +00:00
|
|
|
float orientation = float(touch->pointers[index].orientation)
|
|
|
|
* precalculated.orientationScale;
|
|
|
|
|
|
|
|
bool vertical = abs(orientation) <= M_PI / 8;
|
|
|
|
|
Native input dispatch rewrite work in progress.
The old dispatch mechanism has been left in place and continues to
be used by default for now. To enable native input dispatch,
edit the ENABLE_NATIVE_DISPATCH constant in WindowManagerPolicy.
Includes part of the new input event NDK API. Some details TBD.
To wire up input dispatch, as the ViewRoot adds a window to the
window session it receives an InputChannel object as an output
argument. The InputChannel encapsulates the file descriptors for a
shared memory region and two pipe end-points. The ViewRoot then
provides the InputChannel to the InputQueue. Behind the
scenes, InputQueue simply attaches handlers to the native PollLoop object
that underlies the MessageQueue. This way MessageQueue doesn't need
to know anything about input dispatch per-se, it just exposes (in native
code) a PollLoop that other components can use to monitor file descriptor
state changes.
There can be zero or more targets for any given input event. Each
input target is specified by its input channel and some parameters
including flags, an X/Y coordinate offset, and the dispatch timeout.
An input target can request either synchronous dispatch (for foreground apps)
or asynchronous dispatch (fire-and-forget for wallpapers and "outside"
targets). Currently, finding the appropriate input targets for an event
requires a call back into the WindowManagerServer from native code.
In the future this will be refactored to avoid most of these callbacks
except as required to handle pending focus transitions.
End-to-end event dispatch mostly works!
To do: event injection, rate limiting, ANRs, testing, optimization, etc.
Change-Id: I8c36b2b9e0a2d27392040ecda0f51b636456de25
2010-04-23 01:58:52 +00:00
|
|
|
switch (mDisplayOrientation) {
|
2010-06-15 08:31:58 +00:00
|
|
|
case InputReaderPolicyInterface::ROTATION_90: {
|
Native input dispatch rewrite work in progress.
The old dispatch mechanism has been left in place and continues to
be used by default for now. To enable native input dispatch,
edit the ENABLE_NATIVE_DISPATCH constant in WindowManagerPolicy.
Includes part of the new input event NDK API. Some details TBD.
To wire up input dispatch, as the ViewRoot adds a window to the
window session it receives an InputChannel object as an output
argument. The InputChannel encapsulates the file descriptors for a
shared memory region and two pipe end-points. The ViewRoot then
provides the InputChannel to the InputQueue. Behind the
scenes, InputQueue simply attaches handlers to the native PollLoop object
that underlies the MessageQueue. This way MessageQueue doesn't need
to know anything about input dispatch per-se, it just exposes (in native
code) a PollLoop that other components can use to monitor file descriptor
state changes.
There can be zero or more targets for any given input event. Each
input target is specified by its input channel and some parameters
including flags, an X/Y coordinate offset, and the dispatch timeout.
An input target can request either synchronous dispatch (for foreground apps)
or asynchronous dispatch (fire-and-forget for wallpapers and "outside"
targets). Currently, finding the appropriate input targets for an event
requires a call back into the WindowManagerServer from native code.
In the future this will be refactored to avoid most of these callbacks
except as required to handle pending focus transitions.
End-to-end event dispatch mostly works!
To do: event injection, rate limiting, ANRs, testing, optimization, etc.
Change-Id: I8c36b2b9e0a2d27392040ecda0f51b636456de25
2010-04-23 01:58:52 +00:00
|
|
|
float xTemp = x;
|
|
|
|
x = y;
|
2010-06-18 03:52:56 +00:00
|
|
|
y = mDisplayWidth - xTemp;
|
2010-07-15 01:48:53 +00:00
|
|
|
vertical = ! vertical;
|
Native input dispatch rewrite work in progress.
The old dispatch mechanism has been left in place and continues to
be used by default for now. To enable native input dispatch,
edit the ENABLE_NATIVE_DISPATCH constant in WindowManagerPolicy.
Includes part of the new input event NDK API. Some details TBD.
To wire up input dispatch, as the ViewRoot adds a window to the
window session it receives an InputChannel object as an output
argument. The InputChannel encapsulates the file descriptors for a
shared memory region and two pipe end-points. The ViewRoot then
provides the InputChannel to the InputQueue. Behind the
scenes, InputQueue simply attaches handlers to the native PollLoop object
that underlies the MessageQueue. This way MessageQueue doesn't need
to know anything about input dispatch per-se, it just exposes (in native
code) a PollLoop that other components can use to monitor file descriptor
state changes.
There can be zero or more targets for any given input event. Each
input target is specified by its input channel and some parameters
including flags, an X/Y coordinate offset, and the dispatch timeout.
An input target can request either synchronous dispatch (for foreground apps)
or asynchronous dispatch (fire-and-forget for wallpapers and "outside"
targets). Currently, finding the appropriate input targets for an event
requires a call back into the WindowManagerServer from native code.
In the future this will be refactored to avoid most of these callbacks
except as required to handle pending focus transitions.
End-to-end event dispatch mostly works!
To do: event injection, rate limiting, ANRs, testing, optimization, etc.
Change-Id: I8c36b2b9e0a2d27392040ecda0f51b636456de25
2010-04-23 01:58:52 +00:00
|
|
|
break;
|
|
|
|
}
|
2010-06-15 08:31:58 +00:00
|
|
|
case InputReaderPolicyInterface::ROTATION_180: {
|
Native input dispatch rewrite work in progress.
The old dispatch mechanism has been left in place and continues to
be used by default for now. To enable native input dispatch,
edit the ENABLE_NATIVE_DISPATCH constant in WindowManagerPolicy.
Includes part of the new input event NDK API. Some details TBD.
To wire up input dispatch, as the ViewRoot adds a window to the
window session it receives an InputChannel object as an output
argument. The InputChannel encapsulates the file descriptors for a
shared memory region and two pipe end-points. The ViewRoot then
provides the InputChannel to the InputQueue. Behind the
scenes, InputQueue simply attaches handlers to the native PollLoop object
that underlies the MessageQueue. This way MessageQueue doesn't need
to know anything about input dispatch per-se, it just exposes (in native
code) a PollLoop that other components can use to monitor file descriptor
state changes.
There can be zero or more targets for any given input event. Each
input target is specified by its input channel and some parameters
including flags, an X/Y coordinate offset, and the dispatch timeout.
An input target can request either synchronous dispatch (for foreground apps)
or asynchronous dispatch (fire-and-forget for wallpapers and "outside"
targets). Currently, finding the appropriate input targets for an event
requires a call back into the WindowManagerServer from native code.
In the future this will be refactored to avoid most of these callbacks
except as required to handle pending focus transitions.
End-to-end event dispatch mostly works!
To do: event injection, rate limiting, ANRs, testing, optimization, etc.
Change-Id: I8c36b2b9e0a2d27392040ecda0f51b636456de25
2010-04-23 01:58:52 +00:00
|
|
|
x = mDisplayWidth - x;
|
|
|
|
y = mDisplayHeight - y;
|
|
|
|
break;
|
|
|
|
}
|
2010-06-15 08:31:58 +00:00
|
|
|
case InputReaderPolicyInterface::ROTATION_270: {
|
Native input dispatch rewrite work in progress.
The old dispatch mechanism has been left in place and continues to
be used by default for now. To enable native input dispatch,
edit the ENABLE_NATIVE_DISPATCH constant in WindowManagerPolicy.
Includes part of the new input event NDK API. Some details TBD.
To wire up input dispatch, as the ViewRoot adds a window to the
window session it receives an InputChannel object as an output
argument. The InputChannel encapsulates the file descriptors for a
shared memory region and two pipe end-points. The ViewRoot then
provides the InputChannel to the InputQueue. Behind the
scenes, InputQueue simply attaches handlers to the native PollLoop object
that underlies the MessageQueue. This way MessageQueue doesn't need
to know anything about input dispatch per-se, it just exposes (in native
code) a PollLoop that other components can use to monitor file descriptor
state changes.
There can be zero or more targets for any given input event. Each
input target is specified by its input channel and some parameters
including flags, an X/Y coordinate offset, and the dispatch timeout.
An input target can request either synchronous dispatch (for foreground apps)
or asynchronous dispatch (fire-and-forget for wallpapers and "outside"
targets). Currently, finding the appropriate input targets for an event
requires a call back into the WindowManagerServer from native code.
In the future this will be refactored to avoid most of these callbacks
except as required to handle pending focus transitions.
End-to-end event dispatch mostly works!
To do: event injection, rate limiting, ANRs, testing, optimization, etc.
Change-Id: I8c36b2b9e0a2d27392040ecda0f51b636456de25
2010-04-23 01:58:52 +00:00
|
|
|
float xTemp = x;
|
2010-06-18 03:52:56 +00:00
|
|
|
x = mDisplayHeight - y;
|
Native input dispatch rewrite work in progress.
The old dispatch mechanism has been left in place and continues to
be used by default for now. To enable native input dispatch,
edit the ENABLE_NATIVE_DISPATCH constant in WindowManagerPolicy.
Includes part of the new input event NDK API. Some details TBD.
To wire up input dispatch, as the ViewRoot adds a window to the
window session it receives an InputChannel object as an output
argument. The InputChannel encapsulates the file descriptors for a
shared memory region and two pipe end-points. The ViewRoot then
provides the InputChannel to the InputQueue. Behind the
scenes, InputQueue simply attaches handlers to the native PollLoop object
that underlies the MessageQueue. This way MessageQueue doesn't need
to know anything about input dispatch per-se, it just exposes (in native
code) a PollLoop that other components can use to monitor file descriptor
state changes.
There can be zero or more targets for any given input event. Each
input target is specified by its input channel and some parameters
including flags, an X/Y coordinate offset, and the dispatch timeout.
An input target can request either synchronous dispatch (for foreground apps)
or asynchronous dispatch (fire-and-forget for wallpapers and "outside"
targets). Currently, finding the appropriate input targets for an event
requires a call back into the WindowManagerServer from native code.
In the future this will be refactored to avoid most of these callbacks
except as required to handle pending focus transitions.
End-to-end event dispatch mostly works!
To do: event injection, rate limiting, ANRs, testing, optimization, etc.
Change-Id: I8c36b2b9e0a2d27392040ecda0f51b636456de25
2010-04-23 01:58:52 +00:00
|
|
|
y = xTemp;
|
2010-07-15 01:48:53 +00:00
|
|
|
vertical = ! vertical;
|
Native input dispatch rewrite work in progress.
The old dispatch mechanism has been left in place and continues to
be used by default for now. To enable native input dispatch,
edit the ENABLE_NATIVE_DISPATCH constant in WindowManagerPolicy.
Includes part of the new input event NDK API. Some details TBD.
To wire up input dispatch, as the ViewRoot adds a window to the
window session it receives an InputChannel object as an output
argument. The InputChannel encapsulates the file descriptors for a
shared memory region and two pipe end-points. The ViewRoot then
provides the InputChannel to the InputQueue. Behind the
scenes, InputQueue simply attaches handlers to the native PollLoop object
that underlies the MessageQueue. This way MessageQueue doesn't need
to know anything about input dispatch per-se, it just exposes (in native
code) a PollLoop that other components can use to monitor file descriptor
state changes.
There can be zero or more targets for any given input event. Each
input target is specified by its input channel and some parameters
including flags, an X/Y coordinate offset, and the dispatch timeout.
An input target can request either synchronous dispatch (for foreground apps)
or asynchronous dispatch (fire-and-forget for wallpapers and "outside"
targets). Currently, finding the appropriate input targets for an event
requires a call back into the WindowManagerServer from native code.
In the future this will be refactored to avoid most of these callbacks
except as required to handle pending focus transitions.
End-to-end event dispatch mostly works!
To do: event injection, rate limiting, ANRs, testing, optimization, etc.
Change-Id: I8c36b2b9e0a2d27392040ecda0f51b636456de25
2010-04-23 01:58:52 +00:00
|
|
|
break;
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
2010-07-15 01:48:53 +00:00
|
|
|
float touchMajor, touchMinor, toolMajor, toolMinor;
|
|
|
|
if (vertical) {
|
|
|
|
touchMajor = float(touch->pointers[index].touchMajor) * precalculated.yScale;
|
|
|
|
touchMinor = float(touch->pointers[index].touchMinor) * precalculated.xScale;
|
|
|
|
toolMajor = float(touch->pointers[index].toolMajor) * precalculated.yScale;
|
|
|
|
toolMinor = float(touch->pointers[index].toolMinor) * precalculated.xScale;
|
|
|
|
} else {
|
|
|
|
touchMajor = float(touch->pointers[index].touchMajor) * precalculated.xScale;
|
|
|
|
touchMinor = float(touch->pointers[index].touchMinor) * precalculated.yScale;
|
|
|
|
toolMajor = float(touch->pointers[index].toolMajor) * precalculated.xScale;
|
|
|
|
toolMinor = float(touch->pointers[index].toolMinor) * precalculated.yScale;
|
|
|
|
}
|
|
|
|
|
Native input dispatch rewrite work in progress.
The old dispatch mechanism has been left in place and continues to
be used by default for now. To enable native input dispatch,
edit the ENABLE_NATIVE_DISPATCH constant in WindowManagerPolicy.
Includes part of the new input event NDK API. Some details TBD.
To wire up input dispatch, as the ViewRoot adds a window to the
window session it receives an InputChannel object as an output
argument. The InputChannel encapsulates the file descriptors for a
shared memory region and two pipe end-points. The ViewRoot then
provides the InputChannel to the InputQueue. Behind the
scenes, InputQueue simply attaches handlers to the native PollLoop object
that underlies the MessageQueue. This way MessageQueue doesn't need
to know anything about input dispatch per-se, it just exposes (in native
code) a PollLoop that other components can use to monitor file descriptor
state changes.
There can be zero or more targets for any given input event. Each
input target is specified by its input channel and some parameters
including flags, an X/Y coordinate offset, and the dispatch timeout.
An input target can request either synchronous dispatch (for foreground apps)
or asynchronous dispatch (fire-and-forget for wallpapers and "outside"
targets). Currently, finding the appropriate input targets for an event
requires a call back into the WindowManagerServer from native code.
In the future this will be refactored to avoid most of these callbacks
except as required to handle pending focus transitions.
End-to-end event dispatch mostly works!
To do: event injection, rate limiting, ANRs, testing, optimization, etc.
Change-Id: I8c36b2b9e0a2d27392040ecda0f51b636456de25
2010-04-23 01:58:52 +00:00
|
|
|
pointerIds[pointerCount] = int32_t(id);
|
|
|
|
|
|
|
|
pointerCoords[pointerCount].x = x;
|
|
|
|
pointerCoords[pointerCount].y = y;
|
|
|
|
pointerCoords[pointerCount].pressure = pressure;
|
|
|
|
pointerCoords[pointerCount].size = size;
|
2010-07-15 01:48:53 +00:00
|
|
|
pointerCoords[pointerCount].touchMajor = touchMajor;
|
|
|
|
pointerCoords[pointerCount].touchMinor = touchMinor;
|
|
|
|
pointerCoords[pointerCount].toolMajor = toolMajor;
|
|
|
|
pointerCoords[pointerCount].toolMinor = toolMinor;
|
|
|
|
pointerCoords[pointerCount].orientation = orientation;
|
Native input dispatch rewrite work in progress.
The old dispatch mechanism has been left in place and continues to
be used by default for now. To enable native input dispatch,
edit the ENABLE_NATIVE_DISPATCH constant in WindowManagerPolicy.
Includes part of the new input event NDK API. Some details TBD.
To wire up input dispatch, as the ViewRoot adds a window to the
window session it receives an InputChannel object as an output
argument. The InputChannel encapsulates the file descriptors for a
shared memory region and two pipe end-points. The ViewRoot then
provides the InputChannel to the InputQueue. Behind the
scenes, InputQueue simply attaches handlers to the native PollLoop object
that underlies the MessageQueue. This way MessageQueue doesn't need
to know anything about input dispatch per-se, it just exposes (in native
code) a PollLoop that other components can use to monitor file descriptor
state changes.
There can be zero or more targets for any given input event. Each
input target is specified by its input channel and some parameters
including flags, an X/Y coordinate offset, and the dispatch timeout.
An input target can request either synchronous dispatch (for foreground apps)
or asynchronous dispatch (fire-and-forget for wallpapers and "outside"
targets). Currently, finding the appropriate input targets for an event
requires a call back into the WindowManagerServer from native code.
In the future this will be refactored to avoid most of these callbacks
except as required to handle pending focus transitions.
End-to-end event dispatch mostly works!
To do: event injection, rate limiting, ANRs, testing, optimization, etc.
Change-Id: I8c36b2b9e0a2d27392040ecda0f51b636456de25
2010-04-23 01:58:52 +00:00
|
|
|
|
2010-07-16 22:01:56 +00:00
|
|
|
if (id == changedId) {
|
|
|
|
motionEventAction |= pointerCount << AMOTION_EVENT_ACTION_POINTER_INDEX_SHIFT;
|
|
|
|
}
|
|
|
|
|
Native input dispatch rewrite work in progress.
The old dispatch mechanism has been left in place and continues to
be used by default for now. To enable native input dispatch,
edit the ENABLE_NATIVE_DISPATCH constant in WindowManagerPolicy.
Includes part of the new input event NDK API. Some details TBD.
To wire up input dispatch, as the ViewRoot adds a window to the
window session it receives an InputChannel object as an output
argument. The InputChannel encapsulates the file descriptors for a
shared memory region and two pipe end-points. The ViewRoot then
provides the InputChannel to the InputQueue. Behind the
scenes, InputQueue simply attaches handlers to the native PollLoop object
that underlies the MessageQueue. This way MessageQueue doesn't need
to know anything about input dispatch per-se, it just exposes (in native
code) a PollLoop that other components can use to monitor file descriptor
state changes.
There can be zero or more targets for any given input event. Each
input target is specified by its input channel and some parameters
including flags, an X/Y coordinate offset, and the dispatch timeout.
An input target can request either synchronous dispatch (for foreground apps)
or asynchronous dispatch (fire-and-forget for wallpapers and "outside"
targets). Currently, finding the appropriate input targets for an event
requires a call back into the WindowManagerServer from native code.
In the future this will be refactored to avoid most of these callbacks
except as required to handle pending focus transitions.
End-to-end event dispatch mostly works!
To do: event injection, rate limiting, ANRs, testing, optimization, etc.
Change-Id: I8c36b2b9e0a2d27392040ecda0f51b636456de25
2010-04-23 01:58:52 +00:00
|
|
|
pointerCount += 1;
|
|
|
|
}
|
|
|
|
|
|
|
|
// Check edge flags by looking only at the first pointer since the flags are
|
|
|
|
// global to the event.
|
|
|
|
int32_t motionEventEdgeFlags = 0;
|
2010-07-15 01:48:53 +00:00
|
|
|
if (motionEventAction == AMOTION_EVENT_ACTION_DOWN) {
|
Native input dispatch rewrite work in progress.
The old dispatch mechanism has been left in place and continues to
be used by default for now. To enable native input dispatch,
edit the ENABLE_NATIVE_DISPATCH constant in WindowManagerPolicy.
Includes part of the new input event NDK API. Some details TBD.
To wire up input dispatch, as the ViewRoot adds a window to the
window session it receives an InputChannel object as an output
argument. The InputChannel encapsulates the file descriptors for a
shared memory region and two pipe end-points. The ViewRoot then
provides the InputChannel to the InputQueue. Behind the
scenes, InputQueue simply attaches handlers to the native PollLoop object
that underlies the MessageQueue. This way MessageQueue doesn't need
to know anything about input dispatch per-se, it just exposes (in native
code) a PollLoop that other components can use to monitor file descriptor
state changes.
There can be zero or more targets for any given input event. Each
input target is specified by its input channel and some parameters
including flags, an X/Y coordinate offset, and the dispatch timeout.
An input target can request either synchronous dispatch (for foreground apps)
or asynchronous dispatch (fire-and-forget for wallpapers and "outside"
targets). Currently, finding the appropriate input targets for an event
requires a call back into the WindowManagerServer from native code.
In the future this will be refactored to avoid most of these callbacks
except as required to handle pending focus transitions.
End-to-end event dispatch mostly works!
To do: event injection, rate limiting, ANRs, testing, optimization, etc.
Change-Id: I8c36b2b9e0a2d27392040ecda0f51b636456de25
2010-04-23 01:58:52 +00:00
|
|
|
if (pointerCoords[0].x <= 0) {
|
2010-07-15 01:48:53 +00:00
|
|
|
motionEventEdgeFlags |= AMOTION_EVENT_EDGE_FLAG_LEFT;
|
Native input dispatch rewrite work in progress.
The old dispatch mechanism has been left in place and continues to
be used by default for now. To enable native input dispatch,
edit the ENABLE_NATIVE_DISPATCH constant in WindowManagerPolicy.
Includes part of the new input event NDK API. Some details TBD.
To wire up input dispatch, as the ViewRoot adds a window to the
window session it receives an InputChannel object as an output
argument. The InputChannel encapsulates the file descriptors for a
shared memory region and two pipe end-points. The ViewRoot then
provides the InputChannel to the InputQueue. Behind the
scenes, InputQueue simply attaches handlers to the native PollLoop object
that underlies the MessageQueue. This way MessageQueue doesn't need
to know anything about input dispatch per-se, it just exposes (in native
code) a PollLoop that other components can use to monitor file descriptor
state changes.
There can be zero or more targets for any given input event. Each
input target is specified by its input channel and some parameters
including flags, an X/Y coordinate offset, and the dispatch timeout.
An input target can request either synchronous dispatch (for foreground apps)
or asynchronous dispatch (fire-and-forget for wallpapers and "outside"
targets). Currently, finding the appropriate input targets for an event
requires a call back into the WindowManagerServer from native code.
In the future this will be refactored to avoid most of these callbacks
except as required to handle pending focus transitions.
End-to-end event dispatch mostly works!
To do: event injection, rate limiting, ANRs, testing, optimization, etc.
Change-Id: I8c36b2b9e0a2d27392040ecda0f51b636456de25
2010-04-23 01:58:52 +00:00
|
|
|
} else if (pointerCoords[0].x >= orientedWidth) {
|
2010-07-15 01:48:53 +00:00
|
|
|
motionEventEdgeFlags |= AMOTION_EVENT_EDGE_FLAG_RIGHT;
|
Native input dispatch rewrite work in progress.
The old dispatch mechanism has been left in place and continues to
be used by default for now. To enable native input dispatch,
edit the ENABLE_NATIVE_DISPATCH constant in WindowManagerPolicy.
Includes part of the new input event NDK API. Some details TBD.
To wire up input dispatch, as the ViewRoot adds a window to the
window session it receives an InputChannel object as an output
argument. The InputChannel encapsulates the file descriptors for a
shared memory region and two pipe end-points. The ViewRoot then
provides the InputChannel to the InputQueue. Behind the
scenes, InputQueue simply attaches handlers to the native PollLoop object
that underlies the MessageQueue. This way MessageQueue doesn't need
to know anything about input dispatch per-se, it just exposes (in native
code) a PollLoop that other components can use to monitor file descriptor
state changes.
There can be zero or more targets for any given input event. Each
input target is specified by its input channel and some parameters
including flags, an X/Y coordinate offset, and the dispatch timeout.
An input target can request either synchronous dispatch (for foreground apps)
or asynchronous dispatch (fire-and-forget for wallpapers and "outside"
targets). Currently, finding the appropriate input targets for an event
requires a call back into the WindowManagerServer from native code.
In the future this will be refactored to avoid most of these callbacks
except as required to handle pending focus transitions.
End-to-end event dispatch mostly works!
To do: event injection, rate limiting, ANRs, testing, optimization, etc.
Change-Id: I8c36b2b9e0a2d27392040ecda0f51b636456de25
2010-04-23 01:58:52 +00:00
|
|
|
}
|
|
|
|
if (pointerCoords[0].y <= 0) {
|
2010-07-15 01:48:53 +00:00
|
|
|
motionEventEdgeFlags |= AMOTION_EVENT_EDGE_FLAG_TOP;
|
Native input dispatch rewrite work in progress.
The old dispatch mechanism has been left in place and continues to
be used by default for now. To enable native input dispatch,
edit the ENABLE_NATIVE_DISPATCH constant in WindowManagerPolicy.
Includes part of the new input event NDK API. Some details TBD.
To wire up input dispatch, as the ViewRoot adds a window to the
window session it receives an InputChannel object as an output
argument. The InputChannel encapsulates the file descriptors for a
shared memory region and two pipe end-points. The ViewRoot then
provides the InputChannel to the InputQueue. Behind the
scenes, InputQueue simply attaches handlers to the native PollLoop object
that underlies the MessageQueue. This way MessageQueue doesn't need
to know anything about input dispatch per-se, it just exposes (in native
code) a PollLoop that other components can use to monitor file descriptor
state changes.
There can be zero or more targets for any given input event. Each
input target is specified by its input channel and some parameters
including flags, an X/Y coordinate offset, and the dispatch timeout.
An input target can request either synchronous dispatch (for foreground apps)
or asynchronous dispatch (fire-and-forget for wallpapers and "outside"
targets). Currently, finding the appropriate input targets for an event
requires a call back into the WindowManagerServer from native code.
In the future this will be refactored to avoid most of these callbacks
except as required to handle pending focus transitions.
End-to-end event dispatch mostly works!
To do: event injection, rate limiting, ANRs, testing, optimization, etc.
Change-Id: I8c36b2b9e0a2d27392040ecda0f51b636456de25
2010-04-23 01:58:52 +00:00
|
|
|
} else if (pointerCoords[0].y >= orientedHeight) {
|
2010-07-15 01:48:53 +00:00
|
|
|
motionEventEdgeFlags |= AMOTION_EVENT_EDGE_FLAG_BOTTOM;
|
Native input dispatch rewrite work in progress.
The old dispatch mechanism has been left in place and continues to
be used by default for now. To enable native input dispatch,
edit the ENABLE_NATIVE_DISPATCH constant in WindowManagerPolicy.
Includes part of the new input event NDK API. Some details TBD.
To wire up input dispatch, as the ViewRoot adds a window to the
window session it receives an InputChannel object as an output
argument. The InputChannel encapsulates the file descriptors for a
shared memory region and two pipe end-points. The ViewRoot then
provides the InputChannel to the InputQueue. Behind the
scenes, InputQueue simply attaches handlers to the native PollLoop object
that underlies the MessageQueue. This way MessageQueue doesn't need
to know anything about input dispatch per-se, it just exposes (in native
code) a PollLoop that other components can use to monitor file descriptor
state changes.
There can be zero or more targets for any given input event. Each
input target is specified by its input channel and some parameters
including flags, an X/Y coordinate offset, and the dispatch timeout.
An input target can request either synchronous dispatch (for foreground apps)
or asynchronous dispatch (fire-and-forget for wallpapers and "outside"
targets). Currently, finding the appropriate input targets for an event
requires a call back into the WindowManagerServer from native code.
In the future this will be refactored to avoid most of these callbacks
except as required to handle pending focus transitions.
End-to-end event dispatch mostly works!
To do: event injection, rate limiting, ANRs, testing, optimization, etc.
Change-Id: I8c36b2b9e0a2d27392040ecda0f51b636456de25
2010-04-23 01:58:52 +00:00
|
|
|
}
|
|
|
|
}
|
|
|
|
|
|
|
|
nsecs_t downTime = device->touchScreen.downTime;
|
2010-07-15 01:48:53 +00:00
|
|
|
mDispatcher->notifyMotion(when, device->id, AINPUT_SOURCE_TOUCHSCREEN, policyFlags,
|
Native input dispatch rewrite work in progress.
The old dispatch mechanism has been left in place and continues to
be used by default for now. To enable native input dispatch,
edit the ENABLE_NATIVE_DISPATCH constant in WindowManagerPolicy.
Includes part of the new input event NDK API. Some details TBD.
To wire up input dispatch, as the ViewRoot adds a window to the
window session it receives an InputChannel object as an output
argument. The InputChannel encapsulates the file descriptors for a
shared memory region and two pipe end-points. The ViewRoot then
provides the InputChannel to the InputQueue. Behind the
scenes, InputQueue simply attaches handlers to the native PollLoop object
that underlies the MessageQueue. This way MessageQueue doesn't need
to know anything about input dispatch per-se, it just exposes (in native
code) a PollLoop that other components can use to monitor file descriptor
state changes.
There can be zero or more targets for any given input event. Each
input target is specified by its input channel and some parameters
including flags, an X/Y coordinate offset, and the dispatch timeout.
An input target can request either synchronous dispatch (for foreground apps)
or asynchronous dispatch (fire-and-forget for wallpapers and "outside"
targets). Currently, finding the appropriate input targets for an event
requires a call back into the WindowManagerServer from native code.
In the future this will be refactored to avoid most of these callbacks
except as required to handle pending focus transitions.
End-to-end event dispatch mostly works!
To do: event injection, rate limiting, ANRs, testing, optimization, etc.
Change-Id: I8c36b2b9e0a2d27392040ecda0f51b636456de25
2010-04-23 01:58:52 +00:00
|
|
|
motionEventAction, globalMetaState(), motionEventEdgeFlags,
|
|
|
|
pointerCount, pointerIds, pointerCoords,
|
|
|
|
0, 0, downTime);
|
|
|
|
}
|
|
|
|
|
|
|
|
void InputReader::onTrackballStateChanged(nsecs_t when,
|
|
|
|
InputDevice* device) {
|
|
|
|
static const uint32_t DELTA_FIELDS =
|
|
|
|
InputDevice::TrackballState::Accumulator::FIELD_REL_X
|
|
|
|
| InputDevice::TrackballState::Accumulator::FIELD_REL_Y;
|
|
|
|
|
|
|
|
/* Refresh display properties so we can trackball moves according to display orientation */
|
|
|
|
|
|
|
|
if (! refreshDisplayProperties()) {
|
|
|
|
return;
|
|
|
|
}
|
|
|
|
|
|
|
|
/* Update device state */
|
|
|
|
|
|
|
|
uint32_t fields = device->trackball.accumulator.fields;
|
|
|
|
bool downChanged = fields & InputDevice::TrackballState::Accumulator::FIELD_BTN_MOUSE;
|
2010-06-18 03:52:56 +00:00
|
|
|
bool deltaChanged = fields & DELTA_FIELDS;
|
Native input dispatch rewrite work in progress.
The old dispatch mechanism has been left in place and continues to
be used by default for now. To enable native input dispatch,
edit the ENABLE_NATIVE_DISPATCH constant in WindowManagerPolicy.
Includes part of the new input event NDK API. Some details TBD.
To wire up input dispatch, as the ViewRoot adds a window to the
window session it receives an InputChannel object as an output
argument. The InputChannel encapsulates the file descriptors for a
shared memory region and two pipe end-points. The ViewRoot then
provides the InputChannel to the InputQueue. Behind the
scenes, InputQueue simply attaches handlers to the native PollLoop object
that underlies the MessageQueue. This way MessageQueue doesn't need
to know anything about input dispatch per-se, it just exposes (in native
code) a PollLoop that other components can use to monitor file descriptor
state changes.
There can be zero or more targets for any given input event. Each
input target is specified by its input channel and some parameters
including flags, an X/Y coordinate offset, and the dispatch timeout.
An input target can request either synchronous dispatch (for foreground apps)
or asynchronous dispatch (fire-and-forget for wallpapers and "outside"
targets). Currently, finding the appropriate input targets for an event
requires a call back into the WindowManagerServer from native code.
In the future this will be refactored to avoid most of these callbacks
except as required to handle pending focus transitions.
End-to-end event dispatch mostly works!
To do: event injection, rate limiting, ANRs, testing, optimization, etc.
Change-Id: I8c36b2b9e0a2d27392040ecda0f51b636456de25
2010-04-23 01:58:52 +00:00
|
|
|
|
|
|
|
bool down;
|
|
|
|
if (downChanged) {
|
|
|
|
if (device->trackball.accumulator.btnMouse) {
|
|
|
|
device->trackball.current.down = true;
|
|
|
|
device->trackball.current.downTime = when;
|
|
|
|
down = true;
|
|
|
|
} else {
|
|
|
|
device->trackball.current.down = false;
|
|
|
|
down = false;
|
|
|
|
}
|
|
|
|
} else {
|
|
|
|
down = device->trackball.current.down;
|
|
|
|
}
|
|
|
|
|
|
|
|
/* Apply policy */
|
|
|
|
|
|
|
|
int32_t policyActions = mPolicy->interceptTrackball(when, downChanged, down, deltaChanged);
|
|
|
|
|
|
|
|
uint32_t policyFlags = 0;
|
|
|
|
if (! applyStandardInputDispatchPolicyActions(when, policyActions, & policyFlags)) {
|
|
|
|
return; // event dropped
|
|
|
|
}
|
|
|
|
|
|
|
|
/* Enqueue motion event for dispatch */
|
|
|
|
|
|
|
|
int32_t motionEventAction;
|
|
|
|
if (downChanged) {
|
2010-07-15 01:48:53 +00:00
|
|
|
motionEventAction = down ? AMOTION_EVENT_ACTION_DOWN : AMOTION_EVENT_ACTION_UP;
|
Native input dispatch rewrite work in progress.
The old dispatch mechanism has been left in place and continues to
be used by default for now. To enable native input dispatch,
edit the ENABLE_NATIVE_DISPATCH constant in WindowManagerPolicy.
Includes part of the new input event NDK API. Some details TBD.
To wire up input dispatch, as the ViewRoot adds a window to the
window session it receives an InputChannel object as an output
argument. The InputChannel encapsulates the file descriptors for a
shared memory region and two pipe end-points. The ViewRoot then
provides the InputChannel to the InputQueue. Behind the
scenes, InputQueue simply attaches handlers to the native PollLoop object
that underlies the MessageQueue. This way MessageQueue doesn't need
to know anything about input dispatch per-se, it just exposes (in native
code) a PollLoop that other components can use to monitor file descriptor
state changes.
There can be zero or more targets for any given input event. Each
input target is specified by its input channel and some parameters
including flags, an X/Y coordinate offset, and the dispatch timeout.
An input target can request either synchronous dispatch (for foreground apps)
or asynchronous dispatch (fire-and-forget for wallpapers and "outside"
targets). Currently, finding the appropriate input targets for an event
requires a call back into the WindowManagerServer from native code.
In the future this will be refactored to avoid most of these callbacks
except as required to handle pending focus transitions.
End-to-end event dispatch mostly works!
To do: event injection, rate limiting, ANRs, testing, optimization, etc.
Change-Id: I8c36b2b9e0a2d27392040ecda0f51b636456de25
2010-04-23 01:58:52 +00:00
|
|
|
} else {
|
2010-07-15 01:48:53 +00:00
|
|
|
motionEventAction = AMOTION_EVENT_ACTION_MOVE;
|
Native input dispatch rewrite work in progress.
The old dispatch mechanism has been left in place and continues to
be used by default for now. To enable native input dispatch,
edit the ENABLE_NATIVE_DISPATCH constant in WindowManagerPolicy.
Includes part of the new input event NDK API. Some details TBD.
To wire up input dispatch, as the ViewRoot adds a window to the
window session it receives an InputChannel object as an output
argument. The InputChannel encapsulates the file descriptors for a
shared memory region and two pipe end-points. The ViewRoot then
provides the InputChannel to the InputQueue. Behind the
scenes, InputQueue simply attaches handlers to the native PollLoop object
that underlies the MessageQueue. This way MessageQueue doesn't need
to know anything about input dispatch per-se, it just exposes (in native
code) a PollLoop that other components can use to monitor file descriptor
state changes.
There can be zero or more targets for any given input event. Each
input target is specified by its input channel and some parameters
including flags, an X/Y coordinate offset, and the dispatch timeout.
An input target can request either synchronous dispatch (for foreground apps)
or asynchronous dispatch (fire-and-forget for wallpapers and "outside"
targets). Currently, finding the appropriate input targets for an event
requires a call back into the WindowManagerServer from native code.
In the future this will be refactored to avoid most of these callbacks
except as required to handle pending focus transitions.
End-to-end event dispatch mostly works!
To do: event injection, rate limiting, ANRs, testing, optimization, etc.
Change-Id: I8c36b2b9e0a2d27392040ecda0f51b636456de25
2010-04-23 01:58:52 +00:00
|
|
|
}
|
|
|
|
|
|
|
|
int32_t pointerId = 0;
|
|
|
|
PointerCoords pointerCoords;
|
2010-06-18 03:52:56 +00:00
|
|
|
pointerCoords.x = fields & InputDevice::TrackballState::Accumulator::FIELD_REL_X
|
|
|
|
? device->trackball.accumulator.relX * device->trackball.precalculated.xScale : 0;
|
|
|
|
pointerCoords.y = fields & InputDevice::TrackballState::Accumulator::FIELD_REL_Y
|
|
|
|
? device->trackball.accumulator.relY * device->trackball.precalculated.yScale : 0;
|
Native input dispatch rewrite work in progress.
The old dispatch mechanism has been left in place and continues to
be used by default for now. To enable native input dispatch,
edit the ENABLE_NATIVE_DISPATCH constant in WindowManagerPolicy.
Includes part of the new input event NDK API. Some details TBD.
To wire up input dispatch, as the ViewRoot adds a window to the
window session it receives an InputChannel object as an output
argument. The InputChannel encapsulates the file descriptors for a
shared memory region and two pipe end-points. The ViewRoot then
provides the InputChannel to the InputQueue. Behind the
scenes, InputQueue simply attaches handlers to the native PollLoop object
that underlies the MessageQueue. This way MessageQueue doesn't need
to know anything about input dispatch per-se, it just exposes (in native
code) a PollLoop that other components can use to monitor file descriptor
state changes.
There can be zero or more targets for any given input event. Each
input target is specified by its input channel and some parameters
including flags, an X/Y coordinate offset, and the dispatch timeout.
An input target can request either synchronous dispatch (for foreground apps)
or asynchronous dispatch (fire-and-forget for wallpapers and "outside"
targets). Currently, finding the appropriate input targets for an event
requires a call back into the WindowManagerServer from native code.
In the future this will be refactored to avoid most of these callbacks
except as required to handle pending focus transitions.
End-to-end event dispatch mostly works!
To do: event injection, rate limiting, ANRs, testing, optimization, etc.
Change-Id: I8c36b2b9e0a2d27392040ecda0f51b636456de25
2010-04-23 01:58:52 +00:00
|
|
|
pointerCoords.pressure = 1.0f; // XXX Consider making this 1.0f if down, 0 otherwise.
|
|
|
|
pointerCoords.size = 0;
|
2010-07-15 01:48:53 +00:00
|
|
|
pointerCoords.touchMajor = 0;
|
|
|
|
pointerCoords.touchMinor = 0;
|
|
|
|
pointerCoords.toolMajor = 0;
|
|
|
|
pointerCoords.toolMinor = 0;
|
|
|
|
pointerCoords.orientation = 0;
|
Native input dispatch rewrite work in progress.
The old dispatch mechanism has been left in place and continues to
be used by default for now. To enable native input dispatch,
edit the ENABLE_NATIVE_DISPATCH constant in WindowManagerPolicy.
Includes part of the new input event NDK API. Some details TBD.
To wire up input dispatch, as the ViewRoot adds a window to the
window session it receives an InputChannel object as an output
argument. The InputChannel encapsulates the file descriptors for a
shared memory region and two pipe end-points. The ViewRoot then
provides the InputChannel to the InputQueue. Behind the
scenes, InputQueue simply attaches handlers to the native PollLoop object
that underlies the MessageQueue. This way MessageQueue doesn't need
to know anything about input dispatch per-se, it just exposes (in native
code) a PollLoop that other components can use to monitor file descriptor
state changes.
There can be zero or more targets for any given input event. Each
input target is specified by its input channel and some parameters
including flags, an X/Y coordinate offset, and the dispatch timeout.
An input target can request either synchronous dispatch (for foreground apps)
or asynchronous dispatch (fire-and-forget for wallpapers and "outside"
targets). Currently, finding the appropriate input targets for an event
requires a call back into the WindowManagerServer from native code.
In the future this will be refactored to avoid most of these callbacks
except as required to handle pending focus transitions.
End-to-end event dispatch mostly works!
To do: event injection, rate limiting, ANRs, testing, optimization, etc.
Change-Id: I8c36b2b9e0a2d27392040ecda0f51b636456de25
2010-04-23 01:58:52 +00:00
|
|
|
|
|
|
|
float temp;
|
|
|
|
switch (mDisplayOrientation) {
|
2010-06-15 08:31:58 +00:00
|
|
|
case InputReaderPolicyInterface::ROTATION_90:
|
Native input dispatch rewrite work in progress.
The old dispatch mechanism has been left in place and continues to
be used by default for now. To enable native input dispatch,
edit the ENABLE_NATIVE_DISPATCH constant in WindowManagerPolicy.
Includes part of the new input event NDK API. Some details TBD.
To wire up input dispatch, as the ViewRoot adds a window to the
window session it receives an InputChannel object as an output
argument. The InputChannel encapsulates the file descriptors for a
shared memory region and two pipe end-points. The ViewRoot then
provides the InputChannel to the InputQueue. Behind the
scenes, InputQueue simply attaches handlers to the native PollLoop object
that underlies the MessageQueue. This way MessageQueue doesn't need
to know anything about input dispatch per-se, it just exposes (in native
code) a PollLoop that other components can use to monitor file descriptor
state changes.
There can be zero or more targets for any given input event. Each
input target is specified by its input channel and some parameters
including flags, an X/Y coordinate offset, and the dispatch timeout.
An input target can request either synchronous dispatch (for foreground apps)
or asynchronous dispatch (fire-and-forget for wallpapers and "outside"
targets). Currently, finding the appropriate input targets for an event
requires a call back into the WindowManagerServer from native code.
In the future this will be refactored to avoid most of these callbacks
except as required to handle pending focus transitions.
End-to-end event dispatch mostly works!
To do: event injection, rate limiting, ANRs, testing, optimization, etc.
Change-Id: I8c36b2b9e0a2d27392040ecda0f51b636456de25
2010-04-23 01:58:52 +00:00
|
|
|
temp = pointerCoords.x;
|
|
|
|
pointerCoords.x = pointerCoords.y;
|
|
|
|
pointerCoords.y = - temp;
|
|
|
|
break;
|
|
|
|
|
2010-06-15 08:31:58 +00:00
|
|
|
case InputReaderPolicyInterface::ROTATION_180:
|
Native input dispatch rewrite work in progress.
The old dispatch mechanism has been left in place and continues to
be used by default for now. To enable native input dispatch,
edit the ENABLE_NATIVE_DISPATCH constant in WindowManagerPolicy.
Includes part of the new input event NDK API. Some details TBD.
To wire up input dispatch, as the ViewRoot adds a window to the
window session it receives an InputChannel object as an output
argument. The InputChannel encapsulates the file descriptors for a
shared memory region and two pipe end-points. The ViewRoot then
provides the InputChannel to the InputQueue. Behind the
scenes, InputQueue simply attaches handlers to the native PollLoop object
that underlies the MessageQueue. This way MessageQueue doesn't need
to know anything about input dispatch per-se, it just exposes (in native
code) a PollLoop that other components can use to monitor file descriptor
state changes.
There can be zero or more targets for any given input event. Each
input target is specified by its input channel and some parameters
including flags, an X/Y coordinate offset, and the dispatch timeout.
An input target can request either synchronous dispatch (for foreground apps)
or asynchronous dispatch (fire-and-forget for wallpapers and "outside"
targets). Currently, finding the appropriate input targets for an event
requires a call back into the WindowManagerServer from native code.
In the future this will be refactored to avoid most of these callbacks
except as required to handle pending focus transitions.
End-to-end event dispatch mostly works!
To do: event injection, rate limiting, ANRs, testing, optimization, etc.
Change-Id: I8c36b2b9e0a2d27392040ecda0f51b636456de25
2010-04-23 01:58:52 +00:00
|
|
|
pointerCoords.x = - pointerCoords.x;
|
|
|
|
pointerCoords.y = - pointerCoords.y;
|
|
|
|
break;
|
|
|
|
|
2010-06-15 08:31:58 +00:00
|
|
|
case InputReaderPolicyInterface::ROTATION_270:
|
Native input dispatch rewrite work in progress.
The old dispatch mechanism has been left in place and continues to
be used by default for now. To enable native input dispatch,
edit the ENABLE_NATIVE_DISPATCH constant in WindowManagerPolicy.
Includes part of the new input event NDK API. Some details TBD.
To wire up input dispatch, as the ViewRoot adds a window to the
window session it receives an InputChannel object as an output
argument. The InputChannel encapsulates the file descriptors for a
shared memory region and two pipe end-points. The ViewRoot then
provides the InputChannel to the InputQueue. Behind the
scenes, InputQueue simply attaches handlers to the native PollLoop object
that underlies the MessageQueue. This way MessageQueue doesn't need
to know anything about input dispatch per-se, it just exposes (in native
code) a PollLoop that other components can use to monitor file descriptor
state changes.
There can be zero or more targets for any given input event. Each
input target is specified by its input channel and some parameters
including flags, an X/Y coordinate offset, and the dispatch timeout.
An input target can request either synchronous dispatch (for foreground apps)
or asynchronous dispatch (fire-and-forget for wallpapers and "outside"
targets). Currently, finding the appropriate input targets for an event
requires a call back into the WindowManagerServer from native code.
In the future this will be refactored to avoid most of these callbacks
except as required to handle pending focus transitions.
End-to-end event dispatch mostly works!
To do: event injection, rate limiting, ANRs, testing, optimization, etc.
Change-Id: I8c36b2b9e0a2d27392040ecda0f51b636456de25
2010-04-23 01:58:52 +00:00
|
|
|
temp = pointerCoords.x;
|
|
|
|
pointerCoords.x = - pointerCoords.y;
|
|
|
|
pointerCoords.y = temp;
|
|
|
|
break;
|
|
|
|
}
|
|
|
|
|
2010-07-15 01:48:53 +00:00
|
|
|
mDispatcher->notifyMotion(when, device->id, AINPUT_SOURCE_TRACKBALL, policyFlags,
|
|
|
|
motionEventAction, globalMetaState(), AMOTION_EVENT_EDGE_FLAG_NONE,
|
Native input dispatch rewrite work in progress.
The old dispatch mechanism has been left in place and continues to
be used by default for now. To enable native input dispatch,
edit the ENABLE_NATIVE_DISPATCH constant in WindowManagerPolicy.
Includes part of the new input event NDK API. Some details TBD.
To wire up input dispatch, as the ViewRoot adds a window to the
window session it receives an InputChannel object as an output
argument. The InputChannel encapsulates the file descriptors for a
shared memory region and two pipe end-points. The ViewRoot then
provides the InputChannel to the InputQueue. Behind the
scenes, InputQueue simply attaches handlers to the native PollLoop object
that underlies the MessageQueue. This way MessageQueue doesn't need
to know anything about input dispatch per-se, it just exposes (in native
code) a PollLoop that other components can use to monitor file descriptor
state changes.
There can be zero or more targets for any given input event. Each
input target is specified by its input channel and some parameters
including flags, an X/Y coordinate offset, and the dispatch timeout.
An input target can request either synchronous dispatch (for foreground apps)
or asynchronous dispatch (fire-and-forget for wallpapers and "outside"
targets). Currently, finding the appropriate input targets for an event
requires a call back into the WindowManagerServer from native code.
In the future this will be refactored to avoid most of these callbacks
except as required to handle pending focus transitions.
End-to-end event dispatch mostly works!
To do: event injection, rate limiting, ANRs, testing, optimization, etc.
Change-Id: I8c36b2b9e0a2d27392040ecda0f51b636456de25
2010-04-23 01:58:52 +00:00
|
|
|
1, & pointerId, & pointerCoords,
|
|
|
|
device->trackball.precalculated.xPrecision,
|
|
|
|
device->trackball.precalculated.yPrecision,
|
|
|
|
device->trackball.current.downTime);
|
|
|
|
}
|
|
|
|
|
|
|
|
void InputReader::onConfigurationChanged(nsecs_t when) {
|
|
|
|
// Reset global meta state because it depends on the list of all configured devices.
|
|
|
|
resetGlobalMetaState();
|
|
|
|
|
|
|
|
// Reset virtual keys, just in case.
|
2010-06-15 08:31:58 +00:00
|
|
|
updateExportedVirtualKeyState();
|
Native input dispatch rewrite work in progress.
The old dispatch mechanism has been left in place and continues to
be used by default for now. To enable native input dispatch,
edit the ENABLE_NATIVE_DISPATCH constant in WindowManagerPolicy.
Includes part of the new input event NDK API. Some details TBD.
To wire up input dispatch, as the ViewRoot adds a window to the
window session it receives an InputChannel object as an output
argument. The InputChannel encapsulates the file descriptors for a
shared memory region and two pipe end-points. The ViewRoot then
provides the InputChannel to the InputQueue. Behind the
scenes, InputQueue simply attaches handlers to the native PollLoop object
that underlies the MessageQueue. This way MessageQueue doesn't need
to know anything about input dispatch per-se, it just exposes (in native
code) a PollLoop that other components can use to monitor file descriptor
state changes.
There can be zero or more targets for any given input event. Each
input target is specified by its input channel and some parameters
including flags, an X/Y coordinate offset, and the dispatch timeout.
An input target can request either synchronous dispatch (for foreground apps)
or asynchronous dispatch (fire-and-forget for wallpapers and "outside"
targets). Currently, finding the appropriate input targets for an event
requires a call back into the WindowManagerServer from native code.
In the future this will be refactored to avoid most of these callbacks
except as required to handle pending focus transitions.
End-to-end event dispatch mostly works!
To do: event injection, rate limiting, ANRs, testing, optimization, etc.
Change-Id: I8c36b2b9e0a2d27392040ecda0f51b636456de25
2010-04-23 01:58:52 +00:00
|
|
|
|
2010-06-15 08:31:58 +00:00
|
|
|
// Update input configuration.
|
|
|
|
updateExportedInputConfiguration();
|
Native input dispatch rewrite work in progress.
The old dispatch mechanism has been left in place and continues to
be used by default for now. To enable native input dispatch,
edit the ENABLE_NATIVE_DISPATCH constant in WindowManagerPolicy.
Includes part of the new input event NDK API. Some details TBD.
To wire up input dispatch, as the ViewRoot adds a window to the
window session it receives an InputChannel object as an output
argument. The InputChannel encapsulates the file descriptors for a
shared memory region and two pipe end-points. The ViewRoot then
provides the InputChannel to the InputQueue. Behind the
scenes, InputQueue simply attaches handlers to the native PollLoop object
that underlies the MessageQueue. This way MessageQueue doesn't need
to know anything about input dispatch per-se, it just exposes (in native
code) a PollLoop that other components can use to monitor file descriptor
state changes.
There can be zero or more targets for any given input event. Each
input target is specified by its input channel and some parameters
including flags, an X/Y coordinate offset, and the dispatch timeout.
An input target can request either synchronous dispatch (for foreground apps)
or asynchronous dispatch (fire-and-forget for wallpapers and "outside"
targets). Currently, finding the appropriate input targets for an event
requires a call back into the WindowManagerServer from native code.
In the future this will be refactored to avoid most of these callbacks
except as required to handle pending focus transitions.
End-to-end event dispatch mostly works!
To do: event injection, rate limiting, ANRs, testing, optimization, etc.
Change-Id: I8c36b2b9e0a2d27392040ecda0f51b636456de25
2010-04-23 01:58:52 +00:00
|
|
|
|
2010-06-15 08:31:58 +00:00
|
|
|
// Enqueue configuration changed.
|
|
|
|
mDispatcher->notifyConfigurationChanged(when);
|
Native input dispatch rewrite work in progress.
The old dispatch mechanism has been left in place and continues to
be used by default for now. To enable native input dispatch,
edit the ENABLE_NATIVE_DISPATCH constant in WindowManagerPolicy.
Includes part of the new input event NDK API. Some details TBD.
To wire up input dispatch, as the ViewRoot adds a window to the
window session it receives an InputChannel object as an output
argument. The InputChannel encapsulates the file descriptors for a
shared memory region and two pipe end-points. The ViewRoot then
provides the InputChannel to the InputQueue. Behind the
scenes, InputQueue simply attaches handlers to the native PollLoop object
that underlies the MessageQueue. This way MessageQueue doesn't need
to know anything about input dispatch per-se, it just exposes (in native
code) a PollLoop that other components can use to monitor file descriptor
state changes.
There can be zero or more targets for any given input event. Each
input target is specified by its input channel and some parameters
including flags, an X/Y coordinate offset, and the dispatch timeout.
An input target can request either synchronous dispatch (for foreground apps)
or asynchronous dispatch (fire-and-forget for wallpapers and "outside"
targets). Currently, finding the appropriate input targets for an event
requires a call back into the WindowManagerServer from native code.
In the future this will be refactored to avoid most of these callbacks
except as required to handle pending focus transitions.
End-to-end event dispatch mostly works!
To do: event injection, rate limiting, ANRs, testing, optimization, etc.
Change-Id: I8c36b2b9e0a2d27392040ecda0f51b636456de25
2010-04-23 01:58:52 +00:00
|
|
|
}
|
|
|
|
|
|
|
|
bool InputReader::applyStandardInputDispatchPolicyActions(nsecs_t when,
|
|
|
|
int32_t policyActions, uint32_t* policyFlags) {
|
2010-06-15 08:31:58 +00:00
|
|
|
if (policyActions & InputReaderPolicyInterface::ACTION_APP_SWITCH_COMING) {
|
Native input dispatch rewrite work in progress.
The old dispatch mechanism has been left in place and continues to
be used by default for now. To enable native input dispatch,
edit the ENABLE_NATIVE_DISPATCH constant in WindowManagerPolicy.
Includes part of the new input event NDK API. Some details TBD.
To wire up input dispatch, as the ViewRoot adds a window to the
window session it receives an InputChannel object as an output
argument. The InputChannel encapsulates the file descriptors for a
shared memory region and two pipe end-points. The ViewRoot then
provides the InputChannel to the InputQueue. Behind the
scenes, InputQueue simply attaches handlers to the native PollLoop object
that underlies the MessageQueue. This way MessageQueue doesn't need
to know anything about input dispatch per-se, it just exposes (in native
code) a PollLoop that other components can use to monitor file descriptor
state changes.
There can be zero or more targets for any given input event. Each
input target is specified by its input channel and some parameters
including flags, an X/Y coordinate offset, and the dispatch timeout.
An input target can request either synchronous dispatch (for foreground apps)
or asynchronous dispatch (fire-and-forget for wallpapers and "outside"
targets). Currently, finding the appropriate input targets for an event
requires a call back into the WindowManagerServer from native code.
In the future this will be refactored to avoid most of these callbacks
except as required to handle pending focus transitions.
End-to-end event dispatch mostly works!
To do: event injection, rate limiting, ANRs, testing, optimization, etc.
Change-Id: I8c36b2b9e0a2d27392040ecda0f51b636456de25
2010-04-23 01:58:52 +00:00
|
|
|
mDispatcher->notifyAppSwitchComing(when);
|
|
|
|
}
|
|
|
|
|
2010-06-15 08:31:58 +00:00
|
|
|
if (policyActions & InputReaderPolicyInterface::ACTION_WOKE_HERE) {
|
Native input dispatch rewrite work in progress.
The old dispatch mechanism has been left in place and continues to
be used by default for now. To enable native input dispatch,
edit the ENABLE_NATIVE_DISPATCH constant in WindowManagerPolicy.
Includes part of the new input event NDK API. Some details TBD.
To wire up input dispatch, as the ViewRoot adds a window to the
window session it receives an InputChannel object as an output
argument. The InputChannel encapsulates the file descriptors for a
shared memory region and two pipe end-points. The ViewRoot then
provides the InputChannel to the InputQueue. Behind the
scenes, InputQueue simply attaches handlers to the native PollLoop object
that underlies the MessageQueue. This way MessageQueue doesn't need
to know anything about input dispatch per-se, it just exposes (in native
code) a PollLoop that other components can use to monitor file descriptor
state changes.
There can be zero or more targets for any given input event. Each
input target is specified by its input channel and some parameters
including flags, an X/Y coordinate offset, and the dispatch timeout.
An input target can request either synchronous dispatch (for foreground apps)
or asynchronous dispatch (fire-and-forget for wallpapers and "outside"
targets). Currently, finding the appropriate input targets for an event
requires a call back into the WindowManagerServer from native code.
In the future this will be refactored to avoid most of these callbacks
except as required to handle pending focus transitions.
End-to-end event dispatch mostly works!
To do: event injection, rate limiting, ANRs, testing, optimization, etc.
Change-Id: I8c36b2b9e0a2d27392040ecda0f51b636456de25
2010-04-23 01:58:52 +00:00
|
|
|
*policyFlags |= POLICY_FLAG_WOKE_HERE;
|
|
|
|
}
|
|
|
|
|
2010-06-15 08:31:58 +00:00
|
|
|
if (policyActions & InputReaderPolicyInterface::ACTION_BRIGHT_HERE) {
|
Native input dispatch rewrite work in progress.
The old dispatch mechanism has been left in place and continues to
be used by default for now. To enable native input dispatch,
edit the ENABLE_NATIVE_DISPATCH constant in WindowManagerPolicy.
Includes part of the new input event NDK API. Some details TBD.
To wire up input dispatch, as the ViewRoot adds a window to the
window session it receives an InputChannel object as an output
argument. The InputChannel encapsulates the file descriptors for a
shared memory region and two pipe end-points. The ViewRoot then
provides the InputChannel to the InputQueue. Behind the
scenes, InputQueue simply attaches handlers to the native PollLoop object
that underlies the MessageQueue. This way MessageQueue doesn't need
to know anything about input dispatch per-se, it just exposes (in native
code) a PollLoop that other components can use to monitor file descriptor
state changes.
There can be zero or more targets for any given input event. Each
input target is specified by its input channel and some parameters
including flags, an X/Y coordinate offset, and the dispatch timeout.
An input target can request either synchronous dispatch (for foreground apps)
or asynchronous dispatch (fire-and-forget for wallpapers and "outside"
targets). Currently, finding the appropriate input targets for an event
requires a call back into the WindowManagerServer from native code.
In the future this will be refactored to avoid most of these callbacks
except as required to handle pending focus transitions.
End-to-end event dispatch mostly works!
To do: event injection, rate limiting, ANRs, testing, optimization, etc.
Change-Id: I8c36b2b9e0a2d27392040ecda0f51b636456de25
2010-04-23 01:58:52 +00:00
|
|
|
*policyFlags |= POLICY_FLAG_BRIGHT_HERE;
|
|
|
|
}
|
|
|
|
|
2010-06-15 08:31:58 +00:00
|
|
|
return policyActions & InputReaderPolicyInterface::ACTION_DISPATCH;
|
Native input dispatch rewrite work in progress.
The old dispatch mechanism has been left in place and continues to
be used by default for now. To enable native input dispatch,
edit the ENABLE_NATIVE_DISPATCH constant in WindowManagerPolicy.
Includes part of the new input event NDK API. Some details TBD.
To wire up input dispatch, as the ViewRoot adds a window to the
window session it receives an InputChannel object as an output
argument. The InputChannel encapsulates the file descriptors for a
shared memory region and two pipe end-points. The ViewRoot then
provides the InputChannel to the InputQueue. Behind the
scenes, InputQueue simply attaches handlers to the native PollLoop object
that underlies the MessageQueue. This way MessageQueue doesn't need
to know anything about input dispatch per-se, it just exposes (in native
code) a PollLoop that other components can use to monitor file descriptor
state changes.
There can be zero or more targets for any given input event. Each
input target is specified by its input channel and some parameters
including flags, an X/Y coordinate offset, and the dispatch timeout.
An input target can request either synchronous dispatch (for foreground apps)
or asynchronous dispatch (fire-and-forget for wallpapers and "outside"
targets). Currently, finding the appropriate input targets for an event
requires a call back into the WindowManagerServer from native code.
In the future this will be refactored to avoid most of these callbacks
except as required to handle pending focus transitions.
End-to-end event dispatch mostly works!
To do: event injection, rate limiting, ANRs, testing, optimization, etc.
Change-Id: I8c36b2b9e0a2d27392040ecda0f51b636456de25
2010-04-23 01:58:52 +00:00
|
|
|
}
|
|
|
|
|
|
|
|
void InputReader::resetDisplayProperties() {
|
|
|
|
mDisplayWidth = mDisplayHeight = -1;
|
|
|
|
mDisplayOrientation = -1;
|
|
|
|
}
|
|
|
|
|
|
|
|
bool InputReader::refreshDisplayProperties() {
|
|
|
|
int32_t newWidth, newHeight, newOrientation;
|
|
|
|
if (mPolicy->getDisplayInfo(0, & newWidth, & newHeight, & newOrientation)) {
|
|
|
|
if (newWidth != mDisplayWidth || newHeight != mDisplayHeight) {
|
|
|
|
LOGD("Display size changed from %dx%d to %dx%d, updating device configuration",
|
|
|
|
mDisplayWidth, mDisplayHeight, newWidth, newHeight);
|
|
|
|
|
|
|
|
mDisplayWidth = newWidth;
|
|
|
|
mDisplayHeight = newHeight;
|
|
|
|
|
|
|
|
for (size_t i = 0; i < mDevices.size(); i++) {
|
|
|
|
configureDeviceForCurrentDisplaySize(mDevices.valueAt(i));
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
2010-06-29 23:52:21 +00:00
|
|
|
if (newOrientation != mDisplayOrientation) {
|
|
|
|
LOGD("Display orientation changed to %d", mDisplayOrientation);
|
|
|
|
|
|
|
|
mDisplayOrientation = newOrientation;
|
|
|
|
}
|
Native input dispatch rewrite work in progress.
The old dispatch mechanism has been left in place and continues to
be used by default for now. To enable native input dispatch,
edit the ENABLE_NATIVE_DISPATCH constant in WindowManagerPolicy.
Includes part of the new input event NDK API. Some details TBD.
To wire up input dispatch, as the ViewRoot adds a window to the
window session it receives an InputChannel object as an output
argument. The InputChannel encapsulates the file descriptors for a
shared memory region and two pipe end-points. The ViewRoot then
provides the InputChannel to the InputQueue. Behind the
scenes, InputQueue simply attaches handlers to the native PollLoop object
that underlies the MessageQueue. This way MessageQueue doesn't need
to know anything about input dispatch per-se, it just exposes (in native
code) a PollLoop that other components can use to monitor file descriptor
state changes.
There can be zero or more targets for any given input event. Each
input target is specified by its input channel and some parameters
including flags, an X/Y coordinate offset, and the dispatch timeout.
An input target can request either synchronous dispatch (for foreground apps)
or asynchronous dispatch (fire-and-forget for wallpapers and "outside"
targets). Currently, finding the appropriate input targets for an event
requires a call back into the WindowManagerServer from native code.
In the future this will be refactored to avoid most of these callbacks
except as required to handle pending focus transitions.
End-to-end event dispatch mostly works!
To do: event injection, rate limiting, ANRs, testing, optimization, etc.
Change-Id: I8c36b2b9e0a2d27392040ecda0f51b636456de25
2010-04-23 01:58:52 +00:00
|
|
|
return true;
|
|
|
|
} else {
|
|
|
|
resetDisplayProperties();
|
|
|
|
return false;
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
|
|
|
InputDevice* InputReader::getDevice(int32_t deviceId) {
|
|
|
|
ssize_t index = mDevices.indexOfKey(deviceId);
|
|
|
|
return index >= 0 ? mDevices.valueAt((size_t) index) : NULL;
|
|
|
|
}
|
|
|
|
|
|
|
|
InputDevice* InputReader::getNonIgnoredDevice(int32_t deviceId) {
|
|
|
|
InputDevice* device = getDevice(deviceId);
|
|
|
|
return device && ! device->ignored ? device : NULL;
|
|
|
|
}
|
|
|
|
|
|
|
|
void InputReader::addDevice(nsecs_t when, int32_t deviceId) {
|
|
|
|
uint32_t classes = mEventHub->getDeviceClasses(deviceId);
|
|
|
|
String8 name = mEventHub->getDeviceName(deviceId);
|
|
|
|
InputDevice* device = new InputDevice(deviceId, classes, name);
|
|
|
|
|
|
|
|
if (classes != 0) {
|
|
|
|
LOGI("Device added: id=0x%x, name=%s, classes=%02x", device->id,
|
|
|
|
device->name.string(), device->classes);
|
|
|
|
|
|
|
|
configureDevice(device);
|
|
|
|
} else {
|
|
|
|
LOGI("Device added: id=0x%x, name=%s (ignored non-input device)", device->id,
|
|
|
|
device->name.string());
|
|
|
|
|
|
|
|
device->ignored = true;
|
|
|
|
}
|
|
|
|
|
|
|
|
device->reset();
|
|
|
|
|
|
|
|
mDevices.add(deviceId, device);
|
|
|
|
|
|
|
|
if (! device->ignored) {
|
|
|
|
onConfigurationChanged(when);
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
|
|
|
void InputReader::removeDevice(nsecs_t when, InputDevice* device) {
|
|
|
|
mDevices.removeItem(device->id);
|
|
|
|
|
|
|
|
if (! device->ignored) {
|
|
|
|
LOGI("Device removed: id=0x%x, name=%s, classes=%02x", device->id,
|
|
|
|
device->name.string(), device->classes);
|
|
|
|
|
|
|
|
onConfigurationChanged(when);
|
|
|
|
} else {
|
|
|
|
LOGI("Device removed: id=0x%x, name=%s (ignored non-input device)", device->id,
|
|
|
|
device->name.string());
|
|
|
|
}
|
|
|
|
|
|
|
|
delete device;
|
|
|
|
}
|
|
|
|
|
|
|
|
void InputReader::configureDevice(InputDevice* device) {
|
|
|
|
if (device->isMultiTouchScreen()) {
|
|
|
|
configureAbsoluteAxisInfo(device, ABS_MT_POSITION_X, "X",
|
|
|
|
& device->touchScreen.parameters.xAxis);
|
|
|
|
configureAbsoluteAxisInfo(device, ABS_MT_POSITION_Y, "Y",
|
|
|
|
& device->touchScreen.parameters.yAxis);
|
|
|
|
configureAbsoluteAxisInfo(device, ABS_MT_TOUCH_MAJOR, "Pressure",
|
|
|
|
& device->touchScreen.parameters.pressureAxis);
|
|
|
|
configureAbsoluteAxisInfo(device, ABS_MT_WIDTH_MAJOR, "Size",
|
|
|
|
& device->touchScreen.parameters.sizeAxis);
|
2010-07-15 01:48:53 +00:00
|
|
|
configureAbsoluteAxisInfo(device, ABS_MT_ORIENTATION, "Orientation",
|
|
|
|
& device->touchScreen.parameters.orientationAxis);
|
Native input dispatch rewrite work in progress.
The old dispatch mechanism has been left in place and continues to
be used by default for now. To enable native input dispatch,
edit the ENABLE_NATIVE_DISPATCH constant in WindowManagerPolicy.
Includes part of the new input event NDK API. Some details TBD.
To wire up input dispatch, as the ViewRoot adds a window to the
window session it receives an InputChannel object as an output
argument. The InputChannel encapsulates the file descriptors for a
shared memory region and two pipe end-points. The ViewRoot then
provides the InputChannel to the InputQueue. Behind the
scenes, InputQueue simply attaches handlers to the native PollLoop object
that underlies the MessageQueue. This way MessageQueue doesn't need
to know anything about input dispatch per-se, it just exposes (in native
code) a PollLoop that other components can use to monitor file descriptor
state changes.
There can be zero or more targets for any given input event. Each
input target is specified by its input channel and some parameters
including flags, an X/Y coordinate offset, and the dispatch timeout.
An input target can request either synchronous dispatch (for foreground apps)
or asynchronous dispatch (fire-and-forget for wallpapers and "outside"
targets). Currently, finding the appropriate input targets for an event
requires a call back into the WindowManagerServer from native code.
In the future this will be refactored to avoid most of these callbacks
except as required to handle pending focus transitions.
End-to-end event dispatch mostly works!
To do: event injection, rate limiting, ANRs, testing, optimization, etc.
Change-Id: I8c36b2b9e0a2d27392040ecda0f51b636456de25
2010-04-23 01:58:52 +00:00
|
|
|
} else if (device->isSingleTouchScreen()) {
|
|
|
|
configureAbsoluteAxisInfo(device, ABS_X, "X",
|
|
|
|
& device->touchScreen.parameters.xAxis);
|
|
|
|
configureAbsoluteAxisInfo(device, ABS_Y, "Y",
|
|
|
|
& device->touchScreen.parameters.yAxis);
|
|
|
|
configureAbsoluteAxisInfo(device, ABS_PRESSURE, "Pressure",
|
|
|
|
& device->touchScreen.parameters.pressureAxis);
|
|
|
|
configureAbsoluteAxisInfo(device, ABS_TOOL_WIDTH, "Size",
|
|
|
|
& device->touchScreen.parameters.sizeAxis);
|
2010-07-15 01:48:53 +00:00
|
|
|
device->touchScreen.parameters.orientationAxis.valid = false;
|
Native input dispatch rewrite work in progress.
The old dispatch mechanism has been left in place and continues to
be used by default for now. To enable native input dispatch,
edit the ENABLE_NATIVE_DISPATCH constant in WindowManagerPolicy.
Includes part of the new input event NDK API. Some details TBD.
To wire up input dispatch, as the ViewRoot adds a window to the
window session it receives an InputChannel object as an output
argument. The InputChannel encapsulates the file descriptors for a
shared memory region and two pipe end-points. The ViewRoot then
provides the InputChannel to the InputQueue. Behind the
scenes, InputQueue simply attaches handlers to the native PollLoop object
that underlies the MessageQueue. This way MessageQueue doesn't need
to know anything about input dispatch per-se, it just exposes (in native
code) a PollLoop that other components can use to monitor file descriptor
state changes.
There can be zero or more targets for any given input event. Each
input target is specified by its input channel and some parameters
including flags, an X/Y coordinate offset, and the dispatch timeout.
An input target can request either synchronous dispatch (for foreground apps)
or asynchronous dispatch (fire-and-forget for wallpapers and "outside"
targets). Currently, finding the appropriate input targets for an event
requires a call back into the WindowManagerServer from native code.
In the future this will be refactored to avoid most of these callbacks
except as required to handle pending focus transitions.
End-to-end event dispatch mostly works!
To do: event injection, rate limiting, ANRs, testing, optimization, etc.
Change-Id: I8c36b2b9e0a2d27392040ecda0f51b636456de25
2010-04-23 01:58:52 +00:00
|
|
|
}
|
|
|
|
|
|
|
|
if (device->isTouchScreen()) {
|
|
|
|
device->touchScreen.parameters.useBadTouchFilter =
|
|
|
|
mPolicy->filterTouchEvents();
|
|
|
|
device->touchScreen.parameters.useAveragingTouchFilter =
|
|
|
|
mPolicy->filterTouchEvents();
|
|
|
|
device->touchScreen.parameters.useJumpyTouchFilter =
|
|
|
|
mPolicy->filterJumpyTouchEvents();
|
|
|
|
|
2010-06-29 23:52:21 +00:00
|
|
|
if (device->touchScreen.parameters.pressureAxis.valid) {
|
|
|
|
device->touchScreen.precalculated.pressureOrigin =
|
|
|
|
device->touchScreen.parameters.pressureAxis.minValue;
|
|
|
|
device->touchScreen.precalculated.pressureScale =
|
|
|
|
1.0f / device->touchScreen.parameters.pressureAxis.range;
|
|
|
|
} else {
|
|
|
|
device->touchScreen.precalculated.pressureOrigin = 0;
|
|
|
|
device->touchScreen.precalculated.pressureScale = 1.0f;
|
|
|
|
}
|
|
|
|
|
|
|
|
if (device->touchScreen.parameters.sizeAxis.valid) {
|
|
|
|
device->touchScreen.precalculated.sizeOrigin =
|
|
|
|
device->touchScreen.parameters.sizeAxis.minValue;
|
|
|
|
device->touchScreen.precalculated.sizeScale =
|
|
|
|
1.0f / device->touchScreen.parameters.sizeAxis.range;
|
|
|
|
} else {
|
|
|
|
device->touchScreen.precalculated.sizeOrigin = 0;
|
|
|
|
device->touchScreen.precalculated.sizeScale = 1.0f;
|
|
|
|
}
|
2010-07-15 01:48:53 +00:00
|
|
|
|
|
|
|
if (device->touchScreen.parameters.orientationAxis.valid
|
|
|
|
&& device->touchScreen.parameters.orientationAxis.maxValue > 0) {
|
|
|
|
device->touchScreen.precalculated.orientationScale =
|
|
|
|
M_PI_4 / device->touchScreen.parameters.orientationAxis.maxValue;
|
|
|
|
} else {
|
|
|
|
device->touchScreen.precalculated.orientationScale = 0.0f;
|
|
|
|
}
|
Native input dispatch rewrite work in progress.
The old dispatch mechanism has been left in place and continues to
be used by default for now. To enable native input dispatch,
edit the ENABLE_NATIVE_DISPATCH constant in WindowManagerPolicy.
Includes part of the new input event NDK API. Some details TBD.
To wire up input dispatch, as the ViewRoot adds a window to the
window session it receives an InputChannel object as an output
argument. The InputChannel encapsulates the file descriptors for a
shared memory region and two pipe end-points. The ViewRoot then
provides the InputChannel to the InputQueue. Behind the
scenes, InputQueue simply attaches handlers to the native PollLoop object
that underlies the MessageQueue. This way MessageQueue doesn't need
to know anything about input dispatch per-se, it just exposes (in native
code) a PollLoop that other components can use to monitor file descriptor
state changes.
There can be zero or more targets for any given input event. Each
input target is specified by its input channel and some parameters
including flags, an X/Y coordinate offset, and the dispatch timeout.
An input target can request either synchronous dispatch (for foreground apps)
or asynchronous dispatch (fire-and-forget for wallpapers and "outside"
targets). Currently, finding the appropriate input targets for an event
requires a call back into the WindowManagerServer from native code.
In the future this will be refactored to avoid most of these callbacks
except as required to handle pending focus transitions.
End-to-end event dispatch mostly works!
To do: event injection, rate limiting, ANRs, testing, optimization, etc.
Change-Id: I8c36b2b9e0a2d27392040ecda0f51b636456de25
2010-04-23 01:58:52 +00:00
|
|
|
}
|
|
|
|
|
|
|
|
if (device->isTrackball()) {
|
|
|
|
device->trackball.precalculated.xPrecision = TRACKBALL_MOVEMENT_THRESHOLD;
|
|
|
|
device->trackball.precalculated.yPrecision = TRACKBALL_MOVEMENT_THRESHOLD;
|
|
|
|
device->trackball.precalculated.xScale = 1.0f / TRACKBALL_MOVEMENT_THRESHOLD;
|
|
|
|
device->trackball.precalculated.yScale = 1.0f / TRACKBALL_MOVEMENT_THRESHOLD;
|
|
|
|
}
|
|
|
|
|
|
|
|
configureDeviceForCurrentDisplaySize(device);
|
|
|
|
}
|
|
|
|
|
|
|
|
void InputReader::configureDeviceForCurrentDisplaySize(InputDevice* device) {
|
|
|
|
if (device->isTouchScreen()) {
|
2010-06-29 23:52:21 +00:00
|
|
|
if (device->touchScreen.parameters.xAxis.valid
|
|
|
|
&& device->touchScreen.parameters.yAxis.valid) {
|
|
|
|
device->touchScreen.precalculated.xOrigin =
|
|
|
|
device->touchScreen.parameters.xAxis.minValue;
|
|
|
|
device->touchScreen.precalculated.yOrigin =
|
|
|
|
device->touchScreen.parameters.yAxis.minValue;
|
|
|
|
|
|
|
|
if (mDisplayWidth < 0) {
|
|
|
|
LOGD("Skipping part of touch screen configuration since display size is unknown.");
|
|
|
|
|
|
|
|
device->touchScreen.precalculated.xScale = 1.0f;
|
|
|
|
device->touchScreen.precalculated.yScale = 1.0f;
|
|
|
|
} else {
|
|
|
|
LOGI("Device configured: id=0x%x, name=%s (display size was changed)", device->id,
|
|
|
|
device->name.string());
|
|
|
|
|
|
|
|
device->touchScreen.precalculated.xScale =
|
|
|
|
float(mDisplayWidth) / device->touchScreen.parameters.xAxis.range;
|
|
|
|
device->touchScreen.precalculated.yScale =
|
|
|
|
float(mDisplayHeight) / device->touchScreen.parameters.yAxis.range;
|
|
|
|
|
|
|
|
configureVirtualKeys(device);
|
|
|
|
}
|
Native input dispatch rewrite work in progress.
The old dispatch mechanism has been left in place and continues to
be used by default for now. To enable native input dispatch,
edit the ENABLE_NATIVE_DISPATCH constant in WindowManagerPolicy.
Includes part of the new input event NDK API. Some details TBD.
To wire up input dispatch, as the ViewRoot adds a window to the
window session it receives an InputChannel object as an output
argument. The InputChannel encapsulates the file descriptors for a
shared memory region and two pipe end-points. The ViewRoot then
provides the InputChannel to the InputQueue. Behind the
scenes, InputQueue simply attaches handlers to the native PollLoop object
that underlies the MessageQueue. This way MessageQueue doesn't need
to know anything about input dispatch per-se, it just exposes (in native
code) a PollLoop that other components can use to monitor file descriptor
state changes.
There can be zero or more targets for any given input event. Each
input target is specified by its input channel and some parameters
including flags, an X/Y coordinate offset, and the dispatch timeout.
An input target can request either synchronous dispatch (for foreground apps)
or asynchronous dispatch (fire-and-forget for wallpapers and "outside"
targets). Currently, finding the appropriate input targets for an event
requires a call back into the WindowManagerServer from native code.
In the future this will be refactored to avoid most of these callbacks
except as required to handle pending focus transitions.
End-to-end event dispatch mostly works!
To do: event injection, rate limiting, ANRs, testing, optimization, etc.
Change-Id: I8c36b2b9e0a2d27392040ecda0f51b636456de25
2010-04-23 01:58:52 +00:00
|
|
|
} else {
|
2010-06-29 23:52:21 +00:00
|
|
|
device->touchScreen.precalculated.xOrigin = 0;
|
|
|
|
device->touchScreen.precalculated.xScale = 1.0f;
|
|
|
|
device->touchScreen.precalculated.yOrigin = 0;
|
|
|
|
device->touchScreen.precalculated.yScale = 1.0f;
|
Native input dispatch rewrite work in progress.
The old dispatch mechanism has been left in place and continues to
be used by default for now. To enable native input dispatch,
edit the ENABLE_NATIVE_DISPATCH constant in WindowManagerPolicy.
Includes part of the new input event NDK API. Some details TBD.
To wire up input dispatch, as the ViewRoot adds a window to the
window session it receives an InputChannel object as an output
argument. The InputChannel encapsulates the file descriptors for a
shared memory region and two pipe end-points. The ViewRoot then
provides the InputChannel to the InputQueue. Behind the
scenes, InputQueue simply attaches handlers to the native PollLoop object
that underlies the MessageQueue. This way MessageQueue doesn't need
to know anything about input dispatch per-se, it just exposes (in native
code) a PollLoop that other components can use to monitor file descriptor
state changes.
There can be zero or more targets for any given input event. Each
input target is specified by its input channel and some parameters
including flags, an X/Y coordinate offset, and the dispatch timeout.
An input target can request either synchronous dispatch (for foreground apps)
or asynchronous dispatch (fire-and-forget for wallpapers and "outside"
targets). Currently, finding the appropriate input targets for an event
requires a call back into the WindowManagerServer from native code.
In the future this will be refactored to avoid most of these callbacks
except as required to handle pending focus transitions.
End-to-end event dispatch mostly works!
To do: event injection, rate limiting, ANRs, testing, optimization, etc.
Change-Id: I8c36b2b9e0a2d27392040ecda0f51b636456de25
2010-04-23 01:58:52 +00:00
|
|
|
}
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
|
|
|
void InputReader::configureVirtualKeys(InputDevice* device) {
|
2010-06-29 23:52:21 +00:00
|
|
|
assert(device->touchScreen.parameters.xAxis.valid
|
|
|
|
&& device->touchScreen.parameters.yAxis.valid);
|
|
|
|
|
Native input dispatch rewrite work in progress.
The old dispatch mechanism has been left in place and continues to
be used by default for now. To enable native input dispatch,
edit the ENABLE_NATIVE_DISPATCH constant in WindowManagerPolicy.
Includes part of the new input event NDK API. Some details TBD.
To wire up input dispatch, as the ViewRoot adds a window to the
window session it receives an InputChannel object as an output
argument. The InputChannel encapsulates the file descriptors for a
shared memory region and two pipe end-points. The ViewRoot then
provides the InputChannel to the InputQueue. Behind the
scenes, InputQueue simply attaches handlers to the native PollLoop object
that underlies the MessageQueue. This way MessageQueue doesn't need
to know anything about input dispatch per-se, it just exposes (in native
code) a PollLoop that other components can use to monitor file descriptor
state changes.
There can be zero or more targets for any given input event. Each
input target is specified by its input channel and some parameters
including flags, an X/Y coordinate offset, and the dispatch timeout.
An input target can request either synchronous dispatch (for foreground apps)
or asynchronous dispatch (fire-and-forget for wallpapers and "outside"
targets). Currently, finding the appropriate input targets for an event
requires a call back into the WindowManagerServer from native code.
In the future this will be refactored to avoid most of these callbacks
except as required to handle pending focus transitions.
End-to-end event dispatch mostly works!
To do: event injection, rate limiting, ANRs, testing, optimization, etc.
Change-Id: I8c36b2b9e0a2d27392040ecda0f51b636456de25
2010-04-23 01:58:52 +00:00
|
|
|
device->touchScreen.virtualKeys.clear();
|
|
|
|
|
2010-06-15 08:31:58 +00:00
|
|
|
Vector<InputReaderPolicyInterface::VirtualKeyDefinition> virtualKeyDefinitions;
|
Native input dispatch rewrite work in progress.
The old dispatch mechanism has been left in place and continues to
be used by default for now. To enable native input dispatch,
edit the ENABLE_NATIVE_DISPATCH constant in WindowManagerPolicy.
Includes part of the new input event NDK API. Some details TBD.
To wire up input dispatch, as the ViewRoot adds a window to the
window session it receives an InputChannel object as an output
argument. The InputChannel encapsulates the file descriptors for a
shared memory region and two pipe end-points. The ViewRoot then
provides the InputChannel to the InputQueue. Behind the
scenes, InputQueue simply attaches handlers to the native PollLoop object
that underlies the MessageQueue. This way MessageQueue doesn't need
to know anything about input dispatch per-se, it just exposes (in native
code) a PollLoop that other components can use to monitor file descriptor
state changes.
There can be zero or more targets for any given input event. Each
input target is specified by its input channel and some parameters
including flags, an X/Y coordinate offset, and the dispatch timeout.
An input target can request either synchronous dispatch (for foreground apps)
or asynchronous dispatch (fire-and-forget for wallpapers and "outside"
targets). Currently, finding the appropriate input targets for an event
requires a call back into the WindowManagerServer from native code.
In the future this will be refactored to avoid most of these callbacks
except as required to handle pending focus transitions.
End-to-end event dispatch mostly works!
To do: event injection, rate limiting, ANRs, testing, optimization, etc.
Change-Id: I8c36b2b9e0a2d27392040ecda0f51b636456de25
2010-04-23 01:58:52 +00:00
|
|
|
mPolicy->getVirtualKeyDefinitions(device->name, virtualKeyDefinitions);
|
|
|
|
if (virtualKeyDefinitions.size() == 0) {
|
|
|
|
return;
|
|
|
|
}
|
|
|
|
|
|
|
|
device->touchScreen.virtualKeys.setCapacity(virtualKeyDefinitions.size());
|
|
|
|
|
|
|
|
int32_t touchScreenLeft = device->touchScreen.parameters.xAxis.minValue;
|
|
|
|
int32_t touchScreenTop = device->touchScreen.parameters.yAxis.minValue;
|
|
|
|
int32_t touchScreenWidth = device->touchScreen.parameters.xAxis.range;
|
|
|
|
int32_t touchScreenHeight = device->touchScreen.parameters.yAxis.range;
|
|
|
|
|
|
|
|
for (size_t i = 0; i < virtualKeyDefinitions.size(); i++) {
|
2010-06-15 08:31:58 +00:00
|
|
|
const InputReaderPolicyInterface::VirtualKeyDefinition& virtualKeyDefinition =
|
Native input dispatch rewrite work in progress.
The old dispatch mechanism has been left in place and continues to
be used by default for now. To enable native input dispatch,
edit the ENABLE_NATIVE_DISPATCH constant in WindowManagerPolicy.
Includes part of the new input event NDK API. Some details TBD.
To wire up input dispatch, as the ViewRoot adds a window to the
window session it receives an InputChannel object as an output
argument. The InputChannel encapsulates the file descriptors for a
shared memory region and two pipe end-points. The ViewRoot then
provides the InputChannel to the InputQueue. Behind the
scenes, InputQueue simply attaches handlers to the native PollLoop object
that underlies the MessageQueue. This way MessageQueue doesn't need
to know anything about input dispatch per-se, it just exposes (in native
code) a PollLoop that other components can use to monitor file descriptor
state changes.
There can be zero or more targets for any given input event. Each
input target is specified by its input channel and some parameters
including flags, an X/Y coordinate offset, and the dispatch timeout.
An input target can request either synchronous dispatch (for foreground apps)
or asynchronous dispatch (fire-and-forget for wallpapers and "outside"
targets). Currently, finding the appropriate input targets for an event
requires a call back into the WindowManagerServer from native code.
In the future this will be refactored to avoid most of these callbacks
except as required to handle pending focus transitions.
End-to-end event dispatch mostly works!
To do: event injection, rate limiting, ANRs, testing, optimization, etc.
Change-Id: I8c36b2b9e0a2d27392040ecda0f51b636456de25
2010-04-23 01:58:52 +00:00
|
|
|
virtualKeyDefinitions[i];
|
|
|
|
|
|
|
|
device->touchScreen.virtualKeys.add();
|
|
|
|
InputDevice::VirtualKey& virtualKey =
|
|
|
|
device->touchScreen.virtualKeys.editTop();
|
|
|
|
|
|
|
|
virtualKey.scanCode = virtualKeyDefinition.scanCode;
|
|
|
|
int32_t keyCode;
|
|
|
|
uint32_t flags;
|
|
|
|
if (mEventHub->scancodeToKeycode(device->id, virtualKey.scanCode,
|
|
|
|
& keyCode, & flags)) {
|
2010-07-02 22:37:36 +00:00
|
|
|
LOGW(" VirtualKey %d: could not obtain key code, ignoring", virtualKey.scanCode);
|
Native input dispatch rewrite work in progress.
The old dispatch mechanism has been left in place and continues to
be used by default for now. To enable native input dispatch,
edit the ENABLE_NATIVE_DISPATCH constant in WindowManagerPolicy.
Includes part of the new input event NDK API. Some details TBD.
To wire up input dispatch, as the ViewRoot adds a window to the
window session it receives an InputChannel object as an output
argument. The InputChannel encapsulates the file descriptors for a
shared memory region and two pipe end-points. The ViewRoot then
provides the InputChannel to the InputQueue. Behind the
scenes, InputQueue simply attaches handlers to the native PollLoop object
that underlies the MessageQueue. This way MessageQueue doesn't need
to know anything about input dispatch per-se, it just exposes (in native
code) a PollLoop that other components can use to monitor file descriptor
state changes.
There can be zero or more targets for any given input event. Each
input target is specified by its input channel and some parameters
including flags, an X/Y coordinate offset, and the dispatch timeout.
An input target can request either synchronous dispatch (for foreground apps)
or asynchronous dispatch (fire-and-forget for wallpapers and "outside"
targets). Currently, finding the appropriate input targets for an event
requires a call back into the WindowManagerServer from native code.
In the future this will be refactored to avoid most of these callbacks
except as required to handle pending focus transitions.
End-to-end event dispatch mostly works!
To do: event injection, rate limiting, ANRs, testing, optimization, etc.
Change-Id: I8c36b2b9e0a2d27392040ecda0f51b636456de25
2010-04-23 01:58:52 +00:00
|
|
|
device->touchScreen.virtualKeys.pop(); // drop the key
|
|
|
|
continue;
|
|
|
|
}
|
|
|
|
|
|
|
|
virtualKey.keyCode = keyCode;
|
|
|
|
virtualKey.flags = flags;
|
|
|
|
|
|
|
|
// convert the key definition's display coordinates into touch coordinates for a hit box
|
|
|
|
int32_t halfWidth = virtualKeyDefinition.width / 2;
|
|
|
|
int32_t halfHeight = virtualKeyDefinition.height / 2;
|
|
|
|
|
|
|
|
virtualKey.hitLeft = (virtualKeyDefinition.centerX - halfWidth)
|
|
|
|
* touchScreenWidth / mDisplayWidth + touchScreenLeft;
|
|
|
|
virtualKey.hitRight= (virtualKeyDefinition.centerX + halfWidth)
|
|
|
|
* touchScreenWidth / mDisplayWidth + touchScreenLeft;
|
|
|
|
virtualKey.hitTop = (virtualKeyDefinition.centerY - halfHeight)
|
|
|
|
* touchScreenHeight / mDisplayHeight + touchScreenTop;
|
|
|
|
virtualKey.hitBottom = (virtualKeyDefinition.centerY + halfHeight)
|
|
|
|
* touchScreenHeight / mDisplayHeight + touchScreenTop;
|
|
|
|
|
|
|
|
LOGI(" VirtualKey %d: keyCode=%d hitLeft=%d hitRight=%d hitTop=%d hitBottom=%d",
|
|
|
|
virtualKey.scanCode, virtualKey.keyCode,
|
|
|
|
virtualKey.hitLeft, virtualKey.hitRight, virtualKey.hitTop, virtualKey.hitBottom);
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
|
|
|
void InputReader::configureAbsoluteAxisInfo(InputDevice* device,
|
|
|
|
int axis, const char* name, InputDevice::AbsoluteAxisInfo* out) {
|
|
|
|
if (! mEventHub->getAbsoluteInfo(device->id, axis,
|
|
|
|
& out->minValue, & out->maxValue, & out->flat, &out->fuzz)) {
|
|
|
|
out->range = out->maxValue - out->minValue;
|
|
|
|
if (out->range != 0) {
|
|
|
|
LOGI(" %s: min=%d max=%d flat=%d fuzz=%d",
|
|
|
|
name, out->minValue, out->maxValue, out->flat, out->fuzz);
|
2010-06-29 23:52:21 +00:00
|
|
|
out->valid = true;
|
Native input dispatch rewrite work in progress.
The old dispatch mechanism has been left in place and continues to
be used by default for now. To enable native input dispatch,
edit the ENABLE_NATIVE_DISPATCH constant in WindowManagerPolicy.
Includes part of the new input event NDK API. Some details TBD.
To wire up input dispatch, as the ViewRoot adds a window to the
window session it receives an InputChannel object as an output
argument. The InputChannel encapsulates the file descriptors for a
shared memory region and two pipe end-points. The ViewRoot then
provides the InputChannel to the InputQueue. Behind the
scenes, InputQueue simply attaches handlers to the native PollLoop object
that underlies the MessageQueue. This way MessageQueue doesn't need
to know anything about input dispatch per-se, it just exposes (in native
code) a PollLoop that other components can use to monitor file descriptor
state changes.
There can be zero or more targets for any given input event. Each
input target is specified by its input channel and some parameters
including flags, an X/Y coordinate offset, and the dispatch timeout.
An input target can request either synchronous dispatch (for foreground apps)
or asynchronous dispatch (fire-and-forget for wallpapers and "outside"
targets). Currently, finding the appropriate input targets for an event
requires a call back into the WindowManagerServer from native code.
In the future this will be refactored to avoid most of these callbacks
except as required to handle pending focus transitions.
End-to-end event dispatch mostly works!
To do: event injection, rate limiting, ANRs, testing, optimization, etc.
Change-Id: I8c36b2b9e0a2d27392040ecda0f51b636456de25
2010-04-23 01:58:52 +00:00
|
|
|
return;
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
2010-06-29 23:52:21 +00:00
|
|
|
out->valid = false;
|
Native input dispatch rewrite work in progress.
The old dispatch mechanism has been left in place and continues to
be used by default for now. To enable native input dispatch,
edit the ENABLE_NATIVE_DISPATCH constant in WindowManagerPolicy.
Includes part of the new input event NDK API. Some details TBD.
To wire up input dispatch, as the ViewRoot adds a window to the
window session it receives an InputChannel object as an output
argument. The InputChannel encapsulates the file descriptors for a
shared memory region and two pipe end-points. The ViewRoot then
provides the InputChannel to the InputQueue. Behind the
scenes, InputQueue simply attaches handlers to the native PollLoop object
that underlies the MessageQueue. This way MessageQueue doesn't need
to know anything about input dispatch per-se, it just exposes (in native
code) a PollLoop that other components can use to monitor file descriptor
state changes.
There can be zero or more targets for any given input event. Each
input target is specified by its input channel and some parameters
including flags, an X/Y coordinate offset, and the dispatch timeout.
An input target can request either synchronous dispatch (for foreground apps)
or asynchronous dispatch (fire-and-forget for wallpapers and "outside"
targets). Currently, finding the appropriate input targets for an event
requires a call back into the WindowManagerServer from native code.
In the future this will be refactored to avoid most of these callbacks
except as required to handle pending focus transitions.
End-to-end event dispatch mostly works!
To do: event injection, rate limiting, ANRs, testing, optimization, etc.
Change-Id: I8c36b2b9e0a2d27392040ecda0f51b636456de25
2010-04-23 01:58:52 +00:00
|
|
|
out->minValue = 0;
|
|
|
|
out->maxValue = 0;
|
|
|
|
out->flat = 0;
|
|
|
|
out->fuzz = 0;
|
|
|
|
out->range = 0;
|
2010-06-29 23:52:21 +00:00
|
|
|
LOGI(" %s: unknown axis values, marking as invalid", name);
|
Native input dispatch rewrite work in progress.
The old dispatch mechanism has been left in place and continues to
be used by default for now. To enable native input dispatch,
edit the ENABLE_NATIVE_DISPATCH constant in WindowManagerPolicy.
Includes part of the new input event NDK API. Some details TBD.
To wire up input dispatch, as the ViewRoot adds a window to the
window session it receives an InputChannel object as an output
argument. The InputChannel encapsulates the file descriptors for a
shared memory region and two pipe end-points. The ViewRoot then
provides the InputChannel to the InputQueue. Behind the
scenes, InputQueue simply attaches handlers to the native PollLoop object
that underlies the MessageQueue. This way MessageQueue doesn't need
to know anything about input dispatch per-se, it just exposes (in native
code) a PollLoop that other components can use to monitor file descriptor
state changes.
There can be zero or more targets for any given input event. Each
input target is specified by its input channel and some parameters
including flags, an X/Y coordinate offset, and the dispatch timeout.
An input target can request either synchronous dispatch (for foreground apps)
or asynchronous dispatch (fire-and-forget for wallpapers and "outside"
targets). Currently, finding the appropriate input targets for an event
requires a call back into the WindowManagerServer from native code.
In the future this will be refactored to avoid most of these callbacks
except as required to handle pending focus transitions.
End-to-end event dispatch mostly works!
To do: event injection, rate limiting, ANRs, testing, optimization, etc.
Change-Id: I8c36b2b9e0a2d27392040ecda0f51b636456de25
2010-04-23 01:58:52 +00:00
|
|
|
}
|
|
|
|
|
2010-06-15 08:31:58 +00:00
|
|
|
void InputReader::configureExcludedDevices() {
|
|
|
|
Vector<String8> excludedDeviceNames;
|
|
|
|
mPolicy->getExcludedDeviceNames(excludedDeviceNames);
|
|
|
|
|
|
|
|
for (size_t i = 0; i < excludedDeviceNames.size(); i++) {
|
|
|
|
mEventHub->addExcludedDevice(excludedDeviceNames[i]);
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
Native input dispatch rewrite work in progress.
The old dispatch mechanism has been left in place and continues to
be used by default for now. To enable native input dispatch,
edit the ENABLE_NATIVE_DISPATCH constant in WindowManagerPolicy.
Includes part of the new input event NDK API. Some details TBD.
To wire up input dispatch, as the ViewRoot adds a window to the
window session it receives an InputChannel object as an output
argument. The InputChannel encapsulates the file descriptors for a
shared memory region and two pipe end-points. The ViewRoot then
provides the InputChannel to the InputQueue. Behind the
scenes, InputQueue simply attaches handlers to the native PollLoop object
that underlies the MessageQueue. This way MessageQueue doesn't need
to know anything about input dispatch per-se, it just exposes (in native
code) a PollLoop that other components can use to monitor file descriptor
state changes.
There can be zero or more targets for any given input event. Each
input target is specified by its input channel and some parameters
including flags, an X/Y coordinate offset, and the dispatch timeout.
An input target can request either synchronous dispatch (for foreground apps)
or asynchronous dispatch (fire-and-forget for wallpapers and "outside"
targets). Currently, finding the appropriate input targets for an event
requires a call back into the WindowManagerServer from native code.
In the future this will be refactored to avoid most of these callbacks
except as required to handle pending focus transitions.
End-to-end event dispatch mostly works!
To do: event injection, rate limiting, ANRs, testing, optimization, etc.
Change-Id: I8c36b2b9e0a2d27392040ecda0f51b636456de25
2010-04-23 01:58:52 +00:00
|
|
|
void InputReader::resetGlobalMetaState() {
|
|
|
|
mGlobalMetaState = -1;
|
|
|
|
}
|
|
|
|
|
|
|
|
int32_t InputReader::globalMetaState() {
|
|
|
|
if (mGlobalMetaState == -1) {
|
|
|
|
mGlobalMetaState = 0;
|
|
|
|
for (size_t i = 0; i < mDevices.size(); i++) {
|
|
|
|
InputDevice* device = mDevices.valueAt(i);
|
|
|
|
if (device->isKeyboard()) {
|
|
|
|
mGlobalMetaState |= device->keyboard.current.metaState;
|
|
|
|
}
|
|
|
|
}
|
|
|
|
}
|
|
|
|
return mGlobalMetaState;
|
|
|
|
}
|
|
|
|
|
2010-06-15 08:31:58 +00:00
|
|
|
void InputReader::updateExportedVirtualKeyState() {
|
Native input dispatch rewrite work in progress.
The old dispatch mechanism has been left in place and continues to
be used by default for now. To enable native input dispatch,
edit the ENABLE_NATIVE_DISPATCH constant in WindowManagerPolicy.
Includes part of the new input event NDK API. Some details TBD.
To wire up input dispatch, as the ViewRoot adds a window to the
window session it receives an InputChannel object as an output
argument. The InputChannel encapsulates the file descriptors for a
shared memory region and two pipe end-points. The ViewRoot then
provides the InputChannel to the InputQueue. Behind the
scenes, InputQueue simply attaches handlers to the native PollLoop object
that underlies the MessageQueue. This way MessageQueue doesn't need
to know anything about input dispatch per-se, it just exposes (in native
code) a PollLoop that other components can use to monitor file descriptor
state changes.
There can be zero or more targets for any given input event. Each
input target is specified by its input channel and some parameters
including flags, an X/Y coordinate offset, and the dispatch timeout.
An input target can request either synchronous dispatch (for foreground apps)
or asynchronous dispatch (fire-and-forget for wallpapers and "outside"
targets). Currently, finding the appropriate input targets for an event
requires a call back into the WindowManagerServer from native code.
In the future this will be refactored to avoid most of these callbacks
except as required to handle pending focus transitions.
End-to-end event dispatch mostly works!
To do: event injection, rate limiting, ANRs, testing, optimization, etc.
Change-Id: I8c36b2b9e0a2d27392040ecda0f51b636456de25
2010-04-23 01:58:52 +00:00
|
|
|
int32_t keyCode = -1, scanCode = -1;
|
|
|
|
|
|
|
|
for (size_t i = 0; i < mDevices.size(); i++) {
|
|
|
|
InputDevice* device = mDevices.valueAt(i);
|
|
|
|
if (device->isTouchScreen()) {
|
2010-07-02 22:37:36 +00:00
|
|
|
if (device->touchScreen.currentVirtualKey.status
|
|
|
|
== InputDevice::TouchScreenState::CurrentVirtualKeyState::STATUS_DOWN) {
|
Native input dispatch rewrite work in progress.
The old dispatch mechanism has been left in place and continues to
be used by default for now. To enable native input dispatch,
edit the ENABLE_NATIVE_DISPATCH constant in WindowManagerPolicy.
Includes part of the new input event NDK API. Some details TBD.
To wire up input dispatch, as the ViewRoot adds a window to the
window session it receives an InputChannel object as an output
argument. The InputChannel encapsulates the file descriptors for a
shared memory region and two pipe end-points. The ViewRoot then
provides the InputChannel to the InputQueue. Behind the
scenes, InputQueue simply attaches handlers to the native PollLoop object
that underlies the MessageQueue. This way MessageQueue doesn't need
to know anything about input dispatch per-se, it just exposes (in native
code) a PollLoop that other components can use to monitor file descriptor
state changes.
There can be zero or more targets for any given input event. Each
input target is specified by its input channel and some parameters
including flags, an X/Y coordinate offset, and the dispatch timeout.
An input target can request either synchronous dispatch (for foreground apps)
or asynchronous dispatch (fire-and-forget for wallpapers and "outside"
targets). Currently, finding the appropriate input targets for an event
requires a call back into the WindowManagerServer from native code.
In the future this will be refactored to avoid most of these callbacks
except as required to handle pending focus transitions.
End-to-end event dispatch mostly works!
To do: event injection, rate limiting, ANRs, testing, optimization, etc.
Change-Id: I8c36b2b9e0a2d27392040ecda0f51b636456de25
2010-04-23 01:58:52 +00:00
|
|
|
keyCode = device->touchScreen.currentVirtualKey.keyCode;
|
|
|
|
scanCode = device->touchScreen.currentVirtualKey.scanCode;
|
|
|
|
}
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
2010-06-15 08:31:58 +00:00
|
|
|
{ // acquire exported state lock
|
Native input dispatch rewrite work in progress.
The old dispatch mechanism has been left in place and continues to
be used by default for now. To enable native input dispatch,
edit the ENABLE_NATIVE_DISPATCH constant in WindowManagerPolicy.
Includes part of the new input event NDK API. Some details TBD.
To wire up input dispatch, as the ViewRoot adds a window to the
window session it receives an InputChannel object as an output
argument. The InputChannel encapsulates the file descriptors for a
shared memory region and two pipe end-points. The ViewRoot then
provides the InputChannel to the InputQueue. Behind the
scenes, InputQueue simply attaches handlers to the native PollLoop object
that underlies the MessageQueue. This way MessageQueue doesn't need
to know anything about input dispatch per-se, it just exposes (in native
code) a PollLoop that other components can use to monitor file descriptor
state changes.
There can be zero or more targets for any given input event. Each
input target is specified by its input channel and some parameters
including flags, an X/Y coordinate offset, and the dispatch timeout.
An input target can request either synchronous dispatch (for foreground apps)
or asynchronous dispatch (fire-and-forget for wallpapers and "outside"
targets). Currently, finding the appropriate input targets for an event
requires a call back into the WindowManagerServer from native code.
In the future this will be refactored to avoid most of these callbacks
except as required to handle pending focus transitions.
End-to-end event dispatch mostly works!
To do: event injection, rate limiting, ANRs, testing, optimization, etc.
Change-Id: I8c36b2b9e0a2d27392040ecda0f51b636456de25
2010-04-23 01:58:52 +00:00
|
|
|
AutoMutex _l(mExportedStateLock);
|
|
|
|
|
2010-06-15 08:31:58 +00:00
|
|
|
mExportedVirtualKeyCode = keyCode;
|
|
|
|
mExportedVirtualScanCode = scanCode;
|
|
|
|
} // release exported state lock
|
Native input dispatch rewrite work in progress.
The old dispatch mechanism has been left in place and continues to
be used by default for now. To enable native input dispatch,
edit the ENABLE_NATIVE_DISPATCH constant in WindowManagerPolicy.
Includes part of the new input event NDK API. Some details TBD.
To wire up input dispatch, as the ViewRoot adds a window to the
window session it receives an InputChannel object as an output
argument. The InputChannel encapsulates the file descriptors for a
shared memory region and two pipe end-points. The ViewRoot then
provides the InputChannel to the InputQueue. Behind the
scenes, InputQueue simply attaches handlers to the native PollLoop object
that underlies the MessageQueue. This way MessageQueue doesn't need
to know anything about input dispatch per-se, it just exposes (in native
code) a PollLoop that other components can use to monitor file descriptor
state changes.
There can be zero or more targets for any given input event. Each
input target is specified by its input channel and some parameters
including flags, an X/Y coordinate offset, and the dispatch timeout.
An input target can request either synchronous dispatch (for foreground apps)
or asynchronous dispatch (fire-and-forget for wallpapers and "outside"
targets). Currently, finding the appropriate input targets for an event
requires a call back into the WindowManagerServer from native code.
In the future this will be refactored to avoid most of these callbacks
except as required to handle pending focus transitions.
End-to-end event dispatch mostly works!
To do: event injection, rate limiting, ANRs, testing, optimization, etc.
Change-Id: I8c36b2b9e0a2d27392040ecda0f51b636456de25
2010-04-23 01:58:52 +00:00
|
|
|
}
|
|
|
|
|
|
|
|
bool InputReader::getCurrentVirtualKey(int32_t* outKeyCode, int32_t* outScanCode) const {
|
2010-06-15 08:31:58 +00:00
|
|
|
{ // acquire exported state lock
|
|
|
|
AutoMutex _l(mExportedStateLock);
|
|
|
|
|
|
|
|
*outKeyCode = mExportedVirtualKeyCode;
|
|
|
|
*outScanCode = mExportedVirtualScanCode;
|
|
|
|
return mExportedVirtualKeyCode != -1;
|
|
|
|
} // release exported state lock
|
|
|
|
}
|
|
|
|
|
|
|
|
void InputReader::updateExportedInputConfiguration() {
|
|
|
|
int32_t touchScreenConfig = InputConfiguration::TOUCHSCREEN_NOTOUCH;
|
|
|
|
int32_t keyboardConfig = InputConfiguration::KEYBOARD_NOKEYS;
|
|
|
|
int32_t navigationConfig = InputConfiguration::NAVIGATION_NONAV;
|
|
|
|
|
|
|
|
for (size_t i = 0; i < mDevices.size(); i++) {
|
|
|
|
InputDevice* device = mDevices.valueAt(i);
|
|
|
|
int32_t deviceClasses = device->classes;
|
|
|
|
|
|
|
|
if (deviceClasses & INPUT_DEVICE_CLASS_TOUCHSCREEN) {
|
|
|
|
touchScreenConfig = InputConfiguration::TOUCHSCREEN_FINGER;
|
|
|
|
}
|
|
|
|
if (deviceClasses & INPUT_DEVICE_CLASS_ALPHAKEY) {
|
|
|
|
keyboardConfig = InputConfiguration::KEYBOARD_QWERTY;
|
|
|
|
}
|
|
|
|
if (deviceClasses & INPUT_DEVICE_CLASS_TRACKBALL) {
|
|
|
|
navigationConfig = InputConfiguration::NAVIGATION_TRACKBALL;
|
|
|
|
} else if (deviceClasses & INPUT_DEVICE_CLASS_DPAD) {
|
|
|
|
navigationConfig = InputConfiguration::NAVIGATION_DPAD;
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
|
|
|
{ // acquire exported state lock
|
|
|
|
AutoMutex _l(mExportedStateLock);
|
|
|
|
|
|
|
|
mExportedInputConfiguration.touchScreen = touchScreenConfig;
|
|
|
|
mExportedInputConfiguration.keyboard = keyboardConfig;
|
|
|
|
mExportedInputConfiguration.navigation = navigationConfig;
|
|
|
|
} // release exported state lock
|
|
|
|
}
|
|
|
|
|
|
|
|
void InputReader::getCurrentInputConfiguration(InputConfiguration* outConfiguration) const {
|
|
|
|
{ // acquire exported state lock
|
|
|
|
AutoMutex _l(mExportedStateLock);
|
|
|
|
|
|
|
|
*outConfiguration = mExportedInputConfiguration;
|
|
|
|
} // release exported state lock
|
|
|
|
}
|
|
|
|
|
|
|
|
int32_t InputReader::getCurrentScanCodeState(int32_t deviceId, int32_t deviceClasses,
|
|
|
|
int32_t scanCode) const {
|
|
|
|
{ // acquire exported state lock
|
|
|
|
AutoMutex _l(mExportedStateLock);
|
|
|
|
|
|
|
|
if (mExportedVirtualScanCode == scanCode) {
|
2010-07-15 01:48:53 +00:00
|
|
|
return AKEY_STATE_VIRTUAL;
|
2010-06-15 08:31:58 +00:00
|
|
|
}
|
|
|
|
} // release exported state lock
|
|
|
|
|
|
|
|
return mEventHub->getScanCodeState(deviceId, deviceClasses, scanCode);
|
|
|
|
}
|
|
|
|
|
|
|
|
int32_t InputReader::getCurrentKeyCodeState(int32_t deviceId, int32_t deviceClasses,
|
|
|
|
int32_t keyCode) const {
|
|
|
|
{ // acquire exported state lock
|
|
|
|
AutoMutex _l(mExportedStateLock);
|
|
|
|
|
|
|
|
if (mExportedVirtualKeyCode == keyCode) {
|
2010-07-15 01:48:53 +00:00
|
|
|
return AKEY_STATE_VIRTUAL;
|
2010-06-15 08:31:58 +00:00
|
|
|
}
|
|
|
|
} // release exported state lock
|
|
|
|
|
|
|
|
return mEventHub->getKeyCodeState(deviceId, deviceClasses, keyCode);
|
|
|
|
}
|
|
|
|
|
|
|
|
int32_t InputReader::getCurrentSwitchState(int32_t deviceId, int32_t deviceClasses,
|
|
|
|
int32_t sw) const {
|
|
|
|
return mEventHub->getSwitchState(deviceId, deviceClasses, sw);
|
|
|
|
}
|
Native input dispatch rewrite work in progress.
The old dispatch mechanism has been left in place and continues to
be used by default for now. To enable native input dispatch,
edit the ENABLE_NATIVE_DISPATCH constant in WindowManagerPolicy.
Includes part of the new input event NDK API. Some details TBD.
To wire up input dispatch, as the ViewRoot adds a window to the
window session it receives an InputChannel object as an output
argument. The InputChannel encapsulates the file descriptors for a
shared memory region and two pipe end-points. The ViewRoot then
provides the InputChannel to the InputQueue. Behind the
scenes, InputQueue simply attaches handlers to the native PollLoop object
that underlies the MessageQueue. This way MessageQueue doesn't need
to know anything about input dispatch per-se, it just exposes (in native
code) a PollLoop that other components can use to monitor file descriptor
state changes.
There can be zero or more targets for any given input event. Each
input target is specified by its input channel and some parameters
including flags, an X/Y coordinate offset, and the dispatch timeout.
An input target can request either synchronous dispatch (for foreground apps)
or asynchronous dispatch (fire-and-forget for wallpapers and "outside"
targets). Currently, finding the appropriate input targets for an event
requires a call back into the WindowManagerServer from native code.
In the future this will be refactored to avoid most of these callbacks
except as required to handle pending focus transitions.
End-to-end event dispatch mostly works!
To do: event injection, rate limiting, ANRs, testing, optimization, etc.
Change-Id: I8c36b2b9e0a2d27392040ecda0f51b636456de25
2010-04-23 01:58:52 +00:00
|
|
|
|
2010-06-15 08:31:58 +00:00
|
|
|
bool InputReader::hasKeys(size_t numCodes, const int32_t* keyCodes, uint8_t* outFlags) const {
|
|
|
|
return mEventHub->hasKeys(numCodes, keyCodes, outFlags);
|
Native input dispatch rewrite work in progress.
The old dispatch mechanism has been left in place and continues to
be used by default for now. To enable native input dispatch,
edit the ENABLE_NATIVE_DISPATCH constant in WindowManagerPolicy.
Includes part of the new input event NDK API. Some details TBD.
To wire up input dispatch, as the ViewRoot adds a window to the
window session it receives an InputChannel object as an output
argument. The InputChannel encapsulates the file descriptors for a
shared memory region and two pipe end-points. The ViewRoot then
provides the InputChannel to the InputQueue. Behind the
scenes, InputQueue simply attaches handlers to the native PollLoop object
that underlies the MessageQueue. This way MessageQueue doesn't need
to know anything about input dispatch per-se, it just exposes (in native
code) a PollLoop that other components can use to monitor file descriptor
state changes.
There can be zero or more targets for any given input event. Each
input target is specified by its input channel and some parameters
including flags, an X/Y coordinate offset, and the dispatch timeout.
An input target can request either synchronous dispatch (for foreground apps)
or asynchronous dispatch (fire-and-forget for wallpapers and "outside"
targets). Currently, finding the appropriate input targets for an event
requires a call back into the WindowManagerServer from native code.
In the future this will be refactored to avoid most of these callbacks
except as required to handle pending focus transitions.
End-to-end event dispatch mostly works!
To do: event injection, rate limiting, ANRs, testing, optimization, etc.
Change-Id: I8c36b2b9e0a2d27392040ecda0f51b636456de25
2010-04-23 01:58:52 +00:00
|
|
|
}
|
|
|
|
|
|
|
|
|
|
|
|
// --- InputReaderThread ---
|
|
|
|
|
|
|
|
InputReaderThread::InputReaderThread(const sp<InputReaderInterface>& reader) :
|
|
|
|
Thread(/*canCallJava*/ true), mReader(reader) {
|
|
|
|
}
|
|
|
|
|
|
|
|
InputReaderThread::~InputReaderThread() {
|
|
|
|
}
|
|
|
|
|
|
|
|
bool InputReaderThread::threadLoop() {
|
|
|
|
mReader->loopOnce();
|
|
|
|
return true;
|
|
|
|
}
|
|
|
|
|
|
|
|
} // namespace android
|