SDL2/src/joystick/windows/SDL_rawinputjoystick.c

2034 lines
76 KiB
C
Raw Normal View History

Fixed bug 4477 - Support more than 4 XInput-capable devices on Windows Jimb Esser Add new RawInput controller API, and improved correlation with XInput/WGI Reorder joystick init so drivers can ask the others if they handle a device reliably Do not poll disconnected XInput devices (major perf issue) Fix various cases where incorrect correlation could happen Simple mechanism for propagating unhandled Guide button presses even before guaranteed correlation Correlate by axis motion as well as button presses Fix failing to zero other trigger Fix SDL_HINT_JOYSTICK_HIDAPI not working if set before calling SDL_Init() Add missing device to device names Disable RawInput if we have a mismatch of XInput-capable but not RawInput-capable devices Updated to SDL 2.0.13 code with the following notes: New HID driver: xbox360w - no idea what that is, hopefully urelated SDL_hidapijoystick.c had been refactored to couple data handling logic with device opening logic and device lists caused some problems, yields slightly uglier integration than previously when the 360 HID device driver was just handling the data. SDL_hidapijoystick.c now often pulls the device off of the joystick_hwdata structure for some rumble logic, but it appears that code path is never reached, so probably not a problem. Looks like joystick_hwdata was refactored to not include a mutex in other drivers, maintainers may want to do the same refactor here if that's useful for some reason. Something changed in how devices get names, so getting generic names. Had to fix a (new?) bug where removing an XInput controller caused existing controllers (that moved to a new XInput index) to get identified as 0x045e/0x02fd ("it's probably Bluetooth" in code), rendering the existing HIDAPI_IsDevicePresent and new RAWINPUT_IsDevicePresent unreliable.
2020-03-16 14:25:02 -05:00
/*
Simple DirectMedia Layer
2023-01-09 11:41:41 -06:00
Copyright (C) 2023 Sam Lantinga <slouken@libsdl.org>
Fixed bug 4477 - Support more than 4 XInput-capable devices on Windows Jimb Esser Add new RawInput controller API, and improved correlation with XInput/WGI Reorder joystick init so drivers can ask the others if they handle a device reliably Do not poll disconnected XInput devices (major perf issue) Fix various cases where incorrect correlation could happen Simple mechanism for propagating unhandled Guide button presses even before guaranteed correlation Correlate by axis motion as well as button presses Fix failing to zero other trigger Fix SDL_HINT_JOYSTICK_HIDAPI not working if set before calling SDL_Init() Add missing device to device names Disable RawInput if we have a mismatch of XInput-capable but not RawInput-capable devices Updated to SDL 2.0.13 code with the following notes: New HID driver: xbox360w - no idea what that is, hopefully urelated SDL_hidapijoystick.c had been refactored to couple data handling logic with device opening logic and device lists caused some problems, yields slightly uglier integration than previously when the 360 HID device driver was just handling the data. SDL_hidapijoystick.c now often pulls the device off of the joystick_hwdata structure for some rumble logic, but it appears that code path is never reached, so probably not a problem. Looks like joystick_hwdata was refactored to not include a mutex in other drivers, maintainers may want to do the same refactor here if that's useful for some reason. Something changed in how devices get names, so getting generic names. Had to fix a (new?) bug where removing an XInput controller caused existing controllers (that moved to a new XInput index) to get identified as 0x045e/0x02fd ("it's probably Bluetooth" in code), rendering the existing HIDAPI_IsDevicePresent and new RAWINPUT_IsDevicePresent unreliable.
2020-03-16 14:25:02 -05:00
This software is provided 'as-is', without any express or implied
warranty. In no event will the authors be held liable for any damages
arising from the use of this software.
Permission is granted to anyone to use this software for any purpose,
including commercial applications, and to alter it and redistribute it
freely, subject to the following restrictions:
1. The origin of this software must not be misrepresented; you must not
claim that you wrote the original software. If you use this software
in a product, an acknowledgment in the product documentation would be
appreciated but is not required.
2. Altered source versions must be plainly marked as such, and must not be
misrepresented as being the original software.
3. This notice may not be removed or altered from any source distribution.
*/
/*
RAWINPUT Joystick API for better handling XInput-capable devices on Windows.
XInput is limited to 4 devices.
Windows.Gaming.Input does not get inputs from XBox One controllers when not in the foreground.
DirectInput does not get inputs from XBox One controllers when not in the foreground, nor rumble or accurate triggers.
RawInput does not get rumble or accurate triggers.
So, combine them as best we can!
*/
#include "SDL_internal.h"
Fixed bug 4477 - Support more than 4 XInput-capable devices on Windows Jimb Esser Add new RawInput controller API, and improved correlation with XInput/WGI Reorder joystick init so drivers can ask the others if they handle a device reliably Do not poll disconnected XInput devices (major perf issue) Fix various cases where incorrect correlation could happen Simple mechanism for propagating unhandled Guide button presses even before guaranteed correlation Correlate by axis motion as well as button presses Fix failing to zero other trigger Fix SDL_HINT_JOYSTICK_HIDAPI not working if set before calling SDL_Init() Add missing device to device names Disable RawInput if we have a mismatch of XInput-capable but not RawInput-capable devices Updated to SDL 2.0.13 code with the following notes: New HID driver: xbox360w - no idea what that is, hopefully urelated SDL_hidapijoystick.c had been refactored to couple data handling logic with device opening logic and device lists caused some problems, yields slightly uglier integration than previously when the 360 HID device driver was just handling the data. SDL_hidapijoystick.c now often pulls the device off of the joystick_hwdata structure for some rumble logic, but it appears that code path is never reached, so probably not a problem. Looks like joystick_hwdata was refactored to not include a mutex in other drivers, maintainers may want to do the same refactor here if that's useful for some reason. Something changed in how devices get names, so getting generic names. Had to fix a (new?) bug where removing an XInput controller caused existing controllers (that moved to a new XInput index) to get identified as 0x045e/0x02fd ("it's probably Bluetooth" in code), rendering the existing HIDAPI_IsDevicePresent and new RAWINPUT_IsDevicePresent unreliable.
2020-03-16 14:25:02 -05:00
#if SDL_JOYSTICK_RAWINPUT
#include "../usb_ids.h"
Fixed bug 4477 - Support more than 4 XInput-capable devices on Windows Jimb Esser Add new RawInput controller API, and improved correlation with XInput/WGI Reorder joystick init so drivers can ask the others if they handle a device reliably Do not poll disconnected XInput devices (major perf issue) Fix various cases where incorrect correlation could happen Simple mechanism for propagating unhandled Guide button presses even before guaranteed correlation Correlate by axis motion as well as button presses Fix failing to zero other trigger Fix SDL_HINT_JOYSTICK_HIDAPI not working if set before calling SDL_Init() Add missing device to device names Disable RawInput if we have a mismatch of XInput-capable but not RawInput-capable devices Updated to SDL 2.0.13 code with the following notes: New HID driver: xbox360w - no idea what that is, hopefully urelated SDL_hidapijoystick.c had been refactored to couple data handling logic with device opening logic and device lists caused some problems, yields slightly uglier integration than previously when the 360 HID device driver was just handling the data. SDL_hidapijoystick.c now often pulls the device off of the joystick_hwdata structure for some rumble logic, but it appears that code path is never reached, so probably not a problem. Looks like joystick_hwdata was refactored to not include a mutex in other drivers, maintainers may want to do the same refactor here if that's useful for some reason. Something changed in how devices get names, so getting generic names. Had to fix a (new?) bug where removing an XInput controller caused existing controllers (that moved to a new XInput index) to get identified as 0x045e/0x02fd ("it's probably Bluetooth" in code), rendering the existing HIDAPI_IsDevicePresent and new RAWINPUT_IsDevicePresent unreliable.
2020-03-16 14:25:02 -05:00
#include "../SDL_sysjoystick.h"
#include "../../core/windows/SDL_windows.h"
#include "../../core/windows/SDL_hid.h"
Fixed bug 4477 - Support more than 4 XInput-capable devices on Windows Jimb Esser Add new RawInput controller API, and improved correlation with XInput/WGI Reorder joystick init so drivers can ask the others if they handle a device reliably Do not poll disconnected XInput devices (major perf issue) Fix various cases where incorrect correlation could happen Simple mechanism for propagating unhandled Guide button presses even before guaranteed correlation Correlate by axis motion as well as button presses Fix failing to zero other trigger Fix SDL_HINT_JOYSTICK_HIDAPI not working if set before calling SDL_Init() Add missing device to device names Disable RawInput if we have a mismatch of XInput-capable but not RawInput-capable devices Updated to SDL 2.0.13 code with the following notes: New HID driver: xbox360w - no idea what that is, hopefully urelated SDL_hidapijoystick.c had been refactored to couple data handling logic with device opening logic and device lists caused some problems, yields slightly uglier integration than previously when the 360 HID device driver was just handling the data. SDL_hidapijoystick.c now often pulls the device off of the joystick_hwdata structure for some rumble logic, but it appears that code path is never reached, so probably not a problem. Looks like joystick_hwdata was refactored to not include a mutex in other drivers, maintainers may want to do the same refactor here if that's useful for some reason. Something changed in how devices get names, so getting generic names. Had to fix a (new?) bug where removing an XInput controller caused existing controllers (that moved to a new XInput index) to get identified as 0x045e/0x02fd ("it's probably Bluetooth" in code), rendering the existing HIDAPI_IsDevicePresent and new RAWINPUT_IsDevicePresent unreliable.
2020-03-16 14:25:02 -05:00
#include "../hidapi/SDL_hidapijoystick_c.h"
2021-01-16 20:00:50 -06:00
#ifdef HAVE_XINPUT_H
#define SDL_JOYSTICK_RAWINPUT_XINPUT
2021-01-16 20:00:50 -06:00
#endif
#ifdef HAVE_WINDOWS_GAMING_INPUT_H
#define SDL_JOYSTICK_RAWINPUT_WGI
Fixed bug 4477 - Support more than 4 XInput-capable devices on Windows Jimb Esser Add new RawInput controller API, and improved correlation with XInput/WGI Reorder joystick init so drivers can ask the others if they handle a device reliably Do not poll disconnected XInput devices (major perf issue) Fix various cases where incorrect correlation could happen Simple mechanism for propagating unhandled Guide button presses even before guaranteed correlation Correlate by axis motion as well as button presses Fix failing to zero other trigger Fix SDL_HINT_JOYSTICK_HIDAPI not working if set before calling SDL_Init() Add missing device to device names Disable RawInput if we have a mismatch of XInput-capable but not RawInput-capable devices Updated to SDL 2.0.13 code with the following notes: New HID driver: xbox360w - no idea what that is, hopefully urelated SDL_hidapijoystick.c had been refactored to couple data handling logic with device opening logic and device lists caused some problems, yields slightly uglier integration than previously when the 360 HID device driver was just handling the data. SDL_hidapijoystick.c now often pulls the device off of the joystick_hwdata structure for some rumble logic, but it appears that code path is never reached, so probably not a problem. Looks like joystick_hwdata was refactored to not include a mutex in other drivers, maintainers may want to do the same refactor here if that's useful for some reason. Something changed in how devices get names, so getting generic names. Had to fix a (new?) bug where removing an XInput controller caused existing controllers (that moved to a new XInput index) to get identified as 0x045e/0x02fd ("it's probably Bluetooth" in code), rendering the existing HIDAPI_IsDevicePresent and new RAWINPUT_IsDevicePresent unreliable.
2020-03-16 14:25:02 -05:00
#endif
#ifdef SDL_JOYSTICK_RAWINPUT_XINPUT
#include "../../core/windows/SDL_xinput.h"
#endif
#ifdef SDL_JOYSTICK_RAWINPUT_WGI
#include "../../core/windows/SDL_windows.h"
typedef struct WindowsGamingInputGamepadState WindowsGamingInputGamepadState;
#define GamepadButtons_GUIDE 0x40000000
#define COBJMACROS
#include "windows.gaming.input.h"
2022-04-01 19:32:10 -05:00
#include <roapi.h>
#endif
#if defined(SDL_JOYSTICK_RAWINPUT_XINPUT) || defined(SDL_JOYSTICK_RAWINPUT_WGI)
#define SDL_JOYSTICK_RAWINPUT_MATCHING
#define SDL_JOYSTICK_RAWINPUT_MATCH_AXES
#define SDL_JOYSTICK_RAWINPUT_MATCH_TRIGGERS
#ifdef SDL_JOYSTICK_RAWINPUT_MATCH_TRIGGERS
#define SDL_JOYSTICK_RAWINPUT_MATCH_COUNT 6 // stick + trigger axes
#else
#define SDL_JOYSTICK_RAWINPUT_MATCH_COUNT 4 // stick axes
#endif
#endif
/*#define DEBUG_RAWINPUT*/
Fixed bug 4477 - Support more than 4 XInput-capable devices on Windows Jimb Esser Add new RawInput controller API, and improved correlation with XInput/WGI Reorder joystick init so drivers can ask the others if they handle a device reliably Do not poll disconnected XInput devices (major perf issue) Fix various cases where incorrect correlation could happen Simple mechanism for propagating unhandled Guide button presses even before guaranteed correlation Correlate by axis motion as well as button presses Fix failing to zero other trigger Fix SDL_HINT_JOYSTICK_HIDAPI not working if set before calling SDL_Init() Add missing device to device names Disable RawInput if we have a mismatch of XInput-capable but not RawInput-capable devices Updated to SDL 2.0.13 code with the following notes: New HID driver: xbox360w - no idea what that is, hopefully urelated SDL_hidapijoystick.c had been refactored to couple data handling logic with device opening logic and device lists caused some problems, yields slightly uglier integration than previously when the 360 HID device driver was just handling the data. SDL_hidapijoystick.c now often pulls the device off of the joystick_hwdata structure for some rumble logic, but it appears that code path is never reached, so probably not a problem. Looks like joystick_hwdata was refactored to not include a mutex in other drivers, maintainers may want to do the same refactor here if that's useful for some reason. Something changed in how devices get names, so getting generic names. Had to fix a (new?) bug where removing an XInput controller caused existing controllers (that moved to a new XInput index) to get identified as 0x045e/0x02fd ("it's probably Bluetooth" in code), rendering the existing HIDAPI_IsDevicePresent and new RAWINPUT_IsDevicePresent unreliable.
2020-03-16 14:25:02 -05:00
#ifndef RIDEV_EXINPUTSINK
#define RIDEV_EXINPUTSINK 0x00001000
#define RIDEV_DEVNOTIFY 0x00002000
Fixed bug 4477 - Support more than 4 XInput-capable devices on Windows Jimb Esser Add new RawInput controller API, and improved correlation with XInput/WGI Reorder joystick init so drivers can ask the others if they handle a device reliably Do not poll disconnected XInput devices (major perf issue) Fix various cases where incorrect correlation could happen Simple mechanism for propagating unhandled Guide button presses even before guaranteed correlation Correlate by axis motion as well as button presses Fix failing to zero other trigger Fix SDL_HINT_JOYSTICK_HIDAPI not working if set before calling SDL_Init() Add missing device to device names Disable RawInput if we have a mismatch of XInput-capable but not RawInput-capable devices Updated to SDL 2.0.13 code with the following notes: New HID driver: xbox360w - no idea what that is, hopefully urelated SDL_hidapijoystick.c had been refactored to couple data handling logic with device opening logic and device lists caused some problems, yields slightly uglier integration than previously when the 360 HID device driver was just handling the data. SDL_hidapijoystick.c now often pulls the device off of the joystick_hwdata structure for some rumble logic, but it appears that code path is never reached, so probably not a problem. Looks like joystick_hwdata was refactored to not include a mutex in other drivers, maintainers may want to do the same refactor here if that's useful for some reason. Something changed in how devices get names, so getting generic names. Had to fix a (new?) bug where removing an XInput controller caused existing controllers (that moved to a new XInput index) to get identified as 0x045e/0x02fd ("it's probably Bluetooth" in code), rendering the existing HIDAPI_IsDevicePresent and new RAWINPUT_IsDevicePresent unreliable.
2020-03-16 14:25:02 -05:00
#endif
#ifndef WM_INPUT_DEVICE_CHANGE
#define WM_INPUT_DEVICE_CHANGE 0x00FE
Fixed bug 4477 - Support more than 4 XInput-capable devices on Windows Jimb Esser Add new RawInput controller API, and improved correlation with XInput/WGI Reorder joystick init so drivers can ask the others if they handle a device reliably Do not poll disconnected XInput devices (major perf issue) Fix various cases where incorrect correlation could happen Simple mechanism for propagating unhandled Guide button presses even before guaranteed correlation Correlate by axis motion as well as button presses Fix failing to zero other trigger Fix SDL_HINT_JOYSTICK_HIDAPI not working if set before calling SDL_Init() Add missing device to device names Disable RawInput if we have a mismatch of XInput-capable but not RawInput-capable devices Updated to SDL 2.0.13 code with the following notes: New HID driver: xbox360w - no idea what that is, hopefully urelated SDL_hidapijoystick.c had been refactored to couple data handling logic with device opening logic and device lists caused some problems, yields slightly uglier integration than previously when the 360 HID device driver was just handling the data. SDL_hidapijoystick.c now often pulls the device off of the joystick_hwdata structure for some rumble logic, but it appears that code path is never reached, so probably not a problem. Looks like joystick_hwdata was refactored to not include a mutex in other drivers, maintainers may want to do the same refactor here if that's useful for some reason. Something changed in how devices get names, so getting generic names. Had to fix a (new?) bug where removing an XInput controller caused existing controllers (that moved to a new XInput index) to get identified as 0x045e/0x02fd ("it's probably Bluetooth" in code), rendering the existing HIDAPI_IsDevicePresent and new RAWINPUT_IsDevicePresent unreliable.
2020-03-16 14:25:02 -05:00
#endif
#ifndef WM_INPUT
#define WM_INPUT 0x00FF
Fixed bug 4477 - Support more than 4 XInput-capable devices on Windows Jimb Esser Add new RawInput controller API, and improved correlation with XInput/WGI Reorder joystick init so drivers can ask the others if they handle a device reliably Do not poll disconnected XInput devices (major perf issue) Fix various cases where incorrect correlation could happen Simple mechanism for propagating unhandled Guide button presses even before guaranteed correlation Correlate by axis motion as well as button presses Fix failing to zero other trigger Fix SDL_HINT_JOYSTICK_HIDAPI not working if set before calling SDL_Init() Add missing device to device names Disable RawInput if we have a mismatch of XInput-capable but not RawInput-capable devices Updated to SDL 2.0.13 code with the following notes: New HID driver: xbox360w - no idea what that is, hopefully urelated SDL_hidapijoystick.c had been refactored to couple data handling logic with device opening logic and device lists caused some problems, yields slightly uglier integration than previously when the 360 HID device driver was just handling the data. SDL_hidapijoystick.c now often pulls the device off of the joystick_hwdata structure for some rumble logic, but it appears that code path is never reached, so probably not a problem. Looks like joystick_hwdata was refactored to not include a mutex in other drivers, maintainers may want to do the same refactor here if that's useful for some reason. Something changed in how devices get names, so getting generic names. Had to fix a (new?) bug where removing an XInput controller caused existing controllers (that moved to a new XInput index) to get identified as 0x045e/0x02fd ("it's probably Bluetooth" in code), rendering the existing HIDAPI_IsDevicePresent and new RAWINPUT_IsDevicePresent unreliable.
2020-03-16 14:25:02 -05:00
#endif
#ifndef GIDC_ARRIVAL
#define GIDC_ARRIVAL 1
#define GIDC_REMOVAL 2
Fixed bug 4477 - Support more than 4 XInput-capable devices on Windows Jimb Esser Add new RawInput controller API, and improved correlation with XInput/WGI Reorder joystick init so drivers can ask the others if they handle a device reliably Do not poll disconnected XInput devices (major perf issue) Fix various cases where incorrect correlation could happen Simple mechanism for propagating unhandled Guide button presses even before guaranteed correlation Correlate by axis motion as well as button presses Fix failing to zero other trigger Fix SDL_HINT_JOYSTICK_HIDAPI not working if set before calling SDL_Init() Add missing device to device names Disable RawInput if we have a mismatch of XInput-capable but not RawInput-capable devices Updated to SDL 2.0.13 code with the following notes: New HID driver: xbox360w - no idea what that is, hopefully urelated SDL_hidapijoystick.c had been refactored to couple data handling logic with device opening logic and device lists caused some problems, yields slightly uglier integration than previously when the 360 HID device driver was just handling the data. SDL_hidapijoystick.c now often pulls the device off of the joystick_hwdata structure for some rumble logic, but it appears that code path is never reached, so probably not a problem. Looks like joystick_hwdata was refactored to not include a mutex in other drivers, maintainers may want to do the same refactor here if that's useful for some reason. Something changed in how devices get names, so getting generic names. Had to fix a (new?) bug where removing an XInput controller caused existing controllers (that moved to a new XInput index) to get identified as 0x045e/0x02fd ("it's probably Bluetooth" in code), rendering the existing HIDAPI_IsDevicePresent and new RAWINPUT_IsDevicePresent unreliable.
2020-03-16 14:25:02 -05:00
#endif
static SDL_bool SDL_RAWINPUT_inited = SDL_FALSE;
static int SDL_RAWINPUT_numjoysticks = 0;
static void RAWINPUT_JoystickClose(SDL_Joystick *joystick);
Fixed bug 4477 - Support more than 4 XInput-capable devices on Windows Jimb Esser Add new RawInput controller API, and improved correlation with XInput/WGI Reorder joystick init so drivers can ask the others if they handle a device reliably Do not poll disconnected XInput devices (major perf issue) Fix various cases where incorrect correlation could happen Simple mechanism for propagating unhandled Guide button presses even before guaranteed correlation Correlate by axis motion as well as button presses Fix failing to zero other trigger Fix SDL_HINT_JOYSTICK_HIDAPI not working if set before calling SDL_Init() Add missing device to device names Disable RawInput if we have a mismatch of XInput-capable but not RawInput-capable devices Updated to SDL 2.0.13 code with the following notes: New HID driver: xbox360w - no idea what that is, hopefully urelated SDL_hidapijoystick.c had been refactored to couple data handling logic with device opening logic and device lists caused some problems, yields slightly uglier integration than previously when the 360 HID device driver was just handling the data. SDL_hidapijoystick.c now often pulls the device off of the joystick_hwdata structure for some rumble logic, but it appears that code path is never reached, so probably not a problem. Looks like joystick_hwdata was refactored to not include a mutex in other drivers, maintainers may want to do the same refactor here if that's useful for some reason. Something changed in how devices get names, so getting generic names. Had to fix a (new?) bug where removing an XInput controller caused existing controllers (that moved to a new XInput index) to get identified as 0x045e/0x02fd ("it's probably Bluetooth" in code), rendering the existing HIDAPI_IsDevicePresent and new RAWINPUT_IsDevicePresent unreliable.
2020-03-16 14:25:02 -05:00
typedef struct SDL_RAWINPUT_Device
Fixed bug 4477 - Support more than 4 XInput-capable devices on Windows Jimb Esser Add new RawInput controller API, and improved correlation with XInput/WGI Reorder joystick init so drivers can ask the others if they handle a device reliably Do not poll disconnected XInput devices (major perf issue) Fix various cases where incorrect correlation could happen Simple mechanism for propagating unhandled Guide button presses even before guaranteed correlation Correlate by axis motion as well as button presses Fix failing to zero other trigger Fix SDL_HINT_JOYSTICK_HIDAPI not working if set before calling SDL_Init() Add missing device to device names Disable RawInput if we have a mismatch of XInput-capable but not RawInput-capable devices Updated to SDL 2.0.13 code with the following notes: New HID driver: xbox360w - no idea what that is, hopefully urelated SDL_hidapijoystick.c had been refactored to couple data handling logic with device opening logic and device lists caused some problems, yields slightly uglier integration than previously when the 360 HID device driver was just handling the data. SDL_hidapijoystick.c now often pulls the device off of the joystick_hwdata structure for some rumble logic, but it appears that code path is never reached, so probably not a problem. Looks like joystick_hwdata was refactored to not include a mutex in other drivers, maintainers may want to do the same refactor here if that's useful for some reason. Something changed in how devices get names, so getting generic names. Had to fix a (new?) bug where removing an XInput controller caused existing controllers (that moved to a new XInput index) to get identified as 0x045e/0x02fd ("it's probably Bluetooth" in code), rendering the existing HIDAPI_IsDevicePresent and new RAWINPUT_IsDevicePresent unreliable.
2020-03-16 14:25:02 -05:00
{
2020-04-13 15:24:19 -05:00
SDL_atomic_t refcount;
Fixed bug 4477 - Support more than 4 XInput-capable devices on Windows Jimb Esser Add new RawInput controller API, and improved correlation with XInput/WGI Reorder joystick init so drivers can ask the others if they handle a device reliably Do not poll disconnected XInput devices (major perf issue) Fix various cases where incorrect correlation could happen Simple mechanism for propagating unhandled Guide button presses even before guaranteed correlation Correlate by axis motion as well as button presses Fix failing to zero other trigger Fix SDL_HINT_JOYSTICK_HIDAPI not working if set before calling SDL_Init() Add missing device to device names Disable RawInput if we have a mismatch of XInput-capable but not RawInput-capable devices Updated to SDL 2.0.13 code with the following notes: New HID driver: xbox360w - no idea what that is, hopefully urelated SDL_hidapijoystick.c had been refactored to couple data handling logic with device opening logic and device lists caused some problems, yields slightly uglier integration than previously when the 360 HID device driver was just handling the data. SDL_hidapijoystick.c now often pulls the device off of the joystick_hwdata structure for some rumble logic, but it appears that code path is never reached, so probably not a problem. Looks like joystick_hwdata was refactored to not include a mutex in other drivers, maintainers may want to do the same refactor here if that's useful for some reason. Something changed in how devices get names, so getting generic names. Had to fix a (new?) bug where removing an XInput controller caused existing controllers (that moved to a new XInput index) to get identified as 0x045e/0x02fd ("it's probably Bluetooth" in code), rendering the existing HIDAPI_IsDevicePresent and new RAWINPUT_IsDevicePresent unreliable.
2020-03-16 14:25:02 -05:00
char *name;
char *path;
Fixed bug 4477 - Support more than 4 XInput-capable devices on Windows Jimb Esser Add new RawInput controller API, and improved correlation with XInput/WGI Reorder joystick init so drivers can ask the others if they handle a device reliably Do not poll disconnected XInput devices (major perf issue) Fix various cases where incorrect correlation could happen Simple mechanism for propagating unhandled Guide button presses even before guaranteed correlation Correlate by axis motion as well as button presses Fix failing to zero other trigger Fix SDL_HINT_JOYSTICK_HIDAPI not working if set before calling SDL_Init() Add missing device to device names Disable RawInput if we have a mismatch of XInput-capable but not RawInput-capable devices Updated to SDL 2.0.13 code with the following notes: New HID driver: xbox360w - no idea what that is, hopefully urelated SDL_hidapijoystick.c had been refactored to couple data handling logic with device opening logic and device lists caused some problems, yields slightly uglier integration than previously when the 360 HID device driver was just handling the data. SDL_hidapijoystick.c now often pulls the device off of the joystick_hwdata structure for some rumble logic, but it appears that code path is never reached, so probably not a problem. Looks like joystick_hwdata was refactored to not include a mutex in other drivers, maintainers may want to do the same refactor here if that's useful for some reason. Something changed in how devices get names, so getting generic names. Had to fix a (new?) bug where removing an XInput controller caused existing controllers (that moved to a new XInput index) to get identified as 0x045e/0x02fd ("it's probably Bluetooth" in code), rendering the existing HIDAPI_IsDevicePresent and new RAWINPUT_IsDevicePresent unreliable.
2020-03-16 14:25:02 -05:00
Uint16 vendor_id;
Uint16 product_id;
Uint16 version;
SDL_JoystickGUID guid;
SDL_bool is_xinput;
SDL_bool is_xboxone;
PHIDP_PREPARSED_DATA preparsed_data;
Fixed bug 4477 - Support more than 4 XInput-capable devices on Windows Jimb Esser Add new RawInput controller API, and improved correlation with XInput/WGI Reorder joystick init so drivers can ask the others if they handle a device reliably Do not poll disconnected XInput devices (major perf issue) Fix various cases where incorrect correlation could happen Simple mechanism for propagating unhandled Guide button presses even before guaranteed correlation Correlate by axis motion as well as button presses Fix failing to zero other trigger Fix SDL_HINT_JOYSTICK_HIDAPI not working if set before calling SDL_Init() Add missing device to device names Disable RawInput if we have a mismatch of XInput-capable but not RawInput-capable devices Updated to SDL 2.0.13 code with the following notes: New HID driver: xbox360w - no idea what that is, hopefully urelated SDL_hidapijoystick.c had been refactored to couple data handling logic with device opening logic and device lists caused some problems, yields slightly uglier integration than previously when the 360 HID device driver was just handling the data. SDL_hidapijoystick.c now often pulls the device off of the joystick_hwdata structure for some rumble logic, but it appears that code path is never reached, so probably not a problem. Looks like joystick_hwdata was refactored to not include a mutex in other drivers, maintainers may want to do the same refactor here if that's useful for some reason. Something changed in how devices get names, so getting generic names. Had to fix a (new?) bug where removing an XInput controller caused existing controllers (that moved to a new XInput index) to get identified as 0x045e/0x02fd ("it's probably Bluetooth" in code), rendering the existing HIDAPI_IsDevicePresent and new RAWINPUT_IsDevicePresent unreliable.
2020-03-16 14:25:02 -05:00
HANDLE hDevice;
SDL_Joystick *joystick;
SDL_JoystickID joystick_id;
struct SDL_RAWINPUT_Device *next;
Fixed bug 4477 - Support more than 4 XInput-capable devices on Windows Jimb Esser Add new RawInput controller API, and improved correlation with XInput/WGI Reorder joystick init so drivers can ask the others if they handle a device reliably Do not poll disconnected XInput devices (major perf issue) Fix various cases where incorrect correlation could happen Simple mechanism for propagating unhandled Guide button presses even before guaranteed correlation Correlate by axis motion as well as button presses Fix failing to zero other trigger Fix SDL_HINT_JOYSTICK_HIDAPI not working if set before calling SDL_Init() Add missing device to device names Disable RawInput if we have a mismatch of XInput-capable but not RawInput-capable devices Updated to SDL 2.0.13 code with the following notes: New HID driver: xbox360w - no idea what that is, hopefully urelated SDL_hidapijoystick.c had been refactored to couple data handling logic with device opening logic and device lists caused some problems, yields slightly uglier integration than previously when the 360 HID device driver was just handling the data. SDL_hidapijoystick.c now often pulls the device off of the joystick_hwdata structure for some rumble logic, but it appears that code path is never reached, so probably not a problem. Looks like joystick_hwdata was refactored to not include a mutex in other drivers, maintainers may want to do the same refactor here if that's useful for some reason. Something changed in how devices get names, so getting generic names. Had to fix a (new?) bug where removing an XInput controller caused existing controllers (that moved to a new XInput index) to get identified as 0x045e/0x02fd ("it's probably Bluetooth" in code), rendering the existing HIDAPI_IsDevicePresent and new RAWINPUT_IsDevicePresent unreliable.
2020-03-16 14:25:02 -05:00
} SDL_RAWINPUT_Device;
struct joystick_hwdata
{
SDL_bool is_xinput;
SDL_bool is_xboxone;
PHIDP_PREPARSED_DATA preparsed_data;
ULONG max_data_length;
HIDP_DATA *data;
USHORT *button_indices;
USHORT *axis_indices;
USHORT *hat_indices;
SDL_bool guide_hack;
SDL_bool trigger_hack;
USHORT trigger_hack_index;
#ifdef SDL_JOYSTICK_RAWINPUT_MATCHING
Uint64 match_state; /* Lowest 16 bits for button states, higher 24 for 6 4bit axes */
Uint64 last_state_packet;
#endif
#ifdef SDL_JOYSTICK_RAWINPUT_XINPUT
SDL_bool xinput_enabled;
SDL_bool xinput_correlated;
Uint8 xinput_correlation_id;
Uint8 xinput_correlation_count;
Uint8 xinput_uncorrelate_count;
Uint8 xinput_slot;
#endif
#ifdef SDL_JOYSTICK_RAWINPUT_WGI
SDL_bool wgi_correlated;
Uint8 wgi_correlation_id;
Uint8 wgi_correlation_count;
Uint8 wgi_uncorrelate_count;
WindowsGamingInputGamepadState *wgi_slot;
#endif
Fixed bug 4477 - Support more than 4 XInput-capable devices on Windows Jimb Esser Add new RawInput controller API, and improved correlation with XInput/WGI Reorder joystick init so drivers can ask the others if they handle a device reliably Do not poll disconnected XInput devices (major perf issue) Fix various cases where incorrect correlation could happen Simple mechanism for propagating unhandled Guide button presses even before guaranteed correlation Correlate by axis motion as well as button presses Fix failing to zero other trigger Fix SDL_HINT_JOYSTICK_HIDAPI not working if set before calling SDL_Init() Add missing device to device names Disable RawInput if we have a mismatch of XInput-capable but not RawInput-capable devices Updated to SDL 2.0.13 code with the following notes: New HID driver: xbox360w - no idea what that is, hopefully urelated SDL_hidapijoystick.c had been refactored to couple data handling logic with device opening logic and device lists caused some problems, yields slightly uglier integration than previously when the 360 HID device driver was just handling the data. SDL_hidapijoystick.c now often pulls the device off of the joystick_hwdata structure for some rumble logic, but it appears that code path is never reached, so probably not a problem. Looks like joystick_hwdata was refactored to not include a mutex in other drivers, maintainers may want to do the same refactor here if that's useful for some reason. Something changed in how devices get names, so getting generic names. Had to fix a (new?) bug where removing an XInput controller caused existing controllers (that moved to a new XInput index) to get identified as 0x045e/0x02fd ("it's probably Bluetooth" in code), rendering the existing HIDAPI_IsDevicePresent and new RAWINPUT_IsDevicePresent unreliable.
2020-03-16 14:25:02 -05:00
SDL_RAWINPUT_Device *device;
};
typedef struct joystick_hwdata RAWINPUT_DeviceContext;
Fixed bug 4477 - Support more than 4 XInput-capable devices on Windows Jimb Esser Add new RawInput controller API, and improved correlation with XInput/WGI Reorder joystick init so drivers can ask the others if they handle a device reliably Do not poll disconnected XInput devices (major perf issue) Fix various cases where incorrect correlation could happen Simple mechanism for propagating unhandled Guide button presses even before guaranteed correlation Correlate by axis motion as well as button presses Fix failing to zero other trigger Fix SDL_HINT_JOYSTICK_HIDAPI not working if set before calling SDL_Init() Add missing device to device names Disable RawInput if we have a mismatch of XInput-capable but not RawInput-capable devices Updated to SDL 2.0.13 code with the following notes: New HID driver: xbox360w - no idea what that is, hopefully urelated SDL_hidapijoystick.c had been refactored to couple data handling logic with device opening logic and device lists caused some problems, yields slightly uglier integration than previously when the 360 HID device driver was just handling the data. SDL_hidapijoystick.c now often pulls the device off of the joystick_hwdata structure for some rumble logic, but it appears that code path is never reached, so probably not a problem. Looks like joystick_hwdata was refactored to not include a mutex in other drivers, maintainers may want to do the same refactor here if that's useful for some reason. Something changed in how devices get names, so getting generic names. Had to fix a (new?) bug where removing an XInput controller caused existing controllers (that moved to a new XInput index) to get identified as 0x045e/0x02fd ("it's probably Bluetooth" in code), rendering the existing HIDAPI_IsDevicePresent and new RAWINPUT_IsDevicePresent unreliable.
2020-03-16 14:25:02 -05:00
SDL_RAWINPUT_Device *SDL_RAWINPUT_devices;
static const Uint16 subscribed_devices[] = {
USB_USAGE_GENERIC_GAMEPAD,
Fixed bug 4477 - Support more than 4 XInput-capable devices on Windows Jimb Esser Add new RawInput controller API, and improved correlation with XInput/WGI Reorder joystick init so drivers can ask the others if they handle a device reliably Do not poll disconnected XInput devices (major perf issue) Fix various cases where incorrect correlation could happen Simple mechanism for propagating unhandled Guide button presses even before guaranteed correlation Correlate by axis motion as well as button presses Fix failing to zero other trigger Fix SDL_HINT_JOYSTICK_HIDAPI not working if set before calling SDL_Init() Add missing device to device names Disable RawInput if we have a mismatch of XInput-capable but not RawInput-capable devices Updated to SDL 2.0.13 code with the following notes: New HID driver: xbox360w - no idea what that is, hopefully urelated SDL_hidapijoystick.c had been refactored to couple data handling logic with device opening logic and device lists caused some problems, yields slightly uglier integration than previously when the 360 HID device driver was just handling the data. SDL_hidapijoystick.c now often pulls the device off of the joystick_hwdata structure for some rumble logic, but it appears that code path is never reached, so probably not a problem. Looks like joystick_hwdata was refactored to not include a mutex in other drivers, maintainers may want to do the same refactor here if that's useful for some reason. Something changed in how devices get names, so getting generic names. Had to fix a (new?) bug where removing an XInput controller caused existing controllers (that moved to a new XInput index) to get identified as 0x045e/0x02fd ("it's probably Bluetooth" in code), rendering the existing HIDAPI_IsDevicePresent and new RAWINPUT_IsDevicePresent unreliable.
2020-03-16 14:25:02 -05:00
/* Don't need Joystick for any devices we're handling here (XInput-capable)
USB_USAGE_GENERIC_JOYSTICK,
USB_USAGE_GENERIC_MULTIAXISCONTROLLER,
Fixed bug 4477 - Support more than 4 XInput-capable devices on Windows Jimb Esser Add new RawInput controller API, and improved correlation with XInput/WGI Reorder joystick init so drivers can ask the others if they handle a device reliably Do not poll disconnected XInput devices (major perf issue) Fix various cases where incorrect correlation could happen Simple mechanism for propagating unhandled Guide button presses even before guaranteed correlation Correlate by axis motion as well as button presses Fix failing to zero other trigger Fix SDL_HINT_JOYSTICK_HIDAPI not working if set before calling SDL_Init() Add missing device to device names Disable RawInput if we have a mismatch of XInput-capable but not RawInput-capable devices Updated to SDL 2.0.13 code with the following notes: New HID driver: xbox360w - no idea what that is, hopefully urelated SDL_hidapijoystick.c had been refactored to couple data handling logic with device opening logic and device lists caused some problems, yields slightly uglier integration than previously when the 360 HID device driver was just handling the data. SDL_hidapijoystick.c now often pulls the device off of the joystick_hwdata structure for some rumble logic, but it appears that code path is never reached, so probably not a problem. Looks like joystick_hwdata was refactored to not include a mutex in other drivers, maintainers may want to do the same refactor here if that's useful for some reason. Something changed in how devices get names, so getting generic names. Had to fix a (new?) bug where removing an XInput controller caused existing controllers (that moved to a new XInput index) to get identified as 0x045e/0x02fd ("it's probably Bluetooth" in code), rendering the existing HIDAPI_IsDevicePresent and new RAWINPUT_IsDevicePresent unreliable.
2020-03-16 14:25:02 -05:00
*/
};
#ifdef SDL_JOYSTICK_RAWINPUT_MATCHING
static struct
{
Uint64 last_state_packet;
SDL_Joystick *joystick;
SDL_Joystick *last_joystick;
} guide_button_candidate;
typedef struct WindowsMatchState
{
#ifdef SDL_JOYSTICK_RAWINPUT_MATCH_AXES
SHORT match_axes[SDL_JOYSTICK_RAWINPUT_MATCH_COUNT];
#endif
#ifdef SDL_JOYSTICK_RAWINPUT_XINPUT
WORD xinput_buttons;
#endif
#ifdef SDL_JOYSTICK_RAWINPUT_WGI
Uint32 wgi_buttons;
#endif
SDL_bool any_data;
} WindowsMatchState;
Fixed bug 4477 - Support more than 4 XInput-capable devices on Windows Jimb Esser Add new RawInput controller API, and improved correlation with XInput/WGI Reorder joystick init so drivers can ask the others if they handle a device reliably Do not poll disconnected XInput devices (major perf issue) Fix various cases where incorrect correlation could happen Simple mechanism for propagating unhandled Guide button presses even before guaranteed correlation Correlate by axis motion as well as button presses Fix failing to zero other trigger Fix SDL_HINT_JOYSTICK_HIDAPI not working if set before calling SDL_Init() Add missing device to device names Disable RawInput if we have a mismatch of XInput-capable but not RawInput-capable devices Updated to SDL 2.0.13 code with the following notes: New HID driver: xbox360w - no idea what that is, hopefully urelated SDL_hidapijoystick.c had been refactored to couple data handling logic with device opening logic and device lists caused some problems, yields slightly uglier integration than previously when the 360 HID device driver was just handling the data. SDL_hidapijoystick.c now often pulls the device off of the joystick_hwdata structure for some rumble logic, but it appears that code path is never reached, so probably not a problem. Looks like joystick_hwdata was refactored to not include a mutex in other drivers, maintainers may want to do the same refactor here if that's useful for some reason. Something changed in how devices get names, so getting generic names. Had to fix a (new?) bug where removing an XInput controller caused existing controllers (that moved to a new XInput index) to get identified as 0x045e/0x02fd ("it's probably Bluetooth" in code), rendering the existing HIDAPI_IsDevicePresent and new RAWINPUT_IsDevicePresent unreliable.
2020-03-16 14:25:02 -05:00
static void RAWINPUT_FillMatchState(WindowsMatchState *state, Uint64 match_state)
{
#ifdef SDL_JOYSTICK_RAWINPUT_MATCH_AXES
int ii;
#endif
SDL_bool any_axes_data = SDL_FALSE;
#ifdef SDL_JOYSTICK_RAWINPUT_MATCH_AXES
/* SHORT state->match_axes[4] = {
(match_state & 0x000F0000) >> 4,
(match_state & 0x00F00000) >> 8,
(match_state & 0x0F000000) >> 12,
(match_state & 0xF0000000) >> 16,
}; */
for (ii = 0; ii < 4; ii++) {
state->match_axes[ii] = (SHORT)((match_state & (0x000F0000ull << (ii * 4))) >> (4 + ii * 4));
any_axes_data |= ((Uint32)(state->match_axes[ii] + 0x1000) > 0x2000); /* match_state bit is not 0xF, 0x1, or 0x2 */
Fixed bug 4477 - Support more than 4 XInput-capable devices on Windows Jimb Esser Add new RawInput controller API, and improved correlation with XInput/WGI Reorder joystick init so drivers can ask the others if they handle a device reliably Do not poll disconnected XInput devices (major perf issue) Fix various cases where incorrect correlation could happen Simple mechanism for propagating unhandled Guide button presses even before guaranteed correlation Correlate by axis motion as well as button presses Fix failing to zero other trigger Fix SDL_HINT_JOYSTICK_HIDAPI not working if set before calling SDL_Init() Add missing device to device names Disable RawInput if we have a mismatch of XInput-capable but not RawInput-capable devices Updated to SDL 2.0.13 code with the following notes: New HID driver: xbox360w - no idea what that is, hopefully urelated SDL_hidapijoystick.c had been refactored to couple data handling logic with device opening logic and device lists caused some problems, yields slightly uglier integration than previously when the 360 HID device driver was just handling the data. SDL_hidapijoystick.c now often pulls the device off of the joystick_hwdata structure for some rumble logic, but it appears that code path is never reached, so probably not a problem. Looks like joystick_hwdata was refactored to not include a mutex in other drivers, maintainers may want to do the same refactor here if that's useful for some reason. Something changed in how devices get names, so getting generic names. Had to fix a (new?) bug where removing an XInput controller caused existing controllers (that moved to a new XInput index) to get identified as 0x045e/0x02fd ("it's probably Bluetooth" in code), rendering the existing HIDAPI_IsDevicePresent and new RAWINPUT_IsDevicePresent unreliable.
2020-03-16 14:25:02 -05:00
}
#endif /* SDL_JOYSTICK_RAWINPUT_MATCH_AXES */
#ifdef SDL_JOYSTICK_RAWINPUT_MATCH_TRIGGERS
for (; ii < SDL_JOYSTICK_RAWINPUT_MATCH_COUNT; ii++) {
state->match_axes[ii] = (SHORT)((match_state & (0x000F0000ull << (ii * 4))) >> (4 + ii * 4));
any_axes_data |= (state->match_axes[ii] != SDL_MIN_SINT16);
}
#endif /* SDL_JOYSTICK_RAWINPUT_MATCH_TRIGGERS */
state->any_data = any_axes_data;
Fixed bug 4477 - Support more than 4 XInput-capable devices on Windows Jimb Esser Add new RawInput controller API, and improved correlation with XInput/WGI Reorder joystick init so drivers can ask the others if they handle a device reliably Do not poll disconnected XInput devices (major perf issue) Fix various cases where incorrect correlation could happen Simple mechanism for propagating unhandled Guide button presses even before guaranteed correlation Correlate by axis motion as well as button presses Fix failing to zero other trigger Fix SDL_HINT_JOYSTICK_HIDAPI not working if set before calling SDL_Init() Add missing device to device names Disable RawInput if we have a mismatch of XInput-capable but not RawInput-capable devices Updated to SDL 2.0.13 code with the following notes: New HID driver: xbox360w - no idea what that is, hopefully urelated SDL_hidapijoystick.c had been refactored to couple data handling logic with device opening logic and device lists caused some problems, yields slightly uglier integration than previously when the 360 HID device driver was just handling the data. SDL_hidapijoystick.c now often pulls the device off of the joystick_hwdata structure for some rumble logic, but it appears that code path is never reached, so probably not a problem. Looks like joystick_hwdata was refactored to not include a mutex in other drivers, maintainers may want to do the same refactor here if that's useful for some reason. Something changed in how devices get names, so getting generic names. Had to fix a (new?) bug where removing an XInput controller caused existing controllers (that moved to a new XInput index) to get identified as 0x045e/0x02fd ("it's probably Bluetooth" in code), rendering the existing HIDAPI_IsDevicePresent and new RAWINPUT_IsDevicePresent unreliable.
2020-03-16 14:25:02 -05:00
#ifdef SDL_JOYSTICK_RAWINPUT_XINPUT
/* Match axes by checking if the distance between the high 4 bits of axis and the 4 bits from match_state is 1 or less */
#define XInputAxesMatch(gamepad) ( \
(Uint32)(gamepad.sThumbLX - state->match_axes[0] + 0x1000) <= 0x2fff && \
(Uint32)(~gamepad.sThumbLY - state->match_axes[1] + 0x1000) <= 0x2fff && \
(Uint32)(gamepad.sThumbRX - state->match_axes[2] + 0x1000) <= 0x2fff && \
(Uint32)(~gamepad.sThumbRY - state->match_axes[3] + 0x1000) <= 0x2fff)
/* Explicit
#define XInputAxesMatch(gamepad) (\
SDL_abs((Sint8)((gamepad.sThumbLX & 0xF000) >> 8) - ((match_state & 0x000F0000) >> 12)) <= 0x10 && \
SDL_abs((Sint8)((~gamepad.sThumbLY & 0xF000) >> 8) - ((match_state & 0x00F00000) >> 16)) <= 0x10 && \
SDL_abs((Sint8)((gamepad.sThumbRX & 0xF000) >> 8) - ((match_state & 0x0F000000) >> 20)) <= 0x10 && \
SDL_abs((Sint8)((~gamepad.sThumbRY & 0xF000) >> 8) - ((match_state & 0xF0000000) >> 24)) <= 0x10) */
/* Can only match trigger values if a single trigger has a value. */
#define XInputTriggersMatch(gamepad) ( \
((state->match_axes[4] == SDL_MIN_SINT16) && (state->match_axes[5] == SDL_MIN_SINT16)) || \
((gamepad.bLeftTrigger != 0) && (gamepad.bRightTrigger != 0)) || \
((Uint32)((((int)gamepad.bLeftTrigger * 257) - 32768) - state->match_axes[4]) <= 0x2fff) || \
((Uint32)((((int)gamepad.bRightTrigger * 257) - 32768) - state->match_axes[5]) <= 0x2fff))
state->xinput_buttons =
/* Bitwise map .RLDUWVQTS.KYXBA -> YXBA..WVQTKSRLDU */
(WORD)(match_state << 12 | (match_state & 0x0780) >> 1 | (match_state & 0x0010) << 1 | (match_state & 0x0040) >> 2 | (match_state & 0x7800) >> 11);
/* Explicit
((match_state & (1<<SDL_GAMEPAD_BUTTON_A)) ? XINPUT_GAMEPAD_A : 0) |
((match_state & (1<<SDL_GAMEPAD_BUTTON_B)) ? XINPUT_GAMEPAD_B : 0) |
((match_state & (1<<SDL_GAMEPAD_BUTTON_X)) ? XINPUT_GAMEPAD_X : 0) |
((match_state & (1<<SDL_GAMEPAD_BUTTON_Y)) ? XINPUT_GAMEPAD_Y : 0) |
((match_state & (1<<SDL_GAMEPAD_BUTTON_BACK)) ? XINPUT_GAMEPAD_BACK : 0) |
((match_state & (1<<SDL_GAMEPAD_BUTTON_START)) ? XINPUT_GAMEPAD_START : 0) |
((match_state & (1<<SDL_GAMEPAD_BUTTON_LEFT_STICK)) ? XINPUT_GAMEPAD_LEFT_THUMB : 0) |
((match_state & (1<<SDL_GAMEPAD_BUTTON_RIGHT_STICK)) ? XINPUT_GAMEPAD_RIGHT_THUMB: 0) |
((match_state & (1<<SDL_GAMEPAD_BUTTON_LEFT_SHOULDER)) ? XINPUT_GAMEPAD_LEFT_SHOULDER : 0) |
((match_state & (1<<SDL_GAMEPAD_BUTTON_RIGHT_SHOULDER)) ? XINPUT_GAMEPAD_RIGHT_SHOULDER : 0) |
((match_state & (1<<SDL_GAMEPAD_BUTTON_DPAD_UP)) ? XINPUT_GAMEPAD_DPAD_UP : 0) |
((match_state & (1<<SDL_GAMEPAD_BUTTON_DPAD_DOWN)) ? XINPUT_GAMEPAD_DPAD_DOWN : 0) |
((match_state & (1<<SDL_GAMEPAD_BUTTON_DPAD_LEFT)) ? XINPUT_GAMEPAD_DPAD_LEFT : 0) |
((match_state & (1<<SDL_GAMEPAD_BUTTON_DPAD_RIGHT)) ? XINPUT_GAMEPAD_DPAD_RIGHT : 0);
*/
if (state->xinput_buttons) {
state->any_data = SDL_TRUE;
}
#endif
#ifdef SDL_JOYSTICK_RAWINPUT_WGI
/* Match axes by checking if the distance between the high 4 bits of axis and the 4 bits from match_state is 1 or less */
#define WindowsGamingInputAxesMatch(gamepad) ( \
(Uint16)(((Sint16)(gamepad.LeftThumbstickX * SDL_MAX_SINT16) & 0xF000) - state->match_axes[0] + 0x1000) <= 0x2fff && \
(Uint16)((~(Sint16)(gamepad.LeftThumbstickY * SDL_MAX_SINT16) & 0xF000) - state->match_axes[1] + 0x1000) <= 0x2fff && \
(Uint16)(((Sint16)(gamepad.RightThumbstickX * SDL_MAX_SINT16) & 0xF000) - state->match_axes[2] + 0x1000) <= 0x2fff && \
(Uint16)((~(Sint16)(gamepad.RightThumbstickY * SDL_MAX_SINT16) & 0xF000) - state->match_axes[3] + 0x1000) <= 0x2fff)
#define WindowsGamingInputTriggersMatch(gamepad) ( \
((state->match_axes[4] == SDL_MIN_SINT16) && (state->match_axes[5] == SDL_MIN_SINT16)) || \
((gamepad.LeftTrigger == 0.0f) && (gamepad.RightTrigger == 0.0f)) || \
((Uint16)((((int)(gamepad.LeftTrigger * SDL_MAX_UINT16)) - 32768) - state->match_axes[4]) <= 0x2fff) || \
((Uint16)((((int)(gamepad.RightTrigger * SDL_MAX_UINT16)) - 32768) - state->match_axes[5]) <= 0x2fff))
state->wgi_buttons =
/* Bitwise map .RLD UWVQ TS.K YXBA -> ..QT WVRL DUYX BAKS */
/* RStick/LStick (QT) RShould/LShould (WV) DPad R/L/D/U YXBA bac(K) (S)tart */
(match_state & 0x0180) << 5 | (match_state & 0x0600) << 1 | (match_state & 0x7800) >> 5 | (match_state & 0x000F) << 2 | (match_state & 0x0010) >> 3 | (match_state & 0x0040) >> 6;
/* Explicit
((match_state & (1<<SDL_GAMEPAD_BUTTON_A)) ? GamepadButtons_A : 0) |
((match_state & (1<<SDL_GAMEPAD_BUTTON_B)) ? GamepadButtons_B : 0) |
((match_state & (1<<SDL_GAMEPAD_BUTTON_X)) ? GamepadButtons_X : 0) |
((match_state & (1<<SDL_GAMEPAD_BUTTON_Y)) ? GamepadButtons_Y : 0) |
((match_state & (1<<SDL_GAMEPAD_BUTTON_BACK)) ? GamepadButtons_View : 0) |
((match_state & (1<<SDL_GAMEPAD_BUTTON_START)) ? GamepadButtons_Menu : 0) |
((match_state & (1<<SDL_GAMEPAD_BUTTON_LEFT_STICK)) ? GamepadButtons_LeftThumbstick : 0) |
((match_state & (1<<SDL_GAMEPAD_BUTTON_RIGHT_STICK)) ? GamepadButtons_RightThumbstick: 0) |
((match_state & (1<<SDL_GAMEPAD_BUTTON_LEFT_SHOULDER)) ? GamepadButtons_LeftShoulder: 0) |
((match_state & (1<<SDL_GAMEPAD_BUTTON_RIGHT_SHOULDER)) ? GamepadButtons_RightShoulder: 0) |
((match_state & (1<<SDL_GAMEPAD_BUTTON_DPAD_UP)) ? GamepadButtons_DPadUp : 0) |
((match_state & (1<<SDL_GAMEPAD_BUTTON_DPAD_DOWN)) ? GamepadButtons_DPadDown : 0) |
((match_state & (1<<SDL_GAMEPAD_BUTTON_DPAD_LEFT)) ? GamepadButtons_DPadLeft : 0) |
((match_state & (1<<SDL_GAMEPAD_BUTTON_DPAD_RIGHT)) ? GamepadButtons_DPadRight : 0); */
if (state->wgi_buttons) {
state->any_data = SDL_TRUE;
}
#endif
}
#endif /* SDL_JOYSTICK_RAWINPUT_MATCHING */
#ifdef SDL_JOYSTICK_RAWINPUT_XINPUT
static struct
{
XINPUT_STATE_EX state;
XINPUT_BATTERY_INFORMATION_EX battery;
SDL_bool connected; /* Currently has an active XInput device */
SDL_bool used; /* Is currently mapped to an SDL device */
Uint8 correlation_id;
} xinput_state[XUSER_MAX_COUNT];
static SDL_bool xinput_device_change = SDL_TRUE;
static SDL_bool xinput_state_dirty = SDL_TRUE;
static void RAWINPUT_UpdateXInput()
{
DWORD user_index;
if (xinput_device_change) {
for (user_index = 0; user_index < XUSER_MAX_COUNT; user_index++) {
XINPUT_CAPABILITIES capabilities;
xinput_state[user_index].connected = (XINPUTGETCAPABILITIES(user_index, XINPUT_FLAG_GAMEPAD, &capabilities) == ERROR_SUCCESS) ? SDL_TRUE : SDL_FALSE;
}
xinput_device_change = SDL_FALSE;
xinput_state_dirty = SDL_TRUE;
}
if (xinput_state_dirty) {
xinput_state_dirty = SDL_FALSE;
for (user_index = 0; user_index < SDL_arraysize(xinput_state); ++user_index) {
if (xinput_state[user_index].connected) {
if (XINPUTGETSTATE(user_index, &xinput_state[user_index].state) != ERROR_SUCCESS) {
xinput_state[user_index].connected = SDL_FALSE;
}
xinput_state[user_index].battery.BatteryType = BATTERY_TYPE_UNKNOWN;
if (XINPUTGETBATTERYINFORMATION) {
XINPUTGETBATTERYINFORMATION(user_index, BATTERY_DEVTYPE_GAMEPAD, &xinput_state[user_index].battery);
}
}
}
Fixed bug 4477 - Support more than 4 XInput-capable devices on Windows Jimb Esser Add new RawInput controller API, and improved correlation with XInput/WGI Reorder joystick init so drivers can ask the others if they handle a device reliably Do not poll disconnected XInput devices (major perf issue) Fix various cases where incorrect correlation could happen Simple mechanism for propagating unhandled Guide button presses even before guaranteed correlation Correlate by axis motion as well as button presses Fix failing to zero other trigger Fix SDL_HINT_JOYSTICK_HIDAPI not working if set before calling SDL_Init() Add missing device to device names Disable RawInput if we have a mismatch of XInput-capable but not RawInput-capable devices Updated to SDL 2.0.13 code with the following notes: New HID driver: xbox360w - no idea what that is, hopefully urelated SDL_hidapijoystick.c had been refactored to couple data handling logic with device opening logic and device lists caused some problems, yields slightly uglier integration than previously when the 360 HID device driver was just handling the data. SDL_hidapijoystick.c now often pulls the device off of the joystick_hwdata structure for some rumble logic, but it appears that code path is never reached, so probably not a problem. Looks like joystick_hwdata was refactored to not include a mutex in other drivers, maintainers may want to do the same refactor here if that's useful for some reason. Something changed in how devices get names, so getting generic names. Had to fix a (new?) bug where removing an XInput controller caused existing controllers (that moved to a new XInput index) to get identified as 0x045e/0x02fd ("it's probably Bluetooth" in code), rendering the existing HIDAPI_IsDevicePresent and new RAWINPUT_IsDevicePresent unreliable.
2020-03-16 14:25:02 -05:00
}
}
Fixed bug 4477 - Support more than 4 XInput-capable devices on Windows Jimb Esser Add new RawInput controller API, and improved correlation with XInput/WGI Reorder joystick init so drivers can ask the others if they handle a device reliably Do not poll disconnected XInput devices (major perf issue) Fix various cases where incorrect correlation could happen Simple mechanism for propagating unhandled Guide button presses even before guaranteed correlation Correlate by axis motion as well as button presses Fix failing to zero other trigger Fix SDL_HINT_JOYSTICK_HIDAPI not working if set before calling SDL_Init() Add missing device to device names Disable RawInput if we have a mismatch of XInput-capable but not RawInput-capable devices Updated to SDL 2.0.13 code with the following notes: New HID driver: xbox360w - no idea what that is, hopefully urelated SDL_hidapijoystick.c had been refactored to couple data handling logic with device opening logic and device lists caused some problems, yields slightly uglier integration than previously when the 360 HID device driver was just handling the data. SDL_hidapijoystick.c now often pulls the device off of the joystick_hwdata structure for some rumble logic, but it appears that code path is never reached, so probably not a problem. Looks like joystick_hwdata was refactored to not include a mutex in other drivers, maintainers may want to do the same refactor here if that's useful for some reason. Something changed in how devices get names, so getting generic names. Had to fix a (new?) bug where removing an XInput controller caused existing controllers (that moved to a new XInput index) to get identified as 0x045e/0x02fd ("it's probably Bluetooth" in code), rendering the existing HIDAPI_IsDevicePresent and new RAWINPUT_IsDevicePresent unreliable.
2020-03-16 14:25:02 -05:00
static void RAWINPUT_MarkXInputSlotUsed(Uint8 xinput_slot)
{
if (xinput_slot != XUSER_INDEX_ANY) {
xinput_state[xinput_slot].used = SDL_TRUE;
}
}
static void RAWINPUT_MarkXInputSlotFree(Uint8 xinput_slot)
{
if (xinput_slot != XUSER_INDEX_ANY) {
xinput_state[xinput_slot].used = SDL_FALSE;
}
}
static SDL_bool RAWINPUT_MissingXInputSlot()
{
int ii;
for (ii = 0; ii < SDL_arraysize(xinput_state); ii++) {
if (xinput_state[ii].connected && !xinput_state[ii].used) {
return SDL_TRUE;
}
Fixed bug 4477 - Support more than 4 XInput-capable devices on Windows Jimb Esser Add new RawInput controller API, and improved correlation with XInput/WGI Reorder joystick init so drivers can ask the others if they handle a device reliably Do not poll disconnected XInput devices (major perf issue) Fix various cases where incorrect correlation could happen Simple mechanism for propagating unhandled Guide button presses even before guaranteed correlation Correlate by axis motion as well as button presses Fix failing to zero other trigger Fix SDL_HINT_JOYSTICK_HIDAPI not working if set before calling SDL_Init() Add missing device to device names Disable RawInput if we have a mismatch of XInput-capable but not RawInput-capable devices Updated to SDL 2.0.13 code with the following notes: New HID driver: xbox360w - no idea what that is, hopefully urelated SDL_hidapijoystick.c had been refactored to couple data handling logic with device opening logic and device lists caused some problems, yields slightly uglier integration than previously when the 360 HID device driver was just handling the data. SDL_hidapijoystick.c now often pulls the device off of the joystick_hwdata structure for some rumble logic, but it appears that code path is never reached, so probably not a problem. Looks like joystick_hwdata was refactored to not include a mutex in other drivers, maintainers may want to do the same refactor here if that's useful for some reason. Something changed in how devices get names, so getting generic names. Had to fix a (new?) bug where removing an XInput controller caused existing controllers (that moved to a new XInput index) to get identified as 0x045e/0x02fd ("it's probably Bluetooth" in code), rendering the existing HIDAPI_IsDevicePresent and new RAWINPUT_IsDevicePresent unreliable.
2020-03-16 14:25:02 -05:00
}
return SDL_FALSE;
}
Fixed bug 4477 - Support more than 4 XInput-capable devices on Windows Jimb Esser Add new RawInput controller API, and improved correlation with XInput/WGI Reorder joystick init so drivers can ask the others if they handle a device reliably Do not poll disconnected XInput devices (major perf issue) Fix various cases where incorrect correlation could happen Simple mechanism for propagating unhandled Guide button presses even before guaranteed correlation Correlate by axis motion as well as button presses Fix failing to zero other trigger Fix SDL_HINT_JOYSTICK_HIDAPI not working if set before calling SDL_Init() Add missing device to device names Disable RawInput if we have a mismatch of XInput-capable but not RawInput-capable devices Updated to SDL 2.0.13 code with the following notes: New HID driver: xbox360w - no idea what that is, hopefully urelated SDL_hidapijoystick.c had been refactored to couple data handling logic with device opening logic and device lists caused some problems, yields slightly uglier integration than previously when the 360 HID device driver was just handling the data. SDL_hidapijoystick.c now often pulls the device off of the joystick_hwdata structure for some rumble logic, but it appears that code path is never reached, so probably not a problem. Looks like joystick_hwdata was refactored to not include a mutex in other drivers, maintainers may want to do the same refactor here if that's useful for some reason. Something changed in how devices get names, so getting generic names. Had to fix a (new?) bug where removing an XInput controller caused existing controllers (that moved to a new XInput index) to get identified as 0x045e/0x02fd ("it's probably Bluetooth" in code), rendering the existing HIDAPI_IsDevicePresent and new RAWINPUT_IsDevicePresent unreliable.
2020-03-16 14:25:02 -05:00
static SDL_bool RAWINPUT_XInputSlotMatches(const WindowsMatchState *state, Uint8 slot_idx)
{
if (xinput_state[slot_idx].connected) {
WORD xinput_buttons = xinput_state[slot_idx].state.Gamepad.wButtons;
if ((xinput_buttons & ~XINPUT_GAMEPAD_GUIDE) == state->xinput_buttons
#ifdef SDL_JOYSTICK_RAWINPUT_MATCH_AXES
&& XInputAxesMatch(xinput_state[slot_idx].state.Gamepad)
#endif
#ifdef SDL_JOYSTICK_RAWINPUT_MATCH_TRIGGERS
&& XInputTriggersMatch(xinput_state[slot_idx].state.Gamepad)
#endif
) {
return SDL_TRUE;
}
}
return SDL_FALSE;
}
Fixed bug 4477 - Support more than 4 XInput-capable devices on Windows Jimb Esser Add new RawInput controller API, and improved correlation with XInput/WGI Reorder joystick init so drivers can ask the others if they handle a device reliably Do not poll disconnected XInput devices (major perf issue) Fix various cases where incorrect correlation could happen Simple mechanism for propagating unhandled Guide button presses even before guaranteed correlation Correlate by axis motion as well as button presses Fix failing to zero other trigger Fix SDL_HINT_JOYSTICK_HIDAPI not working if set before calling SDL_Init() Add missing device to device names Disable RawInput if we have a mismatch of XInput-capable but not RawInput-capable devices Updated to SDL 2.0.13 code with the following notes: New HID driver: xbox360w - no idea what that is, hopefully urelated SDL_hidapijoystick.c had been refactored to couple data handling logic with device opening logic and device lists caused some problems, yields slightly uglier integration than previously when the 360 HID device driver was just handling the data. SDL_hidapijoystick.c now often pulls the device off of the joystick_hwdata structure for some rumble logic, but it appears that code path is never reached, so probably not a problem. Looks like joystick_hwdata was refactored to not include a mutex in other drivers, maintainers may want to do the same refactor here if that's useful for some reason. Something changed in how devices get names, so getting generic names. Had to fix a (new?) bug where removing an XInput controller caused existing controllers (that moved to a new XInput index) to get identified as 0x045e/0x02fd ("it's probably Bluetooth" in code), rendering the existing HIDAPI_IsDevicePresent and new RAWINPUT_IsDevicePresent unreliable.
2020-03-16 14:25:02 -05:00
static SDL_bool RAWINPUT_GuessXInputSlot(const WindowsMatchState *state, Uint8 *correlation_id, Uint8 *slot_idx)
{
int user_index;
int match_count;
*slot_idx = 0;
match_count = 0;
for (user_index = 0; user_index < XUSER_MAX_COUNT; ++user_index) {
if (!xinput_state[user_index].used && RAWINPUT_XInputSlotMatches(state, user_index)) {
++match_count;
*slot_idx = (Uint8)user_index;
/* Incrementing correlation_id for any match, as negative evidence for others being correlated */
*correlation_id = ++xinput_state[user_index].correlation_id;
}
}
/* Only return a match if we match exactly one, and we have some non-zero data (buttons or axes) that matched.
Note that we're still invalidating *other* potential correlations if we have more than one match or we have no
data. */
if (match_count == 1 && state->any_data) {
return SDL_TRUE;
}
return SDL_FALSE;
}
#endif /* SDL_JOYSTICK_RAWINPUT_XINPUT */
#ifdef SDL_JOYSTICK_RAWINPUT_WGI
typedef struct WindowsGamingInputGamepadState
{
__x_ABI_CWindows_CGaming_CInput_CIGamepad *gamepad;
struct __x_ABI_CWindows_CGaming_CInput_CGamepadReading state;
RAWINPUT_DeviceContext *correlated_context;
SDL_bool used; /* Is currently mapped to an SDL device */
SDL_bool connected; /* Just used during update to track disconnected */
Uint8 correlation_id;
struct __x_ABI_CWindows_CGaming_CInput_CGamepadVibration vibration;
} WindowsGamingInputGamepadState;
static struct
{
WindowsGamingInputGamepadState **per_gamepad;
int per_gamepad_count;
SDL_bool initialized;
SDL_bool dirty;
SDL_bool need_device_list_update;
int ref_count;
__x_ABI_CWindows_CGaming_CInput_CIGamepadStatics *gamepad_statics;
} wgi_state;
static void RAWINPUT_MarkWindowsGamingInputSlotUsed(WindowsGamingInputGamepadState *wgi_slot, RAWINPUT_DeviceContext *ctx)
{
wgi_slot->used = SDL_TRUE;
wgi_slot->correlated_context = ctx;
}
static void RAWINPUT_MarkWindowsGamingInputSlotFree(WindowsGamingInputGamepadState *wgi_slot)
{
wgi_slot->used = SDL_FALSE;
wgi_slot->correlated_context = NULL;
}
static SDL_bool RAWINPUT_MissingWindowsGamingInputSlot()
{
int ii;
for (ii = 0; ii < wgi_state.per_gamepad_count; ii++) {
if (!wgi_state.per_gamepad[ii]->used) {
return SDL_TRUE;
}
}
return SDL_FALSE;
}
static int RAWINPUT_UpdateWindowsGamingInput()
{
int ii;
if (!wgi_state.gamepad_statics) {
return 0;
}
if (!wgi_state.dirty) {
return 0;
}
wgi_state.dirty = SDL_FALSE;
if (wgi_state.need_device_list_update) {
HRESULT hr;
__FIVectorView_1_Windows__CGaming__CInput__CGamepad *gamepads;
wgi_state.need_device_list_update = SDL_FALSE;
for (ii = 0; ii < wgi_state.per_gamepad_count; ii++) {
wgi_state.per_gamepad[ii]->connected = SDL_FALSE;
}
hr = __x_ABI_CWindows_CGaming_CInput_CIGamepadStatics_get_Gamepads(wgi_state.gamepad_statics, &gamepads);
if (SUCCEEDED(hr)) {
unsigned int num_gamepads;
hr = __FIVectorView_1_Windows__CGaming__CInput__CGamepad_get_Size(gamepads, &num_gamepads);
if (SUCCEEDED(hr)) {
unsigned int i;
for (i = 0; i < num_gamepads; ++i) {
__x_ABI_CWindows_CGaming_CInput_CIGamepad *gamepad;
hr = __FIVectorView_1_Windows__CGaming__CInput__CGamepad_GetAt(gamepads, i, &gamepad);
if (SUCCEEDED(hr)) {
SDL_bool found = SDL_FALSE;
int jj;
for (jj = 0; jj < wgi_state.per_gamepad_count; jj++) {
if (wgi_state.per_gamepad[jj]->gamepad == gamepad) {
found = SDL_TRUE;
wgi_state.per_gamepad[jj]->connected = SDL_TRUE;
break;
}
}
if (!found) {
/* New device, add it */
WindowsGamingInputGamepadState *gamepad_state;
wgi_state.per_gamepad_count++;
wgi_state.per_gamepad = SDL_realloc(wgi_state.per_gamepad, sizeof(wgi_state.per_gamepad[0]) * wgi_state.per_gamepad_count);
if (!wgi_state.per_gamepad) {
return SDL_OutOfMemory();
}
gamepad_state = SDL_calloc(1, sizeof(*gamepad_state));
if (gamepad_state == NULL) {
return SDL_OutOfMemory();
}
wgi_state.per_gamepad[wgi_state.per_gamepad_count - 1] = gamepad_state;
gamepad_state->gamepad = gamepad;
gamepad_state->connected = SDL_TRUE;
} else {
/* Already tracked */
__x_ABI_CWindows_CGaming_CInput_CIGamepad_Release(gamepad);
}
}
}
for (ii = wgi_state.per_gamepad_count - 1; ii >= 0; ii--) {
WindowsGamingInputGamepadState *gamepad_state = wgi_state.per_gamepad[ii];
if (!gamepad_state->connected) {
/* Device missing, must be disconnected */
if (gamepad_state->correlated_context) {
gamepad_state->correlated_context->wgi_correlated = SDL_FALSE;
gamepad_state->correlated_context->wgi_slot = NULL;
}
__x_ABI_CWindows_CGaming_CInput_CIGamepad_Release(gamepad_state->gamepad);
SDL_free(gamepad_state);
wgi_state.per_gamepad[ii] = wgi_state.per_gamepad[wgi_state.per_gamepad_count - 1];
--wgi_state.per_gamepad_count;
}
}
Fixed bug 4477 - Support more than 4 XInput-capable devices on Windows Jimb Esser Add new RawInput controller API, and improved correlation with XInput/WGI Reorder joystick init so drivers can ask the others if they handle a device reliably Do not poll disconnected XInput devices (major perf issue) Fix various cases where incorrect correlation could happen Simple mechanism for propagating unhandled Guide button presses even before guaranteed correlation Correlate by axis motion as well as button presses Fix failing to zero other trigger Fix SDL_HINT_JOYSTICK_HIDAPI not working if set before calling SDL_Init() Add missing device to device names Disable RawInput if we have a mismatch of XInput-capable but not RawInput-capable devices Updated to SDL 2.0.13 code with the following notes: New HID driver: xbox360w - no idea what that is, hopefully urelated SDL_hidapijoystick.c had been refactored to couple data handling logic with device opening logic and device lists caused some problems, yields slightly uglier integration than previously when the 360 HID device driver was just handling the data. SDL_hidapijoystick.c now often pulls the device off of the joystick_hwdata structure for some rumble logic, but it appears that code path is never reached, so probably not a problem. Looks like joystick_hwdata was refactored to not include a mutex in other drivers, maintainers may want to do the same refactor here if that's useful for some reason. Something changed in how devices get names, so getting generic names. Had to fix a (new?) bug where removing an XInput controller caused existing controllers (that moved to a new XInput index) to get identified as 0x045e/0x02fd ("it's probably Bluetooth" in code), rendering the existing HIDAPI_IsDevicePresent and new RAWINPUT_IsDevicePresent unreliable.
2020-03-16 14:25:02 -05:00
}
__FIVectorView_1_Windows__CGaming__CInput__CGamepad_Release(gamepads);
}
} /* need_device_list_update */
for (ii = 0; ii < wgi_state.per_gamepad_count; ii++) {
HRESULT hr = __x_ABI_CWindows_CGaming_CInput_CIGamepad_GetCurrentReading(wgi_state.per_gamepad[ii]->gamepad, &wgi_state.per_gamepad[ii]->state);
if (!SUCCEEDED(hr)) {
wgi_state.per_gamepad[ii]->connected = SDL_FALSE; /* Not used by anything, currently */
Fixed bug 4477 - Support more than 4 XInput-capable devices on Windows Jimb Esser Add new RawInput controller API, and improved correlation with XInput/WGI Reorder joystick init so drivers can ask the others if they handle a device reliably Do not poll disconnected XInput devices (major perf issue) Fix various cases where incorrect correlation could happen Simple mechanism for propagating unhandled Guide button presses even before guaranteed correlation Correlate by axis motion as well as button presses Fix failing to zero other trigger Fix SDL_HINT_JOYSTICK_HIDAPI not working if set before calling SDL_Init() Add missing device to device names Disable RawInput if we have a mismatch of XInput-capable but not RawInput-capable devices Updated to SDL 2.0.13 code with the following notes: New HID driver: xbox360w - no idea what that is, hopefully urelated SDL_hidapijoystick.c had been refactored to couple data handling logic with device opening logic and device lists caused some problems, yields slightly uglier integration than previously when the 360 HID device driver was just handling the data. SDL_hidapijoystick.c now often pulls the device off of the joystick_hwdata structure for some rumble logic, but it appears that code path is never reached, so probably not a problem. Looks like joystick_hwdata was refactored to not include a mutex in other drivers, maintainers may want to do the same refactor here if that's useful for some reason. Something changed in how devices get names, so getting generic names. Had to fix a (new?) bug where removing an XInput controller caused existing controllers (that moved to a new XInput index) to get identified as 0x045e/0x02fd ("it's probably Bluetooth" in code), rendering the existing HIDAPI_IsDevicePresent and new RAWINPUT_IsDevicePresent unreliable.
2020-03-16 14:25:02 -05:00
}
}
return 0;
}
static void RAWINPUT_InitWindowsGamingInput(RAWINPUT_DeviceContext *ctx)
{
wgi_state.need_device_list_update = SDL_TRUE;
wgi_state.ref_count++;
if (!wgi_state.initialized) {
static const IID SDL_IID_IGamepadStatics = { 0x8BBCE529, 0xD49C, 0x39E9, { 0x95, 0x60, 0xE4, 0x7D, 0xDE, 0x96, 0xB7, 0xC8 } };
HRESULT hr;
if (FAILED(WIN_RoInitialize())) {
return;
}
wgi_state.initialized = SDL_TRUE;
wgi_state.dirty = SDL_TRUE;
{
typedef HRESULT(WINAPI * WindowsCreateStringReference_t)(PCWSTR sourceString, UINT32 length, HSTRING_HEADER * hstringHeader, HSTRING * string);
typedef HRESULT(WINAPI * RoGetActivationFactory_t)(HSTRING activatableClassId, REFIID iid, void **factory);
#ifdef __WINRT__
WindowsCreateStringReference_t WindowsCreateStringReferenceFunc = WindowsCreateStringReference;
RoGetActivationFactory_t RoGetActivationFactoryFunc = RoGetActivationFactory;
#else
WindowsCreateStringReference_t WindowsCreateStringReferenceFunc = (WindowsCreateStringReference_t)WIN_LoadComBaseFunction("WindowsCreateStringReference");
RoGetActivationFactory_t RoGetActivationFactoryFunc = (RoGetActivationFactory_t)WIN_LoadComBaseFunction("RoGetActivationFactory");
#endif
if (WindowsCreateStringReferenceFunc && RoGetActivationFactoryFunc) {
PCWSTR pNamespace = L"Windows.Gaming.Input.Gamepad";
HSTRING_HEADER hNamespaceStringHeader;
HSTRING hNamespaceString;
hr = WindowsCreateStringReferenceFunc(pNamespace, (UINT32)SDL_wcslen(pNamespace), &hNamespaceStringHeader, &hNamespaceString);
if (SUCCEEDED(hr)) {
2021-11-26 12:18:08 -06:00
RoGetActivationFactoryFunc(hNamespaceString, &SDL_IID_IGamepadStatics, (void **)&wgi_state.gamepad_statics);
}
}
}
Fixed bug 4477 - Support more than 4 XInput-capable devices on Windows Jimb Esser Add new RawInput controller API, and improved correlation with XInput/WGI Reorder joystick init so drivers can ask the others if they handle a device reliably Do not poll disconnected XInput devices (major perf issue) Fix various cases where incorrect correlation could happen Simple mechanism for propagating unhandled Guide button presses even before guaranteed correlation Correlate by axis motion as well as button presses Fix failing to zero other trigger Fix SDL_HINT_JOYSTICK_HIDAPI not working if set before calling SDL_Init() Add missing device to device names Disable RawInput if we have a mismatch of XInput-capable but not RawInput-capable devices Updated to SDL 2.0.13 code with the following notes: New HID driver: xbox360w - no idea what that is, hopefully urelated SDL_hidapijoystick.c had been refactored to couple data handling logic with device opening logic and device lists caused some problems, yields slightly uglier integration than previously when the 360 HID device driver was just handling the data. SDL_hidapijoystick.c now often pulls the device off of the joystick_hwdata structure for some rumble logic, but it appears that code path is never reached, so probably not a problem. Looks like joystick_hwdata was refactored to not include a mutex in other drivers, maintainers may want to do the same refactor here if that's useful for some reason. Something changed in how devices get names, so getting generic names. Had to fix a (new?) bug where removing an XInput controller caused existing controllers (that moved to a new XInput index) to get identified as 0x045e/0x02fd ("it's probably Bluetooth" in code), rendering the existing HIDAPI_IsDevicePresent and new RAWINPUT_IsDevicePresent unreliable.
2020-03-16 14:25:02 -05:00
}
}
static SDL_bool RAWINPUT_WindowsGamingInputSlotMatches(const WindowsMatchState *state, WindowsGamingInputGamepadState *slot, SDL_bool xinput_correlated)
{
Uint32 wgi_buttons = slot->state.Buttons;
if ((wgi_buttons & 0x3FFF) == state->wgi_buttons
#ifdef SDL_JOYSTICK_RAWINPUT_MATCH_AXES
&& WindowsGamingInputAxesMatch(slot->state)
#endif
#ifdef SDL_JOYSTICK_RAWINPUT_MATCH_TRIGGERS
// Don't try to match WGI triggers if getting values from XInput
&& (xinput_correlated || WindowsGamingInputTriggersMatch(slot->state))
#endif
) {
return SDL_TRUE;
}
return SDL_FALSE;
}
static SDL_bool RAWINPUT_GuessWindowsGamingInputSlot(const WindowsMatchState *state, Uint8 *correlation_id, WindowsGamingInputGamepadState **slot, SDL_bool xinput_correlated)
{
int match_count, user_index;
match_count = 0;
for (user_index = 0; user_index < wgi_state.per_gamepad_count; ++user_index) {
WindowsGamingInputGamepadState *gamepad_state = wgi_state.per_gamepad[user_index];
if (RAWINPUT_WindowsGamingInputSlotMatches(state, gamepad_state, xinput_correlated)) {
++match_count;
*slot = gamepad_state;
/* Incrementing correlation_id for any match, as negative evidence for others being correlated */
*correlation_id = ++gamepad_state->correlation_id;
}
}
/* Only return a match if we match exactly one, and we have some non-zero data (buttons or axes) that matched.
Note that we're still invalidating *other* potential correlations if we have more than one match or we have no
data. */
if (match_count == 1 && state->any_data) {
return SDL_TRUE;
}
return SDL_FALSE;
}
static void RAWINPUT_QuitWindowsGamingInput(RAWINPUT_DeviceContext *ctx)
{
wgi_state.need_device_list_update = SDL_TRUE;
--wgi_state.ref_count;
if (!wgi_state.ref_count && wgi_state.initialized) {
int ii;
for (ii = 0; ii < wgi_state.per_gamepad_count; ii++) {
__x_ABI_CWindows_CGaming_CInput_CIGamepad_Release(wgi_state.per_gamepad[ii]->gamepad);
}
if (wgi_state.per_gamepad) {
SDL_free(wgi_state.per_gamepad);
wgi_state.per_gamepad = NULL;
}
wgi_state.per_gamepad_count = 0;
if (wgi_state.gamepad_statics) {
__x_ABI_CWindows_CGaming_CInput_CIGamepadStatics_Release(wgi_state.gamepad_statics);
wgi_state.gamepad_statics = NULL;
}
WIN_RoUninitialize();
wgi_state.initialized = SDL_FALSE;
}
}
#endif /* SDL_JOYSTICK_RAWINPUT_WGI */
static SDL_RAWINPUT_Device *RAWINPUT_AcquireDevice(SDL_RAWINPUT_Device *device)
{
SDL_AtomicIncRef(&device->refcount);
return device;
}
static void RAWINPUT_ReleaseDevice(SDL_RAWINPUT_Device *device)
{
#ifdef SDL_JOYSTICK_RAWINPUT_XINPUT
if (device->joystick) {
RAWINPUT_DeviceContext *ctx = device->joystick->hwdata;
if (ctx->xinput_enabled && ctx->xinput_correlated) {
RAWINPUT_MarkXInputSlotFree(ctx->xinput_slot);
ctx->xinput_correlated = SDL_FALSE;
}
}
#endif /* SDL_JOYSTICK_RAWINPUT_XINPUT */
if (SDL_AtomicDecRef(&device->refcount)) {
SDL_free(device->preparsed_data);
SDL_free(device->name);
SDL_free(device->path);
SDL_free(device);
}
}
static SDL_RAWINPUT_Device *RAWINPUT_DeviceFromHandle(HANDLE hDevice)
Fixed bug 4477 - Support more than 4 XInput-capable devices on Windows Jimb Esser Add new RawInput controller API, and improved correlation with XInput/WGI Reorder joystick init so drivers can ask the others if they handle a device reliably Do not poll disconnected XInput devices (major perf issue) Fix various cases where incorrect correlation could happen Simple mechanism for propagating unhandled Guide button presses even before guaranteed correlation Correlate by axis motion as well as button presses Fix failing to zero other trigger Fix SDL_HINT_JOYSTICK_HIDAPI not working if set before calling SDL_Init() Add missing device to device names Disable RawInput if we have a mismatch of XInput-capable but not RawInput-capable devices Updated to SDL 2.0.13 code with the following notes: New HID driver: xbox360w - no idea what that is, hopefully urelated SDL_hidapijoystick.c had been refactored to couple data handling logic with device opening logic and device lists caused some problems, yields slightly uglier integration than previously when the 360 HID device driver was just handling the data. SDL_hidapijoystick.c now often pulls the device off of the joystick_hwdata structure for some rumble logic, but it appears that code path is never reached, so probably not a problem. Looks like joystick_hwdata was refactored to not include a mutex in other drivers, maintainers may want to do the same refactor here if that's useful for some reason. Something changed in how devices get names, so getting generic names. Had to fix a (new?) bug where removing an XInput controller caused existing controllers (that moved to a new XInput index) to get identified as 0x045e/0x02fd ("it's probably Bluetooth" in code), rendering the existing HIDAPI_IsDevicePresent and new RAWINPUT_IsDevicePresent unreliable.
2020-03-16 14:25:02 -05:00
{
SDL_RAWINPUT_Device *curr;
for (curr = SDL_RAWINPUT_devices; curr; curr = curr->next) {
if (curr->hDevice == hDevice) {
Fixed bug 4477 - Support more than 4 XInput-capable devices on Windows Jimb Esser Add new RawInput controller API, and improved correlation with XInput/WGI Reorder joystick init so drivers can ask the others if they handle a device reliably Do not poll disconnected XInput devices (major perf issue) Fix various cases where incorrect correlation could happen Simple mechanism for propagating unhandled Guide button presses even before guaranteed correlation Correlate by axis motion as well as button presses Fix failing to zero other trigger Fix SDL_HINT_JOYSTICK_HIDAPI not working if set before calling SDL_Init() Add missing device to device names Disable RawInput if we have a mismatch of XInput-capable but not RawInput-capable devices Updated to SDL 2.0.13 code with the following notes: New HID driver: xbox360w - no idea what that is, hopefully urelated SDL_hidapijoystick.c had been refactored to couple data handling logic with device opening logic and device lists caused some problems, yields slightly uglier integration than previously when the 360 HID device driver was just handling the data. SDL_hidapijoystick.c now often pulls the device off of the joystick_hwdata structure for some rumble logic, but it appears that code path is never reached, so probably not a problem. Looks like joystick_hwdata was refactored to not include a mutex in other drivers, maintainers may want to do the same refactor here if that's useful for some reason. Something changed in how devices get names, so getting generic names. Had to fix a (new?) bug where removing an XInput controller caused existing controllers (that moved to a new XInput index) to get identified as 0x045e/0x02fd ("it's probably Bluetooth" in code), rendering the existing HIDAPI_IsDevicePresent and new RAWINPUT_IsDevicePresent unreliable.
2020-03-16 14:25:02 -05:00
return curr;
}
Fixed bug 4477 - Support more than 4 XInput-capable devices on Windows Jimb Esser Add new RawInput controller API, and improved correlation with XInput/WGI Reorder joystick init so drivers can ask the others if they handle a device reliably Do not poll disconnected XInput devices (major perf issue) Fix various cases where incorrect correlation could happen Simple mechanism for propagating unhandled Guide button presses even before guaranteed correlation Correlate by axis motion as well as button presses Fix failing to zero other trigger Fix SDL_HINT_JOYSTICK_HIDAPI not working if set before calling SDL_Init() Add missing device to device names Disable RawInput if we have a mismatch of XInput-capable but not RawInput-capable devices Updated to SDL 2.0.13 code with the following notes: New HID driver: xbox360w - no idea what that is, hopefully urelated SDL_hidapijoystick.c had been refactored to couple data handling logic with device opening logic and device lists caused some problems, yields slightly uglier integration than previously when the 360 HID device driver was just handling the data. SDL_hidapijoystick.c now often pulls the device off of the joystick_hwdata structure for some rumble logic, but it appears that code path is never reached, so probably not a problem. Looks like joystick_hwdata was refactored to not include a mutex in other drivers, maintainers may want to do the same refactor here if that's useful for some reason. Something changed in how devices get names, so getting generic names. Had to fix a (new?) bug where removing an XInput controller caused existing controllers (that moved to a new XInput index) to get identified as 0x045e/0x02fd ("it's probably Bluetooth" in code), rendering the existing HIDAPI_IsDevicePresent and new RAWINPUT_IsDevicePresent unreliable.
2020-03-16 14:25:02 -05:00
}
return NULL;
}
static void RAWINPUT_AddDevice(HANDLE hDevice)
Fixed bug 4477 - Support more than 4 XInput-capable devices on Windows Jimb Esser Add new RawInput controller API, and improved correlation with XInput/WGI Reorder joystick init so drivers can ask the others if they handle a device reliably Do not poll disconnected XInput devices (major perf issue) Fix various cases where incorrect correlation could happen Simple mechanism for propagating unhandled Guide button presses even before guaranteed correlation Correlate by axis motion as well as button presses Fix failing to zero other trigger Fix SDL_HINT_JOYSTICK_HIDAPI not working if set before calling SDL_Init() Add missing device to device names Disable RawInput if we have a mismatch of XInput-capable but not RawInput-capable devices Updated to SDL 2.0.13 code with the following notes: New HID driver: xbox360w - no idea what that is, hopefully urelated SDL_hidapijoystick.c had been refactored to couple data handling logic with device opening logic and device lists caused some problems, yields slightly uglier integration than previously when the 360 HID device driver was just handling the data. SDL_hidapijoystick.c now often pulls the device off of the joystick_hwdata structure for some rumble logic, but it appears that code path is never reached, so probably not a problem. Looks like joystick_hwdata was refactored to not include a mutex in other drivers, maintainers may want to do the same refactor here if that's useful for some reason. Something changed in how devices get names, so getting generic names. Had to fix a (new?) bug where removing an XInput controller caused existing controllers (that moved to a new XInput index) to get identified as 0x045e/0x02fd ("it's probably Bluetooth" in code), rendering the existing HIDAPI_IsDevicePresent and new RAWINPUT_IsDevicePresent unreliable.
2020-03-16 14:25:02 -05:00
{
#define CHECK(expression) \
{ \
if (!(expression)) \
goto err; \
}
Fixed bug 4477 - Support more than 4 XInput-capable devices on Windows Jimb Esser Add new RawInput controller API, and improved correlation with XInput/WGI Reorder joystick init so drivers can ask the others if they handle a device reliably Do not poll disconnected XInput devices (major perf issue) Fix various cases where incorrect correlation could happen Simple mechanism for propagating unhandled Guide button presses even before guaranteed correlation Correlate by axis motion as well as button presses Fix failing to zero other trigger Fix SDL_HINT_JOYSTICK_HIDAPI not working if set before calling SDL_Init() Add missing device to device names Disable RawInput if we have a mismatch of XInput-capable but not RawInput-capable devices Updated to SDL 2.0.13 code with the following notes: New HID driver: xbox360w - no idea what that is, hopefully urelated SDL_hidapijoystick.c had been refactored to couple data handling logic with device opening logic and device lists caused some problems, yields slightly uglier integration than previously when the 360 HID device driver was just handling the data. SDL_hidapijoystick.c now often pulls the device off of the joystick_hwdata structure for some rumble logic, but it appears that code path is never reached, so probably not a problem. Looks like joystick_hwdata was refactored to not include a mutex in other drivers, maintainers may want to do the same refactor here if that's useful for some reason. Something changed in how devices get names, so getting generic names. Had to fix a (new?) bug where removing an XInput controller caused existing controllers (that moved to a new XInput index) to get identified as 0x045e/0x02fd ("it's probably Bluetooth" in code), rendering the existing HIDAPI_IsDevicePresent and new RAWINPUT_IsDevicePresent unreliable.
2020-03-16 14:25:02 -05:00
SDL_RAWINPUT_Device *device = NULL;
SDL_RAWINPUT_Device *curr, *last;
Fixed bug 4477 - Support more than 4 XInput-capable devices on Windows Jimb Esser Add new RawInput controller API, and improved correlation with XInput/WGI Reorder joystick init so drivers can ask the others if they handle a device reliably Do not poll disconnected XInput devices (major perf issue) Fix various cases where incorrect correlation could happen Simple mechanism for propagating unhandled Guide button presses even before guaranteed correlation Correlate by axis motion as well as button presses Fix failing to zero other trigger Fix SDL_HINT_JOYSTICK_HIDAPI not working if set before calling SDL_Init() Add missing device to device names Disable RawInput if we have a mismatch of XInput-capable but not RawInput-capable devices Updated to SDL 2.0.13 code with the following notes: New HID driver: xbox360w - no idea what that is, hopefully urelated SDL_hidapijoystick.c had been refactored to couple data handling logic with device opening logic and device lists caused some problems, yields slightly uglier integration than previously when the 360 HID device driver was just handling the data. SDL_hidapijoystick.c now often pulls the device off of the joystick_hwdata structure for some rumble logic, but it appears that code path is never reached, so probably not a problem. Looks like joystick_hwdata was refactored to not include a mutex in other drivers, maintainers may want to do the same refactor here if that's useful for some reason. Something changed in how devices get names, so getting generic names. Had to fix a (new?) bug where removing an XInput controller caused existing controllers (that moved to a new XInput index) to get identified as 0x045e/0x02fd ("it's probably Bluetooth" in code), rendering the existing HIDAPI_IsDevicePresent and new RAWINPUT_IsDevicePresent unreliable.
2020-03-16 14:25:02 -05:00
RID_DEVICE_INFO rdi;
UINT size;
char dev_name[MAX_PATH];
HANDLE hFile = INVALID_HANDLE_VALUE;
Fixed bug 4477 - Support more than 4 XInput-capable devices on Windows Jimb Esser Add new RawInput controller API, and improved correlation with XInput/WGI Reorder joystick init so drivers can ask the others if they handle a device reliably Do not poll disconnected XInput devices (major perf issue) Fix various cases where incorrect correlation could happen Simple mechanism for propagating unhandled Guide button presses even before guaranteed correlation Correlate by axis motion as well as button presses Fix failing to zero other trigger Fix SDL_HINT_JOYSTICK_HIDAPI not working if set before calling SDL_Init() Add missing device to device names Disable RawInput if we have a mismatch of XInput-capable but not RawInput-capable devices Updated to SDL 2.0.13 code with the following notes: New HID driver: xbox360w - no idea what that is, hopefully urelated SDL_hidapijoystick.c had been refactored to couple data handling logic with device opening logic and device lists caused some problems, yields slightly uglier integration than previously when the 360 HID device driver was just handling the data. SDL_hidapijoystick.c now often pulls the device off of the joystick_hwdata structure for some rumble logic, but it appears that code path is never reached, so probably not a problem. Looks like joystick_hwdata was refactored to not include a mutex in other drivers, maintainers may want to do the same refactor here if that's useful for some reason. Something changed in how devices get names, so getting generic names. Had to fix a (new?) bug where removing an XInput controller caused existing controllers (that moved to a new XInput index) to get identified as 0x045e/0x02fd ("it's probably Bluetooth" in code), rendering the existing HIDAPI_IsDevicePresent and new RAWINPUT_IsDevicePresent unreliable.
2020-03-16 14:25:02 -05:00
/* Make sure we're not trying to add the same device twice */
if (RAWINPUT_DeviceFromHandle(hDevice)) {
return;
}
Fixed bug 4477 - Support more than 4 XInput-capable devices on Windows Jimb Esser Add new RawInput controller API, and improved correlation with XInput/WGI Reorder joystick init so drivers can ask the others if they handle a device reliably Do not poll disconnected XInput devices (major perf issue) Fix various cases where incorrect correlation could happen Simple mechanism for propagating unhandled Guide button presses even before guaranteed correlation Correlate by axis motion as well as button presses Fix failing to zero other trigger Fix SDL_HINT_JOYSTICK_HIDAPI not working if set before calling SDL_Init() Add missing device to device names Disable RawInput if we have a mismatch of XInput-capable but not RawInput-capable devices Updated to SDL 2.0.13 code with the following notes: New HID driver: xbox360w - no idea what that is, hopefully urelated SDL_hidapijoystick.c had been refactored to couple data handling logic with device opening logic and device lists caused some problems, yields slightly uglier integration than previously when the 360 HID device driver was just handling the data. SDL_hidapijoystick.c now often pulls the device off of the joystick_hwdata structure for some rumble logic, but it appears that code path is never reached, so probably not a problem. Looks like joystick_hwdata was refactored to not include a mutex in other drivers, maintainers may want to do the same refactor here if that's useful for some reason. Something changed in how devices get names, so getting generic names. Had to fix a (new?) bug where removing an XInput controller caused existing controllers (that moved to a new XInput index) to get identified as 0x045e/0x02fd ("it's probably Bluetooth" in code), rendering the existing HIDAPI_IsDevicePresent and new RAWINPUT_IsDevicePresent unreliable.
2020-03-16 14:25:02 -05:00
/* Figure out what kind of device it is */
size = sizeof(rdi);
CHECK(GetRawInputDeviceInfoA(hDevice, RIDI_DEVICEINFO, &rdi, &size) != (UINT)-1);
Fixed bug 4477 - Support more than 4 XInput-capable devices on Windows Jimb Esser Add new RawInput controller API, and improved correlation with XInput/WGI Reorder joystick init so drivers can ask the others if they handle a device reliably Do not poll disconnected XInput devices (major perf issue) Fix various cases where incorrect correlation could happen Simple mechanism for propagating unhandled Guide button presses even before guaranteed correlation Correlate by axis motion as well as button presses Fix failing to zero other trigger Fix SDL_HINT_JOYSTICK_HIDAPI not working if set before calling SDL_Init() Add missing device to device names Disable RawInput if we have a mismatch of XInput-capable but not RawInput-capable devices Updated to SDL 2.0.13 code with the following notes: New HID driver: xbox360w - no idea what that is, hopefully urelated SDL_hidapijoystick.c had been refactored to couple data handling logic with device opening logic and device lists caused some problems, yields slightly uglier integration than previously when the 360 HID device driver was just handling the data. SDL_hidapijoystick.c now often pulls the device off of the joystick_hwdata structure for some rumble logic, but it appears that code path is never reached, so probably not a problem. Looks like joystick_hwdata was refactored to not include a mutex in other drivers, maintainers may want to do the same refactor here if that's useful for some reason. Something changed in how devices get names, so getting generic names. Had to fix a (new?) bug where removing an XInput controller caused existing controllers (that moved to a new XInput index) to get identified as 0x045e/0x02fd ("it's probably Bluetooth" in code), rendering the existing HIDAPI_IsDevicePresent and new RAWINPUT_IsDevicePresent unreliable.
2020-03-16 14:25:02 -05:00
CHECK(rdi.dwType == RIM_TYPEHID);
/* Get the device "name" (HID Path) */
size = SDL_arraysize(dev_name);
CHECK(GetRawInputDeviceInfoA(hDevice, RIDI_DEVICENAME, dev_name, &size) != (UINT)-1);
Fixed bug 4477 - Support more than 4 XInput-capable devices on Windows Jimb Esser Add new RawInput controller API, and improved correlation with XInput/WGI Reorder joystick init so drivers can ask the others if they handle a device reliably Do not poll disconnected XInput devices (major perf issue) Fix various cases where incorrect correlation could happen Simple mechanism for propagating unhandled Guide button presses even before guaranteed correlation Correlate by axis motion as well as button presses Fix failing to zero other trigger Fix SDL_HINT_JOYSTICK_HIDAPI not working if set before calling SDL_Init() Add missing device to device names Disable RawInput if we have a mismatch of XInput-capable but not RawInput-capable devices Updated to SDL 2.0.13 code with the following notes: New HID driver: xbox360w - no idea what that is, hopefully urelated SDL_hidapijoystick.c had been refactored to couple data handling logic with device opening logic and device lists caused some problems, yields slightly uglier integration than previously when the 360 HID device driver was just handling the data. SDL_hidapijoystick.c now often pulls the device off of the joystick_hwdata structure for some rumble logic, but it appears that code path is never reached, so probably not a problem. Looks like joystick_hwdata was refactored to not include a mutex in other drivers, maintainers may want to do the same refactor here if that's useful for some reason. Something changed in how devices get names, so getting generic names. Had to fix a (new?) bug where removing an XInput controller caused existing controllers (that moved to a new XInput index) to get identified as 0x045e/0x02fd ("it's probably Bluetooth" in code), rendering the existing HIDAPI_IsDevicePresent and new RAWINPUT_IsDevicePresent unreliable.
2020-03-16 14:25:02 -05:00
/* Only take XInput-capable devices */
CHECK(SDL_strstr(dev_name, "IG_") != NULL);
#ifdef SDL_JOYSTICK_HIDAPI
/* Don't take devices handled by HIDAPI */
CHECK(!HIDAPI_IsDevicePresent((Uint16)rdi.hid.dwVendorId, (Uint16)rdi.hid.dwProductId, (Uint16)rdi.hid.dwVersionNumber, ""));
#endif
2022-12-01 15:07:03 -06:00
device = (SDL_RAWINPUT_Device *)SDL_calloc(1, sizeof(SDL_RAWINPUT_Device));
CHECK(device);
Fixed bug 4477 - Support more than 4 XInput-capable devices on Windows Jimb Esser Add new RawInput controller API, and improved correlation with XInput/WGI Reorder joystick init so drivers can ask the others if they handle a device reliably Do not poll disconnected XInput devices (major perf issue) Fix various cases where incorrect correlation could happen Simple mechanism for propagating unhandled Guide button presses even before guaranteed correlation Correlate by axis motion as well as button presses Fix failing to zero other trigger Fix SDL_HINT_JOYSTICK_HIDAPI not working if set before calling SDL_Init() Add missing device to device names Disable RawInput if we have a mismatch of XInput-capable but not RawInput-capable devices Updated to SDL 2.0.13 code with the following notes: New HID driver: xbox360w - no idea what that is, hopefully urelated SDL_hidapijoystick.c had been refactored to couple data handling logic with device opening logic and device lists caused some problems, yields slightly uglier integration than previously when the 360 HID device driver was just handling the data. SDL_hidapijoystick.c now often pulls the device off of the joystick_hwdata structure for some rumble logic, but it appears that code path is never reached, so probably not a problem. Looks like joystick_hwdata was refactored to not include a mutex in other drivers, maintainers may want to do the same refactor here if that's useful for some reason. Something changed in how devices get names, so getting generic names. Had to fix a (new?) bug where removing an XInput controller caused existing controllers (that moved to a new XInput index) to get identified as 0x045e/0x02fd ("it's probably Bluetooth" in code), rendering the existing HIDAPI_IsDevicePresent and new RAWINPUT_IsDevicePresent unreliable.
2020-03-16 14:25:02 -05:00
device->hDevice = hDevice;
device->vendor_id = (Uint16)rdi.hid.dwVendorId;
device->product_id = (Uint16)rdi.hid.dwProductId;
device->version = (Uint16)rdi.hid.dwVersionNumber;
device->is_xinput = SDL_TRUE;
device->is_xboxone = SDL_IsJoystickXboxOne(device->vendor_id, device->product_id);
Fixed bug 4477 - Support more than 4 XInput-capable devices on Windows Jimb Esser Add new RawInput controller API, and improved correlation with XInput/WGI Reorder joystick init so drivers can ask the others if they handle a device reliably Do not poll disconnected XInput devices (major perf issue) Fix various cases where incorrect correlation could happen Simple mechanism for propagating unhandled Guide button presses even before guaranteed correlation Correlate by axis motion as well as button presses Fix failing to zero other trigger Fix SDL_HINT_JOYSTICK_HIDAPI not working if set before calling SDL_Init() Add missing device to device names Disable RawInput if we have a mismatch of XInput-capable but not RawInput-capable devices Updated to SDL 2.0.13 code with the following notes: New HID driver: xbox360w - no idea what that is, hopefully urelated SDL_hidapijoystick.c had been refactored to couple data handling logic with device opening logic and device lists caused some problems, yields slightly uglier integration than previously when the 360 HID device driver was just handling the data. SDL_hidapijoystick.c now often pulls the device off of the joystick_hwdata structure for some rumble logic, but it appears that code path is never reached, so probably not a problem. Looks like joystick_hwdata was refactored to not include a mutex in other drivers, maintainers may want to do the same refactor here if that's useful for some reason. Something changed in how devices get names, so getting generic names. Had to fix a (new?) bug where removing an XInput controller caused existing controllers (that moved to a new XInput index) to get identified as 0x045e/0x02fd ("it's probably Bluetooth" in code), rendering the existing HIDAPI_IsDevicePresent and new RAWINPUT_IsDevicePresent unreliable.
2020-03-16 14:25:02 -05:00
/* Get HID Top-Level Collection Preparsed Data */
size = 0;
CHECK(GetRawInputDeviceInfoA(hDevice, RIDI_PREPARSEDDATA, NULL, &size) != (UINT)-1);
2022-12-01 15:07:03 -06:00
device->preparsed_data = (PHIDP_PREPARSED_DATA)SDL_calloc(size, sizeof(BYTE));
CHECK(device->preparsed_data);
CHECK(GetRawInputDeviceInfoA(hDevice, RIDI_PREPARSEDDATA, device->preparsed_data, &size) != (UINT)-1);
hFile = CreateFileA(dev_name, GENERIC_READ, FILE_SHARE_READ | FILE_SHARE_WRITE, NULL, OPEN_EXISTING, 0, NULL);
CHECK(hFile != INVALID_HANDLE_VALUE);
Fixed bug 4477 - Support more than 4 XInput-capable devices on Windows Jimb Esser Add new RawInput controller API, and improved correlation with XInput/WGI Reorder joystick init so drivers can ask the others if they handle a device reliably Do not poll disconnected XInput devices (major perf issue) Fix various cases where incorrect correlation could happen Simple mechanism for propagating unhandled Guide button presses even before guaranteed correlation Correlate by axis motion as well as button presses Fix failing to zero other trigger Fix SDL_HINT_JOYSTICK_HIDAPI not working if set before calling SDL_Init() Add missing device to device names Disable RawInput if we have a mismatch of XInput-capable but not RawInput-capable devices Updated to SDL 2.0.13 code with the following notes: New HID driver: xbox360w - no idea what that is, hopefully urelated SDL_hidapijoystick.c had been refactored to couple data handling logic with device opening logic and device lists caused some problems, yields slightly uglier integration than previously when the 360 HID device driver was just handling the data. SDL_hidapijoystick.c now often pulls the device off of the joystick_hwdata structure for some rumble logic, but it appears that code path is never reached, so probably not a problem. Looks like joystick_hwdata was refactored to not include a mutex in other drivers, maintainers may want to do the same refactor here if that's useful for some reason. Something changed in how devices get names, so getting generic names. Had to fix a (new?) bug where removing an XInput controller caused existing controllers (that moved to a new XInput index) to get identified as 0x045e/0x02fd ("it's probably Bluetooth" in code), rendering the existing HIDAPI_IsDevicePresent and new RAWINPUT_IsDevicePresent unreliable.
2020-03-16 14:25:02 -05:00
{
char *manufacturer_string = NULL;
char *product_string = NULL;
WCHAR string[128];
if (SDL_HidD_GetManufacturerString(hFile, string, sizeof(string))) {
manufacturer_string = WIN_StringToUTF8W(string);
}
if (SDL_HidD_GetProductString(hFile, string, sizeof(string))) {
product_string = WIN_StringToUTF8W(string);
}
device->name = SDL_CreateJoystickName(device->vendor_id, device->product_id, manufacturer_string, product_string);
if (manufacturer_string) {
SDL_free(manufacturer_string);
}
if (product_string) {
SDL_free(product_string);
}
Fixed bug 4477 - Support more than 4 XInput-capable devices on Windows Jimb Esser Add new RawInput controller API, and improved correlation with XInput/WGI Reorder joystick init so drivers can ask the others if they handle a device reliably Do not poll disconnected XInput devices (major perf issue) Fix various cases where incorrect correlation could happen Simple mechanism for propagating unhandled Guide button presses even before guaranteed correlation Correlate by axis motion as well as button presses Fix failing to zero other trigger Fix SDL_HINT_JOYSTICK_HIDAPI not working if set before calling SDL_Init() Add missing device to device names Disable RawInput if we have a mismatch of XInput-capable but not RawInput-capable devices Updated to SDL 2.0.13 code with the following notes: New HID driver: xbox360w - no idea what that is, hopefully urelated SDL_hidapijoystick.c had been refactored to couple data handling logic with device opening logic and device lists caused some problems, yields slightly uglier integration than previously when the 360 HID device driver was just handling the data. SDL_hidapijoystick.c now often pulls the device off of the joystick_hwdata structure for some rumble logic, but it appears that code path is never reached, so probably not a problem. Looks like joystick_hwdata was refactored to not include a mutex in other drivers, maintainers may want to do the same refactor here if that's useful for some reason. Something changed in how devices get names, so getting generic names. Had to fix a (new?) bug where removing an XInput controller caused existing controllers (that moved to a new XInput index) to get identified as 0x045e/0x02fd ("it's probably Bluetooth" in code), rendering the existing HIDAPI_IsDevicePresent and new RAWINPUT_IsDevicePresent unreliable.
2020-03-16 14:25:02 -05:00
}
2022-08-22 21:28:21 -05:00
device->guid = SDL_CreateJoystickGUID(SDL_HARDWARE_BUS_USB, device->vendor_id, device->product_id, device->version, device->name, 'r', 0);
device->path = SDL_strdup(dev_name);
Fixed bug 4477 - Support more than 4 XInput-capable devices on Windows Jimb Esser Add new RawInput controller API, and improved correlation with XInput/WGI Reorder joystick init so drivers can ask the others if they handle a device reliably Do not poll disconnected XInput devices (major perf issue) Fix various cases where incorrect correlation could happen Simple mechanism for propagating unhandled Guide button presses even before guaranteed correlation Correlate by axis motion as well as button presses Fix failing to zero other trigger Fix SDL_HINT_JOYSTICK_HIDAPI not working if set before calling SDL_Init() Add missing device to device names Disable RawInput if we have a mismatch of XInput-capable but not RawInput-capable devices Updated to SDL 2.0.13 code with the following notes: New HID driver: xbox360w - no idea what that is, hopefully urelated SDL_hidapijoystick.c had been refactored to couple data handling logic with device opening logic and device lists caused some problems, yields slightly uglier integration than previously when the 360 HID device driver was just handling the data. SDL_hidapijoystick.c now often pulls the device off of the joystick_hwdata structure for some rumble logic, but it appears that code path is never reached, so probably not a problem. Looks like joystick_hwdata was refactored to not include a mutex in other drivers, maintainers may want to do the same refactor here if that's useful for some reason. Something changed in how devices get names, so getting generic names. Had to fix a (new?) bug where removing an XInput controller caused existing controllers (that moved to a new XInput index) to get identified as 0x045e/0x02fd ("it's probably Bluetooth" in code), rendering the existing HIDAPI_IsDevicePresent and new RAWINPUT_IsDevicePresent unreliable.
2020-03-16 14:25:02 -05:00
CloseHandle(hFile);
hFile = INVALID_HANDLE_VALUE;
device->joystick_id = SDL_GetNextJoystickInstanceID();
Fixed bug 4477 - Support more than 4 XInput-capable devices on Windows Jimb Esser Add new RawInput controller API, and improved correlation with XInput/WGI Reorder joystick init so drivers can ask the others if they handle a device reliably Do not poll disconnected XInput devices (major perf issue) Fix various cases where incorrect correlation could happen Simple mechanism for propagating unhandled Guide button presses even before guaranteed correlation Correlate by axis motion as well as button presses Fix failing to zero other trigger Fix SDL_HINT_JOYSTICK_HIDAPI not working if set before calling SDL_Init() Add missing device to device names Disable RawInput if we have a mismatch of XInput-capable but not RawInput-capable devices Updated to SDL 2.0.13 code with the following notes: New HID driver: xbox360w - no idea what that is, hopefully urelated SDL_hidapijoystick.c had been refactored to couple data handling logic with device opening logic and device lists caused some problems, yields slightly uglier integration than previously when the 360 HID device driver was just handling the data. SDL_hidapijoystick.c now often pulls the device off of the joystick_hwdata structure for some rumble logic, but it appears that code path is never reached, so probably not a problem. Looks like joystick_hwdata was refactored to not include a mutex in other drivers, maintainers may want to do the same refactor here if that's useful for some reason. Something changed in how devices get names, so getting generic names. Had to fix a (new?) bug where removing an XInput controller caused existing controllers (that moved to a new XInput index) to get identified as 0x045e/0x02fd ("it's probably Bluetooth" in code), rendering the existing HIDAPI_IsDevicePresent and new RAWINPUT_IsDevicePresent unreliable.
2020-03-16 14:25:02 -05:00
#ifdef DEBUG_RAWINPUT
SDL_Log("Adding RAWINPUT device '%s' VID 0x%.4x, PID 0x%.4x, version %d, handle 0x%.8x\n", device->name, device->vendor_id, device->product_id, device->version, device->hDevice);
#endif
/* Add it to the list */
RAWINPUT_AcquireDevice(device);
Fixed bug 4477 - Support more than 4 XInput-capable devices on Windows Jimb Esser Add new RawInput controller API, and improved correlation with XInput/WGI Reorder joystick init so drivers can ask the others if they handle a device reliably Do not poll disconnected XInput devices (major perf issue) Fix various cases where incorrect correlation could happen Simple mechanism for propagating unhandled Guide button presses even before guaranteed correlation Correlate by axis motion as well as button presses Fix failing to zero other trigger Fix SDL_HINT_JOYSTICK_HIDAPI not working if set before calling SDL_Init() Add missing device to device names Disable RawInput if we have a mismatch of XInput-capable but not RawInput-capable devices Updated to SDL 2.0.13 code with the following notes: New HID driver: xbox360w - no idea what that is, hopefully urelated SDL_hidapijoystick.c had been refactored to couple data handling logic with device opening logic and device lists caused some problems, yields slightly uglier integration than previously when the 360 HID device driver was just handling the data. SDL_hidapijoystick.c now often pulls the device off of the joystick_hwdata structure for some rumble logic, but it appears that code path is never reached, so probably not a problem. Looks like joystick_hwdata was refactored to not include a mutex in other drivers, maintainers may want to do the same refactor here if that's useful for some reason. Something changed in how devices get names, so getting generic names. Had to fix a (new?) bug where removing an XInput controller caused existing controllers (that moved to a new XInput index) to get identified as 0x045e/0x02fd ("it's probably Bluetooth" in code), rendering the existing HIDAPI_IsDevicePresent and new RAWINPUT_IsDevicePresent unreliable.
2020-03-16 14:25:02 -05:00
for (curr = SDL_RAWINPUT_devices, last = NULL; curr; last = curr, curr = curr->next) {
}
if (last) {
last->next = device;
} else {
SDL_RAWINPUT_devices = device;
}
++SDL_RAWINPUT_numjoysticks;
SDL_PrivateJoystickAdded(device->joystick_id);
Fixed bug 4477 - Support more than 4 XInput-capable devices on Windows Jimb Esser Add new RawInput controller API, and improved correlation with XInput/WGI Reorder joystick init so drivers can ask the others if they handle a device reliably Do not poll disconnected XInput devices (major perf issue) Fix various cases where incorrect correlation could happen Simple mechanism for propagating unhandled Guide button presses even before guaranteed correlation Correlate by axis motion as well as button presses Fix failing to zero other trigger Fix SDL_HINT_JOYSTICK_HIDAPI not working if set before calling SDL_Init() Add missing device to device names Disable RawInput if we have a mismatch of XInput-capable but not RawInput-capable devices Updated to SDL 2.0.13 code with the following notes: New HID driver: xbox360w - no idea what that is, hopefully urelated SDL_hidapijoystick.c had been refactored to couple data handling logic with device opening logic and device lists caused some problems, yields slightly uglier integration than previously when the 360 HID device driver was just handling the data. SDL_hidapijoystick.c now often pulls the device off of the joystick_hwdata structure for some rumble logic, but it appears that code path is never reached, so probably not a problem. Looks like joystick_hwdata was refactored to not include a mutex in other drivers, maintainers may want to do the same refactor here if that's useful for some reason. Something changed in how devices get names, so getting generic names. Had to fix a (new?) bug where removing an XInput controller caused existing controllers (that moved to a new XInput index) to get identified as 0x045e/0x02fd ("it's probably Bluetooth" in code), rendering the existing HIDAPI_IsDevicePresent and new RAWINPUT_IsDevicePresent unreliable.
2020-03-16 14:25:02 -05:00
return;
err:
if (hFile != INVALID_HANDLE_VALUE) {
CloseHandle(hFile);
}
Fixed bug 4477 - Support more than 4 XInput-capable devices on Windows Jimb Esser Add new RawInput controller API, and improved correlation with XInput/WGI Reorder joystick init so drivers can ask the others if they handle a device reliably Do not poll disconnected XInput devices (major perf issue) Fix various cases where incorrect correlation could happen Simple mechanism for propagating unhandled Guide button presses even before guaranteed correlation Correlate by axis motion as well as button presses Fix failing to zero other trigger Fix SDL_HINT_JOYSTICK_HIDAPI not working if set before calling SDL_Init() Add missing device to device names Disable RawInput if we have a mismatch of XInput-capable but not RawInput-capable devices Updated to SDL 2.0.13 code with the following notes: New HID driver: xbox360w - no idea what that is, hopefully urelated SDL_hidapijoystick.c had been refactored to couple data handling logic with device opening logic and device lists caused some problems, yields slightly uglier integration than previously when the 360 HID device driver was just handling the data. SDL_hidapijoystick.c now often pulls the device off of the joystick_hwdata structure for some rumble logic, but it appears that code path is never reached, so probably not a problem. Looks like joystick_hwdata was refactored to not include a mutex in other drivers, maintainers may want to do the same refactor here if that's useful for some reason. Something changed in how devices get names, so getting generic names. Had to fix a (new?) bug where removing an XInput controller caused existing controllers (that moved to a new XInput index) to get identified as 0x045e/0x02fd ("it's probably Bluetooth" in code), rendering the existing HIDAPI_IsDevicePresent and new RAWINPUT_IsDevicePresent unreliable.
2020-03-16 14:25:02 -05:00
if (device) {
if (device->name) {
Fixed bug 4477 - Support more than 4 XInput-capable devices on Windows Jimb Esser Add new RawInput controller API, and improved correlation with XInput/WGI Reorder joystick init so drivers can ask the others if they handle a device reliably Do not poll disconnected XInput devices (major perf issue) Fix various cases where incorrect correlation could happen Simple mechanism for propagating unhandled Guide button presses even before guaranteed correlation Correlate by axis motion as well as button presses Fix failing to zero other trigger Fix SDL_HINT_JOYSTICK_HIDAPI not working if set before calling SDL_Init() Add missing device to device names Disable RawInput if we have a mismatch of XInput-capable but not RawInput-capable devices Updated to SDL 2.0.13 code with the following notes: New HID driver: xbox360w - no idea what that is, hopefully urelated SDL_hidapijoystick.c had been refactored to couple data handling logic with device opening logic and device lists caused some problems, yields slightly uglier integration than previously when the 360 HID device driver was just handling the data. SDL_hidapijoystick.c now often pulls the device off of the joystick_hwdata structure for some rumble logic, but it appears that code path is never reached, so probably not a problem. Looks like joystick_hwdata was refactored to not include a mutex in other drivers, maintainers may want to do the same refactor here if that's useful for some reason. Something changed in how devices get names, so getting generic names. Had to fix a (new?) bug where removing an XInput controller caused existing controllers (that moved to a new XInput index) to get identified as 0x045e/0x02fd ("it's probably Bluetooth" in code), rendering the existing HIDAPI_IsDevicePresent and new RAWINPUT_IsDevicePresent unreliable.
2020-03-16 14:25:02 -05:00
SDL_free(device->name);
}
if (device->path) {
SDL_free(device->path);
}
Fixed bug 4477 - Support more than 4 XInput-capable devices on Windows Jimb Esser Add new RawInput controller API, and improved correlation with XInput/WGI Reorder joystick init so drivers can ask the others if they handle a device reliably Do not poll disconnected XInput devices (major perf issue) Fix various cases where incorrect correlation could happen Simple mechanism for propagating unhandled Guide button presses even before guaranteed correlation Correlate by axis motion as well as button presses Fix failing to zero other trigger Fix SDL_HINT_JOYSTICK_HIDAPI not working if set before calling SDL_Init() Add missing device to device names Disable RawInput if we have a mismatch of XInput-capable but not RawInput-capable devices Updated to SDL 2.0.13 code with the following notes: New HID driver: xbox360w - no idea what that is, hopefully urelated SDL_hidapijoystick.c had been refactored to couple data handling logic with device opening logic and device lists caused some problems, yields slightly uglier integration than previously when the 360 HID device driver was just handling the data. SDL_hidapijoystick.c now often pulls the device off of the joystick_hwdata structure for some rumble logic, but it appears that code path is never reached, so probably not a problem. Looks like joystick_hwdata was refactored to not include a mutex in other drivers, maintainers may want to do the same refactor here if that's useful for some reason. Something changed in how devices get names, so getting generic names. Had to fix a (new?) bug where removing an XInput controller caused existing controllers (that moved to a new XInput index) to get identified as 0x045e/0x02fd ("it's probably Bluetooth" in code), rendering the existing HIDAPI_IsDevicePresent and new RAWINPUT_IsDevicePresent unreliable.
2020-03-16 14:25:02 -05:00
SDL_free(device);
}
#undef CHECK
Fixed bug 4477 - Support more than 4 XInput-capable devices on Windows Jimb Esser Add new RawInput controller API, and improved correlation with XInput/WGI Reorder joystick init so drivers can ask the others if they handle a device reliably Do not poll disconnected XInput devices (major perf issue) Fix various cases where incorrect correlation could happen Simple mechanism for propagating unhandled Guide button presses even before guaranteed correlation Correlate by axis motion as well as button presses Fix failing to zero other trigger Fix SDL_HINT_JOYSTICK_HIDAPI not working if set before calling SDL_Init() Add missing device to device names Disable RawInput if we have a mismatch of XInput-capable but not RawInput-capable devices Updated to SDL 2.0.13 code with the following notes: New HID driver: xbox360w - no idea what that is, hopefully urelated SDL_hidapijoystick.c had been refactored to couple data handling logic with device opening logic and device lists caused some problems, yields slightly uglier integration than previously when the 360 HID device driver was just handling the data. SDL_hidapijoystick.c now often pulls the device off of the joystick_hwdata structure for some rumble logic, but it appears that code path is never reached, so probably not a problem. Looks like joystick_hwdata was refactored to not include a mutex in other drivers, maintainers may want to do the same refactor here if that's useful for some reason. Something changed in how devices get names, so getting generic names. Had to fix a (new?) bug where removing an XInput controller caused existing controllers (that moved to a new XInput index) to get identified as 0x045e/0x02fd ("it's probably Bluetooth" in code), rendering the existing HIDAPI_IsDevicePresent and new RAWINPUT_IsDevicePresent unreliable.
2020-03-16 14:25:02 -05:00
}
static void RAWINPUT_DelDevice(SDL_RAWINPUT_Device *device, SDL_bool send_event)
Fixed bug 4477 - Support more than 4 XInput-capable devices on Windows Jimb Esser Add new RawInput controller API, and improved correlation with XInput/WGI Reorder joystick init so drivers can ask the others if they handle a device reliably Do not poll disconnected XInput devices (major perf issue) Fix various cases where incorrect correlation could happen Simple mechanism for propagating unhandled Guide button presses even before guaranteed correlation Correlate by axis motion as well as button presses Fix failing to zero other trigger Fix SDL_HINT_JOYSTICK_HIDAPI not working if set before calling SDL_Init() Add missing device to device names Disable RawInput if we have a mismatch of XInput-capable but not RawInput-capable devices Updated to SDL 2.0.13 code with the following notes: New HID driver: xbox360w - no idea what that is, hopefully urelated SDL_hidapijoystick.c had been refactored to couple data handling logic with device opening logic and device lists caused some problems, yields slightly uglier integration than previously when the 360 HID device driver was just handling the data. SDL_hidapijoystick.c now often pulls the device off of the joystick_hwdata structure for some rumble logic, but it appears that code path is never reached, so probably not a problem. Looks like joystick_hwdata was refactored to not include a mutex in other drivers, maintainers may want to do the same refactor here if that's useful for some reason. Something changed in how devices get names, so getting generic names. Had to fix a (new?) bug where removing an XInput controller caused existing controllers (that moved to a new XInput index) to get identified as 0x045e/0x02fd ("it's probably Bluetooth" in code), rendering the existing HIDAPI_IsDevicePresent and new RAWINPUT_IsDevicePresent unreliable.
2020-03-16 14:25:02 -05:00
{
SDL_RAWINPUT_Device *curr, *last;
for (curr = SDL_RAWINPUT_devices, last = NULL; curr; last = curr, curr = curr->next) {
if (curr == device) {
if (last) {
last->next = curr->next;
} else {
SDL_RAWINPUT_devices = curr->next;
}
--SDL_RAWINPUT_numjoysticks;
SDL_PrivateJoystickRemoved(device->joystick_id);
Fixed bug 4477 - Support more than 4 XInput-capable devices on Windows Jimb Esser Add new RawInput controller API, and improved correlation with XInput/WGI Reorder joystick init so drivers can ask the others if they handle a device reliably Do not poll disconnected XInput devices (major perf issue) Fix various cases where incorrect correlation could happen Simple mechanism for propagating unhandled Guide button presses even before guaranteed correlation Correlate by axis motion as well as button presses Fix failing to zero other trigger Fix SDL_HINT_JOYSTICK_HIDAPI not working if set before calling SDL_Init() Add missing device to device names Disable RawInput if we have a mismatch of XInput-capable but not RawInput-capable devices Updated to SDL 2.0.13 code with the following notes: New HID driver: xbox360w - no idea what that is, hopefully urelated SDL_hidapijoystick.c had been refactored to couple data handling logic with device opening logic and device lists caused some problems, yields slightly uglier integration than previously when the 360 HID device driver was just handling the data. SDL_hidapijoystick.c now often pulls the device off of the joystick_hwdata structure for some rumble logic, but it appears that code path is never reached, so probably not a problem. Looks like joystick_hwdata was refactored to not include a mutex in other drivers, maintainers may want to do the same refactor here if that's useful for some reason. Something changed in how devices get names, so getting generic names. Had to fix a (new?) bug where removing an XInput controller caused existing controllers (that moved to a new XInput index) to get identified as 0x045e/0x02fd ("it's probably Bluetooth" in code), rendering the existing HIDAPI_IsDevicePresent and new RAWINPUT_IsDevicePresent unreliable.
2020-03-16 14:25:02 -05:00
#ifdef DEBUG_RAWINPUT
2020-12-09 09:50:15 -06:00
SDL_Log("Removing RAWINPUT device '%s' VID 0x%.4x, PID 0x%.4x, version %d, handle %p\n", device->name, device->vendor_id, device->product_id, device->version, device->hDevice);
Fixed bug 4477 - Support more than 4 XInput-capable devices on Windows Jimb Esser Add new RawInput controller API, and improved correlation with XInput/WGI Reorder joystick init so drivers can ask the others if they handle a device reliably Do not poll disconnected XInput devices (major perf issue) Fix various cases where incorrect correlation could happen Simple mechanism for propagating unhandled Guide button presses even before guaranteed correlation Correlate by axis motion as well as button presses Fix failing to zero other trigger Fix SDL_HINT_JOYSTICK_HIDAPI not working if set before calling SDL_Init() Add missing device to device names Disable RawInput if we have a mismatch of XInput-capable but not RawInput-capable devices Updated to SDL 2.0.13 code with the following notes: New HID driver: xbox360w - no idea what that is, hopefully urelated SDL_hidapijoystick.c had been refactored to couple data handling logic with device opening logic and device lists caused some problems, yields slightly uglier integration than previously when the 360 HID device driver was just handling the data. SDL_hidapijoystick.c now often pulls the device off of the joystick_hwdata structure for some rumble logic, but it appears that code path is never reached, so probably not a problem. Looks like joystick_hwdata was refactored to not include a mutex in other drivers, maintainers may want to do the same refactor here if that's useful for some reason. Something changed in how devices get names, so getting generic names. Had to fix a (new?) bug where removing an XInput controller caused existing controllers (that moved to a new XInput index) to get identified as 0x045e/0x02fd ("it's probably Bluetooth" in code), rendering the existing HIDAPI_IsDevicePresent and new RAWINPUT_IsDevicePresent unreliable.
2020-03-16 14:25:02 -05:00
#endif
RAWINPUT_ReleaseDevice(device);
Fixed bug 4477 - Support more than 4 XInput-capable devices on Windows Jimb Esser Add new RawInput controller API, and improved correlation with XInput/WGI Reorder joystick init so drivers can ask the others if they handle a device reliably Do not poll disconnected XInput devices (major perf issue) Fix various cases where incorrect correlation could happen Simple mechanism for propagating unhandled Guide button presses even before guaranteed correlation Correlate by axis motion as well as button presses Fix failing to zero other trigger Fix SDL_HINT_JOYSTICK_HIDAPI not working if set before calling SDL_Init() Add missing device to device names Disable RawInput if we have a mismatch of XInput-capable but not RawInput-capable devices Updated to SDL 2.0.13 code with the following notes: New HID driver: xbox360w - no idea what that is, hopefully urelated SDL_hidapijoystick.c had been refactored to couple data handling logic with device opening logic and device lists caused some problems, yields slightly uglier integration than previously when the 360 HID device driver was just handling the data. SDL_hidapijoystick.c now often pulls the device off of the joystick_hwdata structure for some rumble logic, but it appears that code path is never reached, so probably not a problem. Looks like joystick_hwdata was refactored to not include a mutex in other drivers, maintainers may want to do the same refactor here if that's useful for some reason. Something changed in how devices get names, so getting generic names. Had to fix a (new?) bug where removing an XInput controller caused existing controllers (that moved to a new XInput index) to get identified as 0x045e/0x02fd ("it's probably Bluetooth" in code), rendering the existing HIDAPI_IsDevicePresent and new RAWINPUT_IsDevicePresent unreliable.
2020-03-16 14:25:02 -05:00
return;
}
}
}
static int RAWINPUT_JoystickInit(void)
{
UINT device_count = 0;
SDL_assert(!SDL_RAWINPUT_inited);
if (!WIN_IsWindowsVistaOrGreater()) {
/* According to bug 6400, this doesn't work on Windows XP */
return -1;
}
if (!SDL_GetHintBoolean(SDL_HINT_JOYSTICK_RAWINPUT, SDL_TRUE)) {
return -1;
}
if (WIN_LoadHIDDLL() < 0) {
return -1;
}
SDL_RAWINPUT_inited = SDL_TRUE;
if ((GetRawInputDeviceList(NULL, &device_count, sizeof(RAWINPUTDEVICELIST)) != -1) && device_count > 0) {
PRAWINPUTDEVICELIST devices = NULL;
UINT i;
devices = (PRAWINPUTDEVICELIST)SDL_malloc(sizeof(RAWINPUTDEVICELIST) * device_count);
if (devices) {
if (GetRawInputDeviceList(devices, &device_count, sizeof(RAWINPUTDEVICELIST)) != -1) {
for (i = 0; i < device_count; ++i) {
RAWINPUT_AddDevice(devices[i].hDevice);
}
}
SDL_free(devices);
}
}
return 0;
}
static int RAWINPUT_JoystickGetCount(void)
{
return SDL_RAWINPUT_numjoysticks;
}
SDL_bool
RAWINPUT_IsEnabled()
{
return SDL_RAWINPUT_inited;
}
SDL_bool
RAWINPUT_IsDevicePresent(Uint16 vendor_id, Uint16 product_id, Uint16 version, const char *name)
{
SDL_RAWINPUT_Device *device;
/* If we're being asked about a device, that means another API just detected one, so rescan */
#ifdef SDL_JOYSTICK_RAWINPUT_XINPUT
xinput_device_change = SDL_TRUE;
#endif
#ifdef SDL_JOYSTICK_RAWINPUT_WGI
wgi_state.need_device_list_update = SDL_TRUE;
#endif
device = SDL_RAWINPUT_devices;
while (device) {
if (vendor_id == device->vendor_id && product_id == device->product_id) {
return SDL_TRUE;
}
/* The Xbox 360 wireless controller shows up as product 0 in WGI.
Try to match it to a Raw Input device via name or known product ID. */
if (vendor_id == device->vendor_id && product_id == 0 &&
((name && SDL_strstr(device->name, name) != NULL) ||
(device->vendor_id == USB_VENDOR_MICROSOFT &&
device->product_id == USB_PRODUCT_XBOX360_XUSB_CONTROLLER))) {
return SDL_TRUE;
}
/* The Xbox One controller shows up as a hardcoded raw input VID/PID */
if (name && SDL_strcmp(name, "Xbox One Game Controller") == 0 &&
device->vendor_id == USB_VENDOR_MICROSOFT &&
device->product_id == USB_PRODUCT_XBOX_ONE_XBOXGIP_CONTROLLER) {
return SDL_TRUE;
}
device = device->next;
}
return SDL_FALSE;
}
static void RAWINPUT_PostUpdate(void)
Fixed bug 4477 - Support more than 4 XInput-capable devices on Windows Jimb Esser Add new RawInput controller API, and improved correlation with XInput/WGI Reorder joystick init so drivers can ask the others if they handle a device reliably Do not poll disconnected XInput devices (major perf issue) Fix various cases where incorrect correlation could happen Simple mechanism for propagating unhandled Guide button presses even before guaranteed correlation Correlate by axis motion as well as button presses Fix failing to zero other trigger Fix SDL_HINT_JOYSTICK_HIDAPI not working if set before calling SDL_Init() Add missing device to device names Disable RawInput if we have a mismatch of XInput-capable but not RawInput-capable devices Updated to SDL 2.0.13 code with the following notes: New HID driver: xbox360w - no idea what that is, hopefully urelated SDL_hidapijoystick.c had been refactored to couple data handling logic with device opening logic and device lists caused some problems, yields slightly uglier integration than previously when the 360 HID device driver was just handling the data. SDL_hidapijoystick.c now often pulls the device off of the joystick_hwdata structure for some rumble logic, but it appears that code path is never reached, so probably not a problem. Looks like joystick_hwdata was refactored to not include a mutex in other drivers, maintainers may want to do the same refactor here if that's useful for some reason. Something changed in how devices get names, so getting generic names. Had to fix a (new?) bug where removing an XInput controller caused existing controllers (that moved to a new XInput index) to get identified as 0x045e/0x02fd ("it's probably Bluetooth" in code), rendering the existing HIDAPI_IsDevicePresent and new RAWINPUT_IsDevicePresent unreliable.
2020-03-16 14:25:02 -05:00
{
#ifdef SDL_JOYSTICK_RAWINPUT_MATCHING
SDL_bool unmapped_guide_pressed = SDL_FALSE;
#ifdef SDL_JOYSTICK_RAWINPUT_WGI
if (!wgi_state.dirty) {
int ii;
for (ii = 0; ii < wgi_state.per_gamepad_count; ii++) {
WindowsGamingInputGamepadState *gamepad_state = wgi_state.per_gamepad[ii];
if (!gamepad_state->used && (gamepad_state->state.Buttons & GamepadButtons_GUIDE)) {
unmapped_guide_pressed = SDL_TRUE;
break;
}
}
}
wgi_state.dirty = SDL_TRUE;
#endif
Fixed bug 4477 - Support more than 4 XInput-capable devices on Windows Jimb Esser Add new RawInput controller API, and improved correlation with XInput/WGI Reorder joystick init so drivers can ask the others if they handle a device reliably Do not poll disconnected XInput devices (major perf issue) Fix various cases where incorrect correlation could happen Simple mechanism for propagating unhandled Guide button presses even before guaranteed correlation Correlate by axis motion as well as button presses Fix failing to zero other trigger Fix SDL_HINT_JOYSTICK_HIDAPI not working if set before calling SDL_Init() Add missing device to device names Disable RawInput if we have a mismatch of XInput-capable but not RawInput-capable devices Updated to SDL 2.0.13 code with the following notes: New HID driver: xbox360w - no idea what that is, hopefully urelated SDL_hidapijoystick.c had been refactored to couple data handling logic with device opening logic and device lists caused some problems, yields slightly uglier integration than previously when the 360 HID device driver was just handling the data. SDL_hidapijoystick.c now often pulls the device off of the joystick_hwdata structure for some rumble logic, but it appears that code path is never reached, so probably not a problem. Looks like joystick_hwdata was refactored to not include a mutex in other drivers, maintainers may want to do the same refactor here if that's useful for some reason. Something changed in how devices get names, so getting generic names. Had to fix a (new?) bug where removing an XInput controller caused existing controllers (that moved to a new XInput index) to get identified as 0x045e/0x02fd ("it's probably Bluetooth" in code), rendering the existing HIDAPI_IsDevicePresent and new RAWINPUT_IsDevicePresent unreliable.
2020-03-16 14:25:02 -05:00
#ifdef SDL_JOYSTICK_RAWINPUT_XINPUT
if (!xinput_state_dirty) {
int ii;
for (ii = 0; ii < SDL_arraysize(xinput_state); ii++) {
if (xinput_state[ii].connected && !xinput_state[ii].used && (xinput_state[ii].state.Gamepad.wButtons & XINPUT_GAMEPAD_GUIDE)) {
unmapped_guide_pressed = SDL_TRUE;
break;
}
Fixed bug 4477 - Support more than 4 XInput-capable devices on Windows Jimb Esser Add new RawInput controller API, and improved correlation with XInput/WGI Reorder joystick init so drivers can ask the others if they handle a device reliably Do not poll disconnected XInput devices (major perf issue) Fix various cases where incorrect correlation could happen Simple mechanism for propagating unhandled Guide button presses even before guaranteed correlation Correlate by axis motion as well as button presses Fix failing to zero other trigger Fix SDL_HINT_JOYSTICK_HIDAPI not working if set before calling SDL_Init() Add missing device to device names Disable RawInput if we have a mismatch of XInput-capable but not RawInput-capable devices Updated to SDL 2.0.13 code with the following notes: New HID driver: xbox360w - no idea what that is, hopefully urelated SDL_hidapijoystick.c had been refactored to couple data handling logic with device opening logic and device lists caused some problems, yields slightly uglier integration than previously when the 360 HID device driver was just handling the data. SDL_hidapijoystick.c now often pulls the device off of the joystick_hwdata structure for some rumble logic, but it appears that code path is never reached, so probably not a problem. Looks like joystick_hwdata was refactored to not include a mutex in other drivers, maintainers may want to do the same refactor here if that's useful for some reason. Something changed in how devices get names, so getting generic names. Had to fix a (new?) bug where removing an XInput controller caused existing controllers (that moved to a new XInput index) to get identified as 0x045e/0x02fd ("it's probably Bluetooth" in code), rendering the existing HIDAPI_IsDevicePresent and new RAWINPUT_IsDevicePresent unreliable.
2020-03-16 14:25:02 -05:00
}
}
xinput_state_dirty = SDL_TRUE;
#endif
if (unmapped_guide_pressed) {
if (guide_button_candidate.joystick && !guide_button_candidate.last_joystick) {
SDL_Joystick *joystick = guide_button_candidate.joystick;
RAWINPUT_DeviceContext *ctx = joystick->hwdata;
if (ctx->guide_hack) {
int guide_button = joystick->nbuttons - 1;
2022-12-27 17:05:51 -06:00
SDL_SendJoystickButton(SDL_GetTicksNS(), guide_button_candidate.joystick, guide_button, SDL_PRESSED);
}
guide_button_candidate.last_joystick = guide_button_candidate.joystick;
}
} else if (guide_button_candidate.last_joystick) {
SDL_Joystick *joystick = guide_button_candidate.last_joystick;
RAWINPUT_DeviceContext *ctx = joystick->hwdata;
if (ctx->guide_hack) {
int guide_button = joystick->nbuttons - 1;
2022-12-27 17:05:51 -06:00
SDL_SendJoystickButton(SDL_GetTicksNS(), joystick, guide_button, SDL_RELEASED);
}
guide_button_candidate.last_joystick = NULL;
}
guide_button_candidate.joystick = NULL;
#endif /* SDL_JOYSTICK_RAWINPUT_MATCHING */
}
static void RAWINPUT_JoystickDetect(void)
Fixed bug 4477 - Support more than 4 XInput-capable devices on Windows Jimb Esser Add new RawInput controller API, and improved correlation with XInput/WGI Reorder joystick init so drivers can ask the others if they handle a device reliably Do not poll disconnected XInput devices (major perf issue) Fix various cases where incorrect correlation could happen Simple mechanism for propagating unhandled Guide button presses even before guaranteed correlation Correlate by axis motion as well as button presses Fix failing to zero other trigger Fix SDL_HINT_JOYSTICK_HIDAPI not working if set before calling SDL_Init() Add missing device to device names Disable RawInput if we have a mismatch of XInput-capable but not RawInput-capable devices Updated to SDL 2.0.13 code with the following notes: New HID driver: xbox360w - no idea what that is, hopefully urelated SDL_hidapijoystick.c had been refactored to couple data handling logic with device opening logic and device lists caused some problems, yields slightly uglier integration than previously when the 360 HID device driver was just handling the data. SDL_hidapijoystick.c now often pulls the device off of the joystick_hwdata structure for some rumble logic, but it appears that code path is never reached, so probably not a problem. Looks like joystick_hwdata was refactored to not include a mutex in other drivers, maintainers may want to do the same refactor here if that's useful for some reason. Something changed in how devices get names, so getting generic names. Had to fix a (new?) bug where removing an XInput controller caused existing controllers (that moved to a new XInput index) to get identified as 0x045e/0x02fd ("it's probably Bluetooth" in code), rendering the existing HIDAPI_IsDevicePresent and new RAWINPUT_IsDevicePresent unreliable.
2020-03-16 14:25:02 -05:00
{
RAWINPUT_PostUpdate();
Fixed bug 4477 - Support more than 4 XInput-capable devices on Windows Jimb Esser Add new RawInput controller API, and improved correlation with XInput/WGI Reorder joystick init so drivers can ask the others if they handle a device reliably Do not poll disconnected XInput devices (major perf issue) Fix various cases where incorrect correlation could happen Simple mechanism for propagating unhandled Guide button presses even before guaranteed correlation Correlate by axis motion as well as button presses Fix failing to zero other trigger Fix SDL_HINT_JOYSTICK_HIDAPI not working if set before calling SDL_Init() Add missing device to device names Disable RawInput if we have a mismatch of XInput-capable but not RawInput-capable devices Updated to SDL 2.0.13 code with the following notes: New HID driver: xbox360w - no idea what that is, hopefully urelated SDL_hidapijoystick.c had been refactored to couple data handling logic with device opening logic and device lists caused some problems, yields slightly uglier integration than previously when the 360 HID device driver was just handling the data. SDL_hidapijoystick.c now often pulls the device off of the joystick_hwdata structure for some rumble logic, but it appears that code path is never reached, so probably not a problem. Looks like joystick_hwdata was refactored to not include a mutex in other drivers, maintainers may want to do the same refactor here if that's useful for some reason. Something changed in how devices get names, so getting generic names. Had to fix a (new?) bug where removing an XInput controller caused existing controllers (that moved to a new XInput index) to get identified as 0x045e/0x02fd ("it's probably Bluetooth" in code), rendering the existing HIDAPI_IsDevicePresent and new RAWINPUT_IsDevicePresent unreliable.
2020-03-16 14:25:02 -05:00
}
static SDL_RAWINPUT_Device *RAWINPUT_GetDeviceByIndex(int device_index)
Fixed bug 4477 - Support more than 4 XInput-capable devices on Windows Jimb Esser Add new RawInput controller API, and improved correlation with XInput/WGI Reorder joystick init so drivers can ask the others if they handle a device reliably Do not poll disconnected XInput devices (major perf issue) Fix various cases where incorrect correlation could happen Simple mechanism for propagating unhandled Guide button presses even before guaranteed correlation Correlate by axis motion as well as button presses Fix failing to zero other trigger Fix SDL_HINT_JOYSTICK_HIDAPI not working if set before calling SDL_Init() Add missing device to device names Disable RawInput if we have a mismatch of XInput-capable but not RawInput-capable devices Updated to SDL 2.0.13 code with the following notes: New HID driver: xbox360w - no idea what that is, hopefully urelated SDL_hidapijoystick.c had been refactored to couple data handling logic with device opening logic and device lists caused some problems, yields slightly uglier integration than previously when the 360 HID device driver was just handling the data. SDL_hidapijoystick.c now often pulls the device off of the joystick_hwdata structure for some rumble logic, but it appears that code path is never reached, so probably not a problem. Looks like joystick_hwdata was refactored to not include a mutex in other drivers, maintainers may want to do the same refactor here if that's useful for some reason. Something changed in how devices get names, so getting generic names. Had to fix a (new?) bug where removing an XInput controller caused existing controllers (that moved to a new XInput index) to get identified as 0x045e/0x02fd ("it's probably Bluetooth" in code), rendering the existing HIDAPI_IsDevicePresent and new RAWINPUT_IsDevicePresent unreliable.
2020-03-16 14:25:02 -05:00
{
SDL_RAWINPUT_Device *device = SDL_RAWINPUT_devices;
while (device) {
if (device_index == 0) {
break;
Fixed bug 4477 - Support more than 4 XInput-capable devices on Windows Jimb Esser Add new RawInput controller API, and improved correlation with XInput/WGI Reorder joystick init so drivers can ask the others if they handle a device reliably Do not poll disconnected XInput devices (major perf issue) Fix various cases where incorrect correlation could happen Simple mechanism for propagating unhandled Guide button presses even before guaranteed correlation Correlate by axis motion as well as button presses Fix failing to zero other trigger Fix SDL_HINT_JOYSTICK_HIDAPI not working if set before calling SDL_Init() Add missing device to device names Disable RawInput if we have a mismatch of XInput-capable but not RawInput-capable devices Updated to SDL 2.0.13 code with the following notes: New HID driver: xbox360w - no idea what that is, hopefully urelated SDL_hidapijoystick.c had been refactored to couple data handling logic with device opening logic and device lists caused some problems, yields slightly uglier integration than previously when the 360 HID device driver was just handling the data. SDL_hidapijoystick.c now often pulls the device off of the joystick_hwdata structure for some rumble logic, but it appears that code path is never reached, so probably not a problem. Looks like joystick_hwdata was refactored to not include a mutex in other drivers, maintainers may want to do the same refactor here if that's useful for some reason. Something changed in how devices get names, so getting generic names. Had to fix a (new?) bug where removing an XInput controller caused existing controllers (that moved to a new XInput index) to get identified as 0x045e/0x02fd ("it's probably Bluetooth" in code), rendering the existing HIDAPI_IsDevicePresent and new RAWINPUT_IsDevicePresent unreliable.
2020-03-16 14:25:02 -05:00
}
--device_index;
Fixed bug 4477 - Support more than 4 XInput-capable devices on Windows Jimb Esser Add new RawInput controller API, and improved correlation with XInput/WGI Reorder joystick init so drivers can ask the others if they handle a device reliably Do not poll disconnected XInput devices (major perf issue) Fix various cases where incorrect correlation could happen Simple mechanism for propagating unhandled Guide button presses even before guaranteed correlation Correlate by axis motion as well as button presses Fix failing to zero other trigger Fix SDL_HINT_JOYSTICK_HIDAPI not working if set before calling SDL_Init() Add missing device to device names Disable RawInput if we have a mismatch of XInput-capable but not RawInput-capable devices Updated to SDL 2.0.13 code with the following notes: New HID driver: xbox360w - no idea what that is, hopefully urelated SDL_hidapijoystick.c had been refactored to couple data handling logic with device opening logic and device lists caused some problems, yields slightly uglier integration than previously when the 360 HID device driver was just handling the data. SDL_hidapijoystick.c now often pulls the device off of the joystick_hwdata structure for some rumble logic, but it appears that code path is never reached, so probably not a problem. Looks like joystick_hwdata was refactored to not include a mutex in other drivers, maintainers may want to do the same refactor here if that's useful for some reason. Something changed in how devices get names, so getting generic names. Had to fix a (new?) bug where removing an XInput controller caused existing controllers (that moved to a new XInput index) to get identified as 0x045e/0x02fd ("it's probably Bluetooth" in code), rendering the existing HIDAPI_IsDevicePresent and new RAWINPUT_IsDevicePresent unreliable.
2020-03-16 14:25:02 -05:00
device = device->next;
}
return device;
Fixed bug 4477 - Support more than 4 XInput-capable devices on Windows Jimb Esser Add new RawInput controller API, and improved correlation with XInput/WGI Reorder joystick init so drivers can ask the others if they handle a device reliably Do not poll disconnected XInput devices (major perf issue) Fix various cases where incorrect correlation could happen Simple mechanism for propagating unhandled Guide button presses even before guaranteed correlation Correlate by axis motion as well as button presses Fix failing to zero other trigger Fix SDL_HINT_JOYSTICK_HIDAPI not working if set before calling SDL_Init() Add missing device to device names Disable RawInput if we have a mismatch of XInput-capable but not RawInput-capable devices Updated to SDL 2.0.13 code with the following notes: New HID driver: xbox360w - no idea what that is, hopefully urelated SDL_hidapijoystick.c had been refactored to couple data handling logic with device opening logic and device lists caused some problems, yields slightly uglier integration than previously when the 360 HID device driver was just handling the data. SDL_hidapijoystick.c now often pulls the device off of the joystick_hwdata structure for some rumble logic, but it appears that code path is never reached, so probably not a problem. Looks like joystick_hwdata was refactored to not include a mutex in other drivers, maintainers may want to do the same refactor here if that's useful for some reason. Something changed in how devices get names, so getting generic names. Had to fix a (new?) bug where removing an XInput controller caused existing controllers (that moved to a new XInput index) to get identified as 0x045e/0x02fd ("it's probably Bluetooth" in code), rendering the existing HIDAPI_IsDevicePresent and new RAWINPUT_IsDevicePresent unreliable.
2020-03-16 14:25:02 -05:00
}
static const char *RAWINPUT_JoystickGetDeviceName(int device_index)
Fixed bug 4477 - Support more than 4 XInput-capable devices on Windows Jimb Esser Add new RawInput controller API, and improved correlation with XInput/WGI Reorder joystick init so drivers can ask the others if they handle a device reliably Do not poll disconnected XInput devices (major perf issue) Fix various cases where incorrect correlation could happen Simple mechanism for propagating unhandled Guide button presses even before guaranteed correlation Correlate by axis motion as well as button presses Fix failing to zero other trigger Fix SDL_HINT_JOYSTICK_HIDAPI not working if set before calling SDL_Init() Add missing device to device names Disable RawInput if we have a mismatch of XInput-capable but not RawInput-capable devices Updated to SDL 2.0.13 code with the following notes: New HID driver: xbox360w - no idea what that is, hopefully urelated SDL_hidapijoystick.c had been refactored to couple data handling logic with device opening logic and device lists caused some problems, yields slightly uglier integration than previously when the 360 HID device driver was just handling the data. SDL_hidapijoystick.c now often pulls the device off of the joystick_hwdata structure for some rumble logic, but it appears that code path is never reached, so probably not a problem. Looks like joystick_hwdata was refactored to not include a mutex in other drivers, maintainers may want to do the same refactor here if that's useful for some reason. Something changed in how devices get names, so getting generic names. Had to fix a (new?) bug where removing an XInput controller caused existing controllers (that moved to a new XInput index) to get identified as 0x045e/0x02fd ("it's probably Bluetooth" in code), rendering the existing HIDAPI_IsDevicePresent and new RAWINPUT_IsDevicePresent unreliable.
2020-03-16 14:25:02 -05:00
{
return RAWINPUT_GetDeviceByIndex(device_index)->name;
Fixed bug 4477 - Support more than 4 XInput-capable devices on Windows Jimb Esser Add new RawInput controller API, and improved correlation with XInput/WGI Reorder joystick init so drivers can ask the others if they handle a device reliably Do not poll disconnected XInput devices (major perf issue) Fix various cases where incorrect correlation could happen Simple mechanism for propagating unhandled Guide button presses even before guaranteed correlation Correlate by axis motion as well as button presses Fix failing to zero other trigger Fix SDL_HINT_JOYSTICK_HIDAPI not working if set before calling SDL_Init() Add missing device to device names Disable RawInput if we have a mismatch of XInput-capable but not RawInput-capable devices Updated to SDL 2.0.13 code with the following notes: New HID driver: xbox360w - no idea what that is, hopefully urelated SDL_hidapijoystick.c had been refactored to couple data handling logic with device opening logic and device lists caused some problems, yields slightly uglier integration than previously when the 360 HID device driver was just handling the data. SDL_hidapijoystick.c now often pulls the device off of the joystick_hwdata structure for some rumble logic, but it appears that code path is never reached, so probably not a problem. Looks like joystick_hwdata was refactored to not include a mutex in other drivers, maintainers may want to do the same refactor here if that's useful for some reason. Something changed in how devices get names, so getting generic names. Had to fix a (new?) bug where removing an XInput controller caused existing controllers (that moved to a new XInput index) to get identified as 0x045e/0x02fd ("it's probably Bluetooth" in code), rendering the existing HIDAPI_IsDevicePresent and new RAWINPUT_IsDevicePresent unreliable.
2020-03-16 14:25:02 -05:00
}
static const char *RAWINPUT_JoystickGetDevicePath(int device_index)
{
return RAWINPUT_GetDeviceByIndex(device_index)->path;
}
static int RAWINPUT_JoystickGetDevicePlayerIndex(int device_index)
Fixed bug 4477 - Support more than 4 XInput-capable devices on Windows Jimb Esser Add new RawInput controller API, and improved correlation with XInput/WGI Reorder joystick init so drivers can ask the others if they handle a device reliably Do not poll disconnected XInput devices (major perf issue) Fix various cases where incorrect correlation could happen Simple mechanism for propagating unhandled Guide button presses even before guaranteed correlation Correlate by axis motion as well as button presses Fix failing to zero other trigger Fix SDL_HINT_JOYSTICK_HIDAPI not working if set before calling SDL_Init() Add missing device to device names Disable RawInput if we have a mismatch of XInput-capable but not RawInput-capable devices Updated to SDL 2.0.13 code with the following notes: New HID driver: xbox360w - no idea what that is, hopefully urelated SDL_hidapijoystick.c had been refactored to couple data handling logic with device opening logic and device lists caused some problems, yields slightly uglier integration than previously when the 360 HID device driver was just handling the data. SDL_hidapijoystick.c now often pulls the device off of the joystick_hwdata structure for some rumble logic, but it appears that code path is never reached, so probably not a problem. Looks like joystick_hwdata was refactored to not include a mutex in other drivers, maintainers may want to do the same refactor here if that's useful for some reason. Something changed in how devices get names, so getting generic names. Had to fix a (new?) bug where removing an XInput controller caused existing controllers (that moved to a new XInput index) to get identified as 0x045e/0x02fd ("it's probably Bluetooth" in code), rendering the existing HIDAPI_IsDevicePresent and new RAWINPUT_IsDevicePresent unreliable.
2020-03-16 14:25:02 -05:00
{
return -1;
Fixed bug 4477 - Support more than 4 XInput-capable devices on Windows Jimb Esser Add new RawInput controller API, and improved correlation with XInput/WGI Reorder joystick init so drivers can ask the others if they handle a device reliably Do not poll disconnected XInput devices (major perf issue) Fix various cases where incorrect correlation could happen Simple mechanism for propagating unhandled Guide button presses even before guaranteed correlation Correlate by axis motion as well as button presses Fix failing to zero other trigger Fix SDL_HINT_JOYSTICK_HIDAPI not working if set before calling SDL_Init() Add missing device to device names Disable RawInput if we have a mismatch of XInput-capable but not RawInput-capable devices Updated to SDL 2.0.13 code with the following notes: New HID driver: xbox360w - no idea what that is, hopefully urelated SDL_hidapijoystick.c had been refactored to couple data handling logic with device opening logic and device lists caused some problems, yields slightly uglier integration than previously when the 360 HID device driver was just handling the data. SDL_hidapijoystick.c now often pulls the device off of the joystick_hwdata structure for some rumble logic, but it appears that code path is never reached, so probably not a problem. Looks like joystick_hwdata was refactored to not include a mutex in other drivers, maintainers may want to do the same refactor here if that's useful for some reason. Something changed in how devices get names, so getting generic names. Had to fix a (new?) bug where removing an XInput controller caused existing controllers (that moved to a new XInput index) to get identified as 0x045e/0x02fd ("it's probably Bluetooth" in code), rendering the existing HIDAPI_IsDevicePresent and new RAWINPUT_IsDevicePresent unreliable.
2020-03-16 14:25:02 -05:00
}
static void RAWINPUT_JoystickSetDevicePlayerIndex(int device_index, int player_index)
Fixed bug 4477 - Support more than 4 XInput-capable devices on Windows Jimb Esser Add new RawInput controller API, and improved correlation with XInput/WGI Reorder joystick init so drivers can ask the others if they handle a device reliably Do not poll disconnected XInput devices (major perf issue) Fix various cases where incorrect correlation could happen Simple mechanism for propagating unhandled Guide button presses even before guaranteed correlation Correlate by axis motion as well as button presses Fix failing to zero other trigger Fix SDL_HINT_JOYSTICK_HIDAPI not working if set before calling SDL_Init() Add missing device to device names Disable RawInput if we have a mismatch of XInput-capable but not RawInput-capable devices Updated to SDL 2.0.13 code with the following notes: New HID driver: xbox360w - no idea what that is, hopefully urelated SDL_hidapijoystick.c had been refactored to couple data handling logic with device opening logic and device lists caused some problems, yields slightly uglier integration than previously when the 360 HID device driver was just handling the data. SDL_hidapijoystick.c now often pulls the device off of the joystick_hwdata structure for some rumble logic, but it appears that code path is never reached, so probably not a problem. Looks like joystick_hwdata was refactored to not include a mutex in other drivers, maintainers may want to do the same refactor here if that's useful for some reason. Something changed in how devices get names, so getting generic names. Had to fix a (new?) bug where removing an XInput controller caused existing controllers (that moved to a new XInput index) to get identified as 0x045e/0x02fd ("it's probably Bluetooth" in code), rendering the existing HIDAPI_IsDevicePresent and new RAWINPUT_IsDevicePresent unreliable.
2020-03-16 14:25:02 -05:00
{
}
static SDL_JoystickGUID RAWINPUT_JoystickGetDeviceGUID(int device_index)
Fixed bug 4477 - Support more than 4 XInput-capable devices on Windows Jimb Esser Add new RawInput controller API, and improved correlation with XInput/WGI Reorder joystick init so drivers can ask the others if they handle a device reliably Do not poll disconnected XInput devices (major perf issue) Fix various cases where incorrect correlation could happen Simple mechanism for propagating unhandled Guide button presses even before guaranteed correlation Correlate by axis motion as well as button presses Fix failing to zero other trigger Fix SDL_HINT_JOYSTICK_HIDAPI not working if set before calling SDL_Init() Add missing device to device names Disable RawInput if we have a mismatch of XInput-capable but not RawInput-capable devices Updated to SDL 2.0.13 code with the following notes: New HID driver: xbox360w - no idea what that is, hopefully urelated SDL_hidapijoystick.c had been refactored to couple data handling logic with device opening logic and device lists caused some problems, yields slightly uglier integration than previously when the 360 HID device driver was just handling the data. SDL_hidapijoystick.c now often pulls the device off of the joystick_hwdata structure for some rumble logic, but it appears that code path is never reached, so probably not a problem. Looks like joystick_hwdata was refactored to not include a mutex in other drivers, maintainers may want to do the same refactor here if that's useful for some reason. Something changed in how devices get names, so getting generic names. Had to fix a (new?) bug where removing an XInput controller caused existing controllers (that moved to a new XInput index) to get identified as 0x045e/0x02fd ("it's probably Bluetooth" in code), rendering the existing HIDAPI_IsDevicePresent and new RAWINPUT_IsDevicePresent unreliable.
2020-03-16 14:25:02 -05:00
{
return RAWINPUT_GetDeviceByIndex(device_index)->guid;
Fixed bug 4477 - Support more than 4 XInput-capable devices on Windows Jimb Esser Add new RawInput controller API, and improved correlation with XInput/WGI Reorder joystick init so drivers can ask the others if they handle a device reliably Do not poll disconnected XInput devices (major perf issue) Fix various cases where incorrect correlation could happen Simple mechanism for propagating unhandled Guide button presses even before guaranteed correlation Correlate by axis motion as well as button presses Fix failing to zero other trigger Fix SDL_HINT_JOYSTICK_HIDAPI not working if set before calling SDL_Init() Add missing device to device names Disable RawInput if we have a mismatch of XInput-capable but not RawInput-capable devices Updated to SDL 2.0.13 code with the following notes: New HID driver: xbox360w - no idea what that is, hopefully urelated SDL_hidapijoystick.c had been refactored to couple data handling logic with device opening logic and device lists caused some problems, yields slightly uglier integration than previously when the 360 HID device driver was just handling the data. SDL_hidapijoystick.c now often pulls the device off of the joystick_hwdata structure for some rumble logic, but it appears that code path is never reached, so probably not a problem. Looks like joystick_hwdata was refactored to not include a mutex in other drivers, maintainers may want to do the same refactor here if that's useful for some reason. Something changed in how devices get names, so getting generic names. Had to fix a (new?) bug where removing an XInput controller caused existing controllers (that moved to a new XInput index) to get identified as 0x045e/0x02fd ("it's probably Bluetooth" in code), rendering the existing HIDAPI_IsDevicePresent and new RAWINPUT_IsDevicePresent unreliable.
2020-03-16 14:25:02 -05:00
}
static SDL_JoystickID RAWINPUT_JoystickGetDeviceInstanceID(int device_index)
Fixed bug 4477 - Support more than 4 XInput-capable devices on Windows Jimb Esser Add new RawInput controller API, and improved correlation with XInput/WGI Reorder joystick init so drivers can ask the others if they handle a device reliably Do not poll disconnected XInput devices (major perf issue) Fix various cases where incorrect correlation could happen Simple mechanism for propagating unhandled Guide button presses even before guaranteed correlation Correlate by axis motion as well as button presses Fix failing to zero other trigger Fix SDL_HINT_JOYSTICK_HIDAPI not working if set before calling SDL_Init() Add missing device to device names Disable RawInput if we have a mismatch of XInput-capable but not RawInput-capable devices Updated to SDL 2.0.13 code with the following notes: New HID driver: xbox360w - no idea what that is, hopefully urelated SDL_hidapijoystick.c had been refactored to couple data handling logic with device opening logic and device lists caused some problems, yields slightly uglier integration than previously when the 360 HID device driver was just handling the data. SDL_hidapijoystick.c now often pulls the device off of the joystick_hwdata structure for some rumble logic, but it appears that code path is never reached, so probably not a problem. Looks like joystick_hwdata was refactored to not include a mutex in other drivers, maintainers may want to do the same refactor here if that's useful for some reason. Something changed in how devices get names, so getting generic names. Had to fix a (new?) bug where removing an XInput controller caused existing controllers (that moved to a new XInput index) to get identified as 0x045e/0x02fd ("it's probably Bluetooth" in code), rendering the existing HIDAPI_IsDevicePresent and new RAWINPUT_IsDevicePresent unreliable.
2020-03-16 14:25:02 -05:00
{
return RAWINPUT_GetDeviceByIndex(device_index)->joystick_id;
Fixed bug 4477 - Support more than 4 XInput-capable devices on Windows Jimb Esser Add new RawInput controller API, and improved correlation with XInput/WGI Reorder joystick init so drivers can ask the others if they handle a device reliably Do not poll disconnected XInput devices (major perf issue) Fix various cases where incorrect correlation could happen Simple mechanism for propagating unhandled Guide button presses even before guaranteed correlation Correlate by axis motion as well as button presses Fix failing to zero other trigger Fix SDL_HINT_JOYSTICK_HIDAPI not working if set before calling SDL_Init() Add missing device to device names Disable RawInput if we have a mismatch of XInput-capable but not RawInput-capable devices Updated to SDL 2.0.13 code with the following notes: New HID driver: xbox360w - no idea what that is, hopefully urelated SDL_hidapijoystick.c had been refactored to couple data handling logic with device opening logic and device lists caused some problems, yields slightly uglier integration than previously when the 360 HID device driver was just handling the data. SDL_hidapijoystick.c now often pulls the device off of the joystick_hwdata structure for some rumble logic, but it appears that code path is never reached, so probably not a problem. Looks like joystick_hwdata was refactored to not include a mutex in other drivers, maintainers may want to do the same refactor here if that's useful for some reason. Something changed in how devices get names, so getting generic names. Had to fix a (new?) bug where removing an XInput controller caused existing controllers (that moved to a new XInput index) to get identified as 0x045e/0x02fd ("it's probably Bluetooth" in code), rendering the existing HIDAPI_IsDevicePresent and new RAWINPUT_IsDevicePresent unreliable.
2020-03-16 14:25:02 -05:00
}
static int SDLCALL RAWINPUT_SortValueCaps(const void *A, const void *B)
{
HIDP_VALUE_CAPS *capsA = (HIDP_VALUE_CAPS *)A;
HIDP_VALUE_CAPS *capsB = (HIDP_VALUE_CAPS *)B;
/* Sort by Usage for single values, or UsageMax for range of values */
return (int)capsA->NotRange.Usage - capsB->NotRange.Usage;
}
static int RAWINPUT_JoystickOpen(SDL_Joystick *joystick, int device_index)
Fixed bug 4477 - Support more than 4 XInput-capable devices on Windows Jimb Esser Add new RawInput controller API, and improved correlation with XInput/WGI Reorder joystick init so drivers can ask the others if they handle a device reliably Do not poll disconnected XInput devices (major perf issue) Fix various cases where incorrect correlation could happen Simple mechanism for propagating unhandled Guide button presses even before guaranteed correlation Correlate by axis motion as well as button presses Fix failing to zero other trigger Fix SDL_HINT_JOYSTICK_HIDAPI not working if set before calling SDL_Init() Add missing device to device names Disable RawInput if we have a mismatch of XInput-capable but not RawInput-capable devices Updated to SDL 2.0.13 code with the following notes: New HID driver: xbox360w - no idea what that is, hopefully urelated SDL_hidapijoystick.c had been refactored to couple data handling logic with device opening logic and device lists caused some problems, yields slightly uglier integration than previously when the 360 HID device driver was just handling the data. SDL_hidapijoystick.c now often pulls the device off of the joystick_hwdata structure for some rumble logic, but it appears that code path is never reached, so probably not a problem. Looks like joystick_hwdata was refactored to not include a mutex in other drivers, maintainers may want to do the same refactor here if that's useful for some reason. Something changed in how devices get names, so getting generic names. Had to fix a (new?) bug where removing an XInput controller caused existing controllers (that moved to a new XInput index) to get identified as 0x045e/0x02fd ("it's probably Bluetooth" in code), rendering the existing HIDAPI_IsDevicePresent and new RAWINPUT_IsDevicePresent unreliable.
2020-03-16 14:25:02 -05:00
{
SDL_RAWINPUT_Device *device = RAWINPUT_GetDeviceByIndex(device_index);
RAWINPUT_DeviceContext *ctx;
HIDP_CAPS caps;
HIDP_BUTTON_CAPS *button_caps;
HIDP_VALUE_CAPS *value_caps;
ULONG i;
ctx = (RAWINPUT_DeviceContext *)SDL_calloc(1, sizeof(RAWINPUT_DeviceContext));
if (ctx == NULL) {
return SDL_OutOfMemory();
}
joystick->hwdata = ctx;
Fixed bug 4477 - Support more than 4 XInput-capable devices on Windows Jimb Esser Add new RawInput controller API, and improved correlation with XInput/WGI Reorder joystick init so drivers can ask the others if they handle a device reliably Do not poll disconnected XInput devices (major perf issue) Fix various cases where incorrect correlation could happen Simple mechanism for propagating unhandled Guide button presses even before guaranteed correlation Correlate by axis motion as well as button presses Fix failing to zero other trigger Fix SDL_HINT_JOYSTICK_HIDAPI not working if set before calling SDL_Init() Add missing device to device names Disable RawInput if we have a mismatch of XInput-capable but not RawInput-capable devices Updated to SDL 2.0.13 code with the following notes: New HID driver: xbox360w - no idea what that is, hopefully urelated SDL_hidapijoystick.c had been refactored to couple data handling logic with device opening logic and device lists caused some problems, yields slightly uglier integration than previously when the 360 HID device driver was just handling the data. SDL_hidapijoystick.c now often pulls the device off of the joystick_hwdata structure for some rumble logic, but it appears that code path is never reached, so probably not a problem. Looks like joystick_hwdata was refactored to not include a mutex in other drivers, maintainers may want to do the same refactor here if that's useful for some reason. Something changed in how devices get names, so getting generic names. Had to fix a (new?) bug where removing an XInput controller caused existing controllers (that moved to a new XInput index) to get identified as 0x045e/0x02fd ("it's probably Bluetooth" in code), rendering the existing HIDAPI_IsDevicePresent and new RAWINPUT_IsDevicePresent unreliable.
2020-03-16 14:25:02 -05:00
ctx->device = RAWINPUT_AcquireDevice(device);
device->joystick = joystick;
if (device->is_xinput) {
/* We'll try to get guide button and trigger axes from XInput */
#ifdef SDL_JOYSTICK_RAWINPUT_XINPUT
xinput_device_change = SDL_TRUE;
ctx->xinput_enabled = SDL_GetHintBoolean(SDL_HINT_JOYSTICK_RAWINPUT_CORRELATE_XINPUT, SDL_TRUE);
if (ctx->xinput_enabled && (WIN_LoadXInputDLL() < 0 || XINPUTGETSTATE == NULL)) {
ctx->xinput_enabled = SDL_FALSE;
}
ctx->xinput_slot = XUSER_INDEX_ANY;
#endif
#ifdef SDL_JOYSTICK_RAWINPUT_WGI
RAWINPUT_InitWindowsGamingInput(ctx);
#endif
}
ctx->is_xinput = device->is_xinput;
ctx->is_xboxone = device->is_xboxone;
ctx->preparsed_data = device->preparsed_data;
ctx->max_data_length = SDL_HidP_MaxDataListLength(HidP_Input, ctx->preparsed_data);
ctx->data = (HIDP_DATA *)SDL_malloc(ctx->max_data_length * sizeof(*ctx->data));
if (!ctx->data) {
RAWINPUT_JoystickClose(joystick);
Fixed bug 4477 - Support more than 4 XInput-capable devices on Windows Jimb Esser Add new RawInput controller API, and improved correlation with XInput/WGI Reorder joystick init so drivers can ask the others if they handle a device reliably Do not poll disconnected XInput devices (major perf issue) Fix various cases where incorrect correlation could happen Simple mechanism for propagating unhandled Guide button presses even before guaranteed correlation Correlate by axis motion as well as button presses Fix failing to zero other trigger Fix SDL_HINT_JOYSTICK_HIDAPI not working if set before calling SDL_Init() Add missing device to device names Disable RawInput if we have a mismatch of XInput-capable but not RawInput-capable devices Updated to SDL 2.0.13 code with the following notes: New HID driver: xbox360w - no idea what that is, hopefully urelated SDL_hidapijoystick.c had been refactored to couple data handling logic with device opening logic and device lists caused some problems, yields slightly uglier integration than previously when the 360 HID device driver was just handling the data. SDL_hidapijoystick.c now often pulls the device off of the joystick_hwdata structure for some rumble logic, but it appears that code path is never reached, so probably not a problem. Looks like joystick_hwdata was refactored to not include a mutex in other drivers, maintainers may want to do the same refactor here if that's useful for some reason. Something changed in how devices get names, so getting generic names. Had to fix a (new?) bug where removing an XInput controller caused existing controllers (that moved to a new XInput index) to get identified as 0x045e/0x02fd ("it's probably Bluetooth" in code), rendering the existing HIDAPI_IsDevicePresent and new RAWINPUT_IsDevicePresent unreliable.
2020-03-16 14:25:02 -05:00
return SDL_OutOfMemory();
}
if (SDL_HidP_GetCaps(ctx->preparsed_data, &caps) != HIDP_STATUS_SUCCESS) {
RAWINPUT_JoystickClose(joystick);
return SDL_SetError("Couldn't get device capabilities");
Fixed bug 4477 - Support more than 4 XInput-capable devices on Windows Jimb Esser Add new RawInput controller API, and improved correlation with XInput/WGI Reorder joystick init so drivers can ask the others if they handle a device reliably Do not poll disconnected XInput devices (major perf issue) Fix various cases where incorrect correlation could happen Simple mechanism for propagating unhandled Guide button presses even before guaranteed correlation Correlate by axis motion as well as button presses Fix failing to zero other trigger Fix SDL_HINT_JOYSTICK_HIDAPI not working if set before calling SDL_Init() Add missing device to device names Disable RawInput if we have a mismatch of XInput-capable but not RawInput-capable devices Updated to SDL 2.0.13 code with the following notes: New HID driver: xbox360w - no idea what that is, hopefully urelated SDL_hidapijoystick.c had been refactored to couple data handling logic with device opening logic and device lists caused some problems, yields slightly uglier integration than previously when the 360 HID device driver was just handling the data. SDL_hidapijoystick.c now often pulls the device off of the joystick_hwdata structure for some rumble logic, but it appears that code path is never reached, so probably not a problem. Looks like joystick_hwdata was refactored to not include a mutex in other drivers, maintainers may want to do the same refactor here if that's useful for some reason. Something changed in how devices get names, so getting generic names. Had to fix a (new?) bug where removing an XInput controller caused existing controllers (that moved to a new XInput index) to get identified as 0x045e/0x02fd ("it's probably Bluetooth" in code), rendering the existing HIDAPI_IsDevicePresent and new RAWINPUT_IsDevicePresent unreliable.
2020-03-16 14:25:02 -05:00
}
button_caps = SDL_stack_alloc(HIDP_BUTTON_CAPS, caps.NumberInputButtonCaps);
if (SDL_HidP_GetButtonCaps(HidP_Input, button_caps, &caps.NumberInputButtonCaps, ctx->preparsed_data) != HIDP_STATUS_SUCCESS) {
RAWINPUT_JoystickClose(joystick);
return SDL_SetError("Couldn't get device button capabilities");
}
value_caps = SDL_stack_alloc(HIDP_VALUE_CAPS, caps.NumberInputValueCaps);
if (SDL_HidP_GetValueCaps(HidP_Input, value_caps, &caps.NumberInputValueCaps, ctx->preparsed_data) != HIDP_STATUS_SUCCESS) {
RAWINPUT_JoystickClose(joystick);
return SDL_SetError("Couldn't get device value capabilities");
}
/* Sort the axes by usage, so X comes before Y, etc. */
SDL_qsort(value_caps, caps.NumberInputValueCaps, sizeof(*value_caps), RAWINPUT_SortValueCaps);
for (i = 0; i < caps.NumberInputButtonCaps; ++i) {
HIDP_BUTTON_CAPS *cap = &button_caps[i];
if (cap->UsagePage == USB_USAGEPAGE_BUTTON) {
int count;
if (cap->IsRange) {
count = 1 + (cap->Range.DataIndexMax - cap->Range.DataIndexMin);
} else {
count = 1;
}
joystick->nbuttons += count;
}
}
if (joystick->nbuttons > 0) {
int button_index = 0;
ctx->button_indices = (USHORT *)SDL_malloc(joystick->nbuttons * sizeof(*ctx->button_indices));
if (!ctx->button_indices) {
RAWINPUT_JoystickClose(joystick);
return SDL_OutOfMemory();
}
for (i = 0; i < caps.NumberInputButtonCaps; ++i) {
HIDP_BUTTON_CAPS *cap = &button_caps[i];
if (cap->UsagePage == USB_USAGEPAGE_BUTTON) {
if (cap->IsRange) {
int j, count = 1 + (cap->Range.DataIndexMax - cap->Range.DataIndexMin);
Fixed bug 4477 - Support more than 4 XInput-capable devices on Windows Jimb Esser Add new RawInput controller API, and improved correlation with XInput/WGI Reorder joystick init so drivers can ask the others if they handle a device reliably Do not poll disconnected XInput devices (major perf issue) Fix various cases where incorrect correlation could happen Simple mechanism for propagating unhandled Guide button presses even before guaranteed correlation Correlate by axis motion as well as button presses Fix failing to zero other trigger Fix SDL_HINT_JOYSTICK_HIDAPI not working if set before calling SDL_Init() Add missing device to device names Disable RawInput if we have a mismatch of XInput-capable but not RawInput-capable devices Updated to SDL 2.0.13 code with the following notes: New HID driver: xbox360w - no idea what that is, hopefully urelated SDL_hidapijoystick.c had been refactored to couple data handling logic with device opening logic and device lists caused some problems, yields slightly uglier integration than previously when the 360 HID device driver was just handling the data. SDL_hidapijoystick.c now often pulls the device off of the joystick_hwdata structure for some rumble logic, but it appears that code path is never reached, so probably not a problem. Looks like joystick_hwdata was refactored to not include a mutex in other drivers, maintainers may want to do the same refactor here if that's useful for some reason. Something changed in how devices get names, so getting generic names. Had to fix a (new?) bug where removing an XInput controller caused existing controllers (that moved to a new XInput index) to get identified as 0x045e/0x02fd ("it's probably Bluetooth" in code), rendering the existing HIDAPI_IsDevicePresent and new RAWINPUT_IsDevicePresent unreliable.
2020-03-16 14:25:02 -05:00
for (j = 0; j < count; ++j) {
ctx->button_indices[button_index++] = cap->Range.DataIndexMin + j;
}
} else {
ctx->button_indices[button_index++] = cap->NotRange.DataIndex;
}
}
}
}
if (ctx->is_xinput && joystick->nbuttons == 10) {
ctx->guide_hack = SDL_TRUE;
joystick->nbuttons += 1;
}
for (i = 0; i < caps.NumberInputValueCaps; ++i) {
HIDP_VALUE_CAPS *cap = &value_caps[i];
if (cap->IsRange) {
continue;
}
if (ctx->trigger_hack && cap->NotRange.Usage == USB_USAGE_GENERIC_Z) {
continue;
}
if (cap->NotRange.Usage == USB_USAGE_GENERIC_HAT) {
joystick->nhats += 1;
continue;
}
if (ctx->is_xinput && cap->NotRange.Usage == USB_USAGE_GENERIC_Z) {
continue;
}
joystick->naxes += 1;
}
if (joystick->naxes > 0) {
int axis_index = 0;
ctx->axis_indices = (USHORT *)SDL_malloc(joystick->naxes * sizeof(*ctx->axis_indices));
if (!ctx->axis_indices) {
RAWINPUT_JoystickClose(joystick);
return SDL_OutOfMemory();
}
for (i = 0; i < caps.NumberInputValueCaps; ++i) {
HIDP_VALUE_CAPS *cap = &value_caps[i];
if (cap->IsRange) {
continue;
}
if (cap->NotRange.Usage == USB_USAGE_GENERIC_HAT) {
continue;
}
if (ctx->is_xinput && cap->NotRange.Usage == USB_USAGE_GENERIC_Z) {
ctx->trigger_hack = SDL_TRUE;
ctx->trigger_hack_index = cap->NotRange.DataIndex;
continue;
}
ctx->axis_indices[axis_index++] = cap->NotRange.DataIndex;
}
}
if (ctx->trigger_hack) {
joystick->naxes += 2;
}
if (joystick->nhats > 0) {
int hat_index = 0;
ctx->hat_indices = (USHORT *)SDL_malloc(joystick->nhats * sizeof(*ctx->hat_indices));
if (!ctx->hat_indices) {
RAWINPUT_JoystickClose(joystick);
return SDL_OutOfMemory();
}
for (i = 0; i < caps.NumberInputValueCaps; ++i) {
HIDP_VALUE_CAPS *cap = &value_caps[i];
if (cap->IsRange) {
continue;
}
if (cap->NotRange.Usage != USB_USAGE_GENERIC_HAT) {
continue;
}
ctx->hat_indices[hat_index++] = cap->NotRange.DataIndex;
}
}
2022-02-02 07:58:33 -06:00
joystick->epowerlevel = SDL_JOYSTICK_POWER_UNKNOWN;
Fixed bug 4477 - Support more than 4 XInput-capable devices on Windows Jimb Esser Add new RawInput controller API, and improved correlation with XInput/WGI Reorder joystick init so drivers can ask the others if they handle a device reliably Do not poll disconnected XInput devices (major perf issue) Fix various cases where incorrect correlation could happen Simple mechanism for propagating unhandled Guide button presses even before guaranteed correlation Correlate by axis motion as well as button presses Fix failing to zero other trigger Fix SDL_HINT_JOYSTICK_HIDAPI not working if set before calling SDL_Init() Add missing device to device names Disable RawInput if we have a mismatch of XInput-capable but not RawInput-capable devices Updated to SDL 2.0.13 code with the following notes: New HID driver: xbox360w - no idea what that is, hopefully urelated SDL_hidapijoystick.c had been refactored to couple data handling logic with device opening logic and device lists caused some problems, yields slightly uglier integration than previously when the 360 HID device driver was just handling the data. SDL_hidapijoystick.c now often pulls the device off of the joystick_hwdata structure for some rumble logic, but it appears that code path is never reached, so probably not a problem. Looks like joystick_hwdata was refactored to not include a mutex in other drivers, maintainers may want to do the same refactor here if that's useful for some reason. Something changed in how devices get names, so getting generic names. Had to fix a (new?) bug where removing an XInput controller caused existing controllers (that moved to a new XInput index) to get identified as 0x045e/0x02fd ("it's probably Bluetooth" in code), rendering the existing HIDAPI_IsDevicePresent and new RAWINPUT_IsDevicePresent unreliable.
2020-03-16 14:25:02 -05:00
return 0;
}
static int RAWINPUT_JoystickRumble(SDL_Joystick *joystick, Uint16 low_frequency_rumble, Uint16 high_frequency_rumble)
Fixed bug 4477 - Support more than 4 XInput-capable devices on Windows Jimb Esser Add new RawInput controller API, and improved correlation with XInput/WGI Reorder joystick init so drivers can ask the others if they handle a device reliably Do not poll disconnected XInput devices (major perf issue) Fix various cases where incorrect correlation could happen Simple mechanism for propagating unhandled Guide button presses even before guaranteed correlation Correlate by axis motion as well as button presses Fix failing to zero other trigger Fix SDL_HINT_JOYSTICK_HIDAPI not working if set before calling SDL_Init() Add missing device to device names Disable RawInput if we have a mismatch of XInput-capable but not RawInput-capable devices Updated to SDL 2.0.13 code with the following notes: New HID driver: xbox360w - no idea what that is, hopefully urelated SDL_hidapijoystick.c had been refactored to couple data handling logic with device opening logic and device lists caused some problems, yields slightly uglier integration than previously when the 360 HID device driver was just handling the data. SDL_hidapijoystick.c now often pulls the device off of the joystick_hwdata structure for some rumble logic, but it appears that code path is never reached, so probably not a problem. Looks like joystick_hwdata was refactored to not include a mutex in other drivers, maintainers may want to do the same refactor here if that's useful for some reason. Something changed in how devices get names, so getting generic names. Had to fix a (new?) bug where removing an XInput controller caused existing controllers (that moved to a new XInput index) to get identified as 0x045e/0x02fd ("it's probably Bluetooth" in code), rendering the existing HIDAPI_IsDevicePresent and new RAWINPUT_IsDevicePresent unreliable.
2020-03-16 14:25:02 -05:00
{
#if defined(SDL_JOYSTICK_RAWINPUT_WGI) || defined(SDL_JOYSTICK_RAWINPUT_XINPUT)
RAWINPUT_DeviceContext *ctx = joystick->hwdata;
2021-01-16 20:00:50 -06:00
#endif
2022-05-02 13:12:13 -05:00
SDL_bool rumbled = SDL_FALSE;
#ifdef SDL_JOYSTICK_RAWINPUT_WGI
if (!rumbled && ctx->wgi_correlated) {
WindowsGamingInputGamepadState *gamepad_state = ctx->wgi_slot;
HRESULT hr;
gamepad_state->vibration.LeftMotor = (DOUBLE)low_frequency_rumble / SDL_MAX_UINT16;
gamepad_state->vibration.RightMotor = (DOUBLE)high_frequency_rumble / SDL_MAX_UINT16;
hr = __x_ABI_CWindows_CGaming_CInput_CIGamepad_put_Vibration(gamepad_state->gamepad, gamepad_state->vibration);
if (SUCCEEDED(hr)) {
rumbled = SDL_TRUE;
}
}
#endif
#ifdef SDL_JOYSTICK_RAWINPUT_XINPUT
if (!rumbled && ctx->xinput_correlated) {
XINPUT_VIBRATION XVibration;
if (XINPUTSETSTATE == NULL) {
return SDL_Unsupported();
}
Fixed bug 4477 - Support more than 4 XInput-capable devices on Windows Jimb Esser Add new RawInput controller API, and improved correlation with XInput/WGI Reorder joystick init so drivers can ask the others if they handle a device reliably Do not poll disconnected XInput devices (major perf issue) Fix various cases where incorrect correlation could happen Simple mechanism for propagating unhandled Guide button presses even before guaranteed correlation Correlate by axis motion as well as button presses Fix failing to zero other trigger Fix SDL_HINT_JOYSTICK_HIDAPI not working if set before calling SDL_Init() Add missing device to device names Disable RawInput if we have a mismatch of XInput-capable but not RawInput-capable devices Updated to SDL 2.0.13 code with the following notes: New HID driver: xbox360w - no idea what that is, hopefully urelated SDL_hidapijoystick.c had been refactored to couple data handling logic with device opening logic and device lists caused some problems, yields slightly uglier integration than previously when the 360 HID device driver was just handling the data. SDL_hidapijoystick.c now often pulls the device off of the joystick_hwdata structure for some rumble logic, but it appears that code path is never reached, so probably not a problem. Looks like joystick_hwdata was refactored to not include a mutex in other drivers, maintainers may want to do the same refactor here if that's useful for some reason. Something changed in how devices get names, so getting generic names. Had to fix a (new?) bug where removing an XInput controller caused existing controllers (that moved to a new XInput index) to get identified as 0x045e/0x02fd ("it's probably Bluetooth" in code), rendering the existing HIDAPI_IsDevicePresent and new RAWINPUT_IsDevicePresent unreliable.
2020-03-16 14:25:02 -05:00
XVibration.wLeftMotorSpeed = low_frequency_rumble;
XVibration.wRightMotorSpeed = high_frequency_rumble;
if (XINPUTSETSTATE(ctx->xinput_slot, &XVibration) == ERROR_SUCCESS) {
rumbled = SDL_TRUE;
} else {
return SDL_SetError("XInputSetState() failed");
}
}
#endif /* SDL_JOYSTICK_RAWINPUT_XINPUT */
if (!rumbled) {
#if defined(SDL_JOYSTICK_RAWINPUT_WGI) || defined(SDL_JOYSTICK_RAWINPUT_XINPUT)
return SDL_SetError("Controller isn't correlated yet, try hitting a button first");
#else
return SDL_Unsupported();
#endif
}
return 0;
Fixed bug 4477 - Support more than 4 XInput-capable devices on Windows Jimb Esser Add new RawInput controller API, and improved correlation with XInput/WGI Reorder joystick init so drivers can ask the others if they handle a device reliably Do not poll disconnected XInput devices (major perf issue) Fix various cases where incorrect correlation could happen Simple mechanism for propagating unhandled Guide button presses even before guaranteed correlation Correlate by axis motion as well as button presses Fix failing to zero other trigger Fix SDL_HINT_JOYSTICK_HIDAPI not working if set before calling SDL_Init() Add missing device to device names Disable RawInput if we have a mismatch of XInput-capable but not RawInput-capable devices Updated to SDL 2.0.13 code with the following notes: New HID driver: xbox360w - no idea what that is, hopefully urelated SDL_hidapijoystick.c had been refactored to couple data handling logic with device opening logic and device lists caused some problems, yields slightly uglier integration than previously when the 360 HID device driver was just handling the data. SDL_hidapijoystick.c now often pulls the device off of the joystick_hwdata structure for some rumble logic, but it appears that code path is never reached, so probably not a problem. Looks like joystick_hwdata was refactored to not include a mutex in other drivers, maintainers may want to do the same refactor here if that's useful for some reason. Something changed in how devices get names, so getting generic names. Had to fix a (new?) bug where removing an XInput controller caused existing controllers (that moved to a new XInput index) to get identified as 0x045e/0x02fd ("it's probably Bluetooth" in code), rendering the existing HIDAPI_IsDevicePresent and new RAWINPUT_IsDevicePresent unreliable.
2020-03-16 14:25:02 -05:00
}
static int RAWINPUT_JoystickRumbleTriggers(SDL_Joystick *joystick, Uint16 left_rumble, Uint16 right_rumble)
{
#if defined(SDL_JOYSTICK_RAWINPUT_WGI)
RAWINPUT_DeviceContext *ctx = joystick->hwdata;
if (ctx->wgi_correlated) {
WindowsGamingInputGamepadState *gamepad_state = ctx->wgi_slot;
HRESULT hr;
gamepad_state->vibration.LeftTrigger = (DOUBLE)left_rumble / SDL_MAX_UINT16;
gamepad_state->vibration.RightTrigger = (DOUBLE)right_rumble / SDL_MAX_UINT16;
hr = __x_ABI_CWindows_CGaming_CInput_CIGamepad_put_Vibration(gamepad_state->gamepad, gamepad_state->vibration);
if (!SUCCEEDED(hr)) {
2021-11-26 12:18:08 -06:00
return SDL_SetError("Setting vibration failed: 0x%lx\n", hr);
}
2022-04-22 11:36:39 -05:00
return 0;
} else {
return SDL_SetError("Controller isn't correlated yet, try hitting a button first");
}
#else
return SDL_Unsupported();
#endif
}
static Uint32 RAWINPUT_JoystickGetCapabilities(SDL_Joystick *joystick)
{
Uint32 result = 0;
2022-06-13 14:15:56 -05:00
#if defined(SDL_JOYSTICK_RAWINPUT_XINPUT) || defined(SDL_JOYSTICK_RAWINPUT_WGI)
RAWINPUT_DeviceContext *ctx = joystick->hwdata;
#ifdef SDL_JOYSTICK_RAWINPUT_XINPUT
if (ctx->is_xinput) {
result |= SDL_JOYCAP_RUMBLE;
}
#endif
#ifdef SDL_JOYSTICK_RAWINPUT_WGI
if (ctx->is_xinput) {
result |= SDL_JOYCAP_RUMBLE;
if (ctx->is_xboxone) {
result |= SDL_JOYCAP_RUMBLE_TRIGGERS;
}
}
#endif
2022-06-13 14:15:56 -05:00
#endif /**/
return result;
}
static int RAWINPUT_JoystickSetLED(SDL_Joystick *joystick, Uint8 red, Uint8 green, Uint8 blue)
{
return SDL_Unsupported();
}
static int RAWINPUT_JoystickSendEffect(SDL_Joystick *joystick, const void *data, int size)
{
return SDL_Unsupported();
}
static int RAWINPUT_JoystickSetSensorsEnabled(SDL_Joystick *joystick, SDL_bool enabled)
{
return SDL_Unsupported();
}
static HIDP_DATA *GetData(USHORT index, HIDP_DATA *data, ULONG length)
{
ULONG i;
/* Check to see if the data is at the expected offset */
if (index < length && data[index].DataIndex == index) {
return &data[index];
}
/* Loop through the data to find it */
for (i = 0; i < length; ++i) {
if (data[i].DataIndex == index) {
return &data[i];
}
}
return NULL;
}
/* This is the packet format for Xbox 360 and Xbox One controllers on Windows,
however with this interface there is no rumble support, no guide button,
and the left and right triggers are tied together as a single axis.
We use XInput and Windows.Gaming.Input to make up for these shortcomings.
*/
static void RAWINPUT_HandleStatePacket(SDL_Joystick *joystick, Uint8 *data, int size)
{
RAWINPUT_DeviceContext *ctx = joystick->hwdata;
#ifdef SDL_JOYSTICK_RAWINPUT_MATCHING
/* Map new buttons and axes into game controller controls */
static const int button_map[] = {
SDL_GAMEPAD_BUTTON_A,
SDL_GAMEPAD_BUTTON_B,
SDL_GAMEPAD_BUTTON_X,
SDL_GAMEPAD_BUTTON_Y,
SDL_GAMEPAD_BUTTON_LEFT_SHOULDER,
SDL_GAMEPAD_BUTTON_RIGHT_SHOULDER,
SDL_GAMEPAD_BUTTON_BACK,
SDL_GAMEPAD_BUTTON_START,
SDL_GAMEPAD_BUTTON_LEFT_STICK,
SDL_GAMEPAD_BUTTON_RIGHT_STICK
};
#define HAT_MASK ((1 << SDL_GAMEPAD_BUTTON_DPAD_UP) | (1 << SDL_GAMEPAD_BUTTON_DPAD_DOWN) | (1 << SDL_GAMEPAD_BUTTON_DPAD_LEFT) | (1 << SDL_GAMEPAD_BUTTON_DPAD_RIGHT))
static const int hat_map[] = {
0,
(1 << SDL_GAMEPAD_BUTTON_DPAD_UP),
(1 << SDL_GAMEPAD_BUTTON_DPAD_UP) | (1 << SDL_GAMEPAD_BUTTON_DPAD_RIGHT),
(1 << SDL_GAMEPAD_BUTTON_DPAD_RIGHT),
(1 << SDL_GAMEPAD_BUTTON_DPAD_DOWN) | (1 << SDL_GAMEPAD_BUTTON_DPAD_RIGHT),
(1 << SDL_GAMEPAD_BUTTON_DPAD_DOWN),
(1 << SDL_GAMEPAD_BUTTON_DPAD_DOWN) | (1 << SDL_GAMEPAD_BUTTON_DPAD_LEFT),
(1 << SDL_GAMEPAD_BUTTON_DPAD_LEFT),
(1 << SDL_GAMEPAD_BUTTON_DPAD_UP) | (1 << SDL_GAMEPAD_BUTTON_DPAD_LEFT),
0,
};
Uint64 match_state = ctx->match_state;
/* Update match_state with button bit, then fall through */
2022-12-27 17:05:51 -06:00
#define SDL_SendJoystickButton(timestamp, joystick, button, state) \
if (button < SDL_arraysize(button_map)) { \
Uint64 button_bit = 1ull << button_map[button]; \
match_state = (match_state & ~button_bit) | (button_bit * (state)); \
} \
2022-12-27 17:05:51 -06:00
SDL_SendJoystickButton(timestamp, joystick, button, state)
#ifdef SDL_JOYSTICK_RAWINPUT_MATCH_AXES
/* Grab high 4 bits of value, then fall through */
#define AddAxisToMatchState(axis, value) \
{ \
match_state = (match_state & ~(0xFull << (4 * axis + 16))) | ((value)&0xF000ull) << (4 * axis + 4); \
}
2022-12-27 17:05:51 -06:00
#define SDL_SendJoystickAxis(timestamp, joystick, axis, value) \
if (axis < 4) \
AddAxisToMatchState(axis, value); \
2022-12-27 17:05:51 -06:00
SDL_SendJoystickAxis(timestamp, joystick, axis, value)
#endif
#endif /* SDL_JOYSTICK_RAWINPUT_MATCHING */
ULONG data_length = ctx->max_data_length;
int i;
int nbuttons = joystick->nbuttons - (ctx->guide_hack * 1);
int naxes = joystick->naxes - (ctx->trigger_hack * 2);
int nhats = joystick->nhats;
Uint32 button_mask = 0;
Uint64 timestamp = SDL_GetTicksNS();
if (SDL_HidP_GetData(HidP_Input, ctx->data, &data_length, ctx->preparsed_data, (PCHAR)data, size) != HIDP_STATUS_SUCCESS) {
return;
}
for (i = 0; i < nbuttons; ++i) {
HIDP_DATA *item = GetData(ctx->button_indices[i], ctx->data, data_length);
if (item && item->On) {
button_mask |= (1 << i);
}
}
for (i = 0; i < nbuttons; ++i) {
2022-12-27 17:05:51 -06:00
SDL_SendJoystickButton(timestamp, joystick, i, (button_mask & (1 << i)) ? SDL_PRESSED : SDL_RELEASED);
}
for (i = 0; i < naxes; ++i) {
HIDP_DATA *item = GetData(ctx->axis_indices[i], ctx->data, data_length);
if (item) {
Sint16 axis = (int)(Uint16)item->RawValue - 0x8000;
2022-12-27 17:05:51 -06:00
SDL_SendJoystickAxis(timestamp, joystick, i, axis);
}
}
for (i = 0; i < nhats; ++i) {
HIDP_DATA *item = GetData(ctx->hat_indices[i], ctx->data, data_length);
if (item) {
Uint8 hat = SDL_HAT_CENTERED;
const Uint8 hat_states[] = {
SDL_HAT_CENTERED,
SDL_HAT_UP,
SDL_HAT_UP | SDL_HAT_RIGHT,
SDL_HAT_RIGHT,
SDL_HAT_DOWN | SDL_HAT_RIGHT,
SDL_HAT_DOWN,
SDL_HAT_DOWN | SDL_HAT_LEFT,
SDL_HAT_LEFT,
SDL_HAT_UP | SDL_HAT_LEFT,
SDL_HAT_CENTERED,
};
ULONG state = item->RawValue;
if (state < SDL_arraysize(hat_states)) {
#ifdef SDL_JOYSTICK_RAWINPUT_MATCHING
match_state = (match_state & ~HAT_MASK) | hat_map[state];
#endif
hat = hat_states[state];
}
2022-12-27 17:05:51 -06:00
SDL_SendJoystickHat(timestamp, joystick, i, hat);
}
}
2022-12-27 17:05:51 -06:00
#ifdef SDL_SendJoystickButton
#undef SDL_SendJoystickButton
#endif
2022-12-27 17:05:51 -06:00
#ifdef SDL_SendJoystickAxis
#undef SDL_SendJoystickAxis
#endif
#ifdef SDL_JOYSTICK_RAWINPUT_MATCH_TRIGGERS
#define AddTriggerToMatchState(axis, value) \
{ \
int match_axis = axis + SDL_JOYSTICK_RAWINPUT_MATCH_COUNT - joystick->naxes; \
AddAxisToMatchState(match_axis, value); \
}
#endif /* SDL_JOYSTICK_RAWINPUT_MATCH_TRIGGERS */
if (ctx->trigger_hack) {
SDL_bool has_trigger_data = SDL_FALSE;
2022-01-20 21:31:54 -06:00
int left_trigger = joystick->naxes - 2;
int right_trigger = joystick->naxes - 1;
#ifdef SDL_JOYSTICK_RAWINPUT_XINPUT
/* Prefer XInput over WindowsGamingInput, it continues to provide data in the background */
if (!has_trigger_data && ctx->xinput_enabled && ctx->xinput_correlated) {
has_trigger_data = SDL_TRUE;
}
#endif /* SDL_JOYSTICK_RAWINPUT_XINPUT */
#ifdef SDL_JOYSTICK_RAWINPUT_WGI
if (!has_trigger_data && ctx->wgi_correlated) {
has_trigger_data = SDL_TRUE;
}
#endif /* SDL_JOYSTICK_RAWINPUT_WGI */
#ifndef SDL_JOYSTICK_RAWINPUT_MATCH_TRIGGERS
if (!has_trigger_data)
#endif
{
HIDP_DATA *item = GetData(ctx->trigger_hack_index, ctx->data, data_length);
if (item) {
Sint16 value = (int)(Uint16)item->RawValue - 0x8000;
Sint16 left_value = (value > 0) ? (value * 2 - 32767) : SDL_MIN_SINT16;
Sint16 right_value = (value < 0) ? (-value * 2 - 32769) : SDL_MIN_SINT16;
#ifdef SDL_JOYSTICK_RAWINPUT_MATCH_TRIGGERS
AddTriggerToMatchState(left_trigger, left_value);
AddTriggerToMatchState(right_trigger, right_value);
if (!has_trigger_data)
#endif /* SDL_JOYSTICK_RAWINPUT_MATCH_TRIGGERS */
{
2022-12-27 17:05:51 -06:00
SDL_SendJoystickAxis(timestamp, joystick, left_trigger, left_value);
SDL_SendJoystickAxis(timestamp, joystick, right_trigger, right_value);
}
}
}
}
#ifdef AddAxisToMatchState
#undef AddAxisToMatchState
#endif
#ifdef AddTriggerToMatchState
#undef AddTriggerToMatchState
#endif
#ifdef SDL_JOYSTICK_RAWINPUT_MATCHING
if (ctx->is_xinput) {
ctx->match_state = match_state;
ctx->last_state_packet = SDL_GetTicks();
}
#endif
}
static void RAWINPUT_UpdateOtherAPIs(SDL_Joystick *joystick)
{
#ifdef SDL_JOYSTICK_RAWINPUT_MATCHING
RAWINPUT_DeviceContext *ctx = joystick->hwdata;
SDL_bool has_trigger_data = SDL_FALSE;
SDL_bool correlated = SDL_FALSE;
WindowsMatchState match_state_xinput;
int guide_button = joystick->nbuttons - 1;
int left_trigger = joystick->naxes - 2;
int right_trigger = joystick->naxes - 1;
#ifdef SDL_JOYSTICK_RAWINPUT_WGI
SDL_bool xinput_correlated;
#endif
RAWINPUT_FillMatchState(&match_state_xinput, ctx->match_state);
#ifdef SDL_JOYSTICK_RAWINPUT_WGI
#ifdef SDL_JOYSTICK_RAWINPUT_XINPUT
xinput_correlated = ctx->xinput_correlated;
#else
xinput_correlated = SDL_FALSE;
#endif
/* Parallel logic to WINDOWS_XINPUT below */
RAWINPUT_UpdateWindowsGamingInput();
if (ctx->wgi_correlated &&
!joystick->low_frequency_rumble && !joystick->high_frequency_rumble &&
!joystick->left_trigger_rumble && !joystick->right_trigger_rumble) {
/* We have been previously correlated, ensure we are still matching, see comments in XINPUT section */
2022-05-02 13:12:13 -05:00
if (RAWINPUT_WindowsGamingInputSlotMatches(&match_state_xinput, ctx->wgi_slot, xinput_correlated)) {
ctx->wgi_uncorrelate_count = 0;
} else {
++ctx->wgi_uncorrelate_count;
/* Only un-correlate if this is consistent over multiple Update() calls - the timing of polling/event
pumping can easily cause this to uncorrelate for a frame. 2 seemed reliable in my testing, but
let's set it to 5 to be safe. An incorrect un-correlation will simply result in lower precision
triggers for a frame. */
if (ctx->wgi_uncorrelate_count >= 5) {
#ifdef DEBUG_RAWINPUT
SDL_Log("UN-Correlated joystick %d to WindowsGamingInput device #%d\n", joystick->instance_id, ctx->wgi_slot);
#endif
RAWINPUT_MarkWindowsGamingInputSlotFree(ctx->wgi_slot);
ctx->wgi_correlated = SDL_FALSE;
ctx->wgi_correlation_count = 0;
/* Force release of Guide button, it can't possibly be down on this device now. */
/* It gets left down if we were actually correlated incorrectly and it was released on the WindowsGamingInput
device but we didn't get a state packet. */
if (ctx->guide_hack) {
2022-12-27 17:05:51 -06:00
SDL_SendJoystickButton(0, joystick, guide_button, SDL_RELEASED);
}
}
}
}
if (!ctx->wgi_correlated) {
SDL_bool new_correlation_count = 0;
if (RAWINPUT_MissingWindowsGamingInputSlot()) {
Uint8 correlation_id;
2021-11-26 12:18:08 -06:00
WindowsGamingInputGamepadState *slot_idx = NULL;
2022-05-02 13:12:13 -05:00
if (RAWINPUT_GuessWindowsGamingInputSlot(&match_state_xinput, &correlation_id, &slot_idx, xinput_correlated)) {
/* we match exactly one WindowsGamingInput device */
/* Probably can do without wgi_correlation_count, just check and clear wgi_slot to NULL, unless we need
even more frames to be sure. */
if (ctx->wgi_correlation_count && ctx->wgi_slot == slot_idx) {
/* was correlated previously, and still the same device */
if (ctx->wgi_correlation_id + 1 == correlation_id) {
/* no one else was correlated in the meantime */
new_correlation_count = ctx->wgi_correlation_count + 1;
if (new_correlation_count == 2) {
/* correlation stayed steady and uncontested across multiple frames, guaranteed match */
ctx->wgi_correlated = SDL_TRUE;
#ifdef DEBUG_RAWINPUT
SDL_Log("Correlated joystick %d to WindowsGamingInput device #%d\n", joystick->instance_id, slot_idx);
#endif
correlated = SDL_TRUE;
RAWINPUT_MarkWindowsGamingInputSlotUsed(ctx->wgi_slot, ctx);
/* If the generalized Guide button was using us, it doesn't need to anymore */
if (guide_button_candidate.joystick == joystick) {
guide_button_candidate.joystick = NULL;
}
if (guide_button_candidate.last_joystick == joystick) {
guide_button_candidate.last_joystick = NULL;
}
}
} else {
/* someone else also possibly correlated to this device, start over */
new_correlation_count = 1;
}
} else {
/* new possible correlation */
new_correlation_count = 1;
ctx->wgi_slot = slot_idx;
}
ctx->wgi_correlation_id = correlation_id;
} else {
/* Match multiple WindowsGamingInput devices, or none (possibly due to no buttons pressed) */
}
}
ctx->wgi_correlation_count = new_correlation_count;
} else {
correlated = SDL_TRUE;
}
#endif /* SDL_JOYSTICK_RAWINPUT_WGI */
#ifdef SDL_JOYSTICK_RAWINPUT_XINPUT
/* Parallel logic to WINDOWS_GAMING_INPUT above */
if (ctx->xinput_enabled) {
RAWINPUT_UpdateXInput();
if (ctx->xinput_correlated &&
!joystick->low_frequency_rumble && !joystick->high_frequency_rumble) {
/* We have been previously correlated, ensure we are still matching */
/* This is required to deal with two (mostly) un-preventable mis-correlation situations:
A) Since the HID data stream does not provide an initial state (but polling XInput does), if we open
5 controllers (#1-4 XInput mapped, #5 is not), and controller 1 had the A button down (and we don't
know), and the user presses A on controller #5, we'll see exactly 1 controller with A down (#5) and
exactly 1 XInput device with A down (#1), and incorrectly correlate. This code will then un-correlate
when A is released from either controller #1 or #5.
B) Since the app may not open all controllers, we could have a similar situation where only controller #5
is opened, and the user holds A on controllers #1 and #5 simultaneously - again we see only 1 controller
with A down and 1 XInput device with A down, and incorrectly correlate. This should be very unusual
(only when apps do not open all controllers, yet are listening to Guide button presses, yet
for some reason want to ignore guide button presses on the un-opened controllers, yet users are
pressing buttons on the unopened controllers), and will resolve itself when either button is released
and we un-correlate. We could prevent this by processing the state packets for *all* controllers,
even un-opened ones, as that would allow more precise correlation.
*/
if (RAWINPUT_XInputSlotMatches(&match_state_xinput, ctx->xinput_slot)) {
ctx->xinput_uncorrelate_count = 0;
} else {
++ctx->xinput_uncorrelate_count;
/* Only un-correlate if this is consistent over multiple Update() calls - the timing of polling/event
pumping can easily cause this to uncorrelate for a frame. 2 seemed reliable in my testing, but
let's set it to 5 to be safe. An incorrect un-correlation will simply result in lower precision
triggers for a frame. */
if (ctx->xinput_uncorrelate_count >= 5) {
#ifdef DEBUG_RAWINPUT
SDL_Log("UN-Correlated joystick %d to XInput device #%d\n", joystick->instance_id, ctx->xinput_slot);
#endif
RAWINPUT_MarkXInputSlotFree(ctx->xinput_slot);
ctx->xinput_correlated = SDL_FALSE;
ctx->xinput_correlation_count = 0;
/* Force release of Guide button, it can't possibly be down on this device now. */
/* It gets left down if we were actually correlated incorrectly and it was released on the XInput
device but we didn't get a state packet. */
if (ctx->guide_hack) {
2022-12-27 17:05:51 -06:00
SDL_SendJoystickButton(0, joystick, guide_button, SDL_RELEASED);
}
}
}
}
if (!ctx->xinput_correlated) {
Uint8 new_correlation_count = 0;
if (RAWINPUT_MissingXInputSlot()) {
Uint8 correlation_id = 0;
Uint8 slot_idx = 0;
if (RAWINPUT_GuessXInputSlot(&match_state_xinput, &correlation_id, &slot_idx)) {
/* we match exactly one XInput device */
/* Probably can do without xinput_correlation_count, just check and clear xinput_slot to ANY, unless
we need even more frames to be sure */
if (ctx->xinput_correlation_count && ctx->xinput_slot == slot_idx) {
/* was correlated previously, and still the same device */
if (ctx->xinput_correlation_id + 1 == correlation_id) {
/* no one else was correlated in the meantime */
new_correlation_count = ctx->xinput_correlation_count + 1;
if (new_correlation_count == 2) {
/* correlation stayed steady and uncontested across multiple frames, guaranteed match */
ctx->xinput_correlated = SDL_TRUE;
#ifdef DEBUG_RAWINPUT
SDL_Log("Correlated joystick %d to XInput device #%d\n", joystick->instance_id, slot_idx);
#endif
correlated = SDL_TRUE;
RAWINPUT_MarkXInputSlotUsed(ctx->xinput_slot);
/* If the generalized Guide button was using us, it doesn't need to anymore */
if (guide_button_candidate.joystick == joystick) {
guide_button_candidate.joystick = NULL;
}
if (guide_button_candidate.last_joystick == joystick) {
guide_button_candidate.last_joystick = NULL;
}
}
} else {
/* someone else also possibly correlated to this device, start over */
new_correlation_count = 1;
}
} else {
/* new possible correlation */
new_correlation_count = 1;
ctx->xinput_slot = slot_idx;
}
ctx->xinput_correlation_id = correlation_id;
} else {
/* Match multiple XInput devices, or none (possibly due to no buttons pressed) */
}
}
ctx->xinput_correlation_count = new_correlation_count;
} else {
correlated = SDL_TRUE;
}
}
#endif /* SDL_JOYSTICK_RAWINPUT_XINPUT */
/* Poll for trigger data once (not per-state-packet) */
#ifdef SDL_JOYSTICK_RAWINPUT_XINPUT
/* Prefer XInput over WindowsGamingInput, it continues to provide data in the background */
if (!has_trigger_data && ctx->xinput_enabled && ctx->xinput_correlated) {
RAWINPUT_UpdateXInput();
if (xinput_state[ctx->xinput_slot].connected) {
XINPUT_BATTERY_INFORMATION_EX *battery_info = &xinput_state[ctx->xinput_slot].battery;
Uint64 timestamp;
if (ctx->guide_hack || ctx->trigger_hack) {
timestamp = SDL_GetTicksNS();
2022-12-04 11:39:55 -06:00
} else {
/* timestamp won't be used */
timestamp = 0;
}
if (ctx->guide_hack) {
2022-12-27 17:05:51 -06:00
SDL_SendJoystickButton(timestamp, joystick, guide_button, (xinput_state[ctx->xinput_slot].state.Gamepad.wButtons & XINPUT_GAMEPAD_GUIDE) ? SDL_PRESSED : SDL_RELEASED);
}
if (ctx->trigger_hack) {
2022-12-27 17:05:51 -06:00
SDL_SendJoystickAxis(timestamp, joystick, left_trigger, ((int)xinput_state[ctx->xinput_slot].state.Gamepad.bLeftTrigger * 257) - 32768);
SDL_SendJoystickAxis(timestamp, joystick, right_trigger, ((int)xinput_state[ctx->xinput_slot].state.Gamepad.bRightTrigger * 257) - 32768);
}
has_trigger_data = SDL_TRUE;
if (battery_info->BatteryType != BATTERY_TYPE_UNKNOWN &&
battery_info->BatteryType != BATTERY_TYPE_DISCONNECTED) {
SDL_JoystickPowerLevel ePowerLevel = SDL_JOYSTICK_POWER_UNKNOWN;
if (battery_info->BatteryType == BATTERY_TYPE_WIRED) {
ePowerLevel = SDL_JOYSTICK_POWER_WIRED;
} else {
switch (battery_info->BatteryLevel) {
case BATTERY_LEVEL_EMPTY:
ePowerLevel = SDL_JOYSTICK_POWER_EMPTY;
break;
case BATTERY_LEVEL_LOW:
ePowerLevel = SDL_JOYSTICK_POWER_LOW;
break;
case BATTERY_LEVEL_MEDIUM:
ePowerLevel = SDL_JOYSTICK_POWER_MEDIUM;
break;
default:
case BATTERY_LEVEL_FULL:
ePowerLevel = SDL_JOYSTICK_POWER_FULL;
break;
}
}
2022-12-27 17:05:51 -06:00
SDL_SendJoystickBatteryLevel(joystick, ePowerLevel);
}
}
}
#endif /* SDL_JOYSTICK_RAWINPUT_XINPUT */
#ifdef SDL_JOYSTICK_RAWINPUT_WGI
if (!has_trigger_data && ctx->wgi_correlated) {
RAWINPUT_UpdateWindowsGamingInput(); /* May detect disconnect / cause uncorrelation */
if (ctx->wgi_correlated) { /* Still connected */
struct __x_ABI_CWindows_CGaming_CInput_CGamepadReading *state = &ctx->wgi_slot->state;
Uint64 timestamp;
if (ctx->guide_hack || ctx->trigger_hack) {
timestamp = SDL_GetTicksNS();
2022-12-04 11:39:55 -06:00
} else {
/* timestamp won't be used */
timestamp = 0;
}
if (ctx->guide_hack) {
2022-12-27 17:05:51 -06:00
SDL_SendJoystickButton(timestamp, joystick, guide_button, (state->Buttons & GamepadButtons_GUIDE) ? SDL_PRESSED : SDL_RELEASED);
}
if (ctx->trigger_hack) {
2022-12-27 17:05:51 -06:00
SDL_SendJoystickAxis(timestamp, joystick, left_trigger, ((int)(state->LeftTrigger * SDL_MAX_UINT16)) - 32768);
SDL_SendJoystickAxis(timestamp, joystick, right_trigger, ((int)(state->RightTrigger * SDL_MAX_UINT16)) - 32768);
}
has_trigger_data = SDL_TRUE;
}
}
#endif /* SDL_JOYSTICK_RAWINPUT_WGI */
if (!correlated) {
if (!guide_button_candidate.joystick ||
(ctx->last_state_packet && (!guide_button_candidate.last_state_packet ||
ctx->last_state_packet >= guide_button_candidate.last_state_packet))) {
guide_button_candidate.joystick = joystick;
guide_button_candidate.last_state_packet = ctx->last_state_packet;
}
}
#endif /* SDL_JOYSTICK_RAWINPUT_MATCHING */
}
static void RAWINPUT_JoystickUpdate(SDL_Joystick *joystick)
Fixed bug 4477 - Support more than 4 XInput-capable devices on Windows Jimb Esser Add new RawInput controller API, and improved correlation with XInput/WGI Reorder joystick init so drivers can ask the others if they handle a device reliably Do not poll disconnected XInput devices (major perf issue) Fix various cases where incorrect correlation could happen Simple mechanism for propagating unhandled Guide button presses even before guaranteed correlation Correlate by axis motion as well as button presses Fix failing to zero other trigger Fix SDL_HINT_JOYSTICK_HIDAPI not working if set before calling SDL_Init() Add missing device to device names Disable RawInput if we have a mismatch of XInput-capable but not RawInput-capable devices Updated to SDL 2.0.13 code with the following notes: New HID driver: xbox360w - no idea what that is, hopefully urelated SDL_hidapijoystick.c had been refactored to couple data handling logic with device opening logic and device lists caused some problems, yields slightly uglier integration than previously when the 360 HID device driver was just handling the data. SDL_hidapijoystick.c now often pulls the device off of the joystick_hwdata structure for some rumble logic, but it appears that code path is never reached, so probably not a problem. Looks like joystick_hwdata was refactored to not include a mutex in other drivers, maintainers may want to do the same refactor here if that's useful for some reason. Something changed in how devices get names, so getting generic names. Had to fix a (new?) bug where removing an XInput controller caused existing controllers (that moved to a new XInput index) to get identified as 0x045e/0x02fd ("it's probably Bluetooth" in code), rendering the existing HIDAPI_IsDevicePresent and new RAWINPUT_IsDevicePresent unreliable.
2020-03-16 14:25:02 -05:00
{
RAWINPUT_UpdateOtherAPIs(joystick);
Fixed bug 4477 - Support more than 4 XInput-capable devices on Windows Jimb Esser Add new RawInput controller API, and improved correlation with XInput/WGI Reorder joystick init so drivers can ask the others if they handle a device reliably Do not poll disconnected XInput devices (major perf issue) Fix various cases where incorrect correlation could happen Simple mechanism for propagating unhandled Guide button presses even before guaranteed correlation Correlate by axis motion as well as button presses Fix failing to zero other trigger Fix SDL_HINT_JOYSTICK_HIDAPI not working if set before calling SDL_Init() Add missing device to device names Disable RawInput if we have a mismatch of XInput-capable but not RawInput-capable devices Updated to SDL 2.0.13 code with the following notes: New HID driver: xbox360w - no idea what that is, hopefully urelated SDL_hidapijoystick.c had been refactored to couple data handling logic with device opening logic and device lists caused some problems, yields slightly uglier integration than previously when the 360 HID device driver was just handling the data. SDL_hidapijoystick.c now often pulls the device off of the joystick_hwdata structure for some rumble logic, but it appears that code path is never reached, so probably not a problem. Looks like joystick_hwdata was refactored to not include a mutex in other drivers, maintainers may want to do the same refactor here if that's useful for some reason. Something changed in how devices get names, so getting generic names. Had to fix a (new?) bug where removing an XInput controller caused existing controllers (that moved to a new XInput index) to get identified as 0x045e/0x02fd ("it's probably Bluetooth" in code), rendering the existing HIDAPI_IsDevicePresent and new RAWINPUT_IsDevicePresent unreliable.
2020-03-16 14:25:02 -05:00
}
static void RAWINPUT_JoystickClose(SDL_Joystick *joystick)
Fixed bug 4477 - Support more than 4 XInput-capable devices on Windows Jimb Esser Add new RawInput controller API, and improved correlation with XInput/WGI Reorder joystick init so drivers can ask the others if they handle a device reliably Do not poll disconnected XInput devices (major perf issue) Fix various cases where incorrect correlation could happen Simple mechanism for propagating unhandled Guide button presses even before guaranteed correlation Correlate by axis motion as well as button presses Fix failing to zero other trigger Fix SDL_HINT_JOYSTICK_HIDAPI not working if set before calling SDL_Init() Add missing device to device names Disable RawInput if we have a mismatch of XInput-capable but not RawInput-capable devices Updated to SDL 2.0.13 code with the following notes: New HID driver: xbox360w - no idea what that is, hopefully urelated SDL_hidapijoystick.c had been refactored to couple data handling logic with device opening logic and device lists caused some problems, yields slightly uglier integration than previously when the 360 HID device driver was just handling the data. SDL_hidapijoystick.c now often pulls the device off of the joystick_hwdata structure for some rumble logic, but it appears that code path is never reached, so probably not a problem. Looks like joystick_hwdata was refactored to not include a mutex in other drivers, maintainers may want to do the same refactor here if that's useful for some reason. Something changed in how devices get names, so getting generic names. Had to fix a (new?) bug where removing an XInput controller caused existing controllers (that moved to a new XInput index) to get identified as 0x045e/0x02fd ("it's probably Bluetooth" in code), rendering the existing HIDAPI_IsDevicePresent and new RAWINPUT_IsDevicePresent unreliable.
2020-03-16 14:25:02 -05:00
{
RAWINPUT_DeviceContext *ctx = joystick->hwdata;
Fixed bug 4477 - Support more than 4 XInput-capable devices on Windows Jimb Esser Add new RawInput controller API, and improved correlation with XInput/WGI Reorder joystick init so drivers can ask the others if they handle a device reliably Do not poll disconnected XInput devices (major perf issue) Fix various cases where incorrect correlation could happen Simple mechanism for propagating unhandled Guide button presses even before guaranteed correlation Correlate by axis motion as well as button presses Fix failing to zero other trigger Fix SDL_HINT_JOYSTICK_HIDAPI not working if set before calling SDL_Init() Add missing device to device names Disable RawInput if we have a mismatch of XInput-capable but not RawInput-capable devices Updated to SDL 2.0.13 code with the following notes: New HID driver: xbox360w - no idea what that is, hopefully urelated SDL_hidapijoystick.c had been refactored to couple data handling logic with device opening logic and device lists caused some problems, yields slightly uglier integration than previously when the 360 HID device driver was just handling the data. SDL_hidapijoystick.c now often pulls the device off of the joystick_hwdata structure for some rumble logic, but it appears that code path is never reached, so probably not a problem. Looks like joystick_hwdata was refactored to not include a mutex in other drivers, maintainers may want to do the same refactor here if that's useful for some reason. Something changed in how devices get names, so getting generic names. Had to fix a (new?) bug where removing an XInput controller caused existing controllers (that moved to a new XInput index) to get identified as 0x045e/0x02fd ("it's probably Bluetooth" in code), rendering the existing HIDAPI_IsDevicePresent and new RAWINPUT_IsDevicePresent unreliable.
2020-03-16 14:25:02 -05:00
#ifdef SDL_JOYSTICK_RAWINPUT_MATCHING
if (guide_button_candidate.joystick == joystick) {
guide_button_candidate.joystick = NULL;
}
if (guide_button_candidate.last_joystick == joystick) {
guide_button_candidate.last_joystick = NULL;
}
#endif
if (ctx) {
Fixed bug 4477 - Support more than 4 XInput-capable devices on Windows Jimb Esser Add new RawInput controller API, and improved correlation with XInput/WGI Reorder joystick init so drivers can ask the others if they handle a device reliably Do not poll disconnected XInput devices (major perf issue) Fix various cases where incorrect correlation could happen Simple mechanism for propagating unhandled Guide button presses even before guaranteed correlation Correlate by axis motion as well as button presses Fix failing to zero other trigger Fix SDL_HINT_JOYSTICK_HIDAPI not working if set before calling SDL_Init() Add missing device to device names Disable RawInput if we have a mismatch of XInput-capable but not RawInput-capable devices Updated to SDL 2.0.13 code with the following notes: New HID driver: xbox360w - no idea what that is, hopefully urelated SDL_hidapijoystick.c had been refactored to couple data handling logic with device opening logic and device lists caused some problems, yields slightly uglier integration than previously when the 360 HID device driver was just handling the data. SDL_hidapijoystick.c now often pulls the device off of the joystick_hwdata structure for some rumble logic, but it appears that code path is never reached, so probably not a problem. Looks like joystick_hwdata was refactored to not include a mutex in other drivers, maintainers may want to do the same refactor here if that's useful for some reason. Something changed in how devices get names, so getting generic names. Had to fix a (new?) bug where removing an XInput controller caused existing controllers (that moved to a new XInput index) to get identified as 0x045e/0x02fd ("it's probably Bluetooth" in code), rendering the existing HIDAPI_IsDevicePresent and new RAWINPUT_IsDevicePresent unreliable.
2020-03-16 14:25:02 -05:00
SDL_RAWINPUT_Device *device;
#ifdef SDL_JOYSTICK_RAWINPUT_XINPUT
xinput_device_change = SDL_TRUE;
if (ctx->xinput_enabled) {
if (ctx->xinput_correlated) {
RAWINPUT_MarkXInputSlotFree(ctx->xinput_slot);
}
WIN_UnloadXInputDLL();
}
#endif
#ifdef SDL_JOYSTICK_RAWINPUT_WGI
RAWINPUT_QuitWindowsGamingInput(ctx);
#endif
device = ctx->device;
Fixed bug 4477 - Support more than 4 XInput-capable devices on Windows Jimb Esser Add new RawInput controller API, and improved correlation with XInput/WGI Reorder joystick init so drivers can ask the others if they handle a device reliably Do not poll disconnected XInput devices (major perf issue) Fix various cases where incorrect correlation could happen Simple mechanism for propagating unhandled Guide button presses even before guaranteed correlation Correlate by axis motion as well as button presses Fix failing to zero other trigger Fix SDL_HINT_JOYSTICK_HIDAPI not working if set before calling SDL_Init() Add missing device to device names Disable RawInput if we have a mismatch of XInput-capable but not RawInput-capable devices Updated to SDL 2.0.13 code with the following notes: New HID driver: xbox360w - no idea what that is, hopefully urelated SDL_hidapijoystick.c had been refactored to couple data handling logic with device opening logic and device lists caused some problems, yields slightly uglier integration than previously when the 360 HID device driver was just handling the data. SDL_hidapijoystick.c now often pulls the device off of the joystick_hwdata structure for some rumble logic, but it appears that code path is never reached, so probably not a problem. Looks like joystick_hwdata was refactored to not include a mutex in other drivers, maintainers may want to do the same refactor here if that's useful for some reason. Something changed in how devices get names, so getting generic names. Had to fix a (new?) bug where removing an XInput controller caused existing controllers (that moved to a new XInput index) to get identified as 0x045e/0x02fd ("it's probably Bluetooth" in code), rendering the existing HIDAPI_IsDevicePresent and new RAWINPUT_IsDevicePresent unreliable.
2020-03-16 14:25:02 -05:00
if (device) {
SDL_assert(device->joystick == joystick);
device->joystick = NULL;
RAWINPUT_ReleaseDevice(device);
Fixed bug 4477 - Support more than 4 XInput-capable devices on Windows Jimb Esser Add new RawInput controller API, and improved correlation with XInput/WGI Reorder joystick init so drivers can ask the others if they handle a device reliably Do not poll disconnected XInput devices (major perf issue) Fix various cases where incorrect correlation could happen Simple mechanism for propagating unhandled Guide button presses even before guaranteed correlation Correlate by axis motion as well as button presses Fix failing to zero other trigger Fix SDL_HINT_JOYSTICK_HIDAPI not working if set before calling SDL_Init() Add missing device to device names Disable RawInput if we have a mismatch of XInput-capable but not RawInput-capable devices Updated to SDL 2.0.13 code with the following notes: New HID driver: xbox360w - no idea what that is, hopefully urelated SDL_hidapijoystick.c had been refactored to couple data handling logic with device opening logic and device lists caused some problems, yields slightly uglier integration than previously when the 360 HID device driver was just handling the data. SDL_hidapijoystick.c now often pulls the device off of the joystick_hwdata structure for some rumble logic, but it appears that code path is never reached, so probably not a problem. Looks like joystick_hwdata was refactored to not include a mutex in other drivers, maintainers may want to do the same refactor here if that's useful for some reason. Something changed in how devices get names, so getting generic names. Had to fix a (new?) bug where removing an XInput controller caused existing controllers (that moved to a new XInput index) to get identified as 0x045e/0x02fd ("it's probably Bluetooth" in code), rendering the existing HIDAPI_IsDevicePresent and new RAWINPUT_IsDevicePresent unreliable.
2020-03-16 14:25:02 -05:00
}
SDL_free(ctx->data);
SDL_free(ctx->button_indices);
SDL_free(ctx->axis_indices);
SDL_free(ctx->hat_indices);
SDL_free(ctx);
Fixed bug 4477 - Support more than 4 XInput-capable devices on Windows Jimb Esser Add new RawInput controller API, and improved correlation with XInput/WGI Reorder joystick init so drivers can ask the others if they handle a device reliably Do not poll disconnected XInput devices (major perf issue) Fix various cases where incorrect correlation could happen Simple mechanism for propagating unhandled Guide button presses even before guaranteed correlation Correlate by axis motion as well as button presses Fix failing to zero other trigger Fix SDL_HINT_JOYSTICK_HIDAPI not working if set before calling SDL_Init() Add missing device to device names Disable RawInput if we have a mismatch of XInput-capable but not RawInput-capable devices Updated to SDL 2.0.13 code with the following notes: New HID driver: xbox360w - no idea what that is, hopefully urelated SDL_hidapijoystick.c had been refactored to couple data handling logic with device opening logic and device lists caused some problems, yields slightly uglier integration than previously when the 360 HID device driver was just handling the data. SDL_hidapijoystick.c now often pulls the device off of the joystick_hwdata structure for some rumble logic, but it appears that code path is never reached, so probably not a problem. Looks like joystick_hwdata was refactored to not include a mutex in other drivers, maintainers may want to do the same refactor here if that's useful for some reason. Something changed in how devices get names, so getting generic names. Had to fix a (new?) bug where removing an XInput controller caused existing controllers (that moved to a new XInput index) to get identified as 0x045e/0x02fd ("it's probably Bluetooth" in code), rendering the existing HIDAPI_IsDevicePresent and new RAWINPUT_IsDevicePresent unreliable.
2020-03-16 14:25:02 -05:00
joystick->hwdata = NULL;
}
}
SDL_bool
RAWINPUT_RegisterNotifications(HWND hWnd)
Fixed bug 4477 - Support more than 4 XInput-capable devices on Windows Jimb Esser Add new RawInput controller API, and improved correlation with XInput/WGI Reorder joystick init so drivers can ask the others if they handle a device reliably Do not poll disconnected XInput devices (major perf issue) Fix various cases where incorrect correlation could happen Simple mechanism for propagating unhandled Guide button presses even before guaranteed correlation Correlate by axis motion as well as button presses Fix failing to zero other trigger Fix SDL_HINT_JOYSTICK_HIDAPI not working if set before calling SDL_Init() Add missing device to device names Disable RawInput if we have a mismatch of XInput-capable but not RawInput-capable devices Updated to SDL 2.0.13 code with the following notes: New HID driver: xbox360w - no idea what that is, hopefully urelated SDL_hidapijoystick.c had been refactored to couple data handling logic with device opening logic and device lists caused some problems, yields slightly uglier integration than previously when the 360 HID device driver was just handling the data. SDL_hidapijoystick.c now often pulls the device off of the joystick_hwdata structure for some rumble logic, but it appears that code path is never reached, so probably not a problem. Looks like joystick_hwdata was refactored to not include a mutex in other drivers, maintainers may want to do the same refactor here if that's useful for some reason. Something changed in how devices get names, so getting generic names. Had to fix a (new?) bug where removing an XInput controller caused existing controllers (that moved to a new XInput index) to get identified as 0x045e/0x02fd ("it's probably Bluetooth" in code), rendering the existing HIDAPI_IsDevicePresent and new RAWINPUT_IsDevicePresent unreliable.
2020-03-16 14:25:02 -05:00
{
RAWINPUTDEVICE rid[SDL_arraysize(subscribed_devices)];
int i;
Fixed bug 4477 - Support more than 4 XInput-capable devices on Windows Jimb Esser Add new RawInput controller API, and improved correlation with XInput/WGI Reorder joystick init so drivers can ask the others if they handle a device reliably Do not poll disconnected XInput devices (major perf issue) Fix various cases where incorrect correlation could happen Simple mechanism for propagating unhandled Guide button presses even before guaranteed correlation Correlate by axis motion as well as button presses Fix failing to zero other trigger Fix SDL_HINT_JOYSTICK_HIDAPI not working if set before calling SDL_Init() Add missing device to device names Disable RawInput if we have a mismatch of XInput-capable but not RawInput-capable devices Updated to SDL 2.0.13 code with the following notes: New HID driver: xbox360w - no idea what that is, hopefully urelated SDL_hidapijoystick.c had been refactored to couple data handling logic with device opening logic and device lists caused some problems, yields slightly uglier integration than previously when the 360 HID device driver was just handling the data. SDL_hidapijoystick.c now often pulls the device off of the joystick_hwdata structure for some rumble logic, but it appears that code path is never reached, so probably not a problem. Looks like joystick_hwdata was refactored to not include a mutex in other drivers, maintainers may want to do the same refactor here if that's useful for some reason. Something changed in how devices get names, so getting generic names. Had to fix a (new?) bug where removing an XInput controller caused existing controllers (that moved to a new XInput index) to get identified as 0x045e/0x02fd ("it's probably Bluetooth" in code), rendering the existing HIDAPI_IsDevicePresent and new RAWINPUT_IsDevicePresent unreliable.
2020-03-16 14:25:02 -05:00
for (i = 0; i < SDL_arraysize(subscribed_devices); i++) {
rid[i].usUsagePage = USB_USAGEPAGE_GENERIC_DESKTOP;
rid[i].usUsage = subscribed_devices[i];
rid[i].dwFlags = RIDEV_DEVNOTIFY | RIDEV_INPUTSINK; /* Receive messages when in background, including device add/remove */
rid[i].hwndTarget = hWnd;
}
if (!RegisterRawInputDevices(rid, SDL_arraysize(rid), sizeof(RAWINPUTDEVICE))) {
SDL_SetError("Couldn't register for raw input events");
return SDL_FALSE;
}
return SDL_TRUE;
}
int RAWINPUT_UnregisterNotifications()
{
int i;
RAWINPUTDEVICE rid[SDL_arraysize(subscribed_devices)];
for (i = 0; i < SDL_arraysize(subscribed_devices); i++) {
rid[i].usUsagePage = USB_USAGEPAGE_GENERIC_DESKTOP;
rid[i].usUsage = subscribed_devices[i];
rid[i].dwFlags = RIDEV_REMOVE;
rid[i].hwndTarget = NULL;
}
if (!RegisterRawInputDevices(rid, SDL_arraysize(rid), sizeof(RAWINPUTDEVICE))) {
return SDL_SetError("Couldn't unregister for raw input events");
}
return 0;
}
LRESULT CALLBACK
RAWINPUT_WindowProc(HWND hWnd, UINT msg, WPARAM wParam, LPARAM lParam)
{
LRESULT result = -1;
if (SDL_RAWINPUT_inited) {
SDL_LockJoysticks();
switch (msg) {
case WM_INPUT_DEVICE_CHANGE:
{
HANDLE hDevice = (HANDLE)lParam;
switch (wParam) {
case GIDC_ARRIVAL:
RAWINPUT_AddDevice(hDevice);
break;
case GIDC_REMOVAL:
{
SDL_RAWINPUT_Device *device;
device = RAWINPUT_DeviceFromHandle(hDevice);
if (device) {
RAWINPUT_DelDevice(device, SDL_TRUE);
Fixed bug 4477 - Support more than 4 XInput-capable devices on Windows Jimb Esser Add new RawInput controller API, and improved correlation with XInput/WGI Reorder joystick init so drivers can ask the others if they handle a device reliably Do not poll disconnected XInput devices (major perf issue) Fix various cases where incorrect correlation could happen Simple mechanism for propagating unhandled Guide button presses even before guaranteed correlation Correlate by axis motion as well as button presses Fix failing to zero other trigger Fix SDL_HINT_JOYSTICK_HIDAPI not working if set before calling SDL_Init() Add missing device to device names Disable RawInput if we have a mismatch of XInput-capable but not RawInput-capable devices Updated to SDL 2.0.13 code with the following notes: New HID driver: xbox360w - no idea what that is, hopefully urelated SDL_hidapijoystick.c had been refactored to couple data handling logic with device opening logic and device lists caused some problems, yields slightly uglier integration than previously when the 360 HID device driver was just handling the data. SDL_hidapijoystick.c now often pulls the device off of the joystick_hwdata structure for some rumble logic, but it appears that code path is never reached, so probably not a problem. Looks like joystick_hwdata was refactored to not include a mutex in other drivers, maintainers may want to do the same refactor here if that's useful for some reason. Something changed in how devices get names, so getting generic names. Had to fix a (new?) bug where removing an XInput controller caused existing controllers (that moved to a new XInput index) to get identified as 0x045e/0x02fd ("it's probably Bluetooth" in code), rendering the existing HIDAPI_IsDevicePresent and new RAWINPUT_IsDevicePresent unreliable.
2020-03-16 14:25:02 -05:00
}
break;
Fixed bug 4477 - Support more than 4 XInput-capable devices on Windows Jimb Esser Add new RawInput controller API, and improved correlation with XInput/WGI Reorder joystick init so drivers can ask the others if they handle a device reliably Do not poll disconnected XInput devices (major perf issue) Fix various cases where incorrect correlation could happen Simple mechanism for propagating unhandled Guide button presses even before guaranteed correlation Correlate by axis motion as well as button presses Fix failing to zero other trigger Fix SDL_HINT_JOYSTICK_HIDAPI not working if set before calling SDL_Init() Add missing device to device names Disable RawInput if we have a mismatch of XInput-capable but not RawInput-capable devices Updated to SDL 2.0.13 code with the following notes: New HID driver: xbox360w - no idea what that is, hopefully urelated SDL_hidapijoystick.c had been refactored to couple data handling logic with device opening logic and device lists caused some problems, yields slightly uglier integration than previously when the 360 HID device driver was just handling the data. SDL_hidapijoystick.c now often pulls the device off of the joystick_hwdata structure for some rumble logic, but it appears that code path is never reached, so probably not a problem. Looks like joystick_hwdata was refactored to not include a mutex in other drivers, maintainers may want to do the same refactor here if that's useful for some reason. Something changed in how devices get names, so getting generic names. Had to fix a (new?) bug where removing an XInput controller caused existing controllers (that moved to a new XInput index) to get identified as 0x045e/0x02fd ("it's probably Bluetooth" in code), rendering the existing HIDAPI_IsDevicePresent and new RAWINPUT_IsDevicePresent unreliable.
2020-03-16 14:25:02 -05:00
}
default:
break;
}
}
result = 0;
break;
case WM_INPUT:
{
Uint8 data[sizeof(RAWINPUTHEADER) + sizeof(RAWHID) + USB_PACKET_LENGTH];
UINT buffer_size = SDL_arraysize(data);
if ((int)GetRawInputData((HRAWINPUT)lParam, RID_INPUT, data, &buffer_size, sizeof(RAWINPUTHEADER)) > 0) {
PRAWINPUT raw_input = (PRAWINPUT)data;
SDL_RAWINPUT_Device *device = RAWINPUT_DeviceFromHandle(raw_input->header.hDevice);
if (device) {
SDL_Joystick *joystick = device->joystick;
if (joystick) {
RAWINPUT_HandleStatePacket(joystick, raw_input->data.hid.bRawData, raw_input->data.hid.dwSizeHid);
Fixed bug 4477 - Support more than 4 XInput-capable devices on Windows Jimb Esser Add new RawInput controller API, and improved correlation with XInput/WGI Reorder joystick init so drivers can ask the others if they handle a device reliably Do not poll disconnected XInput devices (major perf issue) Fix various cases where incorrect correlation could happen Simple mechanism for propagating unhandled Guide button presses even before guaranteed correlation Correlate by axis motion as well as button presses Fix failing to zero other trigger Fix SDL_HINT_JOYSTICK_HIDAPI not working if set before calling SDL_Init() Add missing device to device names Disable RawInput if we have a mismatch of XInput-capable but not RawInput-capable devices Updated to SDL 2.0.13 code with the following notes: New HID driver: xbox360w - no idea what that is, hopefully urelated SDL_hidapijoystick.c had been refactored to couple data handling logic with device opening logic and device lists caused some problems, yields slightly uglier integration than previously when the 360 HID device driver was just handling the data. SDL_hidapijoystick.c now often pulls the device off of the joystick_hwdata structure for some rumble logic, but it appears that code path is never reached, so probably not a problem. Looks like joystick_hwdata was refactored to not include a mutex in other drivers, maintainers may want to do the same refactor here if that's useful for some reason. Something changed in how devices get names, so getting generic names. Had to fix a (new?) bug where removing an XInput controller caused existing controllers (that moved to a new XInput index) to get identified as 0x045e/0x02fd ("it's probably Bluetooth" in code), rendering the existing HIDAPI_IsDevicePresent and new RAWINPUT_IsDevicePresent unreliable.
2020-03-16 14:25:02 -05:00
}
}
}
}
result = 0;
break;
Fixed bug 4477 - Support more than 4 XInput-capable devices on Windows Jimb Esser Add new RawInput controller API, and improved correlation with XInput/WGI Reorder joystick init so drivers can ask the others if they handle a device reliably Do not poll disconnected XInput devices (major perf issue) Fix various cases where incorrect correlation could happen Simple mechanism for propagating unhandled Guide button presses even before guaranteed correlation Correlate by axis motion as well as button presses Fix failing to zero other trigger Fix SDL_HINT_JOYSTICK_HIDAPI not working if set before calling SDL_Init() Add missing device to device names Disable RawInput if we have a mismatch of XInput-capable but not RawInput-capable devices Updated to SDL 2.0.13 code with the following notes: New HID driver: xbox360w - no idea what that is, hopefully urelated SDL_hidapijoystick.c had been refactored to couple data handling logic with device opening logic and device lists caused some problems, yields slightly uglier integration than previously when the 360 HID device driver was just handling the data. SDL_hidapijoystick.c now often pulls the device off of the joystick_hwdata structure for some rumble logic, but it appears that code path is never reached, so probably not a problem. Looks like joystick_hwdata was refactored to not include a mutex in other drivers, maintainers may want to do the same refactor here if that's useful for some reason. Something changed in how devices get names, so getting generic names. Had to fix a (new?) bug where removing an XInput controller caused existing controllers (that moved to a new XInput index) to get identified as 0x045e/0x02fd ("it's probably Bluetooth" in code), rendering the existing HIDAPI_IsDevicePresent and new RAWINPUT_IsDevicePresent unreliable.
2020-03-16 14:25:02 -05:00
}
SDL_UnlockJoysticks();
}
if (result >= 0) {
return result;
}
return CallWindowProc(DefWindowProc, hWnd, msg, wParam, lParam);
Fixed bug 4477 - Support more than 4 XInput-capable devices on Windows Jimb Esser Add new RawInput controller API, and improved correlation with XInput/WGI Reorder joystick init so drivers can ask the others if they handle a device reliably Do not poll disconnected XInput devices (major perf issue) Fix various cases where incorrect correlation could happen Simple mechanism for propagating unhandled Guide button presses even before guaranteed correlation Correlate by axis motion as well as button presses Fix failing to zero other trigger Fix SDL_HINT_JOYSTICK_HIDAPI not working if set before calling SDL_Init() Add missing device to device names Disable RawInput if we have a mismatch of XInput-capable but not RawInput-capable devices Updated to SDL 2.0.13 code with the following notes: New HID driver: xbox360w - no idea what that is, hopefully urelated SDL_hidapijoystick.c had been refactored to couple data handling logic with device opening logic and device lists caused some problems, yields slightly uglier integration than previously when the 360 HID device driver was just handling the data. SDL_hidapijoystick.c now often pulls the device off of the joystick_hwdata structure for some rumble logic, but it appears that code path is never reached, so probably not a problem. Looks like joystick_hwdata was refactored to not include a mutex in other drivers, maintainers may want to do the same refactor here if that's useful for some reason. Something changed in how devices get names, so getting generic names. Had to fix a (new?) bug where removing an XInput controller caused existing controllers (that moved to a new XInput index) to get identified as 0x045e/0x02fd ("it's probably Bluetooth" in code), rendering the existing HIDAPI_IsDevicePresent and new RAWINPUT_IsDevicePresent unreliable.
2020-03-16 14:25:02 -05:00
}
static void RAWINPUT_JoystickQuit(void)
Fixed bug 4477 - Support more than 4 XInput-capable devices on Windows Jimb Esser Add new RawInput controller API, and improved correlation with XInput/WGI Reorder joystick init so drivers can ask the others if they handle a device reliably Do not poll disconnected XInput devices (major perf issue) Fix various cases where incorrect correlation could happen Simple mechanism for propagating unhandled Guide button presses even before guaranteed correlation Correlate by axis motion as well as button presses Fix failing to zero other trigger Fix SDL_HINT_JOYSTICK_HIDAPI not working if set before calling SDL_Init() Add missing device to device names Disable RawInput if we have a mismatch of XInput-capable but not RawInput-capable devices Updated to SDL 2.0.13 code with the following notes: New HID driver: xbox360w - no idea what that is, hopefully urelated SDL_hidapijoystick.c had been refactored to couple data handling logic with device opening logic and device lists caused some problems, yields slightly uglier integration than previously when the 360 HID device driver was just handling the data. SDL_hidapijoystick.c now often pulls the device off of the joystick_hwdata structure for some rumble logic, but it appears that code path is never reached, so probably not a problem. Looks like joystick_hwdata was refactored to not include a mutex in other drivers, maintainers may want to do the same refactor here if that's useful for some reason. Something changed in how devices get names, so getting generic names. Had to fix a (new?) bug where removing an XInput controller caused existing controllers (that moved to a new XInput index) to get identified as 0x045e/0x02fd ("it's probably Bluetooth" in code), rendering the existing HIDAPI_IsDevicePresent and new RAWINPUT_IsDevicePresent unreliable.
2020-03-16 14:25:02 -05:00
{
if (!SDL_RAWINPUT_inited) {
Fixed bug 4477 - Support more than 4 XInput-capable devices on Windows Jimb Esser Add new RawInput controller API, and improved correlation with XInput/WGI Reorder joystick init so drivers can ask the others if they handle a device reliably Do not poll disconnected XInput devices (major perf issue) Fix various cases where incorrect correlation could happen Simple mechanism for propagating unhandled Guide button presses even before guaranteed correlation Correlate by axis motion as well as button presses Fix failing to zero other trigger Fix SDL_HINT_JOYSTICK_HIDAPI not working if set before calling SDL_Init() Add missing device to device names Disable RawInput if we have a mismatch of XInput-capable but not RawInput-capable devices Updated to SDL 2.0.13 code with the following notes: New HID driver: xbox360w - no idea what that is, hopefully urelated SDL_hidapijoystick.c had been refactored to couple data handling logic with device opening logic and device lists caused some problems, yields slightly uglier integration than previously when the 360 HID device driver was just handling the data. SDL_hidapijoystick.c now often pulls the device off of the joystick_hwdata structure for some rumble logic, but it appears that code path is never reached, so probably not a problem. Looks like joystick_hwdata was refactored to not include a mutex in other drivers, maintainers may want to do the same refactor here if that's useful for some reason. Something changed in how devices get names, so getting generic names. Had to fix a (new?) bug where removing an XInput controller caused existing controllers (that moved to a new XInput index) to get identified as 0x045e/0x02fd ("it's probably Bluetooth" in code), rendering the existing HIDAPI_IsDevicePresent and new RAWINPUT_IsDevicePresent unreliable.
2020-03-16 14:25:02 -05:00
return;
}
while (SDL_RAWINPUT_devices) {
RAWINPUT_DelDevice(SDL_RAWINPUT_devices, SDL_FALSE);
}
WIN_UnloadHIDDLL();
Fixed bug 4477 - Support more than 4 XInput-capable devices on Windows Jimb Esser Add new RawInput controller API, and improved correlation with XInput/WGI Reorder joystick init so drivers can ask the others if they handle a device reliably Do not poll disconnected XInput devices (major perf issue) Fix various cases where incorrect correlation could happen Simple mechanism for propagating unhandled Guide button presses even before guaranteed correlation Correlate by axis motion as well as button presses Fix failing to zero other trigger Fix SDL_HINT_JOYSTICK_HIDAPI not working if set before calling SDL_Init() Add missing device to device names Disable RawInput if we have a mismatch of XInput-capable but not RawInput-capable devices Updated to SDL 2.0.13 code with the following notes: New HID driver: xbox360w - no idea what that is, hopefully urelated SDL_hidapijoystick.c had been refactored to couple data handling logic with device opening logic and device lists caused some problems, yields slightly uglier integration than previously when the 360 HID device driver was just handling the data. SDL_hidapijoystick.c now often pulls the device off of the joystick_hwdata structure for some rumble logic, but it appears that code path is never reached, so probably not a problem. Looks like joystick_hwdata was refactored to not include a mutex in other drivers, maintainers may want to do the same refactor here if that's useful for some reason. Something changed in how devices get names, so getting generic names. Had to fix a (new?) bug where removing an XInput controller caused existing controllers (that moved to a new XInput index) to get identified as 0x045e/0x02fd ("it's probably Bluetooth" in code), rendering the existing HIDAPI_IsDevicePresent and new RAWINPUT_IsDevicePresent unreliable.
2020-03-16 14:25:02 -05:00
SDL_RAWINPUT_numjoysticks = 0;
SDL_RAWINPUT_inited = SDL_FALSE;
}
static SDL_bool RAWINPUT_JoystickGetGamepadMapping(int device_index, SDL_GamepadMapping *out)
{
return SDL_FALSE;
}
Fixed bug 4477 - Support more than 4 XInput-capable devices on Windows Jimb Esser Add new RawInput controller API, and improved correlation with XInput/WGI Reorder joystick init so drivers can ask the others if they handle a device reliably Do not poll disconnected XInput devices (major perf issue) Fix various cases where incorrect correlation could happen Simple mechanism for propagating unhandled Guide button presses even before guaranteed correlation Correlate by axis motion as well as button presses Fix failing to zero other trigger Fix SDL_HINT_JOYSTICK_HIDAPI not working if set before calling SDL_Init() Add missing device to device names Disable RawInput if we have a mismatch of XInput-capable but not RawInput-capable devices Updated to SDL 2.0.13 code with the following notes: New HID driver: xbox360w - no idea what that is, hopefully urelated SDL_hidapijoystick.c had been refactored to couple data handling logic with device opening logic and device lists caused some problems, yields slightly uglier integration than previously when the 360 HID device driver was just handling the data. SDL_hidapijoystick.c now often pulls the device off of the joystick_hwdata structure for some rumble logic, but it appears that code path is never reached, so probably not a problem. Looks like joystick_hwdata was refactored to not include a mutex in other drivers, maintainers may want to do the same refactor here if that's useful for some reason. Something changed in how devices get names, so getting generic names. Had to fix a (new?) bug where removing an XInput controller caused existing controllers (that moved to a new XInput index) to get identified as 0x045e/0x02fd ("it's probably Bluetooth" in code), rendering the existing HIDAPI_IsDevicePresent and new RAWINPUT_IsDevicePresent unreliable.
2020-03-16 14:25:02 -05:00
SDL_JoystickDriver SDL_RAWINPUT_JoystickDriver = {
Fixed bug 4477 - Support more than 4 XInput-capable devices on Windows Jimb Esser Add new RawInput controller API, and improved correlation with XInput/WGI Reorder joystick init so drivers can ask the others if they handle a device reliably Do not poll disconnected XInput devices (major perf issue) Fix various cases where incorrect correlation could happen Simple mechanism for propagating unhandled Guide button presses even before guaranteed correlation Correlate by axis motion as well as button presses Fix failing to zero other trigger Fix SDL_HINT_JOYSTICK_HIDAPI not working if set before calling SDL_Init() Add missing device to device names Disable RawInput if we have a mismatch of XInput-capable but not RawInput-capable devices Updated to SDL 2.0.13 code with the following notes: New HID driver: xbox360w - no idea what that is, hopefully urelated SDL_hidapijoystick.c had been refactored to couple data handling logic with device opening logic and device lists caused some problems, yields slightly uglier integration than previously when the 360 HID device driver was just handling the data. SDL_hidapijoystick.c now often pulls the device off of the joystick_hwdata structure for some rumble logic, but it appears that code path is never reached, so probably not a problem. Looks like joystick_hwdata was refactored to not include a mutex in other drivers, maintainers may want to do the same refactor here if that's useful for some reason. Something changed in how devices get names, so getting generic names. Had to fix a (new?) bug where removing an XInput controller caused existing controllers (that moved to a new XInput index) to get identified as 0x045e/0x02fd ("it's probably Bluetooth" in code), rendering the existing HIDAPI_IsDevicePresent and new RAWINPUT_IsDevicePresent unreliable.
2020-03-16 14:25:02 -05:00
RAWINPUT_JoystickInit,
RAWINPUT_JoystickGetCount,
RAWINPUT_JoystickDetect,
RAWINPUT_JoystickGetDeviceName,
RAWINPUT_JoystickGetDevicePath,
Fixed bug 4477 - Support more than 4 XInput-capable devices on Windows Jimb Esser Add new RawInput controller API, and improved correlation with XInput/WGI Reorder joystick init so drivers can ask the others if they handle a device reliably Do not poll disconnected XInput devices (major perf issue) Fix various cases where incorrect correlation could happen Simple mechanism for propagating unhandled Guide button presses even before guaranteed correlation Correlate by axis motion as well as button presses Fix failing to zero other trigger Fix SDL_HINT_JOYSTICK_HIDAPI not working if set before calling SDL_Init() Add missing device to device names Disable RawInput if we have a mismatch of XInput-capable but not RawInput-capable devices Updated to SDL 2.0.13 code with the following notes: New HID driver: xbox360w - no idea what that is, hopefully urelated SDL_hidapijoystick.c had been refactored to couple data handling logic with device opening logic and device lists caused some problems, yields slightly uglier integration than previously when the 360 HID device driver was just handling the data. SDL_hidapijoystick.c now often pulls the device off of the joystick_hwdata structure for some rumble logic, but it appears that code path is never reached, so probably not a problem. Looks like joystick_hwdata was refactored to not include a mutex in other drivers, maintainers may want to do the same refactor here if that's useful for some reason. Something changed in how devices get names, so getting generic names. Had to fix a (new?) bug where removing an XInput controller caused existing controllers (that moved to a new XInput index) to get identified as 0x045e/0x02fd ("it's probably Bluetooth" in code), rendering the existing HIDAPI_IsDevicePresent and new RAWINPUT_IsDevicePresent unreliable.
2020-03-16 14:25:02 -05:00
RAWINPUT_JoystickGetDevicePlayerIndex,
RAWINPUT_JoystickSetDevicePlayerIndex,
RAWINPUT_JoystickGetDeviceGUID,
RAWINPUT_JoystickGetDeviceInstanceID,
RAWINPUT_JoystickOpen,
RAWINPUT_JoystickRumble,
RAWINPUT_JoystickRumbleTriggers,
RAWINPUT_JoystickGetCapabilities,
RAWINPUT_JoystickSetLED,
RAWINPUT_JoystickSendEffect,
RAWINPUT_JoystickSetSensorsEnabled,
Fixed bug 4477 - Support more than 4 XInput-capable devices on Windows Jimb Esser Add new RawInput controller API, and improved correlation with XInput/WGI Reorder joystick init so drivers can ask the others if they handle a device reliably Do not poll disconnected XInput devices (major perf issue) Fix various cases where incorrect correlation could happen Simple mechanism for propagating unhandled Guide button presses even before guaranteed correlation Correlate by axis motion as well as button presses Fix failing to zero other trigger Fix SDL_HINT_JOYSTICK_HIDAPI not working if set before calling SDL_Init() Add missing device to device names Disable RawInput if we have a mismatch of XInput-capable but not RawInput-capable devices Updated to SDL 2.0.13 code with the following notes: New HID driver: xbox360w - no idea what that is, hopefully urelated SDL_hidapijoystick.c had been refactored to couple data handling logic with device opening logic and device lists caused some problems, yields slightly uglier integration than previously when the 360 HID device driver was just handling the data. SDL_hidapijoystick.c now often pulls the device off of the joystick_hwdata structure for some rumble logic, but it appears that code path is never reached, so probably not a problem. Looks like joystick_hwdata was refactored to not include a mutex in other drivers, maintainers may want to do the same refactor here if that's useful for some reason. Something changed in how devices get names, so getting generic names. Had to fix a (new?) bug where removing an XInput controller caused existing controllers (that moved to a new XInput index) to get identified as 0x045e/0x02fd ("it's probably Bluetooth" in code), rendering the existing HIDAPI_IsDevicePresent and new RAWINPUT_IsDevicePresent unreliable.
2020-03-16 14:25:02 -05:00
RAWINPUT_JoystickUpdate,
RAWINPUT_JoystickClose,
RAWINPUT_JoystickQuit,
RAWINPUT_JoystickGetGamepadMapping
Fixed bug 4477 - Support more than 4 XInput-capable devices on Windows Jimb Esser Add new RawInput controller API, and improved correlation with XInput/WGI Reorder joystick init so drivers can ask the others if they handle a device reliably Do not poll disconnected XInput devices (major perf issue) Fix various cases where incorrect correlation could happen Simple mechanism for propagating unhandled Guide button presses even before guaranteed correlation Correlate by axis motion as well as button presses Fix failing to zero other trigger Fix SDL_HINT_JOYSTICK_HIDAPI not working if set before calling SDL_Init() Add missing device to device names Disable RawInput if we have a mismatch of XInput-capable but not RawInput-capable devices Updated to SDL 2.0.13 code with the following notes: New HID driver: xbox360w - no idea what that is, hopefully urelated SDL_hidapijoystick.c had been refactored to couple data handling logic with device opening logic and device lists caused some problems, yields slightly uglier integration than previously when the 360 HID device driver was just handling the data. SDL_hidapijoystick.c now often pulls the device off of the joystick_hwdata structure for some rumble logic, but it appears that code path is never reached, so probably not a problem. Looks like joystick_hwdata was refactored to not include a mutex in other drivers, maintainers may want to do the same refactor here if that's useful for some reason. Something changed in how devices get names, so getting generic names. Had to fix a (new?) bug where removing an XInput controller caused existing controllers (that moved to a new XInput index) to get identified as 0x045e/0x02fd ("it's probably Bluetooth" in code), rendering the existing HIDAPI_IsDevicePresent and new RAWINPUT_IsDevicePresent unreliable.
2020-03-16 14:25:02 -05:00
};
#endif /* SDL_JOYSTICK_RAWINPUT */