HIDAPI: fix bug that caused non-HID class parts of composite devices to have windows HID functions called on them.

Sam Lantinga 2019-03-12 14:44:12 -07:00
parent cddb285cdb
commit 1a8e3a0270
1 changed files with 12 additions and 5 deletions

View File

@ -309,7 +309,6 @@ struct hid_device_info HID_API_EXPORT * HID_API_CALL hid_enumerate(unsigned shor
SP_DEVICE_INTERFACE_DETAIL_DATA_A *device_interface_detail_data = NULL;
HDEVINFO device_info_set = INVALID_HANDLE_VALUE;
int device_index = 0;
int i;
if (hid_init() < 0)
return NULL;
@ -373,12 +372,16 @@ struct hid_device_info HID_API_EXPORT * HID_API_CALL hid_enumerate(unsigned shor
/* Make sure this device is of Setup Class "HIDClass" and has a
driver bound to it. */
for (i = 0; ; i++) {
/* In the main HIDAPI tree this is a loop which will erroneously open
devices that aren't HID class. Please preserve this delta if we ever
update to take new changes */
{
char driver_name[256];
/* Populate devinfo_data. This function will return failure
when there are no more interfaces left. */
res = SetupDiEnumDeviceInfo(device_info_set, i, &devinfo_data);
res = SetupDiEnumDeviceInfo(device_info_set, device_index, &devinfo_data);
if (!res)
goto cont;
@ -391,8 +394,12 @@ struct hid_device_info HID_API_EXPORT * HID_API_CALL hid_enumerate(unsigned shor
/* See if there's a driver bound. */
res = SetupDiGetDeviceRegistryPropertyA(device_info_set, &devinfo_data,
SPDRP_DRIVER, NULL, (PBYTE)driver_name, sizeof(driver_name), NULL);
if (res)
break;
if (!res)
goto cont;
}
else
{
goto cont;
}
}