Mouse tracking and mouse capture didn't work well together. Mouse capture was
also buggy in that it could release capture prematurely.
Mouse capture is now counted and tracked better, multiple requests to set
capture will only call SetCapture once. ReleaseCapture will only be called once
the same number of releases have been made as sets.
MW_MOUSELEAVE messages are now ignored if the mouse is captured.
Mouse tracking is renabled when mouse capture is released.
While the mouse is captured enter and leave events are genereated based on
tracking of the mouse inside MouseMove.
Fixes#301
GetIconInfo
(https://msdn.microsoft.com/en-us/library/windows/desktop/ms648070(v=vs.85).aspx)
creates bitmaps that must be deleted after the call to
CreateIconIndirect, which copies the bitmaps to the icon it creates.
bmpIcon created by Bitmap.GetHicon is now destroyed after being used.
The return value of SetCursor was used to retrieve the last cursor, as
it could have been set by another library (Some UI libraries change the
cursor using the .net Cursor) this could of leaked the cursor we created
and now lost track of. We now delete the handle we had set, not the one
returned by SetCursor.
Fixes warnings [#61] by disabling unused field warnings for two structures in
HidProtocol. These fields aren't currently used by OpenTK but the stuctures are
used in native marshalling so must match the documented structures perfectly.
WinFactory.Dispose() could crash with a NRE when the joystick driver has
not been initialized. Fixed by checking for null before disposing the
input driver.
Due to the way we segregate axes from buttons, the easiest approach is
to retrieve the current button state via HidP_GetUsages().
Axes, buttons and hats are now allocated sequentially based on their
order of appearance in the device capability reports.
Marked arrays with [In] and [Out] as required. Fixes issues with
axis/button capabilities returning empty results.
Added bindings for link collection node transversal.
Polling joyGetCaps is very very slow, so we should not do that every
time GetCapabilities is called. Instead, we should call joyGetCaps once
and cache the results.
We need to find a different way to implement hotplugging.
Since we do not use Raw Input for joystick devices, it does not make
sense to expose the IJoystickDriver2 interface through WinRawInput. This
keeps allows joystick and keyboard/mouse implementations to reside in
distinct, modular classes, instantiated separately through WinFactory.
The IGamePadDriver interface is now implemented using
MappedGamePadDriver on all platforms. This minimizes code duplication
and simplifies all input driver implementations, since we only need
backends for IJoystickDriver2.
