Overlays are shown with _currentShakePos offset too, so no need to
compensate the mousepointer position for overlays. This compensation
was done inconsistently in draw/undraw mouse causing the mouse trails.
Bug is: "IRIX: X BadMatch when trying to start any 640x480 game".
40e019efd4 introduced resetting the pixel depth
when unloading modes for SDL output. This was required to make mode listing
for OpenGL work on Win8+.
This causes issues on non-Win32 platforms though. SDL might not give us a valid
a pixel depth, causing the resetting to fail. A particular example is X11 on
IRIX, when only 16bit output modes work. Initially SDL tells us that the pixel
depth is 32bit. Trying to set this up causes a crash though.
Since there is no way to validate SDL's return value, we simply limit the reset
to platforms where it is actually required, i.e. Win32.
As a side effect we get better coordinates from mouse move events in
fullscreen, i.e. we do not get actual window coordinates but something close
to actual viewport coordinates. The easily noticable issue is that mouse
coordinates inside black bars are outside of the viewport. For example, the
x coordinate can get negative when there's a black bar at the left side.
We set a custom video mode that might have a different BPP from the default. To ensure that other graphics managers will get the proper results when listing available video modes, we need to restore the initial BPP when unloading.
This fixes an issue when switching to OpenGL on Windows 8 and later. On those OSes, fullscreen OpenGL only has a 32bpp mode. It is correctly listed in the options but we call SLD_SetVideoMode later in the surface graphics manager. When we list the mode again after a switch, the internally selected BPP is still 16 and we fail to find any available fullscreen modes.
Sadly this also requires us to extend GraphicsManager for this SDL specific
feature. However, since that's only used in the SDL backend and Tizen it
should be fine for now...
Formerly the function created a SDL_Surface by hand. Instead now it uses
SDL_CreateRGBSurface (which is used in the SDL backend anyway and yields
the same results).
This should fix PS3 port compilation.
This removes the need to convert the parameter to copyRectToScreen to
"const byte *", which is commonly used in games, which use Graphics::Surface
to store their graphics data.
All uses of the old target scale API actually wanted to disallow scaling of
the mouse cursor. This commit adapts our API to this and thus simplifies
backend implementations.
Some backends, most notable the Wii and Android, did some implementation of
the cursor target scale, which I didn't adapt yet. I added a TODO for the
porters there.
* Note: This really should be cleaned up and all backends that have a custom loadGFXMode() really should setup their own _videoMode.hardware* settings. Assuming the overlay will always match the hardware is flawed logic on devices with fixed hardware screens.
* Just done for the GPH backend for now to cut down on the regresion risk to other backends near the 1.4.0 release cycle.
The bug in question is "SDL/OpenGL: Crash when switching renderer backend". To
fix it I added a stupid graphics state copying to the SDL backend, in case the
graphics manager is switched. The implementation of this is considered a pure
workaround, no one should ever do it like this in reality... I just want to
die when looking at this... Not sure why I actually committed it.
Anyway it at least makes the OpenGL backend testable for those who do not
want to fiddle with the config file directly.
This gets rid of the hacks, where SdlEventSource added events with custom type
numbers to pass SDL_VIDEOEXPOSE and SDL_VIDEORESIZE to the graphics manager.
Furthermore it get rids of the uninituitive and hard to trace way of assigning
the proper mouse coordinates to mouse related events. Formerly it passed the
real screen coordinates through the even dispatching api to the graphics
manager (at least hopefully ;-) and let that handle creating a new event with
the proper coordinates. Now instead SdlEventSource handles the proper
coordinate setup itself.
Since this is a behavior change and I can not test all the SDL based small
devices ports this commit might break compilation for them and more serve it
might also break mouse position behavior. If any of that occurs I am sorry
about it.
