David White 2012-09-12 13:22:52 PDT
Background: X has two main 'buffers' for copy/paste:
PRIMARY -- normally used for implicit selection of text, with middle-click to
paste
CLIPBOARD -- works with an explicit copy/paste like on other platforms.
Currently SDL2 only provides access to PRIMARY. Since CLIPBOARD is much closer
to functionality of other platforms, SDL should provide access to that instead.
The attached patch makes it so that SDL_SetClipboardText() sets both PRIMARY
and CLIPBOARD and SDL_GetClipboardText() reads from CLIPBOARD instead of
primary.
Hello Sam,
while profiling ScummVM I noticed it was making use of the generic
BlitNToN blitter, which struck me as odd because it should be a very
classical codepath.
After investigating, I saw that in the blit op chooser:
{ 0x0000F800,0x000007E0,0x0000001F, 4, 0x00FF0000,0x0000FF00,0x000000FF,
0, NULL, Blit_RGB565_ARGB8888, SET_ALPHA },
{ 0x0000F800,0x000007E0,0x0000001F, 4, 0x000000FF,0x0000FF00,0x00FF0000,
0, NULL, Blit_RGB565_ABGR8888, SET_ALPHA },
{ 0x0000F800,0x000007E0,0x0000001F, 4, 0xFF000000,0x00FF0000,0x0000FF00,
0, NULL, Blit_RGB565_RGBA8888, SET_ALPHA },
{ 0x0000F800,0x000007E0,0x0000001F, 4, 0x0000FF00,0x00FF0000,0xFF000000,
0, NULL, Blit_RGB565_BGRA8888, SET_ALPHA },
Couldn't the optimized versions be used for NO_ALPHA too? I take it
that the resulting alpha component can be undefined as it should never
be used.
I tried this (see attached patch) and it worked perfectly (and
therefore faster) on ScummVM but there might be a trick (I'm not
expert at the semantics of SDL, ie NO_ALPHA, SET_ALPHA and COPY_ALPHA
there).
What do you think?
Cheers,
Bertrand
Brad Smith 2012-08-01 20:10:19 PDT
The attached patch from the OpenBSD ports tree is to increase the number of
uhid devices to scan for joysticks. It's somewhat easy to exhaust the default
number of devices which are scanned.
Matthias Bentrup 2012-08-09 12:53:17 PDT
With OpenGL 4.3 the ARB added a new context flag for context reset isolation
and renamed the existing ES2 profile bit to ES profile bit, as it can be used
to request GLES 3 compatible contexts, too.
This patch adds these changes to SDL on Linux and Windows.
Also SDL lacks the ability to create shared contexts. This patch also adds a
new GL attribute to enable context sharing. As casting a GL context to int is
not portable, I added only a boolean attribute
SDL_GL_SHARE_WITH_CURRENT_CONTEXT, which makes the new context share resources
with the context current on the creating thread.
This required a small public API change: SDL_GL_SetSwapInterval() now accepts
negative values, and SDL_GL_GetSwapInterval() doesn't report errors anymore
(if it can't work, it'll return zero as a reasonable default).
If you need to test for errors, such as a lack of swap_control_tear support,
check the results of SDL_GL_SetSwapInterval() when you set your desired
value.
From Scott Percival
Okay, I think I have something for this. Tested it on GL and GLES
machines, it seems to work okay.
- Add a new SDL GL attribute SDL_GL_CONTEXT_EGL:
- Only useful for the X11 video driver at the moment
- Set to 1 for an EGL context, 0 to use the default for the video driver
- Default is 0, unless library is built for EGL only
- Should be set after SDL init, but before window/context
creation (i.e. same place you'd specify attributes for major/minor GL
version)
- After a lot of agony pondering the least-terrible way to go about
it, made it so that X11_GL_LoadLibrary and X11_GLES_LoadLibrary check
SDL_GL_CONTEXT_EGL. If no GL context exists yet, and the attribute
choice doesn't match with the checking function, then it changes all
the function pointers in the video driver and passes control on to the
new LoadLibrary method.
- Likewise, make X11_CreateWindow check this attribute before firing
off a call to X11_GL_GetVisual/X11_GLES_GetVisual
- Added a sanity check to the start of X11_GL_LoadLibrary
- Tidied up SDL_x11opengles.h
- Moved ownership of the gles_data structure over to
X11_GLES_LoadLibrary/UnloadLibrary
- Should incorporate the 3 fixes posted by Andre Heider
This is obviously quite a bit to take in, but is (at least) a proof of
concept for the approach I think EGL/GLX mingling should take. Any
comments/criticism is much appreciated.
Both options default to "yes" via configure, and having libs/headers
for both installed is not unusual.
We default to OpenGL on this compile time combination, but can enforce
OpenGLES via setting the envvar SDL_VIDEO_X11_GLES.
This will be further refined based on community feedback.
Contributed by Andre Heider
We cannot use SDL_VIDEO_GL_DRIVER for both EGL and GLES1/2, so rename
the envvar for EGL to SDL_VIDEO_EGL_DRIVER and keep SDL_VIDEO_GL_DRIVER
for GLES1/2.
Contributed by Andre Heider
Using SDL_UpdateWindowRects () with SDL_Rect instances, that do not clip
to the bounds of the SDL_Window, causes a BadValue error on X11, if shared
memory is enabled:
X Error of failed request: BadValue (integer parameter out of range for
operation)
Major opcode of failed request: 142 (MIT-SHM)
Minor opcode of failed request: 3 (X_ShmPutImage)
Value in failed request: 0x5
Serial number of failed request: 67
Current serial number in output stream: 70
Attached is a test program to recreate the behaviour along with a simple
patch for SDL_x11framebuffer.c, which reduces the rect area to the window
bounds.
I am not sure, if XPutImage() is more robust here, so I applied it to the
non-shared memory version, too.
Here is my first rough attempt. "testrelative" feels right to me, but I'd like it someone else tested this, especially compared to Linux/OSX. The "Ctrl+r" to switch between relative and normal mouse movements seems to work flawlessly. With relative mouse movement, the only way to change focus is via keyboard. I'm not sure if that is the correct approach, but that would seem to be the most useful mode for games. Still, if my assumption is wrong, I can fix that no problem.
When subclassing SDLUikitAppDelegate you usually override the -postFinishLaunch
method so that you can (for example) call a different function than SDL_main.
As it is now, however, overriding -postFinishLaunch will bar out the
SDL_SetHint for the idle timer. Attached a simple patch the moves the hint
callback to the -application:didiFinishLaunchingWithOptions:
