Add OpenGL 3.X context creation support

Matthias Bentrup 2011-10-30 03:58:24 PDT

I've updated the context creation patch to include the bugfixes by Martin
Schreiber and also included a profile bit to request a ES2 compatible profile.

The wgl context creation may use 2 call to wglChoosePixelFormat if no
acceleration attribute is selected, this should work around a bug with buggy
AMD drivers (see #1254).

Fixed bug 1424 - Handling of alpha channel in Altivec accelerated blit functions

evilbite 2012-02-19 09:38:21 PST

There is only one Altivec accelerated blit function
(ConvertAltivec32to32_prefetch() or ConvertAltivec32to32_noprefetch(),
depending on the CPU used) that is supposed to handle all alpha combinations.
This works as follows for every pixel line:
1. Blit single pixels until an aligned address is reached
2. Accelerated blit as far as possible
3. Blit single remaining pixels
Part 2. is set up correctly to handle different combinations of the alpha
channels of the participating surfaces. Parts 1. and 3. only do a simple copy
of all the pixel's components from souce to destination. But when the source
surface has no alpha channel (Amask is 0, e.g. the video surface) the surface's
alpha value must be used instead. Otherwise crap (uninitialized data) is being
copied to the destiniation's alpha channel.

The attached patch is a quick'n'dirty solution to the problem. A more
sophisticated solution might require separate functions for different
combinations of the alpha channels of the participating surfaces.

Fixed bug 1424 - Handling of alpha channel in Altivec accelerated blit functions

evilbite 2012-02-19 09:38:21 PST

There is only one Altivec accelerated blit function
(ConvertAltivec32to32_prefetch() or ConvertAltivec32to32_noprefetch(),
depending on the CPU used) that is supposed to handle all alpha combinations.
This works as follows for every pixel line:
1. Blit single pixels until an aligned address is reached
2. Accelerated blit as far as possible
3. Blit single remaining pixels
Part 2. is set up correctly to handle different combinations of the alpha
channels of the participating surfaces. Parts 1. and 3. only do a simple copy
of all the pixel's components from souce to destination. But when the source
surface has no alpha channel (Amask is 0, e.g. the video surface) the surface's
alpha value must be used instead. Otherwise crap (uninitialized data) is being
copied to the destiniation's alpha channel.

The attached patch is a quick'n'dirty solution to the problem. A more
sophisticated solution might require separate functions for different
combinations of the alpha channels of the participating surfaces.

Fixed bug 1427 - integer passed to XChangeProperty() causes crash

Julian Coleman 2012-02-20 06:51:12 PST

In src/video/x11/SDL_x11video.c, the result of getpid(), i.e., a pid_t is
passed to:

XChangeProperty(..., 32, ...)

However, using 32 here means that Xlib treats the value as a long, and pid_t is
an int. So, we get a bus error inside Xlib.
The fix is to make sure that anything passed to XChangeProperty() is aligned
correctly. Note, that the other calls to XChangeProperty() pass long values
here.
The proposed patch makes a union of the pid_t return type from getpid() and a
dummy long. This has been tested to fix the bus error crash on NetBSD/sparc64.

Fix for double free when quitting on iOS

Tim Angus 2012-02-20 09:40:35 PST

As alluded to in the email thread "SDL2 error on iOS (doublefree)", I believe
the original cause of this bug is confusion over the purpose of
SDL_VideoDisplay::current_mode. It looks as though it is a weak reference to
another mode, albeit with value semantics. The iOS port treated it as a strong
reference however and claimed ownership, which is why things blew up. All the
patch really does it to stop treating current_mode as a strong reference.

To prevent this happening again it might be an idea to change current_mode to
be a pointer type rather than a value. This would certainly make its semantics
much more obvious. Failing that, a comment in the struct indicating its weak
reference properties might be wise.

Fixed bug 1419 - SDL_libgl2D.c breaks ndk-build

Philip Taylor 2012-02-15 10:43:47 PST

render/nds/SDL_libgl2D.c unconditionally includes NDS-only code. SDL's
Android.mk compiles source files matching

$(wildcard $(LOCAL_PATH)/src/render/*/*.c)

which includes that file, causing build errors when running ndk-build.

FreeBSD compile fix

Robert Millan

src/joystick/bsd/SDL_sysjoystick.c makes the invalid assumption that
__FreeBSD_kernel__ implies presence of "ucr_data" struct member. This
breaks recent versions of FreeBSD 10-CURRENT, FreeBSD 9-STABLE and
Debian GNU/kFreeBSD "wheezy/sid".

FreeBSD compile fix

Robert Millan

src/joystick/bsd/SDL_sysjoystick.c makes the invalid assumption that
__FreeBSD_kernel__ implies presence of "ucr_data" struct member. This
breaks recent versions of FreeBSD 10-CURRENT, FreeBSD 9-STABLE and
Debian GNU/kFreeBSD "wheezy/sid".

Quick android Makefile fix from Gabriel Jacobo

Missed a file in Frank's patch