It causes problems on modern Windows, when a player ALT-Tab's out of a window.
When they ALT-Tab back, the keystate gets confused.
Fixes Bugzilla #659.
--HG--
branch : SDL-1.2
From http://gcc.gnu.org/bugzilla/show_bug.cgi?id=50111:
Andrew Pinski 2011-08-18 02:25:15 UTC
I really doubt this is a bug in GCC but rather SDL's inline-asm.
"=X" (mmx_trace)
'X'
Any operand whatsoever is allowed.
They most likely want:
y
Any MMX register.
--HG--
branch : SDL-1.2
This should reduce unnecessary hardware interrupts. Apparently improves the
ALSA target on some systems.
Fixes Bugzilla #1170.
Thanks to Paul Cercueil for the patch!
--HG--
branch : SDL-1.2
The license on IOHIDLib.h says we have to distribute a copy of the APSL with
every copy of SDL, and include a "prominent notice" in the documentation, and
that's annoying for something that's only there because Apple forgot to copy
a file in an ancient SDK.
--HG--
branch : SDL-1.2
The new behavior is triggered by setting the environment variable
SDL_FBCON_DONT_CLEAR.
This is useful in certain circumstances (specifically: a game launcher on an
embedded device can leave a "now loading!" screen on the framebuffer while
another process is starting up).
Fixes Bugzilla #1251.
Thanks to Paul Cercueil for the patch!
--HG--
branch : SDL-1.2
I'm not sure if there was a way to make CGLSetFullScreen() work, but the
Cocoa path works fine on all Intel Macs.
Please note that the "official" way to do fullscreen in 10.7 is just to make
a window the size of the display mode, and OS X is now smart enough to
recognize this as "fullscreen", but this helps earlier versions of the OS in
any case.
--HG--
branch : SDL-1.2
"Centering" is a little off-center, vertically. Apparently in non-fullscreen
apps, this is more aestetically pleasing or something.
Fixes positioning on fullscreen video modes on Mac OS X.
--HG--
branch : SDL-1.2
Mac OS X Lion (10.7) always returns NULL from CGDisplayBaseAddress(), so all
the direct framebuffer access code has been replaced with something a little
higher level. This also necessitated taking out the scary SDL_DOUBLEBUF
codepath that was trying to time out vsync by hand. :)
--HG--
branch : SDL-1.2
Newer kernels seem to report bogus axes in the higher ranges, for example
with a standard PlayStation 3 controller plugged in via USB.
--HG--
branch : SDL-1.2