Scott Percival 2011-07-03 06:41:51 PDT
This submission is aimed at making life easier for OpenGL ES capable devices
running a X11 stack (e.g. Maemo, Meego, TrimSlice, other ARM SoC boards not
running Android). SDL's Pandora support already has the neccesary GLES-to-X11
glue code, however it's all ghetto'd off in Makefile.pandora and not very
flexible.
The patch:
- adds an awesome --enable-video-opengles option to configure
- re-modifies the opengles and opengles2 SDL_renderers to use function pointers
- no idea why this was removed?
- for SDL_Renderers, links in libGLESv1_CM, libGLES_CM (for PowerVR fans) or
libGLESv2 at runtime
- links in libEGL.so at runtime - the old code made an assumption that
eglFunctions could be pulled from the active GLES library, PowerVR for one
doesn't let you do that with their libGLESv2
- allows you to pick which of GLES v1 or v2 to load via
SDL_GL_CONTEXT_MAJOR_VERSION
So far I've tested this on a Nokia N900 (OMAP 3430/SGX 530 running Maemo 5) and
a Toshiba AC100 (Tegra 2 running Ubuntu 10.10). I haven't tested it on... well,
everything that isn't those two, such as a Pandora, iOS or Android device. The
Pandora specific code should be kept intact (fingers crossed), and nothing
painfully drastic has been added to the SDL_renderers. The library loading
sequence in SDL_x11opengles has been updated to accomodate both NVIDIA's
propensity to let developers get away with murder and PowerVR's alternative of
punishing every missed step.
The test apps work okay with GLES or GLES2 as the renderer. For some reason
alpha blending doesn't seem to work on the Tegra 2; last week NVIDIA pushed out
a new set of X11 GLES drivers, so I'll try and investigate once I upgrade
those. Also, this patch adds things to configure.in, include/SDL_config.h.in
and test/configure.in. I didn't know what the policy was re. committing
generated spaghetti from autotools, so ./autogen.sh has to be run again. Sorry.
I think that's about everything, let me know if there's anything I've
overlooked.
t.grundner@goto3d.de 2011-09-01 03:59:17 PDT
I figured out what is going on. GCC 4.5.2 assumes the stack is 16 byte aligned
by default. Therefore there are no AND alignment corrections necessary if we
wish to align a stack variable to a 16 byte boundary. That is bad if your OS
ABI is not 16 byte aligned. Windows 32 bit stacks are 4 byte aligned. This
results in the above mentioned SIGSEGV. This is also no problem if I compile
both SDL.dll and my app with MingW because MinGW/GCC inserts a
andl $-16, %esp
instruction right in the beginning of the main function. So at least the stack
of the thread calling the main function is 16 byte aligned. But as soon as I
start to use the SDL.dll from an application not compiled by MinGW there is no
ANDL safing my app.
However there is a GCC option that can change the default stack alignment:
-mpreferred-stack-boundary=num
Setting num=2 assumes a the stack is aligned to a 4 byte boundary. This results
in GCC inserting the necessary
andl $-16, %esp
into SDL_FillRect. Rebuilding SDL with
./configure "CFLAGS=-mpreferred-stack-boundary=2 -g -O3"
solved the problem.
IMHO this should also be a problem on Solaris.
The following links contain further information:
http://gcc.gnu.org/onlinedocs/gcc-4.5.2/gcc/i386-and-x86_002d64-Options.html#i386-and-x86_002d64-Optionshttp://www.agner.org/optimize/calling_conventions.pdf
The trick is that if you're using configure and you don't want to have SDL_config.h and SDL_revision.h to show up as modified, you need to configure and build from a separate directory.
You also need to include SDL_revision.h directly if you want to use the SDL_REVISION constant, as a side effect of these changes.
--HG--
rename : include/SDL_config.h.default => include/SDL_config.h
