From: Glenn Maynard
To: sdl@libsdl.org
Subject: [SDL] SDL_SetVideoMode() failing and not setting an error (patch)
Running an OpenGL SDL application off 1.2.7, at SDL_InitSubSystem(SDL_INIT_VIDEO) time:
Warning: Unable to initialize AAlib mouseUsing AAlib driver: Slang driver 1.0 (slang)
SDL_SetVideoMode then fails; SDL_GetError() returns "".
The installation problem is straightforward: X (the higher priority
driver) isn't running, so SDL is falling back on aalib. However,
no error is being set when aalib fails to initialize. This also
happens with the svgalib driver.
SDL_video.c line ~653 sets mode to NULL, since aalib didn't return an
OpenGL surface. Line ~711 ("failed setting a video mode") returns NULL.
The attached patch sets an error. It's a horrible, useless error
message--it should really say eg. "aalib does not support OpenGL"; this
should probably be done earlier in the individual drivers, too.
From: Stephane Marchesin
Subject: [SDL] [Patch] inlining memcpy functions
I (finally) did some benchmarking of the misc mmx & sse blitting
functions, and found a little bottleneck in the memcpy ones : you get
~10% more performance on small surface blitting if you inline them.
From: Tyler Montbriand
Subject: [SDL] Detecting Opteron CPU features
I can now get SDL_cpuinfo.c to detect the AMD Opteron's RDTSC, MMX, MMXEXT,
3DNOW, 3DNOWEXT, SSE, and SSE2 instruction set extensions under Linux. It
took one #ifdef'ed block of new asm code to account for the 64-bit flags
register, but the other two blocks worked fine without modification, just
needed to modify the #ifdef's a bit.
From: Tyler Montbriand
Subject: [SDL] Opteron MMX patches for SDL_blit.c and SDL_blit_A.c
The inline MMX assembly in SDL_blit.c and SDL_blit_A.c compiles and runs fine
unmodified under AMD Opteron. The inline assembly in SDL_yuv_mmx.c and
SDL_blit_N.c unfortunately isn't directly compatible.
I've included diffs from SDL_blit.c and SDL_blit_A.c that allow the MMX
assembly to be compiled when USE_ASMBLIT, __x86_64__, and __GNUC__ are all
defined. All I had to modify was typedefs, the inline assembly itself wasn't
touched.
From: Tyler Montbriand
Subject: [SDL] Opteron MMX patches for SDL_RLEaccel.c
The inline MMX assembly in SDL_RLEaccel.c compiles unmodified if the typedefs
are adjusted to accept __x86_64__. The diff from SDL-1.2.7 is attached.
Hi folks,
based on Eric Wing's patch, I created the attached patch which fixes
the OpenGL coordinate inversion bug in SDL. It works fine over here on
10.3 with Ryan's test program (which I also attached).
There is another change in it: I removed the "- 1" in the two lines
using CGDisplayPixelsHigh()... while I understand from a logical point
of view why they *should* be correct, I checked the actual values
computed that way, and they were off-by-one. After removing the " - 1",
the returned mouse coordinates are correct. I checked this by moving
the mouse to the screen top/bottom in fullscreen mode, BTW. With the
change, the proper values 0 and 479 are returned (in 640x480 mode).
Sam, you may still want to test on 10.1, it's very simple using Ryan's
minimal test code :-)
Cheers,
Max
(Here is the reproduction case for revision history's sake...)
/*
* To compile:
* gcc -o test test.c `sdl-config --cflags` `sdl-config --libs` -framework OpenGL
*
* --ryan.
*/
#include <stdio.h>
#include "SDL.h"
#include "SDL_opengl.h"
int main(int argc, char **argv)
{
Uint32 flags = SDL_OPENGL /* | SDL_FULLSCREEN */;
SDL_Surface *screen;
SDL_Event event;
int done = 0;
GLfloat ratio;
SDL_Init(SDL_INIT_VIDEO);
SDL_ShowCursor(0);
if ((argv[1]) && (strcmp(argv[1], "--grab") == 0))
SDL_WM_GrabInput(SDL_GRAB_ON);
screen = SDL_SetVideoMode(640, 480, 0, flags);
if (!screen)
return(42);
ratio = ((GLfloat) screen->w) / ((GLfloat) screen->h);
glClearColor( 0.0f, 0.0f, 0.0f, 0.0f );
glClearDepth( 1.0f );
glEnable( GL_DEPTH_TEST );
glDepthFunc( GL_LEQUAL );
glViewport( 0, 0, screen->w, screen->h);
glMatrixMode( GL_PROJECTION );
glLoadIdentity();
gluPerspective( 45.0f, ratio, 0.1f, 100.0f );
glMatrixMode( GL_MODELVIEW );
glLoadIdentity();
glClear(GL_COLOR_BUFFER_BIT | GL_DEPTH_BUFFER_BIT);
SDL_GL_SwapBuffers();
// eh, close enough.
#define MAX_X 6.12
#define MAX_Y 4.50
while (!done)
{
int x, y;
GLfloat glx, gly;
if (!SDL_WaitEvent(&event))
break;
switch (event.type)
{
case SDL_KEYUP:
if (event.key.keysym.sym == SDLK_ESCAPE)
done = 1;
break;
}
SDL_GetMouseState(&x, &y);
glx = ((((GLfloat) x) / ((GLfloat) screen->w)) - 0.5f) * MAX_X;
gly = ((((GLfloat) y) / ((GLfloat) screen->h)) - 0.5f) * MAX_Y;
glClear(GL_COLOR_BUFFER_BIT | GL_DEPTH_BUFFER_BIT);
glLoadIdentity();
glTranslatef(glx,-gly,-6.0f);
glBegin(GL_TRIANGLES);
glColor3f(1,0,0); glVertex3f( 0.00f, 0.25f, 0.00f);
glColor3f(0,1,0); glVertex3f(-0.25f, -0.25f, 0.00f);
glColor3f(0,0,1); glVertex3f( 0.25f, -0.25f, 0.00f);
glEnd();
SDL_GL_SwapBuffers();
}
SDL_Quit();
return(0);
}
/* end of test.c ... */
From: Hayashi Naoyuki
Subject: Re: Tru64 cdrom and pthread option fix
I wrote
/* Some CD-ROM drives cannot play the first 150 frames. */
in src/cdrom/osf/SDL_syscdrom.c
and
Some CD-ROM drives(ex. TEAC CD-532E) cannot play first 150 frames(aka gap).
but this is not right.
It is written in MMC that LBA = 4500*M + 75*S + F - 150.
From: Hayashi Naoyuki
Subject: Tru64 cdrom and pthread option fix
SDL-1.2.7/src/cdrom/osf/SDL_syscdrom.c Fixes
Some CD-ROM drives(ex. TEAC CD-532E) cannot play
first 150 frames(aka gap).
CheckDrive() fix.
The end of comment is "* /" in AddDrive().
SDL-1.2.7/configure.in Fixes
Add *-*-osf pthread option and
delete line SDL_LIBS="$SDL_LIBS -lrt".
From: Alan Swanson
Subject: Re: [SDL] Fatal signal when initiaize with USB joystick on 2.6.2 kern
On Fri, 2004-03-05 at 15:09, Sam Lantinga wrote:
> > Fred, how does the attached patch work for you? Do all your axes work?
>
> I think you meant the two lines to be reversed.
> I checked a more robust version of this into CVS. Fred, can you see if
> it works?
You've misread the code. :-/
For coef[0] and coef[1], it is (values[2]+values[1]) / 2 and then add
or subtract values[4] to the answer.
From: "Mike Gorchak"
Subject: Misc fixes again
I've added custom WM info for the QNX (same as generic for now), which allows to compile all applications, which included SDL_syswm.h directly, otherwise DISABLE_X11 appearing to be not defined and X11 info was used instead, which is wrong.
From: Bartosz Fenski aka fEnIo
Subject: outdated entry in INSTALL file
I've just found in INSTALL file:
If you are cross-compiling from Linux to Win32, you should read
the file README.Win32
But you don't shipped this file with SDL. I've found in CVS logs that
this file has been deleted and this information is now in FAQ.
From: Martin_Storsj
Subject: Dynamic loading of ALSA
I recently discovered that SDL can dynamically load ESD and aRts, and
made a patch which adds this same functionality to ALSA.
The update for configure.in isn't too good (it should e.g. look for
libasound.so in other directories than /usr/lib), because I'm not too
good at shellscripting and autoconf.
The reason for using dlfcn.h and dlopen instead of SDL_LoadLibrary and
SDL_LoadFunction is that libasound uses versioned symbols, and it is
necessary to load the correct version using dlvsym. This isn't probably
any real portability issue, because ALSA is linux-only.
From: Martin_Storsj
Subject: Slight bug in ESD and aRts
When I experimented with the ALSA-patch, I found a slight bug in the
Load{ESD,ARTS}Library-functions. The check of whether a function pointer
was correctly loaded looks like this right now:
*esd_functions[i].func = SDL_LoadFunction(esd_handle,
esd_functions[i].name);
if ( ! esd_functions[i].func )
Isn't that supposed to be ( ! *esd_functions[i].func )?
From: "Mike Gorchak"
Subject: Just a misc. fixes
1) I've added a small fix to the .qpg file - SDLqpg.diff. Added autoconf variables like the @SDL_BINARY_AGE@ instead of directly using "...so.7".
call SDL_PumpEvents() so the X11 driver sets its this->hidden->switch_waiting,
then set a fullscreen OpenGL window (which makes the X11 driver tear down and
create a new window instead of just resizing the existing one), poll for
events, and the newly-created window will think it needs to pop back to a
window. Fixed by resetting switch_waiting to zero if X11_CreateWindow() had
to tear down a previous window.
From: "Ryan C. Gordon"
Subject: Re: MacOS X bugs...
This isn't an ideal patch (trying to open a joystick that has previously
been unplugged will report success, but it'll just never give any input,
etc), but it handles the worst case of deadlock in the event subsystem.
From: "Juergen \"George\" Sawinski"
Subject: Problems compiling libsdl with gcc 3.3
I attached another patch (configure.in.patch) that might resolv issues
for some ppl finding the correct place of libartsc.so.*.
(here, the output of "artsc-config --libs" contains two "-L.*",
therefore the respecting arts_lib_spec contains the wrong path).
On BeOS is should setup it the same way, but it only does when Tracker
wasn't restarted.
I checked code and it looks like a hack to me :(
It looks for env variable and than comapres it to default when OpenTracker
was started after boot, and wasn't restarted. That's probably ok, for that
exact case. Unfortunetly that variable isn't always like that. For
example, after Tracker crashes and is restarted, env variable most
probably is different (depends on how Tracker was restarted, by what
application, etc... for example: i have launcher application from which i
can restart Tracker, and after that nev variable points to that
application's directory, not Tracker's).
ut2004 makes a 2D window for the splash screen, which sets the screen
surface's pixels field. Then we tear down that video mode and create a GL
context, and the Quartz target isn't resetting the pixels field to NULL.
When you just create a GL window, the structure is memset'd to zero the
first time through, so unless you hit ut2004's codepath, you won't see the
bug. :)
Without this patch, quitting a windowed ut2003/ut2004 game makes the OS dump a
warning about a bogus free() to stderr, but it doesn't actually crash. All we
need to do is explicitly initialize the current->pixels field.
