From: "Mike Gorchak"
Subject: Re: SDL 1.2.6
- minor changes about shared library building under QNX6 into README.QNX
- added forgotten libSDLmain.a into distribution, SDL.qpg.in
- added header guards to the all headers.
- fixed fullscreen double buffered mode.
- fixed Photon crashes after/during using fullscreen OpenGL modes.
- added GL_MakeCurrent function.
- added SDL_VIDEOEXPOSE event, when OpenGL window have been resized
- added more HAVE_OPENGL checks to avoid dead code compilation without using OpenGL
- finished code reorganization (began into previous patches).
From: "Mike Gorchak"
Subject: Patches for tests and QNX6
Here more fixes for the QNX6 in sdlqnx.diff file:
- Spellchecked README.QNX (thanks to Julian Kinraid)
- Fixed bugs in fullscreen mode: window region wasn't on top by default, so \
it caused some artifacts to be appeared on the screen, prevent window conten\
ts default filler in Photon while in fullscreen mode, it damages the screen.
- Added support for the SDL_VIDEO_WINDOW_POS, SDL_VIDEO_CENTERED env variabl\
es.
- Some minor code restructurization.
From: Stephane Marchesin
Subject: Re: [SDL] [patch] MMX alpha blit patches with MMX detection
I think everything is correct now. I've done as much testing as I could,
but some real-world testing wouldn't hurt, I think.
The patch is here : http://icps.u-strasbg.fr/~marchesin/sdl_mmxblit.patch
If you do byte-by-byte comparison of the output between C and MMX
functions, you'll notice that the results for 555 and 565 RGB alpha
blits aren't exactly the same. This is because MMX functions for 555 and
565 RGB have an higher accuracy. If you want the exact same behaviour
that's possible by masking the three lower alpha bits in the MMX
functions. Just ask !
I removed one MMX function because after I fixed it to match its C
equivalent, it revealed to be slower than the C version on a PIII
(although a bit faster on an Athlon XP).
I've also added MMX and PIII replacements for SDL_memcpy. Those provide
some speed up in testvidinfo -benchmark (at least for me, under linux &
X11).
From: "Mattias Engdeg?rd"
Subject: bug in SDL_GetRGB/GetRGBA
There's an embarrassing bug in GetRGB/GetRGBA which apparently has been there
for years. It incorrectly converts colours with < 8 bits/channel.
It came to my attention today in #sdl.
What it does now is (for each channel):
rv = (pixel & fmt->Rmask) >> fmt->Rshift;
*r = (rv << fmt->Rloss) + (rv >> (8 - fmt->Rloss));
which is wrong; the last line should be
*r = (rv << fmt->Rloss) + (rv >> (8 - (fmt->Rloss << 1)));
From: Peter Naulls
Subject: RISC OS patches
This patch includes a patch to configure.in for RISC OS cross compiling
which was somehow missed from last time.
It also corrects BPP setting and checking behaviour.
From: Darrell Walisser
Subject: Re: Updated projects?
There was one additional patch I forgot to send. I needed to add a few
exports to the exports file.
From: Darrell Walisser
Subject: Re: Updated projects?
>> Did you get a chance to look at my "Custom Cocoa" demo? I have a few
>> minor patches that enable SDL/Cocoa integration, and a project
>> template.
>
> I didn't yet, but go ahead and send me the patches. :)
>
I updated the patch for current CVS. There are a lot of changes, but I
don't think I've broken anything. This patch also improves the behavior
of window minimize/deminimize.
From: Max Horn
Subject: SDL OSX fullscreen FIX
the attached patch fixes the fullscreen problems on SDL/OSX. The cause
was that click events are bounded by winRect. Now, winRect is set to
the size of the video surface. But if you e.g. request a 640x420
surface, you might get a 640x480 "real" surface. Still,
SDL_VideoSurface->h will be set to 420! Thus, the upper 60 pixels in my
example received no mouse down events.
My fix simply disables this clipping when in full screen mode - after
all, all clicks then should be inside the screen surface. Higher SDL
functions ensure that the coordinates then are clipped to 640x420. It
works fine in all my tests here. I don't know if it's the right thing
to do in multi screen scenarios, though.
From: "Mike Gorchak"
Subject: New patches for QNX6
Here my patches for the SDL/QNX:
QNXSDL.diff - diff to non-QNX related sources:
- updated BUGS file, I think QNX6 is now will be officially supported
- configure.in - added shared library support for QNX, and removed dependency between the ALSA and QNX6.
- SDL_audio.c - added QNX NTO sound bootstrap insted of ALSA's.
- SDL_sysaudio.h - the same.
- SDL_nto_audio.c - the same.
- SDL_video.c - right now, QNX doesn't offer any method to obtain pointers to the OpenGL functions by function name, so they must be hardcoded in library, otherwise OpenGL will not be supported.
- testsprite.c - fixed: do not draw vertical red line if we are in non-double-buffered mode.
sdlqnxph.tar.gz - archive of the ./src/video/photon/* . Too many changes in code to make diffs :) :
+ Added stub for support hide/unhide window event
+ Added full YUV overlays support.
+ Added window maximize support.
+ Added mouse wheel events.
+ Added support for some specific key codes in Unicode mode (like ESC).
+ Added more checks to the all memory allocation code.
+ Added SDL_DOUBLEBUF support in all fullscreen modes.
+ Added fallback to window mode, if desired fullscreen mode is not supported.
+ Added stub support for the GL_LoadLibrary and GL_GetProcAddress functions.
+ Added resizable window support without caption.
! Fixed bug in the Ph_EV_EXPOSE event handler, when rectangles to update is 0 and when width or height of the rectangle is 0.
! Fixed bug in the event handler code. Events has not been passed to the window widget handler.
! Fixed codes for Win keys (Super/Hyper/Menu).
! Fixed memory leak, when deallocation palette.
! Fixed palette emulation code bugs.
! Fixed fullscreen and hwsurface handling.
! Fixed CLOSE button bug. First event was passed to the handler, but second terminated the application. Now all events passed to the application correctly.
- Removed all printfs in code, now SDL_SetError used instead of them.
- Disabled ToggleFullScreen function.
README.QNX - updated README.QNX file. Added much more issues.
From: Max Horn
Subject: SDL whitespace patch
this patch is lowest priority: it only contains indention fixes. In
some places, instead of tabs, whitespaces where used for indention. But
of course that only "looks good" if you use the exact same tab width as
the author of those files. I tried to replace such cases with tabs for
a more uniform indention. I only did that for a few files.
So, no code changes in this, only whitespaces transformed.
From: Shawn Kirst
Subject: SDL-1.2.5 patch to add ARB_multisample support
Attached is a patch I have written for SDL-1.2.5 that adds ARB_multisample
support. I only have the X11 and Win32 video patched. The Win32 patch also
adds support for WGL_ARB_pixel_format, as it was required for getting a
multisample capable pixel format. No additional GL header files are required
to compile on either platform (though you need an up-to-date glx.h for X11).
Requesting a multisample pixel format is made possible using
SDL_GL_SetAttribute with the two new SDL_GLattr's I've added
(SDL_GL_SAMPLE_BUFFERS and SDL_GL_SAMPLES). I've been using SDL in my
projects for quite a while now, so I am happy to contribute back to the
project. Now you can have and control FSAA in your SDL/GL apps at the
application level!
since the X11 target needs to query for extensions in there...hopefully this
other platforms don't need to have a current context before symbol lookup
works or something strange like that... --ryan.
From: Alan Swanson
Subject: Re: [SDL] New XFree 4.3 Video Mode Patch
I have a wee amendment that moves the qsort in set_best_resolution
to only occur after failing to find an exact match only. This would
make absolutely sure we get a user set mode.
While I've never had any problems for my normal resolutions (1280x1024,
1024x768, 800x600 & 640,480) while closely examining the output from
qsort I've noticed it doesn't seem to sort the modes fully. These is
one definite wrong at 1152x768 and a few that just look wrong to me.
From a program (attached) I made to examine this more easily. X has
sorted its mode list using the same method as ours (plus frequency),
and our user modes get inserted without any other movement.
On the patch I've made I've also changed cmpmodes to sort on vertical
resolution and then horizontal. Ie vertical is now most significant
bit.
From: Pasi K?rkk?inen
Subject: [PATCH] fix SDL OpenGL segfault with DRI/Mesa drivers and Glew
Hello!
The attached patch fixes a bug in SDL which causes SDL to crash in
X11_GL_Shutdown() if you are using DRI/Mesa drivers AND glew
(http://glew.sf.net).
The bug is caused by a namespace collision affecting dlsym() to fetch wrong
pointer for glXReleaseBuffersMESA() (uninitialized pointer from glew because
the extension is NOT supported by the driver) and then SDL calling it in
X11_GL_Shutdown().
SDL should check if the glXReleaseBuffersMESA() is really supported by the
driver (from the extensions string) before calling it.
Attached patch adds extension string parsing to check if
glXReleaseBuffersMESA() is really supported (and this way
prevents the segfault).
Availability of the extensions should be _always_ checked from the
extensions string rather than using dlsym()!
Please add it to the next version of SDL.
Thanks to gltron and author of glew to help fixing this.
From: Alan Swanson
Subject: [SDL] New XFree 4.3 Video Mode Patch
The current patch to fix the issues with XFree 4.3 it is a bit of
overkill to a simple problem. Default screen settings should be set in
X, not selected by SDL with environment variables. Any program or user
using non-standard or unset display modes get what they deserve :-)
If you look at the unsorted list of modes returned by X, here's mine;
1280 x 1024 @ 85.0 >
1024 x 768 @ 100.3 > USER
800 x 600 @ 125.5 > SET
640 x 480 @ 124.9 >
1280 x 1024 @ 75.0 ]
1280 x 1024 @ 60.0 ]
1280 x 960 @ 85.0 ] X11
1280 x 960 @ 60.0 ] AUTO
1152 x 864 @ 75.0 ]
1152 x 768 @ 54.8 ]
960 x 720 @ 120.0 ]
...
640 x 400 @ 85.1 ] 256k
576 x 432 @ 150.0 ] 249k PIXEL
640 x 350 @ 85.1 ] 224k COUNT
576 x 384 @ 109.6 ] 221k
...
The user set modes come first followed by X set modes which are ordered
by decreasing number of pixels and refresh.
The reason why every other library or program not using SDL was working
is due to SDL scanning the modes in reverse getting X11 provided modes
modes with the lowest refresh.
The solution is to scan forward for the first user set mode or highest X
mode. The qsort still keeps user set modes above higher refresh modes
added by X.
For the best match we still reverse search for the nearest larger size
and then try to find a higher version of it.
From: "alan buckley" <alan_baa@hotmail.com>
Subject: Modification for RISC OS SDL port
My name is Alan Buckley and I have been working on the RISC OS SDL port.
I've attached a modification to SDL_WimpVideo.c to this email that ensures
the Window is displayed immediately when the video mode is set to a windowed
mode.
