felix
Building SDL 2.0.5, or even the Mercurial snapshot (r10608) with GCC 6.2.1 and --enable-video-directfb generates a number of compiler diagnostics and fails.
Simon Hug
The software renderer produces incorrect results when blending textures at an angle with certain blend modes. It seems that there were some edge cases that weren't considered when the SW_RenderCopyEx function was last changed. Or another bug possibly covered up the problem. (More on that in another bug report.)
Most of the issues come from the fact that the rotating function sets a black colorkey. This is problematic because black is most likely appearing in the surface and the final blit will ignore these pixels. Unless a colorkey is already set (the software renderer currently never sets one), it's very hard to find a free color. Of course it could scan over the whole image until one is found, but that seems inefficient.
The following blend modes have issues when drawn at an angle.
NONE: The black pixels get ignored, making them essentially transparent. This breaks the 'dstRGBA = srcRGBA' definition of the NONE blend mode.
MOD: Again, the black pixels get ignored. This also breaks the 'dstRGB = dstRGB * srcRGB' definition of the MOD blend mode, where black pixels would make the destination black as well. A white colorkey will work though, with some preparations.
BLEND: There are some issues when blending a texture with a translucent RGBA target texture. I - uh - forgot what the problem here exactly is.
This patch fixes the issues mentioned above. It mainly changes the code so it tries to do things without the colorkey and removes the automatic format conversion part from the SDLgfx_rotateSurface function. Getting the format right is something the caller has to do now and the required code has been added to the SW_RenderCopyEx function.
There's a small change to the SW_CreateTexture function. RLE encoding a surface with an alpha mask can be a lossy process. Depending on how the user uses the RGBA channels, this may be undesired. The change that surfaces with an alpha mask don't get encoded makes the software renderer consistent with the other renderers.
The SW_RenderCopyEx function now does these steps: Lock the source surface if necessary. Create a clone of the source by using the pixel buffer directly. Check the format and set a flag if a conversion is necessary. Check if scaling or cropping is necessary and set the flag for that as well. Check if color and alpha modulation has to be done before the rotate. Check if the source is an opaque surface. If not, it creates a mask surface that is necessary for the NONE blend mode. If any of the flags were set, a new surface is created and the source will be converted, scaled, cropped, and modulated. The rest of the function stays somewhat the same. The mask also needs to be rotated of course and then there is the NONE blend mode...
It's surprisingly hard to get the pixel from a rotated surface to the destination buffer without affecting the pixel outside the rotated area. I found a way to do this with three blits which is pretty hard on the performance. Perhaps someone has an idea how to do this faster?
As mentioned above, the SDLgfx_rotateSurface now only takes 8-bit paletted or 32-bit with alpha mask surfaces. It additionally sets the new surfaces up for the MOD blend mode.
I shortly tested the 8-bit path of SDLgfx_rotateSurface and it seemed to work so far. This path is not used by the software renderer anyway.
Vitaly Novichkov
Okay, when I researched code and algorithm, I tried to replace condition "while(dst >= target)" with "while(dst > target)" and crashes are gone.
Seems on some moments it tries to write into the place before memory block begin, therefore phantom crashes appearing after some moments.
Sylvain 2016-11-07 08:49:34 UTC
when rotated +90 or -90, some transparent lines appears, though there is no Alpha or ColorKey.
if you set a dummy colorkey, it will remove the line ...
if you set a some alpha mod, the +90/-90 get transparent but not the 0/180 ...
x414e54
It is a bit of a pain to update the library or rely on whatever version the user has on their computer for default mappings.
So providing an easily updatable text file via SDL_GameControllerAddMappingsFromFile is still currently the most viable way. However using this replaces all mappings provided by the SDL_HINT_GAMECONTROLLERCONFIG environment variable which may have come from the user's custom Steam mapping.
There should be an easy way for games to supply extra game controller mappings to fill in the differences between SDL versions without it clobbering the SDL_HINT_GAMECONTROLLERCONFIG environment variable.
Internally the mappings could use a priority system and if the priority is lower then it will not overwrite the mappings.
For now it just assumes SDL_HINT_GAMECONTROLLERCONFIG is the highest priority, the default hardcoded are the lowest and anything set via the API is medium.
Sam
I've discovered that when building on powerpc64le (and probably powerpc64) SDL's configure script fails to detect dynamic library support, causing it to build a static library. This causes link failures due to undefined symbols later when packages link with -lSDL.
This seems to be because the included autotools package is too old to detect powerpc64le. This change corrects the problem for me but newer versions of autotools should handle it without a patch
Mark Pizzolato
On Windows with Visual Studio, when building SDL as a static library using the x86 (32bit) mode, several intrinsic operations are implemented in code in SDL_stdlib.c.
One of these, _allshr() is not properly implemented and fails for some input. As a result, some operations on 64bit data elements (long long) don't always work.
I classified this bug as a blocker since things absolutely don't work when the affected code is invoked. The affected code is only invoked when SDL is compiled in x86 mode on Visual Studio when building a SDL as a static library. This build environment isn't common, and hence the bug hasn't been noticed previously.
I reopened#2537 and mentioned this problem and provided a fix. That fix is provided again here along with test code which could be added to some of the SDL test code. This test code verifies that the x86 intrinsic routines produce the same results as the native x64 instructions which these routines emulate under the Microsoft compiler. The point of the tests is to make sure that Visual Studio x86 code produces the same results as Visual Studio x64 code. Some of the arguments (or boundary conditions) may produce different results on other compiler environments, so the tests really shouldn't be run on all compilers. The test driver only actually exercised code when the compiler defines _MSC_VER, so the driver can generically be invoked without issue.
Ozkan Sezer
On systems with old glibc, such mine with glibc-2.8, the following warning
is issued and is fixed easily by defining _GNU_SOURCE:
/home/me/SDL2-2.0.5/src/video/x11/SDL_x11modes.c: In function 'CalculateXRandRRefreshRate':
/home/me/SDL2-2.0.5/src/video/x11/SDL_x11modes.c:263: warning: implicit declaration of function 'round'
/home/me/SDL2-2.0.5/src/video/x11/SDL_x11modes.c:263: warning: incompatible implicit declaration of built-in function 'round'
x414e54
I have implemented Drag and Drop and Clipboard support for Wayland.
Drag and dropping files from nautilus to the testdropfile application seems to work and also copy and paste.
This no longer uses a script to generate code for every possible type
conversion or resampler. This caused a bloat in binary size and and compile
times. Now we use a handful of more generic functions and assume staying in
the CPU cache is the most important thing anyhow.
This shrinks the size of the final build (in this case: macOS X amd64, -Os to
optimize for size) by 15%. When compiling on a single core, build times drop
by about 15% too (although the previous cost was largely hidden by multicore
builds).
Alex Baines
I realized overnight that my patch probably broke text input events with UIM, and I confirmed that it does. Can't believe I overlooked that... I've been making stupid mistakes in these patches recently, sorry.
Anyway, *this* one seems to fix it properly. Knowing my luck it probably breaks something else.
Patch uses XkbFreeKeyboard to free the memory returned by XkbGetMap.
Earlier implementation called XkbFreeClientMap which frees all the maps
but not data->xkb structure itself, XkbFreeKeyboard will free maps and
the structure.
Philipp Wiesemann
Sam Lantinga
Thomas Perl
Ryan C. Gordon
Alex Baines