Jona
The following explains why this bug was happening:
This crash was caused because the audio session was being set as active [session setActive:YES error:&err] when the audio device was actually being CLOSED. Certain cases the audio session being set to active would fail and the method would return right away. Because of the way the error was handled we never removed the SDLInterruptionListener thus leaking it. Later when an interruption was received the THIS_ object would contain a pointer to an already released device causing the crash.
The fix:
When only one device remained open and it was being closed we needed to set the audio session as NOT active and completely ignore the returned error to successfully release the SDLInterruptionListener. I think the user assumed that the open_playback_devices and open_capture_devices would equal 0 when all of them where closed but the truth is that at the end of the closing process that the open devices count is decremented.
(It gets upset at the -2147483648, thinking this should be an unsigned value
because 2147483648 is too large for an int32, so the negative sign upsets the
compiler.)
The concern is that a massive int sample, like 0x7FFFFFFF, won't fit in a
float32, which doesn't have enough bits to hold a whole number this large,
just to divide it to get a value between 0 and 1.
Previously we would convert to double, to get more bits, do the division, and
cast back to a float, but this is expensive.
Casting to double is more accurate, but it's 2x to 3x slower. Shifting out
the least significant byte of an int32, so it'll definitely fit in a float,
and dividing by 0x7FFFFF is still accurate to about 5 decimal places, and the
difference doesn't appear to be perceptable.
Fixes problems launching Firewatch on Linux (which statically links SDL but
also dynamically loads a system-wide copy from a plugin shared library) with
a newer SDL build.
The change makes sure that SDL_vsnprintf() nul terminates if it is
using _vsnprintf() for the job.
I made this patch for Watcom, whose _vsnprintf() doesn't guarantee
nul termination. The preprocessor check can be extended to windows
in general too, if required.
Closes bug #3769.
Daniel Gibson
Sorry, but it seems like Microsoft didn't fix the issue properly.
I just updated my Win10 machine, it now is Version 1803, Build 17134.1
I tested with SDL2 2.0.7 (my workaround was released with 2.0.8) and still got
lots of events that directly undid the prior "real" events - just like before.
(See simple testcase in attachement)
By default it sets SDL_HINT_MOUSE_RELATIVE_MODE_WARP - which triggered (and on my machine still triggers) the buggy behavior. You can start it with -raw, then it'll not set that hint and the events will be as expected.
The easiest way to see the difference is looking at the window title, which shows accumulated X and Y values: If you just move your mouse to the right, in -raw mode the number just increases. In non-raw mode (using mouse warping) it stays around 0.
I also had a WinAPI-only testcase: https://gist.github.com/DanielGibson/b5b033c67b9137f0280af9fc53352c68
It just calls SetCursorPos(320,240); on each WM_MOUSEMOVE event, and it also
logs all those events to a mouseevents.log textfile.
This log indeed looks a bit different since the latest Win10 update: It seems like all those events with x=320 y=240 do arrive - but only after I stopped moving the mouse - even though the cursor seems to be moved back every frame (or so).
So moving the mouse to the right gives X coordinates like
330, 325, 333, 340, 330, ...
and then when stopping movement I get lots of events with X coordinate 320
Olli-Samuli Lehmus
If one creates a window with the SDL_WINDOW_FULLSCREEN_DESKTOP flag, and creates a render target with SDL_SetHint(SDL_HINT_RENDER_SCALE_QUALITY, "linear"), and afterwards sets SDL_SetHint(SDL_HINT_RENDER_SCALE_QUALITY, "nearest"), after minimizing the window, the scale quality hint is lost on the render target. Textures however do keep their interpolation modes.
Christian Herzig
pthread_mutex_trylock() and by the way, pthread_mutex_lock() do not set errno.
Pthread-methods directly return error code as int. See related man-pages for
details.
Michael Sartain
This is a quick pass at adding Linux RealtimeKit thread priority support to SDL.
It allows me to bump the thread priority to high without root privileges or setting any caps, etc.
rtkit readme here:
http://git.0pointer.net/rtkit.git/tree/README
Simon Hug
I just wanted to fix a simple compiler warning in SDL_ShowMessageBox on Windows (which Sam fixed recently) and ended up finding some issues.
Attached patch fixes these issues:
- Because Windows only reports the lower 16 bits of the control identifier that was pushed, the button IDs used by SDL (C type int, most likely 32 bits) can get cut off.
- The documentation states (somewhat ambiguously) that the button ID will be -1 if the dialog was closed, but the current code sets 0. For SDL 2.1, I think this should be a return code of SDL_ShowMessageBox itself. That will free up the button ID and it seems a more appropriate place for signaling this event.
- Ampersands in controls will create mnemonics on Windows (underlined letters that, if combined with the Alt key, will push the button). I was thinking of adding a hint or flag to let the users enable it, but that might have unexpected results.
- When the size of the text gets calculated, it doesn't use the same parameters as the static control. This can cut off text or wrap it weirdly.
- On Windows, the Tab key is used to switch between control groups and sometimes between buttons in dialogs. This didn't seem to work correctly.
Attached patch also adds:
- Icons. Just the system ones that can be loaded with the ordinals IDI_ERROR, IDI_WARNING and IDI_INFORMATION.
- A button limit of 2^16 - 101.
- Some more specific error messages, but they never reach the user because how SDL_ShowMessageBox handles them if an implementation returns with an error.
This is commented out in SDLActivity.java, with the note #CURSORIMPLEENTATION because it requires API 24, which is higher than the minimum required SDK
Ozkan Sezer
The following patch defines _WIN32_WINNT_WIN7 if it is not already
defined in core/windows/SDL_windows.c, similar to what is already
there for _WIN32_WINNT_VISTA.
Eric Wasylishen
This bug was reintroduced by https://hg.libsdl.org/SDL/rev/fcf24b38a28a
The steps to reproduce are the same: run the "testrelative" SDL demo with "--info all",
connect a USB mouse with a scroll wheel, and roll the scroll wheel one "notch". You'll get log output like:
testdraw2[1644:67222] INFO: SDL EVENT: Mouse: wheel scrolled 0 in x and 0 in y (reversed: 1) in window 1
As far as I can tell macOS doesn't have an API for getting the number of "wheel notches"; I get a deltaY of 0.100006 for one "notch", and it's heavily accelerated (if you roll the wheel quickly you'll get large deltas). So NSEvent's deltaY is only meant to be used for scrolling a scroll view, with the given distance in points, not something like selecting an item in a game.
Here's a temporary patch that at restores the foor/ceil in Cocoa_HandleMouseWheel.
Not ideal, but at least it restores the ability to scroll one notch of a mousewheel.
Ozkan Sezer 2018-03-02 20:02:37 UTC
http://hg.libsdl.org/SDL/rev/d702b0c54e52 resulted in an error and
two warnings when compiled with mingw.
1. Error from SDL_windowstaskdialog.h:
In file included from src/video/windows/SDL_windowsmessagebox.c:29:0:
src/video/windows/SDL_windowstaskdialog.h:23:54: error: expected ')' before 'HWND'
This is fixed by removing unnecessary annotations:
2. Warning from SDL_assert.c:
src/SDL_assert.c: In function 'SDL_ExitProcess':
src/SDL_assert.c:138:1: warning: 'noreturn' function does return
Indeed ExitProcess() is prototyped with DECLSPEC_NORETURN, but
TerminateProcess() is not. This can be rectified by adding an
exit() call in there. Do NOTE, however, that requires building
with a libc:
3. Warning from SDL_windowsmessagebox.c:
src/video/windows/SDL_windowsmessagebox.c: In function 'WIN_ShowMessageBox':
src/video/windows/SDL_windowsmessagebox.c:513:9: warning: 'nCancelButton' may be used uninitialized in this function
My lazy solution was manually initializing nCancelButton to 0.
