We are currently migrating Bugzilla to GitHub issues.
Any changes made to the bug tracker now will be lost, so please do not post new bugs or make changes to them.
When we're done, all bug URLs will redirect to their equivalent location on the new bug tracker.

Bug 1343

Summary: SDL_AudioInit, SDL_AudioQuit and again SDL_AudioInit crashes
Product: SDL Reporter: Ellie <etc0de>
Component: audioAssignee: Ryan C. Gordon <icculus>
Status: RESOLVED FIXED QA Contact: Sam Lantinga <slouken>
Severity: critical    
Priority: P2 CC: aschiffler
Version: HG 2.0   
Hardware: x86   
OS: Windows (All)   

Description Ellie 2011-12-14 20:20:47 UTC 
In an attempt to reopen audio at some point using another device, I ran  into this crash which many windows users can reproduce (probably on machines without proper float 32bit support) and which I can reproduce myself on Linux in wine emulation:


#include <windows.h>

#include "SDL.h"

void audiocallback(void *intentionally_unused, Uint8 *stream, int len) {
    memset(stream, 0, (unsigned int)len);
}

int main(int argc, char** argv) {
    SDL_Init(SDL_INIT_TIMER);
    SDL_AudioInit("waveout");

    SDL_AudioSpec fmt;

    memset(&fmt,0,sizeof(fmt));
    fmt.freq = 48000;
    fmt.format = AUDIO_F32SYS;
    fmt.channels = 2;
    fmt.samples = 4096;
    fmt.callback = audiocallback;
    fmt.userdata = NULL;

    if (SDL_OpenAudio(&fmt, NULL) < 0) {
        SDL_AudioQuit();
        printf("We will crash right now:\n");fflush(stdout);
        SDL_AudioInit("directsound");
    }
    printf("Or did we not?\n");fflush(stdout);

    return 0;
}


The only workaround is to use SDL_OpenAudio() again without using SDL_AudioQuit() in between. Since I assume that will result in memory leaks or even worse problems (does it even close the old device properly?), that is pretty odd.
Comment 1 Ryan C. Gordon 2013-07-12 00:39:08 UTC 
This doesn't crash for me on Windows now; it's possible we fixed the issue at some point.

(It's worth noting that the test program doesn't fail the SDL_OpenAudio() call now, but if I set fmt.channels to 88 or something to force it to fail, the crash doesn't happen on the quit/reinit, either.)

Can you confirm this is still giving you problems with the latest version of SDL? I'm happy to look into this further if you can still reproduce it.

--ryan.
Comment 2 Ellie 2013-07-12 07:42:20 UTC 
I had this mainly tested in wine because float32 audio appeared to be unavailable which forced an audio reopen. On Windows, the example code usually worked because SDL_OpenAudio doesn't return an error (in which case this bug can NOT be reproduced).

Sadly, I'm currently blocked by https://bugzilla.redhat.com/show_bug.cgi?id=983936 from wine testing.
Comment 3 Andreas Schiffler 2014-07-08 03:15:54 UTC 
Recommend to close this bug as no-repro.

The test audio_initQuitAudio (which is part of the testautomation suite) and can be run with:
  ./testautomation.exe --filter audio_initQuitAudio
used to fail (and was disabled because of this bug).

The test passes now on cygwin as well as VS on my Win7 box.

$ ./testautomation.exe --filter audio_initQuitAudio
INFO:  07/07/14 20:15:02: ::::: Test Run /w seed '3LKSXORVDKCBUK75' started
INFO:  07/07/14 20:15:02: Filtering: running only test 'audio_initQuitAudio' in suite 'Audio'
INFO:  07/07/14 20:15:02: ===== Test Suite 1: 'Audio' started
INFO:  07/07/14 20:15:02: ===== Test Case 1.1: 'audio_enumerateAndNameAudioDevices' skipped
INFO:  07/07/14 20:15:02: ===== Test Case 1.2: 'audio_enumerateAndNameAudioDevicesNegativeTests' skipped
INFO:  07/07/14 20:15:02: ===== Test Case 1.3: 'audio_printAudioDrivers' skipped
INFO:  07/07/14 20:15:02: ===== Test Case 1.4: 'audio_printCurrentAudioDriver' skipped
INFO:  07/07/14 20:15:02: ===== Test Case 1.5: 'audio_buildAudioCVT' skipped
INFO:  07/07/14 20:15:02: ===== Test Case 1.6: 'audio_buildAudioCVTNegative' skipped
INFO:  07/07/14 20:15:02: ===== Test Case 1.7: 'audio_getAudioStatus' skipped
INFO:  07/07/14 20:15:02: ===== Test Case 1.8: 'audio_openCloseAndGetAudioStatus' skipped
INFO:  07/07/14 20:15:02: ===== Test Case 1.9: 'audio_lockUnlockOpenAudioDevice' skipped
INFO:  07/07/14 20:15:02: ===== Test Case 1.10: 'audio_convertAudio' skipped
INFO:  07/07/14 20:15:02: ===== Test Case 1.11: 'audio_openCloseAudioDeviceConnected' skipped
INFO:  07/07/14 20:15:02: ===== Test Case 1.12: 'audio_quitInitAudioSubSystem' skipped
INFO:  07/07/14 20:15:02: Force run of disabled test since test filter was set
INFO:  07/07/14 20:15:02: ----- Test Case 1.13: 'audio_initQuitAudio' started
INFO:  07/07/14 20:15:02: Test Description: 'Init and quit audio drivers directly.'
INFO:  07/07/14 20:15:02: Test Iteration 1: execKey 2135562624452638555
INFO:  07/07/14 20:15:02: Assert 'Call to SDL_InitSubSystem(SDL_INIT_AUDIO)': Pass
INFO:  07/07/14 20:15:02: Assert 'Check result from SDL_InitSubSystem(SDL_INIT_AUDIO)': Passed
INFO:  07/07/14 20:15:02: Assert 'Call to SDL_QuitSubSystem(SDL_INIT_AUDIO)': Pass
INFO:  07/07/14 20:15:02: Assert 'Call to SDL_GetNumAudioDrivers': Pass
INFO:  07/07/14 20:15:02: Assert 'Validate number of audio drivers; expected: >0 got: 4': Passed
INFO:  07/07/14 20:15:02: Assert 'Call to SDL_GetAudioDriver(0)': Pass
INFO:  07/07/14 20:15:02: Assert 'Audio driver name is not NULL': Passed
INFO:  07/07/14 20:15:02: Assert 'Audio driver name is not empty; got: directsound': Passed
INFO:  07/07/14 20:15:02: Assert 'Call to SDL_AudioInit('directsound')': Pass
INFO:  07/07/14 20:15:02: Assert 'Validate result value; expected: 0 got: 0': Passed
INFO:  07/07/14 20:15:02: Assert 'Call to SDL_AudioQuit()': Pass
INFO:  07/07/14 20:15:02: Assert 'Call to SDL_GetAudioDriver(1)': Pass
INFO:  07/07/14 20:15:02: Assert 'Audio driver name is not NULL': Passed
INFO:  07/07/14 20:15:02: Assert 'Audio driver name is not empty; got: winmm': Passed
INFO:  07/07/14 20:15:02: Assert 'Call to SDL_AudioInit('winmm')': Pass
INFO:  07/07/14 20:15:02: Assert 'Validate result value; expected: 0 got: 0': Passed
INFO:  07/07/14 20:15:02: Assert 'Call to SDL_AudioQuit()': Pass
INFO:  07/07/14 20:15:02: Assert 'Call to SDL_GetAudioDriver(2)': Pass
INFO:  07/07/14 20:15:02: Assert 'Audio driver name is not NULL': Passed
INFO:  07/07/14 20:15:02: Assert 'Audio driver name is not empty; got: disk': Passed
INFO:  07/07/14 20:15:02: Assert 'Call to SDL_AudioInit('disk')': Pass
INFO:  07/07/14 20:15:02: Assert 'Validate result value; expected: 0 got: 0': Passed
INFO:  07/07/14 20:15:02: Assert 'Call to SDL_AudioQuit()': Pass
INFO:  07/07/14 20:15:02: Assert 'Call to SDL_GetAudioDriver(3)': Pass
INFO:  07/07/14 20:15:02: Assert 'Audio driver name is not NULL': Passed
INFO:  07/07/14 20:15:02: Assert 'Audio driver name is not empty; got: dummy': Passed
INFO:  07/07/14 20:15:02: Assert 'Call to SDL_AudioInit('dummy')': Pass
INFO:  07/07/14 20:15:02: Assert 'Validate result value; expected: 0 got: 0': Passed
INFO:  07/07/14 20:15:02: Assert 'Call to SDL_AudioQuit()': Pass
INFO:  07/07/14 20:15:02: Assert 'Call to SDL_AudioInit(NULL)': Pass
INFO:  07/07/14 20:15:02: Assert 'Validate result value; expected: 0 got: 0': Passed
INFO:  07/07/14 20:15:02: Assert 'Call to SDL_AudioQuit()': Pass
INFO:  07/07/14 20:15:02: Assert 'Call to SDL_InitSubSystem(SDL_INIT_AUDIO)': Pass
INFO:  07/07/14 20:15:02: Assert 'Check result from SDL_InitSubSystem(SDL_INIT_AUDIO)': Passed
INFO:  07/07/14 20:15:02: Assert Summary: Total=34 Passed=34 Failed=0
INFO:  07/07/14 20:15:02: Total Test runtime: 0.0 sec
INFO:  07/07/14 20:15:02: >>> Test 'audio_initQuitAudio': Passed
INFO:  07/07/14 20:15:02: ===== Test Case 1.14: 'audio_initOpenCloseQuitAudio' skipped
INFO:  07/07/14 20:15:02: Total Suite runtime: 0.0 sec
INFO:  07/07/14 20:15:02: Suite Summary: Total=1 Passed=1 Failed=0 Skipped=0
INFO:  07/07/14 20:15:02: >>> Suite 'Audio': Passed
INFO:  07/07/14 20:15:02: Total Run runtime: 0.0 sec
INFO:  07/07/14 20:15:02: Run Summary: Total=1 Passed=1 Failed=0 Skipped=0
INFO:  07/07/14 20:15:02: >>> Run /w seed '3LKSXORVDKCBUK75': Passed
Comment 4 Sam Lantinga 2014-07-08 04:37:39 UTC 
Thanks for confirming, Andreas.