MIxxx 2.1 crash on startup

[Daniel Schürmann]

This was tested on Ubuntu Trusty 64 bit

Debug [Main]: BaseTrackPlayerImpl::slotLoadTrack "[Sampler59]"
[New Thread 0x7ffea27d4700 (LWP 16179)]
Debug [Main]: BaseTrackPlayerImpl::slotLoadTrack "[Sampler60]"

Program received signal SIGSEGV, Segmentation fault.
[Switching to Thread 0x7ffee17fa700 (LWP 16127)]
EngineMaster::processChannels (this=this@entry=0x14493c0, iBufferSize=8192)
    at src/engine/enginemaster.cpp:351
351 EngineChannel* pChannel = pChannelInfo->m_pChannel;
(gdb) bt
#0 EngineMaster::processChannels (this=this@entry=0x14493c0, iBufferSize=8192)
    at src/engine/enginemaster.cpp:351
#1 0x00000000007f0086 in EngineMaster::process (this=0x14493c0,
    iBufferSize=iBufferSize@entry=8192) at src/engine/enginemaster.cpp:397
#2 0x0000000000b2a4cb in SoundManager::onDeviceOutputCallback (
    this=<optimized out>, iFramesPerBuffer=iFramesPerBuffer@entry=4096)
    at src/soundio/soundmanager.cpp:597
#3 0x0000000000b20159 in SoundDevicePortAudio::callbackProcessClkRef (
    this=0x1e2bdb0, framesPerBuffer=4096, out=0x13770c70, in=<optimized out>,
    timeInfo=<optimized out>, statusFlags=<optimized out>)
    at src/soundio/sounddeviceportaudio.cpp:941
#4 0x00007ffff7675e63 in ?? ()
   from /usr/lib/x86_64-linux-gnu/libportaudio.so.2
#5 0x00007ffff7677f0b in ?? ()
   from /usr/lib/x86_64-linux-gnu/libportaudio.so.2
#6 0x00007ffff768066e in ?? ()
   from /usr/lib/x86_64-linux-gnu/libportaudio.so.2
#7 0x00007ffff2a19184 in start_thread (arg=0x7ffee17fa700)
    at pthread_create.c:312
#8 0x00007ffff0aba03d in clone ()
    at ../sysdeps/unix/sysv/linux/x86_64/clone.S:111
(gdb) A

[Daniel Schürmann (daschuer)]

(gdb) print pChannelInfo
$1 = (EngineMaster::ChannelInfo *) 0x0

i is 4

[Be (be.ing)]

Maybe there is something going wrong with the initialization of samplers?

[Be (be.ing)]

Have you been able to reproduce this?

[Daniel Schürmann (daschuer)]

No, I will have a look tonight.

[Daniel Schürmann (daschuer)]

Yes, this looks like a race condition around m_activeChannels. We need to skip processing if we change anything with the channels.

[Be (be.ing)]

I think I see what you mean. BaseTrackPlayerImpl::BaseTrackPlayerImpl calls EngineMaster::addChannel. PlayerManager::addSamplerInner creates a new Sampler instance, which is a subclass of BaseTrackPlayerImpl. PlayerManager::slotNumSamplersControlChanged calls PlayerManager::addSamplerInner when the [Master], num_samplers ControlObject changes. How can we work around this?

[Be (be.ing)]

Augh, the lack of comment editing on Launchpad is annoying. PlayerManager::slotNumSamplersControlChanged is called from the main thread, so EngineMaster's m_activeChannels member variable may be accessed by both threads at once.

Have you figured out how to reproduce this reliably?

[Daniel Schürmann (daschuer)]

No, I have not managed to reproduce it at all.
The backtrace was a lucky catch, while debugging an other issue.

[Daniel Schürmann (daschuer)]

https://github.com/mixxxdj/mixxx/pull/1564

[Swiftb0y (swiftb0y)]

Mixxx now uses GitHub for bug tracking. This bug has been migrated to:
https://github.com/mixxxdj/mixxx/issues/9188