Diemo Schwarz

Hi hackers, I get Max7 to livelock when I delete an object with which my dsp object is registered.

Max hangs in the main thread doing object_free -> z_dsp_freewrapper -> dspchain_setbroken,

with my code being blocked in dsp_perform at a systread_mutex_lock called by object_detach.

Is this a bug, or am I not supposed to detach within perform? If so, I'd rather work around the deadlock by checking the systhread_mutex_lock, but which one is it?

Hi hackers, I get Max7 to livelock when I delete an object with which my dsp object is registered.
Max hangs in the main thread doing object_free -> z_dsp_freewrapper -> dspchain_setbroken,
with my code being blocked in dsp_perform at a systread_mutex_lock called by object_detach.
I can't find out who's holding the lock.

am-i-allowed-to-object_detach-in-dsp-perform-livelock

You should defer any object creation/deletion/attachment/detachment to the main thread.

Thanks, it works deferred! That would be worth a mention in the sdk, wouldn't it?

And while I have your attention, why isn't there an ATOMIC_COMPARE_SWAPxxx wrapper for pointers, in ext_atomic.h, i.e. OSAtomicCompareAndSwapPtrBarrier on Mac?

Thanks, it works deferred!  That would be worth a mention in the sdk, wouldn't it?

I don't have a definitive answer to your question, but presumably because there's no direct equivalent to OSAtomicCompareAndSwapPtrBarrier() on Windows. I'll see about updating the SDK wrt the previous question.

Am I allowed to object_detach in dsp perform? (livelock)