This is important for fixed sized processing.

This is the new default. The size of the buffer will be based on the
periodSizeInFrames and periodSizeInMilliseconds config variable. To go
back to the previous system, where the data callback could specify any
number of frames, set the noFixedSizedCallback config variable to true.

This flag is used to tell miniaudio that it should assume the output of
the node is silence. This is useful for special nodes that want to
output to a file and don't want the output of the node to contribute to
the final mix.

Public issue https://github.com/mackron/miniaudio/issues/388

Public issue https://github.com/mackron/miniaudio/issues/386

Public issue https://github.com/mackron/miniaudio/issues/385

Public issue https://github.com/mackron/miniaudio/issues/385

Public issue https://github.com/mackron/miniaudio/issues/385

This was due to context-level mainloop being accessed from different
audio threads at the same time. I've fixed this by giving each device
their own pa_mainloop and pa_context objects.

I considered the idea of having only a single context-level mainloop
and just using a mutex for mutal exclusion, but that would involve a
lock in the audio thread's data loop which I wasn't happy about. Also,
I wasn't sure which thread PulseAudio callbacks would get fired from
since the main loop would be iterated on different audio threads.

Public issue https://github.com/mackron/miniaudio/issues/376

Public issue https://github.com/mackron/miniaudio/issues/334

This replaces the stop callback. The new callback supports different
event types, not all of which are supported on all backends.

This commit also fixes a bug where the stop callback is not fired.

Public issue https://github.com/mackron/miniaudio/issues/351

Public issue https://github.com/mackron/miniaudio/issues/384

  * Add support for setting the directional attenuation factor. This is
    useful for allowing other effects to control the directional
    attenuation, such as HRTF effects.

  * Add a function for retrieving the index of the listener that a
    sound will be spatialized against based on the current state. A
    function to retrieve the direction to that listener has also been
    added. This is useful for when an effect needs to know information
    about the listener for some kind of spatialization.

  * A bug has been fixed where silence is output if a node has zero
    inputs.

  * A bug has been fixed where a bad channel map can be used for
    spatialization.

This option is intentionally undocumented and should be used at your
own risk.