It turns out pa_mainloop_prepare was treating the msec timeout as usec in PulseAudio versions before 13.0, so the value actually passed to ppoll() was very short indeed. This is fixed in 13.0 and the result was to make it quite hard for the suspend/resume thread to acquire the loop mutex, which was almost always contended. Fix this by adding a couple of yields, and lowering the timeout drastically, since it obviously worked fine with a much lower effective value anyway.

Use override

Report failure to resample, rather than allow the exception to be propagated - there is no useful way for anyone else to catch it from here

Log some sample-rate setup info

These are too much, even in debug mode

Update DynamicJACK for more recent API use

Avoid theoretical buffer overflow

Don't create record/playback stream at all if only the other one is requested; fix crash on dtor of driver that failed to start

Comments

A bit more logging for PortAudio driver

Add necessary file

Logging from audio I/O

One more bit of debug output

Constrain channel count to device limits (necessary when recording from mono mic)

Return error string on startup if specific (or only) implementation failed

Docs

Windows build fixes

Update to current bqresample API

Minor comment tweaks

Reorder getSourceSamples and putSamples args and add channel count. This makes a more consistent API and removes the possibility of a race condition in checking channel count in an application callback source whose system target may change.