FlexASIO never waits for ASIOOutputReady()

[dechamps]

While reviewing some FlexASIO code I stumbled upon this:

FlexASIO/src/flexasio/FlexASIO/flexasio.cpp

Lines 998 to 1003 in 57e14e8

	std::unique_lock outputReadyLock(outputReadyMutex);
	if (!outputReady) {
	if (IsLoggingEnabled()) Log() << "Waiting for the ASIO Host Application to signal OutputReady";
	outputReadyCondition.wait(outputReadyLock, [&] { return outputReady; });
	outputReady = false;
	}

I suspect this is subtly wrong. With this code outputReady only gets reset if it happened to get set after we started waiting for the condition. If outputReady was already true before we entered this block, then it will stay true forever. That doesn't seem right. The reset should happen unconditionally.

The consequence of outputReady being stuck at true is that ASIO output buffers will be transferred to PortAudio before the ASIO host application is done writing them, leading to corrupted audio output.

One reason why this went unnoticed until now is probably because this only has consequences in ASIO Host Applications that support OutputReady and return from bufferSwitch before they call it. I suspect such applications are rare.

[dechamps]

This is fixed in FlexASIO 1.10.

[dechamps]

In retrospect, given the investigation in #235, it turns out this bug was worse than I originally thought: indeed the initial value of outputReady is true, which means we would never wait for outputReady, period. The wait code was basically dead until the fix. Thankfully, it looks most ASIO applications call outputReady before returning before bufferSwitch anyway, which means this bug would have no consequences for them.