AudioNodeOutput->bus(r) returns m_inPlacebus instead of m_internalBus

[dug9]

When working with 6 channel source audioBus, SampledAudioNode goes through a bit of code to prepare itself to receive all the channels of the input:
bool mess = false; if (dstChannelCount != srcChannelCount) { mess = true; output(0)->setNumberOfChannels(r, srcChannelCount); dstChannelCount = srcChannelCount; dstBus = output(0)->bus(r); } if (mess) printf("1 dstBus nchan %d\n", dstBus->numberOfChannels());
AudioNodeOutput->setNumberOfChannels works great, resetting m_internalBus to a fresh bus with 6 channels in my case.
Then the problem: output(0)->bus(r) returns something with 2 channels, and code below -confused- bombs.
AudioBus * AudioNodeOutput::bus(ContextRenderLock & r) const { // only legal during rendering because an in-place bus might have been supplied to pull ASSERT(r.context()); //return m_inPlaceBus ? m_inPlaceBus : m_internalBus.get(); // HAS BEEN RETURNING INPLACEBUS WITH M_CHAN 2 return m_internalBus.get(); }
When I modify the function to return m_internalBus and ignor m_inPlaceBus, then my scene runs normally, sound out, and with channelSplitter, channelMerger I'm able to have fun swapping channels around.
Q1. do I need m_internalBus or can I keep it commented out and live happily ever after?
Q2. if I need it, is there something I can do from my application code to clear it before associating AudioBus with SampledAudioNode?
Thanks.

[meshula]

Thanks for the notes.

In the case of a SampledAudioNode, the internalBus will always be what you want. The inPlace is an optimization for pass through nodes (ie. a gain node set to 1.0 could in theory just pass on a pointer to an upstream bus and thus save copying a buffer).

Your patch should work for now, but I will review the logic around the in place bus. It is a remnant from the original WebKit implementation, and might be best removed at this point.