Hi all,

I have come up with a second way of fixing the regression (the first one was a few commits before that):

https://github.com/HinTak/seeed-voicecard/commit/19067f3333d7759bdf626e62bc0...

As far as I understand it, the breakage came from the somewhat unsual design of the SoC device: it has both recording / ADC and a playback / DAC codecs. (a pair of ADC ones, actually) to do real-time echo cancellation / noise filtering applications. The recording codecs use the playback codec's clock; and the playback codec's trigger needs to be called, before the clock can start.

I am not entirely happy with the rest of the vendor's code in terms of what code is where of event sequences, but I'll leave that for now.

Would appreciate comments, etc.