This has been maddening - I’m guessing I’m doing something stupid but whatever it is, it’s eluding me.
I start with what I suspect is a canonical deployment scenario: rpi 3, speakers on 3.5mm input rather than hdmi and snoball mic. During initial configuration, I have combo keyboard and trackpad connected on usb and a monitor on hdmi. Run through setup without issue, configure some skills and all is well. Specifically I can talk to the device, be heard, and hear the outcome of mycroft handling my directives. Then to prepare for moving the device to its intended location, I sudo shutdown, then disconnect the USB keyboard/trackpad and hdmi, move it to its new location and power it up. And it neither hears me nor will speak aloud. ssh to the device and I can do speaker and mic tests, which work successfully. In the microft cli I can type in directives and see the expected responses, but no audio is played on the speakers. The visual indicator for the mic in the cli shows either an invariant value or nothing at all.
I reproduced that scenario a few times on the stable 7-20 image and yesterday on the 9-26 one. Back when I was working with the 7-20 image, I thought that perhaps the removal of the USB keyboard/trackpad might be confusing mycroft about the addressing of the mic and so worked out a way to configure picroft without ever having either hdmi or keyboard/trackpad connected: install image and connect to laptop via direct ethernet connection; the pi will auto assign itself a link local address (169.254.*); work out what that address is, then ssh to it and run setup. When I did that, I would get through setup without issue - in particular the mic and speaker tests were successful - but once setup was complete I would end up in the same state as above: mycroft would neither receive audio input nor emit any audio output but would work in the cli in responding to typed directives.
I have a device running the 9-26 image which I will keep in this state to facilitate any desired log capture or debugging analysis, but given that this is very much an expected deployment scenario (the first one I described, not the ethernet/link local address one), I wonder if there’s a test description of it that I could review to see if there’s anything I’m doing differently?