Featured Articles
Announcing Pi
Technical Specifications
Raspberry Pi FAQs
Win a Pi
Raspberry Pi Wishlist
Actions
Forum Thread Details
  • Replies 19 replies
  • Subscribers 669 subscribers
  • Views 2191 views
  • Users 0 members are here
Related

Is it just me, or do you end up installing raspbian over and over again on a complex project?

easyejl

I don't even think this project is all that complex. A raspberry pi 3b+ with a Waveshare WM8960 sound card, a generic 16 channel servo card, running google assistant with hotword recognition, running servos to move a puppet (and have the mouth move more or less in time with the speech).

 

I keep running into audio issues, where installing one thing or another ends up messing up the asound or pulseaudio config and then I lose output to the speakers. I got that card to simplify things as it has a tiny onboard amp and speakers, as well as built in microphone. But i'm now up to having reinstalled raspbian at least 8 or 10 times trying to get all the components to work together. I know I could scrap that card for this, add a usb microphone, and powered speakers but I don't have a usb mic handy, and I wanted this to all fit inside a moose head puppet. If I need another power supply on top of the pi + servo board (to drive the speakers) it seems like its gone a little too far image

  • Sometimes, a system configuration management tool like these could be useful (or they could be overkill). I think SBC projects are often right at the edge of the need for these things. We could teach each other how to make new packages for Raspbian so that installation is a lot easier after we finish our adventures with the RPi. I was going to offer a how-to blog entry on this. However, my link is a page from the Debian Administrator's Handbook. image This is a wheel we needn't recreate!

  • For everyone's laughs, i'll share this weekend. I installed fresh buster on the pi, set up samba, set up a windows share. Got sound recording with usb mic + playback via headphone jack + powered speakers, and used rsync to back it up. Installed pulse audio, configured it, backed it up. Then I installed only the deprecated google assistant and got it working fairly easily. so I used rsync to back it up again. Installed the drivers for my sound card hat, only had to play around a little bit to get it all working, backed it up.

     

    So I guess the plan with taking intermediate backups had me being just cautious enough with everything else that it all more or less worked the first time. I spent more time waiting on rsync than actually installing or modifying configs image

  • in reply to easyejl

    Glad to hear you had some success.

     

    Sometime between when you did your first install (not working) and now (working) some repository may have changed. This can be frustrating unless you develop the understanding things changing for the better is not such a bad thing:)

  • in reply to colporteur

    most of these were older, the google samples I had to go backwards through tags on their github to find the one I wanted. I think maybe my subconscious got a better understanding of asound.conf and how pulseaudio works.

     

    Also this time around I realized that if you have the google assistant running and exit out "poorly" (still not sure the exact scenario) it locks/kills the pipe to output to that sound device. Starting it again does not fix that. So I imagine that may have been what happened to me in the past, when I thought it wasn't working.

  • in reply to easyejl

    Yes a good showing of poorly written software given out as "examples". But people will take and reuse in discriminatory in production and wonder why not working as they imagined. Such is the state of software in the wild. I think found out with Google!

  • in reply to clem57

    I've seen it so many times over the years, but somehow I think it gets worse as we go. The more open source software there is, the more libraries people use for small sections of them (like with typescript using Moment just for formatting the date for screen) and things end up with enormous dependency grids. Not everything stays self contained, so you end up with accidentally breaking one piece by installing something else. At work we've just gone through a pain in the neck over a new release failing to build at first, then failing to have everything work correctly primarily because a reference was added to an outside service that cause a chain of dependency updates.

     

    ok part of it also is the way our build/release process is for that particular product image

  • in reply to clem57

    and with the google assistant in particular, as they increased versions they at one point took away the hotword capability being built in, and you had to use an external speech recognition package to capture the "Hey Google" part and then turn it over to the assistant. I have no idea why they'd purposefully hamstring their SDK that way, but they did.

  • Former Member
    Former Member in reply to clem57

    Amen,    As a rule, people should supply example code that is was successfully executable - from simple to sophisticated.  How else did they test before project/product release?

     

    My other pet peeve (especially on stackoverflow) are suggestions to try something when the one making the suggestion hasn't done so themselves.  You know that is true because the suggestion fails to work! image

  • in reply to Former Member

    Also provide a complete version specific list of dependencies. That way it's not a run to realize it was done with libpython-dev 3.4 and you installed with python 3.7 so don't have it