Assumes your Pi has access to Wifi, with internet access (during setup). (But all setup methods do currently.) USB networking still enabled for troubleshooting.

Configure the SD card before first boot of the Pi

Flash the image from here (10-24 dated image) using Etcher or similar. (Be sure to click the ... on Dropbox and download the .zip file.)
Mount the card again, and in the boot directory create a teslausb_setup_variables.conf file to export the same environment varibles normally needed for setup (including archive, Wifi, and push notifications (if desired).) A sample conf file is in this repo. I.e. file should contain at a minimum (replace with your own values):
```
export archiveserver=Nautilus
export sharename=SailfishCam
export shareuser=sailfish
export sharepassword=pa$$w0rd
export campercent=100
export SSID=your_ssid
export WIFIPASS=your_wifi_password
# export pushover_enabled=false
# export pushover_user_key=user_key
# export pushover_app_key=app_key
```
(Pushover should be working but commented out by default.)

Boot it in your Pi, give it a bit, watching for a series of flashes (2, 3, 4, 5, maybe 6) and then a reboot and/or the CAM to become available on your PC/Mac.
The Pi should be available at teslausb.local over Wifi (if it works) or USB networking (if it doesn't). Takes about 5 minutes for me. You should see in /boot the TESLAUSB_SETUP_FINISHED and WIFI_ENABLED files as markers of success too.
Currently doesn't create the TeslaCam folder, so you'll need to do that before taking to your car.
If plugged into just a power source, or your car, give it a few minutes until the LED starts pulsing steadily which means the archive loop is running and you're good to go.

What happens under the covers

When the Pi boots the first time:

A /boot/teslausb-headless-setup.log file will be created and stages logged. This takes the place of the "STOP" commands
Marker files will be created in boot like TESLA_USB_SETUP_STARTED and TESLA_USB_SETUP_FINISHED to track progress.
(Working on a progress system so the script can pick back up if needed. This is probably useful for the general/old way of setup too.)
Wifi is detected by looking for /boot/WIFI_ENABLED and if not, creates the wpa_supplicant.conf file in place and reboots.
The Pi LED will flash patterns (2, 3, 4, 5, maybe 6) as it gets to each stage (labeled in the setup-teslausb-headless script).
- 10 flashes means setup failed!
- After the final stage and reboot the LED will go back to normal. Remember, the step to remount the filesystem takes a few minutes.

At this point the next boot should start the Dashcam/music drives like normal. If you're watching the LED it will start flashing every 1 second, which is the archive loop running.

NOTE: Don't delete the TESLAUSB_SETUP_FINISHED or WIFI_ENABLED files. This is how the system knows setup is complete.

Image builder source and patches

For now the image creation work is at:

Modified pi-gen rtgoodwin's fork of pi-gen in (whatever current branch I'm working at the time).
headless-patch branch of rtgoodwin fork https://github.com/rtgoodwin/teslausb/tree/headless-patch/headless-scripts

Image creation TODOs

Patch the hostname to teslausb
Make it so if someone deletes the TESLAUSB_SETUP_FINISHED file it's handled gracefully.
I still see some errors during pi-gen about locale, may need to be fixed? stage0/01-locale/debconf en_US.UTF-8
Cache the remount packages? Might mess with first boot like rsyslog
Any other steps to move into the base image?
Aspirational TODO: Remove more packages and set services to stopped to make the boot process faster?
NOTE: I moved all script downloads and variable creation to the initial setup. At this point, I'm designing it to pull the setup scripts dynamically, since development is still ongoing. If/when we reach a good frozen state, we can generate an image that is ready to run. I think it'll also be pretty tricky to do some of the remounting and creating the backing files etc. on the image creation side. Open to suggestions/contributions there though! At the very least we could bake in stable first stage headlessBuild scripts for Mac/Linux/Windows.

Modifications to pi-gen builder from master

Built image on a Raspi running Stretch, for maximum Pi-ception.

Add SKIP and SKIP_IMAGES files to stage3, 4, and 5 (if present).
Add a stage6. (There are stages0-5, but may be a stage5 in some cases. This will help keep a clean merge later.)
Copy the prerun.sh from stage2. Be SURE to mark chmod +x it.
Remove or rename the EXPORT_NOOBS files in all stages. We don't need a NOOBS image built.
In stage6, create a 00-tweaks folder, with a 00-patches folder and patch inside to patch cmdline.txt to remove the resize and add the needed modules. The build process uses quilt for patching. Note: the path for any patching you do at this stage is stage6/rootfs/FILEPATH where rootfs represents the Pi's /. So, cmdline.txt is stage6/rootfs/boot/cmdline.txt.
Add a patch for the config.txt file.
Add a file called series in the patches directory with the name of each .diff file in the order you want them applied.
Add a files folder in stage6 with modified rc.local. The modified rc.local will handle pulling down the setup-teslausb-headless file the first time. (Still working on build logic here.) Files are moved into final locations in a 00-run.sh script and the install command. See the script for details.
(Yes at this point you could suggest that just putting the end state files in place instead of patching would be good, but why not be idiomatic? :) )
Add a script to flash LEDs
Run sudo ./build.sh from the pi-gen directory.
If you get a failure, it's almost certainly after stage2, so you can add SKIP files in stage2-stage5 present) and rerun sudo CLEAN=1 ./build.sh