I’ve been following the directions in the L4T dev guide. I took the example rootfs, applied the binaries, customized it a bit more (added some drivers and other packages), and then created a disk image. I used the command:
I used that 8G because I wanted the resulting image to be as small as possible, as I understood that the initial setup would re-expand the card to use available space.
On first boot I’m presented with the mostly normal sequence, although I am asked something that doesn’t happen with the “stock” image - “Please enter the desired size of the APP partition in Megabytes”. I was trying a 32GB card, it said the max size 30513.
Then it goes through a few other setup steps and then sort of hangs. Mouse still works, I can click on the network and power icons (although the power off command doesn’t seem to work). Any way I can find out why it’s unhappy?
I got the “please enter app partition size” with the latest image I created, but it booted fine eventually. It seems the first boot resizing step has changed in the latest release if using the GUI.
I mastered the image using the same steps you described, although I didn’t install any new drivers.
The first, official, l4t image I used hung at unattended upgrades. You can try booting without a network cable and see if that fixes it for you. It did when I encountered the issue.
If all else fails, you could try booting with a serial console and see if you can complete first boot like that (no HDMI/DP connected). Instructions are in this forum’s faq thread, I believe.
I am unaware of any way to get the GUI installer to be more verbose (it would be nice) but someone from Nvidia might know. I don’t think virtual consoles are available before first boot completes, so that option is out.