I've been playing a bit with a small netbook that has a WonderMedia WM8650 CPU (armv5tejl) with 256MB of RAM. This netbook came with android 2 by default, with kernel 2.6.32, and thanks to all the efforts put in this github and all the people involved in this mailing list, it's possible to have up to kernel 4.14 (some people are getting to kernel 6.1).

Now, I've been playing with this github, which is basically a script to build the kernel from linux-vtwm (and also does some things to bump the kernel to version 6.1), build a Debian 12 rootfs using multistrap, add some udev rules, systemd services and first-time config scripts, and pack everything to two compressed files that are ment to be placed onto an SD card for the netbook to boot.

Now, with everything I was able to learn from the previous githubs and mailing lists, I was able to get Debian 11 with Linux 4.14 (from linux-vtwm, with some modifications to dts to add pwm support) working, using the bookconfig repo with some modifications (no kernel/boot partition is built. udev rule and systemd services have some modifications to work with my netbook), just to make things easier. The thing is that everything works perfectly fine up until I want to add a GUI. For the moment, I want to install i3 for the wm, and of course Xorg as the X window server.

So far, I've tried two paths:
1. Install Xorg, i3 and nodm while creating the rootfs (using multistrap): This, results in a totally bootable system (only after the second boot. After finishing setup, I got a black screen, and I wasn't able to change to any tty. On the serial console I had "systemd-journald[7746]: Failed to send WATCHDOG=1 notification message: Connection refused". Unplugging the system was the only way to get out of there), with working GUI. and that's it. You cannot install new packages because the system has a hard time updating the package database. It always has something to complain about:
These are some of the possible error outputs. They vary from package name. I was able to work (some times) past that error, but installing any package that requires more than two or three dependencies, ends up in one of the previous errors.

2. Install a basic system (packages in .config file), and then, manually install (using apt) xorg, i3 and nodm: In this case, I wasn't able to complete the installation, because on multiple occasions, I got segmentation faults while installing one of the many dependencies of xorg. It was apparently caused by a package that got corrupt (while downloading, I think). Deleting the package on cache and running "sudo apt --fix-missing install" solved that issue, but then near 99% it failed again with another segmentation fault and then I wasn't able to fix the error* (except now, while writing this thread. It had a segmentation fault and some issues with dpkg but it installed succesfully and also i3 and nodm installed after fighting for a little bit, but again, issues with segmentation fault while running apt update occurred. log attached)

I'm kind of lost on what could the culprit be. It could be the kernel, the rootfs, the ram, or I don't know why. What makes fixing this (for me) is the inconsistency between every try. I've probably copied more than 50 rootfs's to my sdcard, and none of them shared the exact same error, all of them had a little difference, primarily on package names.

If any of the config files (kernel, boot, etc) are needed, tell me and I'll upload them.

*between segmentation faults, I always rebooted the netbook using "systemctl reboot". This is because sometimes the screen would get all the characters corrupted (even new ones) or it would cause unexpected behavior from all commands executed (once I wasn't able to execute systemctl reboot)

Attached Files

	option1 (2nd boot).txt (12.3 KB, 1 views)
	option2.txt (84.6 KB, 9 views)
	multistrap.conf.txt (320 Bytes, 2 views)

Hi,

seems like your have some bad files under /var/lib/apt/. I suggest deleting them.
However I don't know what the underlying problem was that caused them to become
corrupted.

Evo2.

Yeah, the issue is with the files under /bar/lib/apt, but as you say, there is no apparent reason for them to get bad.

I have to say, that I continued with my testing a little bit, and got some interesting results. You see, if you boot the kernel with the basic bootargs (them being: root=/dev/mmcblk0p2 rw rootwait console=ttyWMT0,115200) the kernel automatically grabs 128MB of all the available 256MB on hardware. There is no reason on why that happens (at least I wasn't able to find any reason), but, not using all the memory got some good things happening, primarily, no segfaults (for the most part). The issue then is that apt ends up running out of mem and the install fails for big packages.
Now, for all the previous tests, I also had "mem=256M" on the bootargs to use the full 256MB of ram, but it came to my mind the idea that maybe linux is counting from 0 to whatever number is mem set to, but by setting mem to 256M, I'm making linux count from 0 to 256MB, making linux use 257MB of RAM. I have to test that theory, but in mind it makes sense (coming from programming C for pic mucus) I haven't looked at any of the kernel code on how it counts memory when using the mem argument, but I figured that it would be easier to just try changing the value and see what happens