I was able to set up 13.04 KDE (with compiz) and LXDE on my 32GB TF700. I used 0.8.0 as a base. It's possible to freely experiment with the rabits software stack provided the following packages are pinned:
Pinned packages (/var/lib/synaptic/preferences file). Here's
the list of installed packages (dpkg --get-selections) on my system.
Here are my impressions:
LXDE is very fast. KDE with compiz takes around one minute to boot, but after that its performance is acceptable (in fact, I use it as my main DE). KDE uses ~310MB of RAM after booting.
Brightness doesn't work automatically, though it's possible to control it via /sys/class/backlight/pwm-backlight/brightness. Max value is 255, min is 3. If you set brightness to 2 or lower then the screen turns off without a way to turn it back on. Only a reset fixes this.
The screen doesn't turn off on lid close. This is due to the touchscreen and the touchpad being active and firing touch events when the screen touches the dock. This
script can be used to fix the issue. It disables the offending input devices on lid close and re-enables on lid open. Don't close the lid quickly if you use this script! Not only script doesn't work properly in this case due some delays in lid-close event processing, it also exposes a bug somewhere in the xserver stack which leads to a xserver crash quite often. Just don't close the lid quickly and you'll be fine. My "procedure" is as follows: close the lid so that there's ~1cm between the tablet and the dock, close lid slowly until the screen turns off, wait ~1 second after that and finally close the lid full.
I occasionally experience severlal second long lock-ups, especially during heavy IO load. Does anyone experience this too? Maybe there's a workaround?
I have one a bit unrelated question: what's the rationale for the UKSM patch being applied? The samepage merging is only useful when there many duplicate pages actually exist, which almost universally is when one runs several identical VMs on the same host. This is not the case here. I think it's not worth to waste precious memory and CPU bandwidth for little benefit - at least on my system there was only ~20-30MB savings when using all 1GB ram. A better solution would be to simply swap out unused memory. The main mmc memory is comparatively fast - ~15MB seq read/write, 4MB/s 4k read, 3MB/s 4k write - this should work reasonably well.
I get the feeling that the initrd scripts should fsck the rootfs image being loaded. Every time the system locks up or reboots (which still does happen), there's a certain risk of filesystem corruption. If occasional errors are not fixed, there's huge risk for data loss, or mysteriuos bugs. I've spent at least an hour debugging why I can't recreate /etc/mtab file - the culprit was filesystem corruption during an earlier reboot.