[NiLuJe]

@geekmaster: Yeah, if I understood correctly, mmcblk0p4 is a FAT32 part, then there's loop/0 (mounted off mmcblk0p4 with an offset) that's mounted on /mnt/base-us, and then there's the fuse mount from base-us to us. I assume it's related to being able to export it over USB mass storage (AFAIU, because it has to be unmounted from the host OS to be exportable?), but I have no idea why they go to such lengths :?. (Is it still that tortuous on the Touch?).

@allowingtoo: Does that still work? (being able to trigger an update process @ boot?).

The reboot file is used *only* when you want to trigger a reboot (technically, a framework restart), if you didn't change anything, there's no reason to do that. Did you (change anything/try to trigger a reboot/actually triggered a reboot)?

And, no, installing basic stuff should not mess the system up like this. I assume every hack install ended succesfully? When exacty did stuff start to go wrong?

The only thing I can think of is something going wonky because of a full filesystem somewhere during an install process... Problem is we only really tinker with /tmp and the userstore, and /tmp is a tmpfs mount, so, not persistent, and not likely to wreak havoc like that, and the userstore should have plenty of free space, especially on a 'new' device, and (I assume) anything going wrong like that would have triggered an update process failure (most likely U006). There's very, very few changes made to the actual rootfs (an init script, a bunch of symlinks, and a file), and a few directories created on the /var/local partition. (I'm talking about the fonts hack here, because it's the biggest, most invasive/complex install script), so I can't see anything going wrong there (or at least, not *that* wrong ^^).