Quote:
|
Originally Posted by BACbKA
this is a nice package, but I fail to see it as "unbrickable". Recently a script was shown doing a cd to a directory that didn't exist ( that failed ), followed by an rm -rf * . (which had / as the cwd). For that kind of mistakes, as well as some fs / master config files being overwritten, the irex-promised unbricking is needed.
|
You're right that it is still possible to hose the system to the point that it requires a return trip back to Eindhoven, but the chance of a bad /mnt/settings preventing the boot from completing is far lower. Running 'rm -rf /' on any Linux system is a bad idea and will set you back on any machine, especially if it takes out your /boot and /bin directories. Even Apple has messed up shell expansion in install scripts before -- they did a "rm -rf /Volumes/$DRIVE, where $DRIVE was a drive name selected from the gui. If the drive had a leading space in the name it would execute 'rm -rf /Volumes/', effectively wiping ALL the drives currently mounted.
But, from my perspective, baring any serious screwups this patch makes my unit far less likely to need to visit its home. I can rebuild my kdrive-Xfbdev to include the evdev patches without worry that I won't be able to get to my device if the X server doesn't start.
As for reflashing at home, the xscale PXA255 CPU has three serial ports, one of which almost always has a boot loader console in most implementations. I see in the dmesg output that the three of them are detected and initialized, but haven't pulled my device apart to find the ports are level shifted to RS232 or are still 3V TTL.