On 01/19/2010 09:18 PM, peter green wrote: > A while back fsck got more bitchy (treated as an error rather than a > warning) about timestamps in the future, I filed a bug report about this > change in behaviour that breaks but the this seems to have been cut off -- what is the bug report number? Are you referring to http://bugs.debian.org/557636 ? > anyway just recently I decided I needed to fix the boot process of my > armel qemu system which was suffering from the issue so I hacked up the > following script and saved it as /etc/rcS.d/S06ntpdate.sh > > #!/bin/sh > #temporary tmpfs to let us bring up network with root readonly > mount -t tmpfs none /etc/network/run > ifup eth0 > #give the interface some time to come up > sleep 10 > /usr/sbin/ntpdate pool.ntp.org > ifdown eth0 > umount /etc/network/run Thanks, this doesn't just affect arm users; i'm now seeing it on old i386 hardware that has failed and/or unreachable BIOS Batteries. :( It's a real problem, i think, because you basically cannot reboot machines like that without running into the problem. And not all of these machines have net access at boot (e.g. a travelling laptop). Are we expected to set the system clock correctly by hand from the bootloader before we let it load the kernel? how frustrating. --dkg
Attachment:
signature.asc
Description: OpenPGP digital signature