[systemd-devel] journal file corruption

[Allin Cottrell]

I'm running systemd-200, in journal-only mode (i.e. no 
rsyslog). My journald.conf is all "factory defaults" apart 
from setting SystemMaxUse=16M (and this limit is not yet met). 
My /var/log/journal/<id> contains 12 files (dated from March 
31 to the present); and of these journalctl --verify tells me 
6 contain "Invalid object contents".

I put some debugging statements into journal-verify.c and the 
problem is the same in all 6 cases: the test that fails (in
journal_file_object_verify()) is

if (le64toh(o->data.entry_offset) <= 0)

And the status of the object in question is

o->data.entry_offset = 0
o->object.flags = 0
o->object.type = 1

Any idea what's happening here?

(Additional info: The machine in question is a laptop, 
rebooted very day. The system is "pure systemd" in the sense 
of having no init scripts other than systemd service files.
halt and reboot are symlinks to systemctl. All shutdowns
over this period have been clean, so far as I can tell.)

-- 
Allin Cottrell
Department of Economics
Wake Forest University, NC