Hi Julien,
I discussed the timestamp idea with Andi Kleen, and, basically, it's
of no use because the precision of the timestamp would suck.
The kernel collects MCE data every 5 minutes by default, then mcelog
runs every 5 minutes, collecting the data collected by the kernel.
[..]
Andi strongly advised against adding any timestamp whatsoever because
it really can't be precise enough to be even remotely useful.
Uhm, if I've got a logfile full of MCE entries after (e.g.) returning
from a weeks vacation a timestamp with a granularity of 5 minutes is
better than none, right? I can understand that Andi doesn't want to
implement this for the given reasons but the cron hackaround is useful,
will help people and should be robust enough (the only problem I could
see would be exceeding the storage size of the variables of the shell
when storing the log messages, but this is highly unlikely). Adding a
hint that the timestamps are only estimates (either in the log or the
README) or letting the user choose if he wants "inaccurate" timestamps
should mitigate any user complaints (if there should be any. I pretty
much doubt it).
So, closing this bug as a feature request which unfortunately can't be
implemented.
I'd opt for reconsidering that, cutting usability because the gathered
information isn't 100% accurate isn't good practice.
best regards,
Michael
--
To UNSUBSCRIBE, email to [EMAIL PROTECTED]
with a subject of "unsubscribe". Trouble? Contact [EMAIL PROTECTED]