Excerpts from James Hunt's message of 2013-03-20 10:57:23 UTC:
> I can see this would be useful, but since the data the logger receives
> is byte-oriented, adding timestamp data certainly wouldn't look very
> pretty. It would also change the semantics of the stanza and break all
> the log tests since they currently assume the log contains only what is
> output from the jobs.
>
> Rather than adding timestamps, my strong preference would be to support
> 'console syslog' (bug 885202) such that timestamps or time offsets are
> added "for free".
>
Except they're lost in early boot, when they might do a lot of good in
finding timing problems.
The tests can be fixed. Reading and looking for newlines to know when
to add a timestamp is not hard. This seems trivial to do as long as we
have easy access to the system clock, which I'm not entirely sure we do.
Excerpts from James Hunt's message of 2013-03-20 10:57:23 UTC:
> I can see this would be useful, but since the data the logger receives
> is byte-oriented, adding timestamp data certainly wouldn't look very
> pretty. It would also change the semantics of the stanza and break all
> the log tests since they currently assume the log contains only what is
> output from the jobs.
>
> Rather than adding timestamps, my strong preference would be to support
> 'console syslog' (bug 885202) such that timestamps or time offsets are
> added "for free".
>
Except they're lost in early boot, when they might do a lot of good in
finding timing problems.
The tests can be fixed. Reading and looking for newlines to know when
to add a timestamp is not hard. This seems trivial to do as long as we
have easy access to the system clock, which I'm not entirely sure we do.