[
https://issues.apache.org/jira/browse/LOG4J2-982?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14375190#comment-14375190
]
Mikhail Mazurskiy commented on LOG4J2-982:
------------------------------------------
It is wall clock time, right? If you adjust time on your system, it will jump.
If something like ntpd is running, it can make it slew. Also think about
virtualization - the drift is much worse there unless you have ntpd running.
Also it looks like you have answered you own question (the link) =)
p.s. as for the bug in JDK - the program, that is doing the measurement, is
wrong. Take a look at this blogpost
http://shipilev.net/blog/2014/nanotrusting-nanotime/
p.p.s. Also some code (e.g. in tests and benchmarks) is already using nanoTime
to do measurements. I just wanted to fix other places.
> Use System.nanoTime() to measure time intervals
> -----------------------------------------------
>
> Key: LOG4J2-982
> URL: https://issues.apache.org/jira/browse/LOG4J2-982
> Project: Log4j 2
> Issue Type: Improvement
> Components: Core, Flume Appender
> Reporter: Mikhail Mazurskiy
> Priority: Minor
> Attachments: LOG4J2-982.patch
>
>
> Unlike {{System.currentTimeMillis()}}, which can jump/slew forward and
> backwards, {{System.nanoTime()}} is a monotonic clock (at least it should be)
> and hence it should be used to measure time intervals (timeouts/delays/etc).
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
