[
https://issues.apache.org/jira/browse/LOG4J2-2575?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
Carter Kozak reopened LOG4J2-2575:
----------------------------------
Assignee: Carter Kozak
> CronExpression.getTimeBefore() returns incorrect result
> -------------------------------------------------------
>
> Key: LOG4J2-2575
> URL: https://issues.apache.org/jira/browse/LOG4J2-2575
> Project: Log4j 2
> Issue Type: Bug
> Components: Core
> Affects Versions: 2.11.2
> Reporter: Nathan Friess
> Assignee: Carter Kozak
> Priority: Critical
> Fix For: 2.12.1
>
> Attachments: cronexpression.patch, cronexpressiontest.patch
>
>
> The method getTimeBefore(targetDate) of
> org.apache.logging.log4j.core.util.CronExpression incorrectly returns the
> targetDate argument instead of one time increment before targetDate when
> running in OpenJDK 1.8.0_191.
> When CronTiggeringPolicy calls this method the targetDate object contains a
> time with milliseconds, but prevFireTime is always computed with zero
> milliseconds which causes the comparison at line 1586:
> {code:java}
> ... while (prevFireTime.compareTo(targetDate) >= 0);{code}
> to be false on the first time through the loop and so targetDate (with zero
> milliseconds) is returned.
>
> Configuration example:
> {code:java}
> <RollingFile name="my_log" fileName="/var/log/my.log"
> filePattern="/var/log/my.log.%d{yyyy-MM-dd}">
> <PatternLayout pattern="[%d{EEE MMM dd HH:mm:ss yyyy}] [%p] %x [%X{user}]
> %m{nolookups}%n" />
> <CronTriggeringPolicy schedule="0 0 0 * * ?" evaluateOnStartup="true" />
> </RollingFile>{code}
>
> Adding some extra debugging, we can see the behaviour:
>
> {noformat}
> DEBUG StatusLogger getTimeBefore(targetDate=2019-03-22T00:00:00.002-0600
> DEBUG StatusLogger in loop preFireTime=2019-03-22T00:00:00.000-0600
> DEBUG StatusLogger return 2019-03-22T00:00:00.000-0600{noformat}
> Sometimes the file rolling will run at exactly zero milliseconds and
> getTimeBefore() will correctly returns the previous day in the above example,
> masking this issue during those runs. I marked this issue as critical
> because in combination with FileRenameAction's procedure of replacing
> existing files, log files will be lost at seemingly random rolls when this
> method returns the current date on one run, but then reverts back to the
> previous date on the next run, causing a filename collision.
> A patch is attached. This issue may be related to LOG4J2-2246 since that bug
> mentions issues with milliseconds in Date objects as well, but I have not
> investigated that part of the file rolling code.
>
>
>
--
This message was sent by Atlassian Jira
(v8.3.4#803005)
