如何在没有堆栈跟踪的情况下创建/抛出异常？ [英] How can an Exception be created/thrown with no stack trace?

问题描述

今天我正在浏览一些日志并遇到一个奇怪的错误。

I was trawling through some logs today and came across a strange error.

以下是它在日志中的显示方式：

Here's how it appears in the log:

2014/09/11 15:23:52.801 [CC3A5FDD16035540B87F1B8C5E806588:<removed>] WARN a.b.c.Ddd - Main failure 
java.lang.NullPointerException: null
2014/09/11 15:23:52.801 [CC3A5FDD16035540B87F1B8C5E806588:<removed>] ...

以下是代码的样子：

} catch (Exception e) {
    Ddd.log.warn("Main failure ", e);
    throw e;
}

如果重要的话，代码在jsp中。在日志中再次重复相同的异常（正如您对 throw e 所期望的那样）。

The code is in a jsp if that is important. The same exception is repeated once more in the log (as you'd expect from the throw e).

我没有原因记录 - 日志中的前一行显示了查询的执行情况。这种情况在4天内仅发生两次，似乎没有对系统造成任何损害。

I have no record of what the cause was - the previous line in the log shows execution of a query. This occurred just twice over a 4-day period and seems not to have caused any harm to the system.

环境：使用Java 5在Tomcat下运行相当繁忙的Web服务。

Environment: Fairly busy web service running under Tomcat with Java 5.

我不是在询问有关调试系统的提示 - 这些错误很快就会消失，甚至可能再也不会发生。关于如何在没有堆栈跟踪的情况下创建任何异常（尤其是NPE），我感到很难过？

I am not asking for tips on debugging the system - these errors are long-gone and may even never happen again. I was just stumped as to how any exception (especially an NPE) could be created without a stack trace?

已添加

正在使用的记录器是 slf4j 驱动 Logback 实例。我相信警告方法是此处。不确定解析为的 Logback 方法，但我相信 Throwable 参数是专门处理的，如果有堆栈的话附加到 Throwable 的跟踪将显示在日志中。

The logger being used is an slf4j driven Logback instance. I believe the warn method is here. Not sure what Logback method that resolves to but I am confident the Throwable parameter is treated specially and if there was an stack trace attached to the Throwable it would appear in the log.

LogBack.xml - 按要求：

LogBack.xml - as requested:

<?xml version="1.0" encoding="UTF-8"?>
<configuration>
  <property name="log.package" value="Package" />
  <property name="log.pattern" value="%date{yyyy/MM/dd HH:mm:ss.SSS} [%X{session}:%X{device}] %level %logger{25} - %msg%n"/> 
  <property name="log.consolePattern" value="%highlight(%-5level) %white(%logger{25}) - %msg%n"/> 
  <if condition='isDefined("catalina.home")'>
    <then>
      <property name="log.dir" value="${catalina.home}/logs" />
    </then>
    <else>
      <property name="log.dir" value="." />
    </else>
  </if>

  <appender name="console" class="ch.qos.logback.core.ConsoleAppender">
    <filter class="ch.qos.logback.classic.filter.ThresholdFilter">
      <level>DEBUG</level>
    </filter>
    <encoder>
      <Pattern>${log.consolePattern}</Pattern>
    </encoder>
  </appender>

  <appender name="rolling" class="ch.qos.logback.core.rolling.RollingFileAppender">
    <file>${log.dir}/${log.package}.log</file>
    <encoder>
      <Pattern>${log.pattern}</Pattern>
    </encoder>
    <rollingPolicy class="ch.qos.logback.core.rolling.TimeBasedRollingPolicy">
      <fileNamePattern>${log.dir}/${log.package}.%d{yyyy-MM-dd}.%i.log.zip</fileNamePattern>
      <timeBasedFileNamingAndTriggeringPolicy class="ch.qos.logback.core.rolling.SizeAndTimeBasedFNATP">
        <!-- or whenever the file size reaches 16MB. -->
        <maxFileSize>16MB</maxFileSize>
      </timeBasedFileNamingAndTriggeringPolicy>
      <!-- Keep no more than 3 months data. -->
      <maxHistory>90</maxHistory>
      <cleanHistoryOnStart>true</cleanHistoryOnStart>
    </rollingPolicy>
  </appender>

  <!-- Default logging levels. -->
  <root level="INFO">
    <appender-ref ref="console"/>
    <appender-ref ref="rolling"/>
  </root>
  <!-- Specific logging levels. -->
  <!-- Generally set to INFO or ERROR but if you need more details, set to DEBUG. -->
  <logger name="org.apache" level="INFO"/>
  <logger name="net.sf.ehcache" level="ERROR"/>
  <logger name="com.zaxxer" level="ERROR"/>
  <logger name="ch.qos" level="ERROR"/>
</configuration>

我手工编辑了日志中会话ID之后的值，以删除客户数据。

I hand-edited out the values after the session ID in the log to remove customer data.

推荐答案

有时候，尤其是在NullPointers（根据我的经验）中，jvm可以优化异常的创建和转换，以及堆栈跟踪丢失（或更准确，从未创建）。我怀疑你的问题与某些java库无关，而是与jvm本身有关。

Sometimes, especially when it comes to NullPointers (in my experience), the jvm can optimize the creation and casting of exceptions, and the stack trace is lost (or more correctly, never created). I suspect that your issue is not related to certain java libs, but rather the jvm itself.

如果在启动jvm-process时添加此参数，您将得到堆栈如果我的怀疑是正确的话，追溯回来。

If you add this argument when starting your jvm-process you will get your stack traces back, if my suspicion is correct.

-XX:-OmitStackTraceInFastThrow

之前已经询问过，请点击此处了解更多详情：

This has been asked before, look here for more details:

• Java中没有StackTrace的没有StackTrace的Java中的NullPointerException


• 重复异常没有堆栈跟踪 - 如何重置？

• NullPointerException in Java with no StackTrace
• Recurring Exception without a stack trace - how to reset?

注意，这适用于sun / oracle jvm

Note, that this applies sun/oracle jvm

这篇关于如何在没有堆栈跟踪的情况下创建/抛出异常？的文章就介绍到这了，希望我们推荐的答案对大家有所帮助，也希望大家多多支持IT屋！