在最近与终结器方法搏斗之后(为了在测试期间处理连接池)，我不得不说终结器缺少很多东西。使用VisualVM来观察以及使用弱引用来跟踪实际的交互，我发现以下事情在Java 8环境中是正确的(Oracle JDK, Ubuntu 15):

Finalize is not called immediately the Finalizer (GC part) individually owns the reference elusively The default Garbage Collector pools unreachable objects Finalize is called in bulk pointing to an implementation detail that there is a certain phase the garbage collector frees the resources. Calling System.gc() often does not result in objects being finalized more often, it just results in the Finalizer getting aware of an unreachable object more rapidly Creating a thread dump almost always result in triggering the finalizer due to high heap overhead during performing the heap dump or some other internal mechanism Finalization seams to be bound by either memory requirements (free up more memory) or by the list of objects being marked for finalization growing of a certain internal limit. So if you have a lot of objects getting finalized the finalization phase will be triggered more often and earlier when compared with only a few There were circumstances a System.gc() triggered a finalize directly but only if the reference was a local and short living. This might be generation related.

最后认为

最后确定方法是不可靠的，但只能用于一件事。您可以确保在垃圾收集之前关闭或释放对象，从而在正确处理涉及生命结束操作的更复杂生命周期的对象时实现故障安全。这是我能想到的一个值得我们去推翻它的原因。

一般来说，最好不要依赖finalize()来做任何清理等。

根据Javadoc(值得一读)，它是:

当垃圾回收确定对象不再有引用时，由垃圾回收器在对象上调用。

正如Joachim指出的，如果对象总是可访问的，那么在程序的生命周期中，这可能永远不会发生。

此外，垃圾收集器不保证在任何特定时间运行。一般来说，我想说的是finalize()可能不是最好的方法，除非有特定的事情需要它。

最终确定方法是不保证的。当对象符合GC条件时调用此方法。在许多情况下，对象可能不会被垃圾收集。

正如在https://wiki.sei.cmu.edu/confluence/display/java/MET12-J.+Do+not+use+finalizers上指出的那样，

There is no fixed time at which finalizers must be executed because time of execution depends on the Java Virtual Machine (JVM). The only guarantee is that any finalizer method that executes will do so sometime after the associated object has become unreachable (detected during the first cycle of garbage collection) and sometime before the garbage collector reclaims the associated object's storage (during the garbage collector's second cycle). Execution of an object's finalizer may be delayed for an arbitrarily long time after the object becomes unreachable. Consequently, invoking time-critical functionality such as closing file handles in an object's finalize() method is problematic.

JDK 18的最新消息

根据openjdk 18发布的JEPS 421, finalize()方法的终结和功能将被标记为已弃用(forRemoval=true)，这意味着永久删除将在jdk 18之后的某个后续版本中进行。

从jdk 18开始，一个新的命令行选项——finalization=disabled禁用了所有地方的终结机制，甚至是jdk内部的声明。

这也与这里的这个问题有关，因为它被计划删除的原因是它包含的一些主要缺陷。其中一个缺陷是，从一个对象变得不可访问到调用它的终结器之间可能会经过很长一段时间。GC也不能保证任何终结器将被调用，这也是事实。

protected void finalize() throws Throwable {} every class inherits the finalize() method from java.lang.Object the method is called by the garbage collector when it determines no more references to the object exist the Object finalize method performs no actions but it may be overridden by any class normally it should be overridden to clean-up non-Java resources ie closing a file if overridding finalize() it is good programming practice to use a try-catch-finally statement and to always call super.finalize(). This is a safety measure to ensure you do not inadvertently miss closing a resource used by the objects calling class protected void finalize() throws Throwable { try { close(); // close open files } finally { super.finalize(); } } any exception thrown by finalize() during garbage collection halts the finalization but is otherwise ignored finalize() is never run more than once on any object

引用自:http://www.janeg.ca/scjp/gc/finalize.html

你也可以看看这篇文章:

对象终结和清理