"double-checked locking"是否真的安全呢

本文的思想来源于：
[b]Jeremy Manson[/b] and [b]Brian Goetz[/b], 《[b]JSR 133 (Java Memory Model) FAQ[/b]》, February 2004, http://www.cs.umd.edu/~pugh/java/memoryModel/jsr-133-faq.html#synchronization

"double-checked locking"经常被用于Singleton模型中实现lazy load，典型的"double-checked locking"代码如下所示：

  // double-checked-locking - don't do this!

		// initialization
		private static Something instance = null;

		public static Something getInstance() {
		  if (instance == null) {
		    synchronized (this) {
		      if (instance == null)
		        instance = new Something();
		    }
		  }
		  return instance;
		}

然后由于reordering的存在，instance 的初始化有可能会被move到getInstance之后执行。这种情况下，
虽然在getInstance()方法中，instance被正确的赋值，但是之后又被初始化为null。下面的程序打印的结果为null。

  public static Something getInstance() {
		  if (instance == null) {
		    synchronized (this) {
		      if (instance == null)
		        instance = new Something();
		    }
		  }
		  return instance;
		}

		private static Something instance = null;

		public static void main(String[] args) {
			Something instance = getInstance();

			System.out.println(instance);
		}

由于volatile的特性（参加另一篇文章《volatile关键词的原理与用途》），这个问题显然可以通过将instance字段声明为volatile来fix。但是更为简单直观的方法为：

  public Class LazySomething{
			// static inner class
			private static class LazySomethingHolder {
			  public static Something something = new Something();
			}

			public static Something getInstance() {
			  return LazySomethingHolder.something;
			}
		}