发生之前的关系中的微妙时刻

Question

This question concerns JMM: https://www.cs.umd.edu/~pugh/java/memoryModel/jsr133.pdf

Could you please explain the following expression from JMM Specification (this is the section 5 on page 13):

More specifically, if two actions share a happens-before relationship, they do not necessarily have to appear to have happened in that order to any code with which they do not share a happens- before relationship. Writes in one thread that are in a data race with reads in another thread may, for example, appear to occur out of order to those reads.

If possible please provide Execution Traces that reflect this issue.

Answer 1

The paragraph preceding the paragraph you quote has stated, that the fact that an action a is in a happens-before relationship with an action b, does not mean that a is really executed before b. For example, in the code snippet

x = 10;
y = 20;

the two write action are in a happens-before relationship because they occur in the same thread. However, they may actually be swapped (by the compiler, by the JVM or even by the processor) because this swapping cannot be observed from within the thread itself.

However, it can be observed by another thread (and this is specifically allowed by the Java memory model). So another thread which reads those two variables, without some form of synchronization, may observe that y is changed to 20 before x is changed to 10. (The absense of synchronization makes sure that the write are in a data race with the reads in the other thread.)