Question

Both of wait() and join() methods when called by thread-1 on thread-2 makes thread-1 wait for the thread-2, either for sometime or till thread-2 completes.

If we are using the overloaded versions of these methods i.e. wait(long timeout) and join(long millis), then

1. In case of wait(long timeout), thread-1 will become runnable either by notify (or notifyall) or even timeout occurs (whichever is first).

2. In case of join(long millis), thread-2 will become runnable either when thread-2 completes or timeout occurs (whichever is first).

So then what is the difference between these two implementations?

Some that I thought are these :-

1. For wait(), we need to have the lock on the object we are waiting on. For join() these are not necessary.
2. After executing wait(), the thread removes lock it obtained and re-gains the lock once it becomes running again. But what about join? Does a thread removes a lock after executing join if this was executed from a synchronized block (or method)?

Answer 1

As you say, the "release" process is quite different – in one case (wait) it’s based on notify(), the other (join) it’s based on the thread completing. They’re entirely different calls which serve entirely different purposes.

In fact, there are explicit warnings not to call wait() on Thread monitors (although I can’t immediately find those warnings), as internal Java code acquires the locks for them (and uses wait/notify itself).

But no, calling join() on Thread doesn’t release the monitor if the currently executing thread owns it.

Basically, you shouldn’t think of them as similar at all – one is for waiting for a thread to terminate; the other is for waiting for co-operative coordination.