Develop Reference

Can a java thread waiting with wait() ,notify itself?

I came across the following e example to implement custom suspend and wait from some website.
// Suspending and resuming a thread the modern way.
class NewThread implements Runnable {
String name; // name of thread
Thread t;
boolean suspendFlag;
NewThread(String threadname) {
name = threadname;
t = new Thread(this, name);
System.out.println("New thread: " + t);
suspendFlag = false;
t.start(); // Start the thread
}
// This is the entry point for thread.
public void run() {
try {
for (int i = 15; i > 0; i--) {
System.out.println(name + ": " + i);
Thread.sleep(200);
synchronized(this) {
while (suspendFlag) {
wait();
}
}
}
} catch (InterruptedException e) {
System.out.println(name + " interrupted.");
}
System.out.println(name + " exiting.");
}
void mysuspend() {
suspendFlag = true;
}
synchronized void myresume() {
suspendFlag = false;
notify();
}
}
class SuspendResume {
public static void main(String args[]) {
NewThread ob1 = new NewThread("One");
NewThread ob2 = new NewThread("Two");
try {
Thread.sleep(1000);
ob1.mysuspend();
System.out.println("Suspending thread One");
Thread.sleep(1000);
ob1.myresume();
...................
I am more concerned about the ob1.mysuspend() and ob1.myresume() calls. When my suspend is called then ob1 will be placed into the blocking queue associated with the runnable object it is using. When ob1 calls myresume, then how does it work as ob1 is already in waiting queue for the same object, can the waiting object enters another synchronised method and then signals notify to itself?How does this work?What am I missing?

The thread is written so that while an instance of NewThread is running, another thread can call mysuspend to suspend that running thread. Again, a thread other than the suspended thread calls myresume to resume the suspended thread.
There also appears to be a data race because mysuspend writes to suspendFlag without any synchronization. That means, the thread that needs to be suspended may not see that write immediately. mysuspend must be declared synchronized, or suspendFlag must be volatile.

This code is flat out broken.
Straight up broken: JMM violation
The mysuspend method (which should be named mySuspend, by the way) updates a field that is then read from another thread, and isn't synchronized. This is an error - and a really nasty one because you cannot reliably test that it is an error. The Java Memory Model (JMM) states that any write to a field may be observable or not, at the discretion of the JVM implementation, unless a so-called Happens-Before/Happens-After relationship is established (there are many ways to do it; usually you do so via synchronized, volatile, or some other concurrency tool built on these primitives, such as the latches and queues in the java.util.concurrent package).
You do not establish such a relationship here, meaning, that suspendFlag = true results in a schroedingers cat variable: The other thread that reads this field may read true or false, the JVM gets to decide what you see. Hence: A bug, and, untestable. bad. Any field that is read/written to by multiple threads needs to be written extremely carefully.
Mark that method synchronized, that's a good first step.
Wait and Notify
You've got it flipped around: You must in fact hold the synchronized lock on x when you invoke wait on x (here, x is this).
To call x.wait() (you are calling this.wait(), effectively), you must first be in a synchronized(x) block. Once the wait 'goes through', the code releases the lock (other synchronized(x) blocks can run). To invoke x.notify() you must also hold that lock.
wait does not return until the lock is re-established.
In other words:
public void foo() {
wait();
}
will fail at runtime. Try it. Guaranteed exception. In the mean time, this:
public void foo() {
synchronized (this) {
// code before wait
wait();
// code after wait
}
}
is executed as if it is written like this:
public void foo() {
synchronized (this) {
// code before wait
release_lock(this);
this.wait();
acquire_lock(this);
// code after wait
}
}
Where acquire_lock is guaranteed to actually take a while (because by definition whatever invoked notify() to wake you up is currently holding it! So wait is always a two-hit thing: You need to be both notified AND the lock needs to be reacquired before your code will continue). Except, of course, acquire_lock and release_lock don't exist, and unlike this hypothetical code, wait() is more atomic than that.

Significance of use of keyword synchronized in the following code

I was reading multi threading in Java from the book Java The Complete Reference by Herbert Schildt. I came across following code [Pg. 252, 7th ed.] that explained the usage of wait() and notify() to suspend and resume threads in modern Java. My question is regarding the significance of the keyword synchronization at two places in following code (in run() method of class NewThread):
// Suspending and resuming a thread the modern way.
class NewThread implements Runnable {
String name;
Thread t;
boolean suspendFlag;
NewThread(String threadname) {
name = threadname;
t = new Thread(this, name);
suspendFlag = false;
t.start();
}
// This is the entry point for thread.
public void run() {
try {
for (int i = 15; i > 0; i--) {
System.out.println(name + ": " + i);
Thread.sleep(200);
synchronized (this) { //First doubt here
while (suspendFlag) {
wait();
}
}
}
} catch (InterruptedException e) {
System.out.println(name + " interrupted.");
}
System.out.println(name + " exiting.");
}
void mysuspend() {
suspendFlag = true;
}
synchronized void myresume() { //Second doubt here
suspendFlag = false;
notify();
}
}
class SuspendResume {
public static void main(String args[]) {
NewThread ob1 = new NewThread("One");
NewThread ob2 = new NewThread("Two");
try {
Thread.sleep(1000);
ob1.mysuspend();
Thread.sleep(1000);
ob1.myresume();
ob2.mysuspend();
Thread.sleep(1000);
ob2.myresume();
} catch (InterruptedException e) {
System.out.println("Main thread Interrupted");
}
//some code
}
My doubt here: I know about the use of keyword synchronization i.e. allowing only one thread to enter a synchronized method on the same object but here we have two threads running on two different objects. So what is the significance of both synchronization keywords used in above code.
I tried running the above code by removing the synchronized keyword at each place differently and simultaneously. I am getting the same error: java.lang.IllegalMonitorStateException: current thread is not owner different number of times and at different line numbers depending upon if I remove both or only one (and which one) synchronization keyword. I looked for the above error and found an explanation for it here but still couldn't connect the answer to my doubt.

The problem that synchronized solves is, it allows the two threads to have a consistent view of the shared suspendFlag variable.
In some real program, a thread might set other shared variables before setting susependFlag=false. If synchronized was not used, then the waiting thread could wake up, and see suspendFlag==false, but not see the other variables set. Or worse, it could see some of them set, but not others.
Without synchronization, Java does not guarantee that different threads will see variables updated in the same order.
I am getting the same error: java.lang.IllegalMonitorStateException: current thread is not owner.
The Java library is trying to help you by forcing you to use synchronized before it will allow you to use wait() and notify(). The rule is simple: You can only call o.wait() or o.notify() or o.notifyAll() from code that is inside a synchronized(o) block. If you break that rule, then the library throws the exception.
When your code calls o.wait() the wait() call temporarily unlocks the monitor lock so that the other thread will be able to synchronize on o and call o.notify(). The o.wait() call is guaranteed to re-lock o before it returns.

synchronized thread execution

my task is to create thread in this order: if A start->start B and C, if B start->start D.
And destroy them in reverse order If D then B. If B and C then A. I hope you get it. I manage to do it but I guess there is better way to do it. Do you have any suggestions?
After your comments i have changed my code and it is much more simply. But now it looks "stupid". I would like to change hardcore of if statements and implementation, any advice? tnx for advice I'm learning with you.
This is my new code:
import java.util.*;
class RobotController implements Runnable{
String name;
public void run() {
Thread t = Thread.currentThread();
System.out.println(t.getName() + " status = " + t.isAlive());
System.out.println(t.getName() + " status = " + t.getState());
}
public static void main(String args[]) throws InterruptedException{
Thread thread_A = new Thread(new RobotController(), "Thread A");
Thread thread_B = new Thread(new RobotController(), "Thread B");
Thread thread_C = new Thread(new RobotController(), "Thread C");
Thread thread_D = new Thread(new RobotController(), "Thread D");
thread_A.start();
thread_A.join();
System.out.println(thread_A.getState());
thread_B.start();
thread_B.join();
System.out.println(thread_B.getState());
thread_C.start();
thread_C.join();
System.out.println(thread_C.getState());
thread_D.start();
System.out.println(thread_D.getState());
}
}

There are some flaws in your code which will make it not to work accordingly sometimes:
You called thread_A.start() and then checked thread_A.isAlive(). Now what if , thread_A is already completed before thread_A.isAlive() condition is checked?.thread_B and thread_C is never started. Your application fails.
Assume that thread_A is not completed and thread_A.isAlive() condition is passed, then starting of thread_B before thread_C is not always guaranteed by Java thread scheduler. Again your application fails.
Assume that thread_B starts before thread_C and if thread_B completes before thread_B.isAlive() is checked then the if condition fails and thread_D is never started. Again your application fails.
Now a point to ponder:
There is no need to check if the thread is alive after its join() method is called. It is an unnecessary runtime overhead.
EDIT
OK, Here is the modified version of code..I hope it would let you understand the dynamics of thread:
class RobotController implements Runnable
{
private final Object lock = new Object();
private void notifyThread()
{
synchronized(lock)
{
lock.notify();
}
}
public void run()
{
synchronized(lock)
{
try
{
System.out.println(Thread.currentThread().getName() + " started");
lock.wait();
System.out.println(Thread.currentThread().getName()+ " stopped");
}
catch (InterruptedException ex)
{
ex.printStackTrace();
}
}
}
public static void main(String args[]) throws InterruptedException
{
RobotController rca = new RobotController();
RobotController rcb = new RobotController();
RobotController rcc = new RobotController();
RobotController rcd = new RobotController();
Thread thread_A = new Thread(rca,"Thread A");
Thread thread_B = new Thread(rcb,"Thread B");
Thread thread_C = new Thread(rcc,"Thread C");
Thread thread_D = new Thread(rcd,"Thread D");
thread_A.start();
while (thread_A.getState() != Thread.State.WAITING)
{
Thread.sleep(100);
}
thread_B.start();
thread_C.start();
while (thread_B.getState() != Thread.State.WAITING && thread_C.getState() != Thread.State.WAITING)
{
Thread.sleep(100);
}
thread_D.start();
while (thread_D.getState() != Thread.State.WAITING)
{
Thread.sleep(100);
}
rcd.notifyThread();
thread_D.join();
rcc.notifyThread();
thread_C.join();
rcb.notifyThread();
thread_B.join();
rca.notifyThread();
}
}
And here is the output:
Thread A started
Thread B started
Thread C started
Thread D started
Thread D stopped
Thread C stopped
Thread B stopped
Thread A stopped

In multi threading, there is no need of synchronization unless the common data is shared by multiple threads.
In your case, you want to start and stop the threads in a particular order. For this, there is join method in Thread class. This link shows good example of join method.

In my opinion, it is quite strange to use synchronized (lock) in your run method to lock your object. The reason is that in each Thread object has the different lock attribute, which is belong to each object. It means you are trying to lock the different objects. Actually, it doesn't make sense.
Basically, the object that you should apply the synchronized are any shared objects. For example, you need to count something and then you create a class object to share it in your class. In this case, it should be locked while being read or written.

I would like to highlight two points here:
Have a look at thread execution life cycle here. It says that, when start() method is called, thread enters in runnable state and not in running state. When thread enters in running state, that means run() method is getting exexuted. CPU/OS decides the priority of which thread should be transferred from runnable to running thread. E.g. if you call start() method for 4 threads, it is not necessary that they will execute in that particular order. (Running the same program in my pc multiple times gives me different outputs.
In your case, when the condition if(thread_A.isAlive()) is executed, it is possible that the thread A may not be in running state. Hence, the control will not go into if which is not correct.
To correct this behavior, in main, a while loop should implemented which waits until the thread becomes alive and so on.
2 . In your program, you have not assigned names to the threads and you are printing the name in the run() method. In this case, JVM assigns the names to threads in order of their execution e.g. first thread to execute will have name as 'Thread-0' and so on. Hence, we will not be able to identify which thread executed first. Assign the names using setName() method.

Difference between notify() and notifyAll()

I know that similar questions have been discussed in this site, but I have not still got further by their aid considering a specific example. I can grasp the difference of notify() and notifyAll() regarding Thread "awakeining" in theory but I cannot perceive how they influence the functionality of program when either of them is used instead of the other. Therefore I set the following code and I would like to know what is the impact of using each one of them. I can say from the start that they give the same output (Sum is printed 3 times).
How do they differ virtually? How could someone modify the program, in order for the applying notify or notifyAll to play a crucial role to its functionality (to give different results)?
Task:
class MyWidget implements Runnable {
private List<Integer> list;
private int sum;
public MyWidget(List<Integer> l) {
list = l;
}
public synchronized int getSum() {
return sum;
}
#Override
public void run() {
synchronized (this) {
int total = 0;
for (Integer i : list)
total += i;
sum = total;
notifyAll();
}
}
}
Thread:
public class MyClient extends Thread {
MyWidget mw;
public MyClient(MyWidget wid) {
mw = wid;
}
public void run() {
synchronized (mw) {
while (mw.getSum() == 0) {
try {
mw.wait();
} catch (InterruptedException e) {
e.printStackTrace();
}
}
System.out.println("Sum calculated from Thread "
+ Thread.currentThread().getId() + " : " + mw.getSum());
}
}
public static void main(String[] args) {
Integer[] array = { 4, 6, 3, 8, 6 };
List<Integer> integers = Arrays.asList(array);
MyWidget wid = new MyWidget(integers);
Thread widThread = new Thread(wid);
Thread t1 = new MyClient(wid);
Thread t2 = new MyClient(wid);
Thread t3 = new MyClient(wid);
widThread.start();
t1.start();
t2.start();
t3.start();
}
}
UPDATE:
I write it explicitly. The result is the same whether one uses notify or notifyAll:
Sum calculated from Thread 12 : 27
Sum calculated from Thread 11 : 27
Sum calculated from Thread 10 : 27
Therefore my question: What is the difference?

The difference is subtler than your example aims to provoke. In the words of Josh Bloch (Effective Java 2nd Ed, Item 69):
... there may be cause to use notifyAll in place of notify. Just as placing the wait invocation in a loop protects against accidental or malicious notifications on a publicly accessible object, using notifyAll in place of notify protects against accidental or malicious waits by an unrelated thread. Such waits could otherwise “swallow” a critical notification, leaving its intended recipient waiting indefinitely.
So the idea is that you must consider other pieces of code entering wait on the same monitor you are waiting on, and those other threads swallowing the notification without reacting in the designed way.
Other pitfalls apply as well, which can result in thread starvation, such as that several threads may wait for different conditions, but notify always happens to wake the same thread, and the one whose condition is not satisfied.
Even though not immediately related to your question, I feel it is important to quote this conclusion as well (emphasis by original author):
In summary, using wait and notify directly is like programming in “concurrency assembly language,” as compared to the higher-level language provided by java.util.concurrent. There is seldom, if ever, a reason to use wait and notify in new code. If you maintain code that uses wait and notify, make sure that it always invokes wait from within a while loop using the standard idiom. The notifyAll method should generally be used in preference to notify. If notify is used, great care must be taken to ensure liveness.

This is made clear in all sorts of docs. The difference is that notify() selects (randomly) one thread, waiting for a given lock, and starts it. notifyAll() instead, restarts all threads waiting for the lock.
Best practice suggests that threads always wait in a loop, exited only when the condition on which they are waiting is satisfied. If all threads do that, then you can always use notifyAll(), guaranteeing that every thread whose wait condition has been satisfied, is restarted.
Edited to add hopefully enlightening code:
This program:
import java.util.concurrent.CountDownLatch;
public class NotifyExample {
static final int N_THREADS = 10;
static final char[] lock = new char[0];
static final CountDownLatch latch = new CountDownLatch(N_THREADS);
public static void main(String[] args) {
for (int i = 0; i < N_THREADS; i++) {
final int id = i;
new Thread() {
#Override public void run() {
synchronized (lock) {
System.out.println("waiting: " + id);
latch.countDown();
try { lock.wait(); }
catch (InterruptedException e) {
System.out.println("interrupted: " + id);
}
System.out.println("awake: " + id);
}
}
}.start();
}
try { latch.await(); }
catch (InterruptedException e) {
System.out.println("latch interrupted");
}
synchronized (lock) { lock.notify(); }
}
}
produced this output, in one example run:
waiting: 0
waiting: 4
waiting: 3
waiting: 6
waiting: 2
waiting: 1
waiting: 7
waiting: 5
waiting: 8
waiting: 9
awake: 0
None of the other 9 threads will ever awaken, unless there are further calls to notify.

notify wakes (any) one thread in the wait set, notifyAll wakes all threads in the waiting set. notifyAll should be used most of the time. If you are not sure which to use, then use notifyAll.
In some cases, all waiting threads can take useful action once the wait finishes. An example would be a set of threads waiting for a certain task to finish; once the task has finished, all waiting threads can continue with their business. In such a case you would use notifyAll() to wake up all waiting threads at the same time.
Another case, for example mutually exclusive locking, only one of the waiting threads can do something useful after being notified (in this case acquire the lock). In such a case, you would rather use notify(). Properly implemented, you could use notifyAll() in this situation as well, but you would unnecessarily wake threads that can't do anything anyway.
Javadocs on notify.
Javadocs on notifyAll.

Once only one thread is waiting to sum to not be zero, there is no difference. If there are several threads waiting, notify will wake up only one of them, and all the other will wait forever.
Run this test to better understand the difference:
public class NotifyTest implements Runnable {
#Override
public void run ()
{
synchronized (NotifyTest.class)
{
System.out.println ("Waiting: " + this);
try
{
NotifyTest.class.wait ();
}
catch (InterruptedException ex)
{
return;
}
System.out.println ("Notified: " + this);
}
}
public static void main (String [] args) throws Exception
{
for (int i = 0; i < 10; i++)
new Thread (new NotifyTest ()).start ();
Thread.sleep (1000L); // Let them go into wait ()
System.out.println ("Doing notify ()");
synchronized (NotifyTest.class)
{
NotifyTest.class.notify ();
}
Thread.sleep (1000L); // Let them print their messages
System.out.println ("Doing notifyAll ()");
synchronized (NotifyTest.class)
{
NotifyTest.class.notifyAll ();
}
}
}

I found what is going on with my program. The three Threads print the result even with the notify(), because they do not manage to enter the waiting state. The calculation in the widThread is performed quickly enough to preempt the entering of the other Threads in the waiting state, since it depends on the condition mw.getSum() == 0 (while loop). The widThread calculates the sum, so that the remaining Threads do not ever "see" its value as 0.
If the while loop is removed and the start of widThread comes after the start of the other Threads, then by notify() only one Thread prints the result and the others are waiting forever, as the theory and the other answers indicate.

Java synchronized block, not all the threads get terminated

So I have the following code:
import java.lang.Thread;
import java.lang.Integer;
class MyThread extends Thread {
private int id;
MyThread(int i){
id = i;
}
public void run() {
while(true){
try{
synchronized(Global.lock){
Global.lock.wait();
if(Global.n == 0) {System.out.println(id); Global.lock.notify(); break;}
--Global.n;
System.out.println("I am thread " + id + "\tn is now " + Global.n);
Global.lock.notify();
}
}
catch(Exception e){break;}
}
}
}
class Global{
public static int n;
public static Object lock = new Object();
}
public class Sync2{
public static final void main(String[] sArgs){
int threadNum = Integer.parseInt(sArgs[0]);
Global.n = Integer.parseInt(sArgs[1]);
MyThread[] threads = new MyThread[threadNum];
for(int i = 0; i < threadNum; ++i){
threads[i] = new MyThread(i);
threads[i].start();
}
synchronized(Global.lock){Global.lock.notify();}
}
}
two parameters are entered: a number n and the number of threads to be created. Every thread decreases n by one and then passes control. All threads should stop when n is 0. It seems to work fine so far, but the only problem is that in most of the cases all threads except one terminate. And one is hanging on. Any idea why?
And yes, this is part of a homework, and that is what I've done so far (I was no provided with the code). I'am also explicitly restricted to use a synchronized block and only wait() and .notify() methods by the task.
EDIT: modified the synchronized block a bit:
synchronized(Global.lock){
Global.lock.notify();
if (Global.n == 0) {break;}
if (Global.next != id) {Global.lock.wait(); continue;}
--Global.n;
System.out.println("I am thread " + id + "\tn is now " + Global.n);
Global.next = ++Global.next % Global.threadNum;
}
now threads act strictly in the order they are created. Its pretty unclear from the task wording, but might be the right thing.

You have a race condition. Think about what happens with a single worker thread. Global.n is set to 1 and then the thread starts. It immediately goes into a wait state. Suppose, though, that notify() had already been called on the main thread. Since the worker thread hasn't yet entered a wait state, it isn't notified. Then, when it finally does call wait(), there are no other threads around to call notify(), it stays in the wait state forever. You need to fix up your logic to avoid this race condition.
Also, do you really want a single worker thread to decrement Global.n more than once? That can easily happen with your while (true) ... loop.
EDIT
You also have another logic problem with a single thread. Suppose it enters the wait state and then the notify() in main is called. It wakes the worker thread which decrements Global.n to 0, calls notify(), and then goes back to waiting. The problem is that notify() didn't wake any other thread because there were no other threads to wake. So the one worker thread will wait forever. I haven't analyzed it fully, but something like this might also happen with more than one worker thread.

You should never have a naked wait() call, as semaphores in java are not cached. wait() should always be nested in some sort of
while (condition that you are waiting on)
obj.wait();