Pipelined I/O: java.io.IOException: Write end dead [1.4.2]

Name: elR10090 Date: 09/05/2001



JDK 1.4.0beta3-b78 fails against a simple test trying to pass
one text line through PipedInputStream/PipedOutputStream line.

The test a connected pair of PipedInputStream & PipedOutputStream,
and starts two threads -- one reading & another writing a string
via the pipeline. But, the reading thread suffers IOException
while expecting to observe EOF on the PipedInputStream:

    >>>> javac PrintStreamTest.java
    >>>> java PrintStreamTest
    test message
    java.io.IOException: Write end dead
     at java.io.PipedInputStream.read(PipedInputStream.java:216)
     at java.io.PipedInputStream.read(PipedInputStream.java:277)
     at java.io.InputStream.read(InputStream.java:91)
     at java.io.InputStreamReader$ConverterFiller.fill(InputStreamReader.java:248)
     at java.io.InputStreamReader.read(InputStreamReader.java:511)
     at java.io.BufferedReader.fill(BufferedReader.java:139)
     at java.io.BufferedReader.readLine(BufferedReader.java:302)
     at java.io.BufferedReader.readLine(BufferedReader.java:365)
     at PrintStreamTest$Redirector.run(PrintStreamTest.java:50)

    >>>> java -version
    java version "1.4.0-beta3"
    Java(TM) 2 Runtime Environment, Standard Edition (build 1.4.0-beta3-b78)
    Java HotSpot(TM) Client VM (build 1.4.0-beta3-b78, mixed mode)

Following is the source code for the "PrintStreamTest.java":

import java.io.*;

public class PrintStreamTest {
    final static int RACE_CONDITION = 10; // seconds

    public static void main(String args[])
        throws IOException, InterruptedException
    {
        PipedInputStream in = new PipedInputStream();
        PipedOutputStream out = new PipedOutputStream(in);

        Redirector redirector = new Redirector(in,System.out);
        redirector.start();

        PrintStream testOut = new PrintStream(out);
        TestThread testThread = new TestThread(testOut);
        testThread.start();
        testThread.join();
        testOut.close();

        redirector.join();
    }

    private static class TestThread extends Thread {
        private PrintStream out;
        public TestThread(PrintStream out) {
            this.out = out;
        }
        public void run() {
            try {
                sleep(RACE_CONDITION * 1000L);
            } catch (InterruptedException exception) {
                exception.printStackTrace(out);
                return;
            };
            out.println("test message");
        }
    }

    private static class Redirector extends Thread {
        private BufferedReader in;
        private PrintStream out;
        public Redirector(InputStream in, PrintStream out) {
            this.in = new BufferedReader(new InputStreamReader(in));
            this.out = out;
        }
        public void run() {
            for (String line; ; )
                try {
                    line = in.readLine();
                    if (line == null)
                        break;
                    out.println(line);
                } catch (IOException exception) {
                    exception.printStackTrace(out);
                    return;
                }
        }
    }
}

Probably, this testcase reproduces the known intermittent failure
filed against HS 1.0 and closed as "will not fix":

    4211703 PipedInputStream fails to read the correct number of bytes

Please note, that the exception doesn't happen if RACE_CONDITION
value is reassigned to less than 2 seconds.

This bug affects the following testbase_nsk test stressing-out logging
into ConsoleHandler (the test will appear with the next "r07" release
of the testbase):

    nsk/logging/stress/threads/cnshsl001

While the simplified testcase "PrintStreamTest.java" demonstrates the
problem only on Solaris/SPARC; the Logging API stress test "cnshsl001"
fails on all platforms -- that is why I marked this bug's O/S and
hardware versions as "generic".

The following simplified testcase "ConsoleHandlerThreadsTest.java"
also demonstrates the exception on all platforms (intermittently
on Windows):

    import java.io.*;
    import java.util.logging.*;

    /**
     * Check if ConsoleHandler can successfully log when <tt>stderr</tt>
     * is reassigned with <tt>setErr()</tt> method of the <tt>System</tt>.
     */
    public class ConsoleHandlerThreadsTest {
        final static int THREADS_NUMBER = 10;

        public static void main(String args[])
            throws IOException, InterruptedException
        {
            PipedInputStream in = new PipedInputStream();
            PipedOutputStream out = new PipedOutputStream(in);

            Redirector redirector = new Redirector(in,System.out);
            redirector.start();

            PrintStream fakeErr = new PrintStream(out);
            PrintStream stdErr = System.err;
            System.setErr(fakeErr);

            Handler handler = new ConsoleHandler();
            handler.setLevel(Level.ALL);
            handler.setFormatter(new TrivialFormatter());

            Logger logger = Logger.getLogger("test.logger000");
            logger.setLevel(Level.ALL);
            logger.addHandler(handler);

            Thread threads[] = new TestThread [ THREADS_NUMBER ];
            for (int i=0; i<threads.length; i++)
                threads[i] = new TestThread(i+1, logger);
            for (int i=0; i<threads.length; i++)
                threads[i].start();
            for (int i=0; i<threads.length; i++)
                threads[i].join();

            logger.removeHandler(handler);
            handler.close();

            System.setErr(stdErr);
            fakeErr.close();

            redirector.join();
        }
    
        private static class TestThread extends Thread {
            private Logger logger;
            private int id;
        
            public TestThread(int id, Logger logger) {
                this.logger = logger;
                this.id = id;
            }
        
            public void run() {
                logger.finest("message #" + id);
            }
        }
    
        private static class Redirector extends Thread {
            private BufferedReader in;
            private PrintStream out;
        
            public Redirector(InputStream in, PrintStream out) {
                this.in = new BufferedReader(new InputStreamReader(in));
                this.out = out;
            }
        
            public void run() {
                for (String line; ; )
                    try {
                        line = in.readLine();
                        if (line == null)
                            break;
                        out.println(line);
                    } catch (IOException exception) {
                        exception.printStackTrace(out);
                        return;
                    }
            }
        }
    
        final static String LINE_SEPARATOR = System.getProperty("line.separator");

        private static class TrivialFormatter extends Formatter {
            public String format(LogRecord record) {
                return record.getMessage() + LINE_SEPARATOR;
            }
        }
    }

======================================================================

Name: elR10090 Date: 09/11/2001



Alexey Gibadullin, ###@###.###

This bug also affects the following tests:

    nsk/logging/stress/threads/cnshml001
    nsk/logging/stress/threads/cnshml002


======================================================================

Name: elR10090 Date: 09/12/2001



Following is the full list of testbase_nsk tests
failing due to this bug:

    nsk/logging/stress/threads/cnshsl001
    nsk/logging/stress/threads/cnshml001
    nsk/logging/stress/threads/cnshml002
    nsk/logging/stress/threads/cnmhml001
    nsk/logging/stress/threads/cnmhml002
    nsk/logging/stress/threads/smmhml001
    nsk/logging/stress/threads/smmhml002

I add this notice to let the automated script preparing
Tonga FAIL_LIST for testbase_nsk executions to include
these test names into the list of expected failures.

======================================================================

Name: elR10090 Date: 11/19/2001



Alexey Gibadullin, ###@###.###

One more test is affected by this bug:

    nsk/logging/stress/threads/smshml002
    
The failure was observed on Linux and Solsparc platforms in
various modes, but not in all runs (approximatly 1 out of 5 cases).



======================================================================

Name: elR10090 Date: 03/09/2004


The failure still looks reproducible against Tiger-b41;
I've observed multiple test fails for various platforms
(Windows, Linux, Solaris/x86, Solaris/SPARC/v8/v9) and
for various VM modes (-server and -client, mode default,
-Xcomp, java_g).

Though, the following two tests fail only for java_g:

    nsk/logging/stress/threads/smshml001
    nsk/logging/stress/threads/smshsl001

I'm adding the test names here to put them to the list
of expected test failures (ie: vm.knownbugs.v15).

======================================================================