minor extensions to tck20 test utilities

• Key: JDO-309
• Type: New Feature
• Status: Resolved
• Resolution: Fixed
• Priority: Trivial
• Assignee: Martin Zaun
• Reporter: Martin Zaun
• Component/s: tck2

Description

JDO_Test should be extended for methods collection error messages.
ThreadExceptionHandler should be extended for clearing stored exceptions.

Comments

[Martin Zaun]

My planned extensions to tck20's ThreadExceptionHandler:

$ svn diff src/java/org/apache/jdo/tck/util/ThreadExceptionHandler.java
Index: src/java/org/apache/jdo/tck/util/ThreadExceptionHandler.java
===================================================================
--- src/java/org/apache/jdo/tck/util/ThreadExceptionHandler.java (revisio
n 375502)
+++ src/java/org/apache/jdo/tck/util/ThreadExceptionHandler.java (working
 copy)
@@ -32,7 +32,7 @@
      * Map of uncaught exceptions. The thread is the key and the uncaught
      * Throwable is the value in the map.
      */
- private Map uncaughtExceptions = new HashMap();
+ private final Map uncaughtExceptions = new HashMap();

     /** Constructor. */
     public ThreadExceptionHandler() {
@@ -53,11 +53,18 @@
     }

     /**
- * Returns all uncaught exception stored in this ThreadGroup.
+ * Returns all uncaught exceptions stored in this ThreadGroup.
      * Each element in the returned set is a Map.Entry with the
      * thread as the key and the uncaught Throwable is the value.
      */
     public Set getAllUncaughtExceptions() {
         return uncaughtExceptions.entrySet();
     }
+
+ /**
+ * Clears all exceptions in this ThreadGroup.
+ */
+ public void clear() {
+ uncaughtExceptions.clear();
+ }
 }

[Martin Zaun]

The planned extensions to tck20's JDO_Test (based on Craig's suggestion):

$ svn diff src/java/org/apache/jdo/tck/JDO_Test.java
Index: src/java/org/apache/jdo/tck/JDO_Test.java
===================================================================
--- src/java/org/apache/jdo/tck/JDO_Test.java (revision 375502)
+++ src/java/org/apache/jdo/tck/JDO_Test.java (working copy)
@@ -128,6 +128,10 @@
     /** identitytype value for datastoreidentity. */
     public static final String DATASTORE_IDENTITY = "datastoreidentity";

+ /** New line.
+ */
+ public static final String NL = System.getProperty("line.separator");
+
     /**
      * String indicating the type of identity used for the current test case.
      * The value is either "applicationidentity" or "datastoreidentity".
@@ -147,6 +151,10 @@
     /** The Properties object for the PersistenceManagerFactory. */
     protected static Properties PMFPropertiesObject;

+ /** A buffer of of error messages.
+ */
+ protected static StringBuffer messages;
+
     /** The PersistenceManagerFactory. */
     protected PersistenceManagerFactory pmf;

@@ -537,13 +545,36 @@
         return props;
     }

+ /** Appends to error messages.
+ */
+ protected static synchronized void appendMessage(String message) {
+ if (messages == null) {
+ messages = new StringBuffer();
+ }
+ messages.append(message);
+ messages.append(NL);
+ }
+
+ /**
+ * Returns collected error messages, or <code>null</code> if there
+ * are none, and clears the buffer.
+ */
+ protected static synchronized String popMessages() {
+ if (messages == null) {
+ return null;
+ }
+ final String msg = messages.toString();
+ messages = null;
+ return msg;
+ }
+
     /**
      * Prints the specified msg (if debug is true), before it aborts the
      * test case.
      */
     public void fail(String assertionFailure, String msg) {
         if (debug) logger.debug(msg);
- fail(assertionFailure + "\n" + msg);
+ fail(assertionFailure + NL + msg);
     }

     // Helper methods to check for supported options

[Craig Russell]

Looks good. Just a comment.

1. Please change name of popMessages to something else. Pop has a specific meaning (remove the top element of a Stack) and I don't think it's appropriate for a Message pad.

How about:

String retrieveMessages() that has the same semantics as your popMessages.

[Martin Zaun]

> How about:
> String retrieveMessages() that has the same semantics as your popMessages.

OK.

[Martin Zaun]

For thread-safety testing, a simple barrier synchronization utility is needed.

[Michael Bouschen]

Hi Martin,

the ThreadExceptionHandler changes look good.

I looked at the RougueBarriere class. It's small, but still takes some time to understand (at least for me :-)). Maybe you can add some comments motivating why we need a holder for the boolean tripped. I think this allows other threads to see that the barrier is tripped, but the current thread can prepare the value for the next run. Maybe you can also rearrange the code in the await method executed when the barrier is reset. I think
    tripped[0] = true;
    lock.notifyAll();
is needed to wake up the other threads, where
    missing = parties;
    tripped = new boolean[1]; // new generation
is used to initialized the next run.

Am I right?

Michael

[Martin Zaun]

Followed Michael's comments on new class RougueBarrier.
Added extension to ThreadExceptionHandler and new RougueBarrier with revision 384047.
(Additions to JDO_Test had been checked in already on another JIRA issue.)

[Martin Zaun]

Removed executable property from new file RogueBarrier.java with revision 386446.