[mdr-users] NullPointerException with IBM JRE

(I tried filing a bug report on this but it seems my
login doesn't have the necessary permission to do so.)

Running under the IBM JRE produces a NullPointerException
while trying to initialize a B-tree repository.

     [java] [org.netbeans.mdr.Logger] Rebooting storage. Reason: 
org.netbeans.mdr.persistence.StorageBadRequestException: Btree repository 
mdr does not exist
     [java] *********** Exception occurred ************ at Wed May 18 
15:11:12 CDT 2005
     [java] org.netbeans.mdr.util.DebugException: Fatal error: Repository 
boot/initialization failed with message: java.lang.NullPointerException: 
null
     [java]     at 
org.netbeans.mdr.NBMDRepositoryImpl.initCheck(NBMDRepositoryImpl.java:753)
     [java]     at 
org.netbeans.mdr.NBMDRepositoryImpl.getExtent(NBMDRepositoryImpl.java:298)
     [java]     at org.netbeans.lib.jmi.uml2mof.Main.init(Main.java:70)
     [java]     at org.netbeans.lib.jmi.uml2mof.Main.main(Main.java:53)
     [java] ANNOTATIONS:
     [java] [null] Exception occurred:

This is with latest sources as of today (May 18 2005)
running on an IBM 1.4.1 JRE for z/OS:

    $ java -version
    java version "1.4.1"
    Java(TM) 2 Runtime Environment, Standard Edition (build 1.4.1)
    Classic VM (build 1.4.1, J2RE 1.4.1 IBM z/OS Persistent Reusable VM 
build cm1411sr2a-20040515 (JIT enabled: jitc))

That's right, I'm running MDR on a mainframe!

I chased the problem and it's failing because some code
in MDRCache is using Map.Entry objects after they've gone
invalid.  The following patch fixes it:

    Index: 
mdr/src/org/netbeans/mdr/persistence/btreeimpl/btreestorage/MDRCache.java
    ===================================================================
    RCS file: 
/cvs/mdr/src/org/netbeans/mdr/persistence/btreeimpl/btreestorage/MDRCache.java,v
    retrieving revision 1.16
    diff -r1.16 MDRCache.java
    455c455
    <         result.addAll(newOnes.entrySet());
    ---
    >         result.addAll(new HashMap(newOnes).entrySet());
    464c464
    <         ArrayList result = new ArrayList(dirty.entrySet());
    ---
    >         ArrayList result = new ArrayList(new 
HashMap(dirty).entrySet());

This program demonstrates the difference between the Sun and IBM
JREs re. Map views:

    import java.util.*;

    public class MapViews
    {
        public static void main(String[] args)
        {
            Map m = new HashMap();
            m.put("key1", "value1");
            m.put("key2", "value2");

            Set s = new HashSet(m.entrySet());
            m.clear();

            System.out.println("Set s = " + s);
        }
    }

With the Sun JRE this code outputs:

    Set s = [key2=value2, key1=value1]

while the IBM JRE outputs:

    Set s = [null=null, null=null]

Both are legal results.  The IBM version makes it clearer
that Map.Entry objects are valid only for the duration of
an iteration over the entrySet().

Mike