Comment 1 Jerome Lanneluc 2008-05-24 14:45:53 EDT

We're missing some context to reproduce. Can you please provide the types Record and FastMap?

Comment 2 Tadas 2008-05-25 03:38:38 EDT

@Override
public synchronized boolean remove(Object key)
{
    _cache.remove(key);
    _mruList.remove(key);
    return _map.remove(key) == key;
}

private FastMap<K,CacheNode> _cache = new FastMap<K,CacheNode>().setKeyComparator(FastComparator.DIRECT);
private FastMap<K,V> _map;
private FastList<K> _mruList = new FastList<K>();

Comment 3 Philipe Mulet 2008-05-26 07:16:00 EDT

Please provide a self-contained and complete testcase,
e.g.
import java.util.*;

public class X {
	public final Object valueOf(Record record) {
		return ((FastMap.Entry<Object, Object>) record).getKey();
	}
	public final void delete(Record record) {
		remove(((FastMap.Entry<Object, Object>) record).getKey());
	}
	public synchronized boolean remove(Object key) {
		return false;
	}
}
class Record {}
abstract class FastMap implements Map {
}


The program above does NOT reproduce any problem.

Comment 4 Philipe Mulet 2008-05-26 09:09:26 EDT

The trace indicates a bound parameterized type for which arguments would be null... this shouldn't occur ever.

Comment 5 Philipe Mulet 2008-05-27 05:12:58 EDT

Tadas - can you pls provide a self-contained and complete testcase ?

Comment 6 Andy Carroll 2008-05-27 11:50:45 EDT

The following java code below doesn't compile under Eclipse 3.4RC2 build I20080523-0100 due to a NullPointerException in CastExpression.checkUnsafeCase:335.

Note that line#382 of Expression.java explicitly passes null as the match argument to the method checkUnsafeCase, which is subsequently dereferenced, resulting in a NullPointerException.

Hope this helps.

- Andy


public class Bug233800<T>
{
   
   public void doesNotCompile(SomeInterface i)
   {
      T t = ((SomeDerivedInterface<T>)i).getItem();
   }
   
   static interface SomeInterface
   {
      
   }
   
   static interface SomeDerivedInterface<T> extends SomeInterface
   {
      T getItem();
   }
}

Comment 7 Philipe Mulet 2008-05-27 13:06:57 EDT

Excellent ! Thanks

Comment 8 Philipe Mulet 2008-05-27 13:10:02 EDT

Added GenericTypeTest#test1335-1337.

This is a regression from 3.4m7 and 3.3.2 (introduced in 3.4rc1).

Comment 9 Philipe Mulet 2008-05-27 13:12:01 EDT

Created attachment 102193 [details]
Proposed patch

Missing trivial null check. Some parameterized type may have null argument, denoting non generic members of parameterized types.
One thing to ensure is that an unchecked warning is emitted.

Comment 10 Philipe Mulet 2008-05-27 13:14:52 EDT

Regression came from fix for bug 228291 missing one null check.
Also need to backport this fix to 3.3.x maintenance, where bug 228291 was already backported (with this NPE weakness).

Comment 11 Jerome Lanneluc 2008-05-27 13:26:51 EDT

Patch looks good: +1 for RC3

Comment 12 Olivier Thomann 2008-05-27 15:25:52 EDT

+1

Comment 13 Philipe Mulet 2008-05-28 04:21:26 EDT

Released for 3.4RC3.
Fixed

Comment 14 Philipe Mulet 2008-05-28 05:42:48 EDT

Released to 3.3.x maintenance.

Comment 15 David Audel 2008-05-30 11:02:39 EDT

Verified for 3.4RC4 using build I20080530-0100