The Berkeley DB Java Edition Package: BDB JE 3.2.13 Change Log

Berkeley DB Java Edition 3.2.13 Change Log

The changes between 3.1.0 and 3.2.13 are described directly below. A full list of all changes since JE 2.1.34 is also available below.

Changes since 3.1.0

Log File On-Disk Format Changes:

JE 3.2.13 has moved to on-disk file format 5.
The change is forward compatible in that JE files created with release 3.1.0 and earlier can be read when opened with JE 3.2.13. The change is not backward compatible in that files created with JE 3.2.13 cannot be read by earlier releases. Note that if an existing environment is opened read/write, a new log file is written by JE 3.2.13 and the environment can no longer be read by earlier releases.[#15195]

New Features:

Added three new configuration properties to limit the amount of I/O performed by JE background activities such as cleaning and checkpointing: je.env.backgroundReadLimit, je.env.backgroundWriteLimit, je.env.backgroundSleepInterval. The amount of I/O performed by JE background threads sometimes spikes, which increases the response time (latency) of user operations that are performed in other threads. By setting these new properties, an application can limit the amount of background I/O in order to reduce the the response time of user operations. [#15158][Metalink 5839909.993]

Tuple bindings now support the java.math.BigInteger type. Like other tuple binding values, BigInteger values are sorted in natural integer order by default, without using a custom comparator. For details please see the Javadoc for:
```
com.sleepycat.bind.tuple.TupleInput.readBigInteger
com.sleepycat.bind.tuple.TupleOutput.writeBigInteger
com.sleepycat.bind.tuple.BigIntegerBinding 
```
[#15244]

The java.math.BigInteger class is now a DPL simple type. Values of this type are sorted in natural integer order by default, and BigInteger can be used as the type of a primary or secondary key field. DPL class evolution will automatically widen all other integer types (byte, char, short, int, long) to a BigInteger -- in an existing class, changing from any of these types to BigInteger is a compatible type change. [#15244]

The JE Deferred Write mode is now explicitly available in the DPL. Deferred Write mode may be set for an entire DPL store using StoreConfig.setDeferredWrite and all index changes may be flushed to disk using EntityStore.sync. [#15249]

When setting the Environment configuration property je.log.memOnly to true, a non-existent environment directory may now be specified and the je.lock file will not be written. Note that memory-only mode is primarily useful with automated tests to avoid file cleanup and it not recommended for production use. If you need a low durability mode for production use, see the Deferred Write mode (Database.setDeferredWrite). [#15255]

General Environment Changes

Fixed a bug where JE incorrectly handled the case where the environment consists of a single log file, there is a checksum error immediately following the file header, and the application repeatedly attempts to open the environment without heeding the DatabaseExceptions thrown by new Environment(). The first two open attempts correctly warned the application that the 00000000.jdb file should be moved away; the third attempt incorrectly threw a ClassCastException. [#15133][Metalink 5802188.992]

Fixed a bug where latches were not released when a LockNotGrantedException was thrown, resulting in a deadlock. This problem only applies to applications which use TransactionConfig.setNoWait(). One symptom of the problem are threads which are waiting on com.sleepycat.je.tree.IN.latch, in this manner:

"Thread-20" prio=5 tid=0x00538990 nid=0x1824400 waiting on condition [0xf181f000..0xf1820ab0]
at sun.misc.Unsafe.park(Native Method)
at java.util.concurrent.locks.LockSupport.park(LockSupport.java:118)
at java.util.concurrent.locks.AbstractQueuedSynchronizer.parkAndCheckInterrupt(AbstractQueuedSynchronizer.java:681)
at java.util.concurrent.locks.AbstractQueuedSynchronizer.acquireQueued(AbstractQueuedSynchronizer.java:711)
at java.util.concurrent.locks.AbstractQueuedSynchronizer.acquire(AbstractQueuedSynchronizer.java:1041)
at java.util.concurrent.locks.ReentrantReadWriteLock$NonfairSync.wlock(ReentrantReadWriteLock.java:342)
at java.util.concurrent.locks.ReentrantReadWriteLock$WriteLock.lock(ReentrantReadWriteLock.java:637)
at com.sleepycat.je.latch.Java5SharedLatchImpl.acquireExclusive(Java5SharedLatchImpl.java:91)
at com.sleepycat.je.tree.IN.latch(IN.java:329)
            ...

[#15142]

A minor change was made to the way that checkpoints are scheduled. As a result, checkpoints are performed at slightly more regular intervals than before. [#15169]

Fix a bug where a monitor (synchronized lock) deadlock could occur under low memory conditions. This problem was discovered during internal testing and has not been reported from the field. [#15176]

Improve log cleaning by improving the counting of obsolete record sizes. The utilization of a given log file is now known more accurately, which means that cleaning can be more efficient. The improvement impacts applications that insert both large and small records, and that delete or update them asymmetrically; for example, both large and small records are inserted but only small records are deleted. A new Environment configuration property was also added: je.cleaner.fetchObsoleteSize; if set to true, this property causes log records to be fetched in order to count their sizes accurately when a database is truncated or removed, and when running recovery. [#15195] Metalink: 5872288.993

Fixed a bug where the wrong comparator could be used in the first insertion of a duplicate record when the application has specified both a custom BtreeComparator and a custom DuplicateComparator. [#15197]

Fixed a race condition which could occur when there are concurrent Environment instantiations and closings. An application could erroneously get access to a invalidated (closed) Environment instance. [#15200]

Fixed a bug in JE's management of its internal utilization statistics which could result in an OutOfMemoryError. The problem was initially reported on this JE Forum thread. This bug sometimes caused OutOfMemoryError when both of the following conditions occurred: 1) A very large data set is used or the applications runs for a very long internal without closing the environment, and 2) the cache is large enough to fit all data records into memory. [#15214]

Fixed a bug which could result in a ConcurrentModificationException and this stack trace when running under Java 1.4.

Exception in thread "Thread-0" java.util.ConcurrentModificationException
at java.util.AbstractList$Itr.checkForComodification(AbstractList.java:448)
at java.util.AbstractList$Itr.next(AbstractList.java:419)
at com.sleepycat.je.latch.SharedLatchImpl.indexOf(SharedLatchImpl.java:246)
at com.sleepycat.je.latch.SharedLatchImpl.acquireExclusiveNoWait(SharedLatchImpl.java:131)
at com.sleepycat.je.tree.IN.latchNoWait(IN.java:352)
...

[#15231]

Fixed a bug which could occur when je.env.sharedLatches=true and the application is executing inserts, which manifested as this intermittent stack trace:

Exception in thread "Thread-0" java.lang.IllegalMonitorStateException
 at java.util.concurrent.locks.ReentrantReadWriteLock$Sync.tryReleaseShared(ReentrantReadWriteLock.java:275)
 at java.util.concurrent.locks.AbstractQueuedSynchronizer.releaseShared(AbstractQueuedSynchronizer.java:1180)
 at java.util.concurrent.locks.ReentrantReadWriteLock$ReadLock.unlock(ReentrantReadWriteLock.java:576)
 at com.sleepycat.je.latch.Java5SharedLatchImpl.releaseIfOwner(Java5SharedLatchImpl.java:227)
 at com.sleepycat.je.tree.IN.releaseLatchIfOwner(IN.java:405)
 at com.sleepycat.je.tree.Tree.searchSplitsAllowed(Tree.java:1701)
 at com.sleepycat.je.tree.Tree.findBinForInsert(Tree.java:3321)
 at com.sleepycat.je.tree.Tree.insert(Tree.java:2475) 
...

[#15233]

Database.sync now flushes the log to disk so that all changes are durable. Before, some changes might not be durable if a crash occurred after the sync and prior to the log buffer naturally being written as further changes were made or during the next checkpoint or Environment.close. [#15254]

Direct Persistence Layer (DPL)

Fixed a bug where the wrong key class was validated when calling EntityStore.getSecondaryIndex for a read-only store with a TO_MANY secondary. The application would see an IllegalArgumentException prefixed with "Wrong secondary key class:". [#15156]

Fixed a bug where the DPL was not working correctly with implicit transactions in an XA environment. The problem was reported in this JE Forum thread.. Now, when an XAEnvironment is used, XA transactions are implicit when a null transaction parameter, or no transaction parameter, is used in the DPL API. [#15206]

Fixed a potential performance problem with secondary indexes that are defined in entity subclasses. When an instance of an entity subclass is stored (via PrimaryIndex.put), we now create any secondary indexes that are defined on that subclass. Before, the secondary indexes were not created until EntityStore.getSubclassIndex was called. This was not technically a bug, since there was no loss of data and secondary indexes were created when they were used. But now the index is created as soon as possible, so that the index does not have to populated from scratch when it is first accessed. [#15247]

API Changes:

For debugging usage, DatabaseEntry.toString() now outputs the data as decimal bytes. The new DatabaseEntry.MAX_DUMP_BYTES (default 100) static field can be changed to limit the number of bytes that are output. [#15248]

Utility Changes:

com.sleepycat.je.util.DbCacheSize is now callable programmatically, in addition to being usable as a command line program. A new option, -r, for recalculating utilization is also available. DbCacheSize is not an officially supported utility and is not documented; for details, please see the DbCacheSize.java source file. [#15217]

Configuration, Documentation, Test and Build Changes:

Fixed Javadoc for EnvironmentMutableConfig.setTxnWriteNoSync(). The Javadoc for this method incorrectly stated that this type of durability would not survive JVM failures. setTxnWriteNoSync style durability does, in fact, survive JVM crashes.

The Javadoc for the com.sleepycat.persist.SecondaryIndex class was enhanced to discuss: One-to-Many versus Many-to-One for Related Entities; Key Placement with Many-to-Many for Related Entities; and Many-to-Many Versus a Relationship Entity.

The Java Collections Tutorial and associated example code was updated to remove mention of StoredIterator and the requirement to close stored iterators. As of JE 3.0.11, iterators over stored collections no longer need to be closed. [#15202]

Several sporadic test failures were fixed and the test suite ("ant test") should now run reliably on all platforms.

Full list of changes since JE 2.1.34

All changes from JE 3.0.11, 3.0.12, 3.1.0 and 3.2.13 are listed here.

For additional details, please see the documentation and Release Notes included in your download package or on our website.

Log File On-Disk Format Changes:

JE 3.2.13 has moved to on-disk file format 5.
The change is forward compatible in that JE files created with release 3.1.0 and earlier can be read when opened with JE 3.2.13. The change is not backward compatible in that files created with JE 3.2.13 cannot be read by earlier releases. Note that if an existing environment is opened read/write, a new log file is written by JE 3.2.13 and the environment can no longer be read by earlier releases.[#15195] (3.2.13)

JE 3.0.11 has moved to on-disk file format 4.

The change is forward compatible in that JE files created with release 2.1.30 and earlier can be read when opened with JE 3.0.11. The change is not backward compatible in that files created with JE 3.0.11 cannot be read by earlier releases. Note that if an earlier environment is opened read/write, a new log file is written by JE 3.0.11 and the environment can no longer be read by earlier releases. (3.0.11)

An upgrade problem was inadvertently introduced between JE 2.0.90, which was released in Nov 1, 2005 and the following JE 2.1.34 release in January 2006. The problem does not apply to you if you have an environment created with JE 2.0.90 or earlier, and have since successfully opened it with JE 2.1.30. Likewise, a 2.0.90 and earlier environment that has been closed cleanly with Environment.close() will not encounter this problem, nor will most environments that have more than 30 Mb of data. The problem manifests as the following stack trace. A workaround is described below.
```
com.sleepycat.je.DatabaseException: JE 3.0.11: searchResult=class com.sleepycat.
je.tree.IN nodeId=479 nEntries=0
       at com.sleepycat.je.tree.Tree.getParentBINForChildLN(Tree.java:1031)
...
Caused by: java.lang.ClassCastException: com.sleepycat.je.tree.IN
       at com.sleepycat.je.tree.Tree.getParentBINForChildLN(Tree.java:1021)
```
(3.0.11)

New Features:

A new Direct Persistence Layer adds a built-in Plain Old Java Object (POJO)-based persistent object model, which provides support for complex object models without compromises in performance. For an introduction to the Direct Persistence Layer API, see Getting Started with the Direct Persistence Layer.
JE 3.0.11 introduced the beta release of the DPL, while JE 3.1 is the GA release. Please see the Release Notes for more information. [#13894]

A new "deferred write" database mode provides better performance for temporary databases or databases which rapidly modify data by reducing the amount of I/O incurred. In a normal durable database, every new version of a data record is written to disk; in deferred write mode, JE avoids disk I/O until cache is exceeded or Database.sync() is called. See the Getting Started Guide, Chapter 3 (Databases) for more information. [#13936](3.0.11)

com.sleepycat.je.util.DbBackup is a new helper class which simplifies backups by managing JE background activity in an open environment. See the Getting Started Guide, Chapter 7 (Backing up and Restoring Berkeley DB Java Edition Applications) and the Javadoc for DbBackup for more information. [#14022](3.0.11)

The com.sleepycat.collections package is now fully compatible with the Java Collections framework. In previous releases, Collections.size() was not supported, and collection iterators had to be explicitly closed. These incompatibilities have been addressed to provide full interoperability with other Java libraries that use the Java Collections Framework interfaces. [#14154] [#12986](3.0.11)

The com.sleepycat.bind.tuple package now has bindings that correctly sort negative floating point numbers. See the following new APIs:

  SortedFloatBinding, SortedDoubleBinding,
  TupleInput.readSortedFloat, TupleInput.readSortedDouble,
  TupleOutput.writeSortedFloat, TupleOutput.writeSortedDouble.

[#14127](3.0.11)

Added a new method, com.sleepycat.je.Database.count(), which has been optimized to execute faster than the count obtained from a database scan. Note that the count may not be accurate in the face of concurrent update operations.(3.1.0)
This functionality is also exposed through the new method com.sleepycat.persist.EntityIndex.count() and is also used by com.sleepycat.collections.StoredMap.size() and com.sleepycat.collections.StoredCollection.size(). See the Javadoc for full details. [#14060](3.1.0)

A new method, EnvironmentConfig.setExceptionListener() provides a notification mechanism for asynchronous exceptions thrown by JE daemon threads. See the Javadoc for more details. [#15018](3.1.0)

Added three new configuration properties to limit the amount of I/O performed by JE background activities such as cleaning and checkpointing: je.env.backgroundReadLimit, je.env.backgroundWriteLimit, je.env.backgroundSleepInterval. The amount of I/O performed by JE background threads sometimes spikes, which increases the response time (latency) of user operations that are performed in other threads. By setting these new properties, an application can limit the amount of background I/O in order to reduce the the response time of user operations. [#15158][Metalink 5839909.993] (3.2.13)

Tuple bindings now support the java.math.BigInteger type. Like other tuple binding values, BigInteger values are sorted in natural integer order by default, without using a custom comparator. For details please see the Javadoc for:
```
com.sleepycat.bind.tuple.TupleInput.readBigInteger
com.sleepycat.bind.tuple.TupleOutput.writeBigInteger
com.sleepycat.bind.tuple.BigIntegerBinding 
```
[#15244] (3.2.13)

The java.math.BigInteger class is now a DPL simple type. Values of this type are sorted in natural integer order by default, and BigInteger can be used as the type of a primary or secondary key field. DPL class evolution will automatically widen all other integer types (byte, char, short, int, long) to a BigInteger -- in an existing class, changing from any of these types to BigInteger is a compatible type change. [#15244] (3.2.13)

The JE Deferred Write mode is now explicitly available in the DPL. Deferred Write mode may be set for an entire DPL store using StoreConfig.setDeferredWrite and all index changes may be flushed to disk using EntityStore.sync. [#15249] (3.2.13)

When setting the Environment configuration property je.log.memOnly to true, a non-existent environment directory may now be specified and the je.lock file will not be written. Note that memory-only mode is primarily useful with automated tests to avoid file cleanup and it not recommended for production use. If you need a low durability mode for production use, see the Deferred Write mode (Database.setDeferredWrite). [#15255] (3.2.13)

General Environment Changes:

Shared latches are now used on all internal nodes of the b-tree, which increases concurrency for many operations. Shared latches are only enabled when je.env.sharedLatches is set to true; the default value is false. The default may change in a future release. [#14168](3.0.11)

Fixed a bug where an internal latch was not taken during the deletion of log cleaning metadata or database metadata. [#14214](3.0.11)

Fixed a bug where internal node latches were not released in a corner case where the database has duplicate records and there are concurrent readers and writers. [#14230](3.0.11)

Calling com.sleepycat.je.Transaction.toString() after the transaction was closed resulted in a NullPointerException. This is now fixed. [#14349](3.0.11)

In a corner case requiring a certain pattern of inserts coupled with btree splits and checkpoints, a data record could be lost after an environment recovery following a non-clean shutdown. This has been fixed. [#14424](3.0.11)

The following NullPointerException could be seen when je.cleaner.threads > 0. This has been fixed.


java.lang.NullPointerException
	at com.sleepycat.je.cleaner.UtilizationProfile.getCheapestFileToClean(UtilizationProfile.java:201)
	at com.sleepycat.je.cleaner.FileSelector.selectFileForCleaning(FileSelector.java:205)
	at com.sleepycat.je.cleaner.FileProcessor.doClean(FileProcessor.java:201)
	at com.sleepycat.je.cleaner.FileProcessor.onWakeup(FileProcessor.java:143)

[#14431](3.0.11)

Cursor.getSearchBothRange() was incorrectly using the custom duplicate key comparator instead of the main btree key comparator, if the database was configured with a custom duplicate comparator. [#14464](3.0.11)

A user reported an OutOfMemoryError in a deployment using fairly large (512K) records. We finally tracked the problem down to the behavior of the standard java.util.zip.Adler32 class, which JE uses to checksum log records. Our research found JVM implementation notes which said that a JVM may temporarily disable garbage collection when executing the native code for Adler32.
One workaround for the user was to use -Dje.disable.java.adler32=true at runtime. This forces JE to its own, slower Adler32 implementation. It's an existing flag that was used to work around a Java 1.4 problem with the Adler32 implementation.
A second workaround provided in (3.1.0) is the je.adler32.chunkSize property. Setting this parameter will cause JE to pass chunks of the log record to the checksumming class so that the GC does not block. 0 means do not chunk.
Note that the Adler32 memory allocation issue is reported in this Sun bug entry to be fixed in various releases of the Sun JVM. [#14149] (3.1.0)

Invalidate the Environment when an Error such as java.lang.OutOfMemoryError is thrown during database operations, so that RunRecoveryException will be thrown if further operations are attempted. [#14300](3.1.0)

Fixed a bug where high concurrency updates on Java 1.6 could result in this exception from JE daemon threads: com.sleepycat.util.RuntimeExceptionWrapper: Channel closed, may be due to thread interrupt. [#14571](3.1.0)

Fixed a bug that caused a NullPointerException when Database.sync() is called on a newly created, empty deferred-write database. [#14781](3.1.0)

Fixed a bug that caused a NullPointerException in the cleaner thread when running under a JDK1.6 beta. The exception was not fatal, and looked like this:

<Cleaner name="Cleaner-1"/> caught exception: java.lang.NullPointerException
java.lang.NullPointerException
        at java.util.TreeMap.getEntry(TreeMap.java:324)
        at java.util.TreeMap.remove(TreeMap.java:580)
        at java.util.TreeSet.remove(TreeSet.java:259)
        at com.sleepycat.je.cleaner.FileSelector.selectFileForCleaning(FileSelector.java:236)
        at com.sleepycat.je.cleaner.FileProcessor.doClean(FileProcessor.java:201)
        at com.sleepycat.je.cleaner.FileProcessor.onWakeup(FileProcessor.java:143)
        at com.sleepycat.je.utilint.DaemonThread.run(DaemonThread.java:189)
        at java.lang.Thread.run(Thread.java:619)
Continuing

[#14877](3.1.0)

Fixed a bug that resulted in poor cache eviction management, and therefore poor performance when there is a high level of log cleaning in conjunction with a high level of cache eviction. This was a performance regression from the JE 2.1.30 release. One symptom of the problem is that the ratio of EnvironmentStats.getNBINSsStripped() to EnvironmentStats.getNNodesSelected() is lower in 3.0 than in 2.1.30. [#15062](3.1.0)

Fixed a bug that resulted in poor cache eviction management and could trigger OutOfMemoryException. A symptom of the problem is that EnvironmentStats.getCacheDataBytes() could become negative. [#15129]

Fixed a bug where JE incorrectly handled the case where the environment consists of a single log file, there is a checksum error immediately following the file header, and the application repeatedly attempts to open the environment without heeding the DatabaseExceptions thrown by new Environment(). The first two open attempts correctly warned the application that the 00000000.jdb file should be moved away; the third attempt incorrectly threw a ClassCastException. [#15133][Metalink 5802188.992] (3.2.13)

"Thread-20" prio=5 tid=0x00538990 nid=0x1824400 waiting on condition [0xf181f000..0xf1820ab0]
at sun.misc.Unsafe.park(Native Method)
at java.util.concurrent.locks.LockSupport.park(LockSupport.java:118)
at java.util.concurrent.locks.AbstractQueuedSynchronizer.parkAndCheckInterrupt(AbstractQueuedSynchronizer.java:681)
at java.util.concurrent.locks.AbstractQueuedSynchronizer.acquireQueued(AbstractQueuedSynchronizer.java:711)
at java.util.concurrent.locks.AbstractQueuedSynchronizer.acquire(AbstractQueuedSynchronizer.java:1041)
at java.util.concurrent.locks.ReentrantReadWriteLock$NonfairSync.wlock(ReentrantReadWriteLock.java:342)
at java.util.concurrent.locks.ReentrantReadWriteLock$WriteLock.lock(ReentrantReadWriteLock.java:637)
at com.sleepycat.je.latch.Java5SharedLatchImpl.acquireExclusive(Java5SharedLatchImpl.java:91)
at com.sleepycat.je.tree.IN.latch(IN.java:329)
            ...

[#15142] (3.2.13)

A minor change was made to the way that checkpoints are scheduled. As a result, checkpoints are performed at slightly more regular intervals than before. [#15169] (3.2.13)

Fix a bug where a monitor (synchronized lock) deadlock could occur under low memory conditions. This problem was discovered during internal testing and has not been reported from the field. [#15176] (3.2.13)

Improve log cleaning by improving the counting of obsolete record sizes. The utilization of a given log file is now known more accurately, which means that cleaning can be more efficient. The improvement impacts applications that insert both large and small records, and that delete or update them asymmetrically; for example, both large and small records are inserted but only small records are deleted. A new Environment configuration property was also added: je.cleaner.fetchObsoleteSize; if set to true, this property causes log records to be fetched in order to count their sizes accurately when a database is truncated or removed, and when running recovery. [#15195] Metalink: 5872288.993 (3.2.13)

Fixed a bug where the wrong comparator could be used in the first insertion of a duplicate record when the application has specified both a custom BtreeComparator and a custom DuplicateComparator. [#15197] (3.2.13)

Fixed a race condition which could occur when there are concurrent Environment instantiations and closings. An application could erroneously get access to a invalidated (closed) Environment instance. [#15200] (3.2.13)

Fixed a bug in JE's management of its internal utilization statistics which could result in an OutOfMemoryError. The problem was initially reported on this JE Forum thread. This bug sometimes caused OutOfMemoryError when both of the following conditions occurred: 1) A very large data set is used or the applications runs for a very long internal without closing the environment, and 2) the cache is large enough to fit all data records into memory. [#15214] (3.2.13)

Fixed a bug which could result in a ConcurrentModificationException and this stack trace when running under Java 1.4.

Exception in thread "Thread-0" java.util.ConcurrentModificationException
at java.util.AbstractList$Itr.checkForComodification(AbstractList.java:448)
at java.util.AbstractList$Itr.next(AbstractList.java:419)
at com.sleepycat.je.latch.SharedLatchImpl.indexOf(SharedLatchImpl.java:246)
at com.sleepycat.je.latch.SharedLatchImpl.acquireExclusiveNoWait(SharedLatchImpl.java:131)
at com.sleepycat.je.tree.IN.latchNoWait(IN.java:352)
...

[#15231] (3.2.13)

Fixed a bug which could occur when je.env.sharedLatches=true and the application is executing inserts, which manifested as this intermittent stack trace:

Exception in thread "Thread-0" java.lang.IllegalMonitorStateException
 at java.util.concurrent.locks.ReentrantReadWriteLock$Sync.tryReleaseShared(ReentrantReadWriteLock.java:275)
 at java.util.concurrent.locks.AbstractQueuedSynchronizer.releaseShared(AbstractQueuedSynchronizer.java:1180)
 at java.util.concurrent.locks.ReentrantReadWriteLock$ReadLock.unlock(ReentrantReadWriteLock.java:576)
 at com.sleepycat.je.latch.Java5SharedLatchImpl.releaseIfOwner(Java5SharedLatchImpl.java:227)
 at com.sleepycat.je.tree.IN.releaseLatchIfOwner(IN.java:405)
 at com.sleepycat.je.tree.Tree.searchSplitsAllowed(Tree.java:1701)
 at com.sleepycat.je.tree.Tree.findBinForInsert(Tree.java:3321)
 at com.sleepycat.je.tree.Tree.insert(Tree.java:2475) 
...

[#15233] (3.2.13)

Database.sync now flushes the log to disk so that all changes are durable. Before, some changes might not be durable if a crash occurred after the sync and prior to the log buffer naturally being written as further changes were made or during the next checkpoint or Environment.close. [#15254] (3.2.13)

API Changes:

Iterators returned by StoredCollection.iterator() no longer need closing. The original StoredIterator is now available using the new StoredCollection.storedIterator() method. See the Release Notes for more information on collection API changes. [#12986](3.0.11)

Database.entry now supports equals() and hashCode(). [#13778](3.0.11)

Key comparators can now be stored as serialized objects using the new DatabaseConfig.setBtreeComparator(Comparator) and DatabaseConfig.setDuplicateComparator(Comparator) methods. This allows a single comparator class to handle more than one key format. [#14422](3.0.11)

SecondaryDatabase.delete has been corrected to delete all keys (and their associated primary records) in the presence of duplicates. Before, only the first of a duplicate set (and its associated primary) was deleted. Please note this behavior change. [#14506](3.0.11)

Added com.sleepycat.je.jca.ra.JEConnection.removeDatabase(), com.sleepycat.je.jca.ra.JEConnection.truncateDatabase()to support database truncation and removal for JCS users. [#15105]

For debugging usage, DatabaseEntry.toString() now outputs the data as decimal bytes. The new DatabaseEntry.MAX_DUMP_BYTES (default 100) static field can be changed to limit the number of bytes that are output. [#15248] (3.2.13)

Utility Changes:

DbVerify will now verify all databases in the environment if the "-s" flag is omitted. In previous versions, "-s" was a mandatory parameter. A new option, "-c" has been added which will check log cleaner metadata. [#14239](3.0.11)

com.sleepycat.je.util.DbCacheSize is now callable programmatically, in addition to being usable as a command line program. A new option, -r, for recalculating utilization is also available. DbCacheSize is not an officially supported utility and is not documented; for details, please see the DbCacheSize.java source file. [#15217] (3.2.13)

Configuration, Documentation, and Build Changes:

Ant 1.6.0 or later is required to build JE from the provided sources.(3.0.11)

The directions in the installation page on how to compile a non-debug jar were updated.(3.0.11)

The "Writing Transactional Applications" guide has been corrected to make it clear that nested transactions are not yet supported in JE. [#14079](3.0.11)

A variety of problems noted by our faithful readers have been fixed.
- The examples now specify a character encoding when constructing a String object.
- The examples use the base interface type when possible, rather than a specific collection class.
- ExampleDatabaseLoad.loadFile() now closes its FileInputStream.
- Two examples in the Concurrency chapter of the "Writing Transactional Applications" manual were updated to correctly return after a commit.
(3.0.11)

In response to popular demand, the JE jarfile in a JE package is now named je-<version>.jar instead of just je.jar. For example, the jar in this release is named je-3.1.0.jar. [#14669](3.1.0)

Fixed Javadoc for EnvironmentMutableConfig.setTxnWriteNoSync(). The Javadoc for this method incorrectly stated that this type of durability would not survive JVM failures. setTxnWriteNoSync style durability does, in fact, survive JVM crashes.

The Javadoc for the com.sleepycat.persist.SecondaryIndex class was enhanced to discuss: One-to-Many versus Many-to-One for Related Entities; Key Placement with Many-to-Many for Related Entities; and Many-to-Many Versus a Relationship Entity.

The Java Collections Tutorial and associated example code was updated to remove mention of StoredIterator and the requirement to close stored iterators. As of JE 3.0.11, iterators over stored collections no longer need to be closed. [#15202] (3.2.13)

Several sporadic test failures were fixed and the test suite ("ant test") should now run reliably on all platforms.

Direct Persistence Layer:

Fixed a bug where "UnsupportedOperationException: Class evolution is not supported" was thrown when performing class enhancement. This problem would occur when creating a store without enhancing classes and then using enhancement later, or when creating a store with class enhancement and then not using enhancement later. [#14651](3.0.12)

Fixed a bug where "ArrayIndexOutOfBoundsException: 2147483646" was thrown when updating a record with a secondary key. The problem occurred when a single instance was used for both the primary key and secondary key fields. [#14665](3.0.12)

Fixed a bug where "IllegalArgumentException: Class is not persistent" was thrown for a class that was correctly made persistent using a proxy class. The problem occurred when one proxy class referenced another. [#14665](3.0.12)

Before, if a custom key class (composite key class) was defined but the @KeyField annotation was not included, an UnsupportedOperationException would occur when trying to put() a record, and the exception message was not meaningful. Now, an IllegalArgumentException will occur when getting the index, with a meaningful exception message. [#14673](3.0.12)
Class evolution is now implemented in the Direct Persistence Layer. For more information please see the release notes and the package specification for the com.sleepycat.persist.evolve package. Also note that if you have deployed a store using the JE 3.0.12 beta release, before changing your persistent classes for the first time you must open and close your store using the JE 3.1 release; see the release notes for more information. [#14762]

Fixed a bug that prevented opening a read-only store when a primary key sequence was used. [#14941]

Fixed a bug that prevented using EntityJoin with read-uncommitted (for example, CursorConfig.READ_UNCOMMITTED). The fix is a change to JE secondary databases that allows a partial DatabaseEntry to be passed as the data parameter to a search method, when using read-uncommitted. [#14966]

Fixed a bug in EntityStore.getStoreNames() which resulted in this stack trace:

java.lang.StringIndexOutOfBoundsException: String index out of range: -9
	at java.lang.String.substring(Unknown Source)
	at com.sleepycat.persist.impl.Store.getStoreNames(Store.java:209)
	at com.sleepycat.persist.EntityStore.getStoreNames(EntityStore.java:186)

[#14978]

Fixed a bug that caused an IllegalArgumentException when calling EntityStore.getSecondaryIndex for a @SecondaryKey that was declared on a superclass of the entity class. [#14978]

The EntityStore.getSubclassIndex method has been added. This method should be used instead of getSecondaryIndex when the secondary key field is in a subclass of an entity class. The index returned by getSubclassIndex is used to access only entities of that subclass. Note that in the previous release, calling getSecondaryIndex in such a situation would fail, and it was necessary to call registerClass as a workaround. [#15102]

A bug was fixed that caused an IllegalArgumentException with the message "Not a key class". This occurred when using a TO_MANY secondary index with a composite key class. The workaround for the bug was to call registerClass for the key class, and this is no longer necessary. [#15103]

An optional verbose option was added to the ClassEnhancer to print the name of each class file that is enhanced. See ClassEnhancer for more information. [#14680](3.0.12)

Class evolution is now implemented in the Direct Persistence Layer. For more information please see the release notes and the package specification for the com.sleepycat.persist.evolve package. Also note that if you have deployed a store using the JE 3.0.12 beta release, before changing your persistent classes for the first time you must open and close your store using the JE 3.1 release; see the release notes for more information. [#14762](3.1.0)

Fixed a bug that prevented opening a read-only store when a primary key sequence was used. [#14941](3.1.0)

Fixed a bug that prevented using EntityJoin with read-uncommitted (for example, CursorConfig.READ_UNCOMMITTED). The fix is a change to JE secondary databases that allows a partial DatabaseEntry to be passed as the data parameter to a search method, when using read-uncommitted. [#14966](3.1.0)

Fixed a bug in EntityStore.getStoreNames() which resulted in this stack trace:

java.lang.StringIndexOutOfBoundsException: String index out of range: -9
	at java.lang.String.substring(Unknown Source)
	at com.sleepycat.persist.impl.Store.getStoreNames(Store.java:209)
	at com.sleepycat.persist.EntityStore.getStoreNames(EntityStore.java:186)

[#14978](3.1.0)

Fixed a bug that caused an IllegalArgumentException when calling EntityStore.getSecondaryIndex for a @SecondaryKey that was declared on a superclass of the entity class. [#14978](3.1.0)

The EntityStore.getSubclassIndex method has been added. This method should be used instead of getSecondaryIndex when the secondary key field is in a subclass of an entity class. The index returned by getSubclassIndex is used to access only entities of that subclass. Note that in the previous release, calling getSecondaryIndex in such a situation would fail, and it was necessary to call registerClass as a workaround. [#15102](3.1.0)

A bug was fixed that caused an IllegalArgumentException with the message "Not a key class". This occurred when using a TO_MANY secondary index with a composite key class. The workaround for the bug was to call registerClass for the key class, and this is no longer necessary. [#15103](3.1.0)

Fixed a bug where the wrong key class was validated when calling EntityStore.getSecondaryIndex for a read-only store with a TO_MANY secondary. The application would see an IllegalArgumentException prefixed with "Wrong secondary key class:". [#15156] (3.2.13)

Fixed a bug where the DPL was not working correctly with implicit transactions in an XA environment. The problem was reported in this JE Forum thread.. Now, when an XAEnvironment is used, XA transactions are implicit when a null transaction parameter, or no transaction parameter, is used in the DPL API. [#15206] (3.2.13)

Fixed a potential performance problem with secondary indexes that are defined in entity subclasses. When an instance of an entity subclass is stored (via PrimaryIndex.put), we now create any secondary indexes that are defined on that subclass. Before, the secondary indexes were not created until EntityStore.getSubclassIndex was called. This was not technically a bug, since there was no loss of data and secondary indexes were created when they were used. But now the index is created as soon as possible, so that the index does not have to populated from scratch when it is first accessed. [#15247] (3.2.13)

Workaround for upgrade issue between JE 2.0.90 and JE 2.1.30

If you are upgrading from a release of JE that is 2.0.90 or earlier, then you should ensure that you call Environment.close() prior to upgrading. If you do not close the environment cleanly, and you have very recently created a new database, you may see the following exception:

com.sleepycat.je.DatabaseException: JE 3.0.11: searchResult=class
com.sleepycat. je.tree.IN nodeId=479 nEntries=0
      at com.sleepycat.je.tree.Tree.getParentBINForChildLN(Tree.java:1031)
...
Caused by: java.lang.ClassCastException: com.sleepycat.je.tree.IN
       at
       com.sleepycat.je.tree.Tree.getParentBINForChildLN(Tree.java:1021)

The data in the environment was not corrupted, and the environment can be successfully opened with the following steps which remove any new files create by JE 3.0.11, and then call Environment.close() prior to upgrading.

To remove any data created by JE 3.0.11:

First take a directory listing of your Environment and examine all the *.jdb files written at the time of the attempted upgrade. For example:

-rwx------   1 cwl      other      19629 May  5 11:26 00000000.jdb
-rw-r--r--   1 cwl      other       3014 May  8 11:36 00000001.jdb <-- new file

Examine all the new files to verify that they use file version 4 by doing:
```
java com.sleepycat.je.util.DbPrintLog -h <envhome> -s <lastfilenumber>
```
where <lastfilenumber> is a hex number of the last file (e.g. '-s 0x1' in the case above). You should see output like this:
```
<DbPrintLog>
<entry lsn="0x1/0x0" type="FileHeader/0" prev="0x0" size="24" cksum="800588407"><FileHeader num="0x1" lastEntryInPrevFileOffset="0x4c87"
logVersion="0x4" time=" 2006-05-08 12:50:20.63"/></entry>

           ...
```
The first file header record shows a logVersion of 0x4. These files were written during the attempted upgrade and can be deleted.

To close the environment cleanly after any JE 3.0.11 data has been removed:

Once the version 4 log files have been deleted, perform a checkpoint using the old (pre-2.0.90) version of JE. You can do this by calling Environment.close() from your application code or by invoking the unadvertised DbRunAction utility:
```
java -cp "je-1.7.1.jar:je-3.0.11.jar" com.sleepycat.je.util.DbRunAction -h <envdir>
```
Since DbRunAction is not present in releases prior to 2.1.30, you should include the latest release in the classpath, but after the old release. That will run a checkpoint on the old environment using the old version of JE. Several lines of output will be produced. Verify that the last log file was written to by taking a directory listing.

Finally, perform another checkpoint using the newer version of JE. After this, the environment will be fully upgraded.
```
java -cp "je-3.0.11.jar" com.sleepycat.je.util.DbRunAction -h <envdir>
```