OPENDJ-2526: NegativeArraySizeException in OnDiskMergeImporter during OpenAM embedded DJ upgrade (DJ 2.6 -> 3.0) Fix: OffHeapBuffer's InputStream read() must return int between 0 and 255.

OPENDJ-2393: Possible index corruption In case of operation cancellation/exception, transactions are rolled back and with them the potential update of indexes' trusted state. That's why in case of transaction rollback we persist the indexes trusted state using another transaction. The number of keys contained in a VLV index is cached. This number can becomes incorrect when transaction updating index keys is aborted. The number of keys is now persisted using a sharded counted. Incorrect PDB RollbackException handling could have cause operations to abort rather than being retried. Refactoring: fix comments, removed dead code (obsoleted by id2children & id2subtree removal) Use specific compact encoding for shared-counter values.

OPENDJ-1802 Make ByteString methods more intentional ByteString.java: Renamed: - valueOf(byte[]) to valueOfBytes(byte[]) - valueOf(byte[], int, int) to valueOfBytes(byte[], int, int) - valueOf(char[]) to valueOfUtf8(char[]) - valueOf(CharSequence) to valueOfUtf8(CharSequence) - valueOf(Object) to valueOfObject(Object) - valueOf(int) to valueOfInt(int) - valueOf(long) to valueOfLong(long)

OPENDJ-1802 Make ByteStringBuilder methods more intentional ByteStringBuilder.java: Changed: - append(byte) to appendByte(int) - append(short) to appendShort(int) - append(int) to appendInt(int) - append(long) to appendLong(long) - append(Object) to appendObject(Object) - append(String) to appendUtf8(String) - append(char[]) to appendUtf8(char[]) - All remaining append() to appendBytes(). I changed appendByte() and appendShort() to accept int parameters. This avoids continuously casting, at the expense of a possible loss of the high bits. ByteStringBuilder.java: appendBytes(), appendUtf8(), appendObject()

OPENDJ-2349: fix deadlocks during subtree deletes and moddn The following changes were made: * ensure that cursors in JE use read committed isolation, not the default which is repeatable read * ensure that records are locked in a consistent order during update operations: dn2id, dn2uri, id2entry, id2childrenCount, and finally the attribute indexes * added support for deleting records directly from cursors * DN2ID: added renameSubtree() method in order to avoid breaking encapsulation when renaming DN keys during moddn operations * ID2Count: split apart updates to parent counts and total counts so that records can be updated in a well defined order. A consequence of this change is that the total count includes base entries * ID2Entry: added openCursor() method to facilitate implementation of subtree deletes and moddn operations.

OPENDJ-1791 Implement a tool like dbtest for the pluggable backend Implement debugging tool for pluggable backend on the same lines as dbtest, adding storage and backend level information. When dumping trees at seen from the backend, also try to print keys and values according to their meaning.

OPENDJ-2016: New on disk merge import strategy based on storage engine.