OPENJPA-1376 commits, it uncovered a few issues, some old and some new.
1. Backward compatibility - Previous, openjpa relies on the native sequence to provide the next value based on the sequence's INCREMENT BY setting. Now that openjpa manages the next sequence value and allocation size, if the db sequence is manually created that does not match the application's increment and allocation size, committing a new entity may result a constraint violation exception when new sequence value is requested that crossed the allocation boundary. This behavior is indirectly documented in the migration consideration section in the user manual, but it would be better if this can be handled automatically or at least to provide a compatibility option to honor previous behavior.
2. Initial value always start with the same value (default to 1) - In NativeJDBCSeq, the _initial, _increment and _allocate are used for the CREATE SEQUENCE generation for SynchronizeMapping function. For existing db sequence, next returns whatever managed by the db sequence. After the commit, next will first return the_nextValue value in NativeJDBCSeq until _maxValue is reached, then it go to db to get the sequence's next value, which may be ways off from the current _maxValue.
3. _nextValue not handled proper when crossing _allocate boundary - When NativeJDBCSeq's _nextValue exceeds the _maxValue boundary, it queries db's sequence for the next incremental value and assign directly as the next return value. It works for incrementBy = 1 but for some value (e.g. 3), the difference of the before and after cross boundary values may NOT be the _increment value.