char attributes are stored as CHAR(1)
StoreCharsAsNumbers now defaults to false on PostgreSQL 9 and later (OPENJPA-2971), MySQL (OPENJPA-3001), MariaDB (OPENJPA-3060), Apache Derby, H2, Microsoft SQL Server (OPENJPA-3033) and Oracle (OPENJPA-3028), as it already did on DB2.
Java char/Character attributes are then written and read as characters and map to CHAR(1) columns instead of INTEGER columns holding the code point.
A pre-existing CHAR(1) column, such as the one in the Jakarta Persistence TCK schema, now works on these databases.
On every database that stores chars as CHAR, a generated column for a char or Character attribute without an explicit length is now CHAR(1) instead of CHAR sized by CharacterColumnSize (255), which Derby rejected.
On Oracle this column is generated as CHAR(1 CHAR), so that it holds any character even though Oracle sizes columns in bytes by default (NLS_LENGTH_SEMANTICS=BYTE).
A byte sized CHAR(1) column, such as the one in the Jakarta Persistence TCK schema, only holds characters that take a single byte in the database character set; others, for example é in an AL32UTF8 database, fail with ORA-12899.
Existing wider CHAR columns keep working.
Schemas created by earlier releases on PostgreSQL, MySQL, MariaDB, Derby, H2, SQL Server or Oracle hold such attributes in INTEGER (on Oracle NUMBER) columns, which now fail validation or return wrong values.
Either migrate them to CHAR(1) or restore the previous mapping with an explicit StoreCharsAsNumbers=true, for example openjpa.jdbc.DBDictionary=derby(StoreCharsAsNumbers=true); an explicitly configured value is always respected.
Sybase and the other databases keep storing chars as numbers by default.
The reverse mapping tool follows the same setting.
On these databases it now maps a CHAR(1) column to a char or Character field instead of a String field.
Set StoreCharsAsNumbers=true while running the tool to keep generating String fields.