How specify nullable primitive array generic result type?

[msqr]

I'm using NullAway 0.13.1 in Java 27 with Spring 7, which has this API:

public interface PreparedStatementCallback<T extends @Nullable Object> {

	T doInPreparedStatement(PreparedStatement ps) throws SQLException, DataAccessException;

}

I have an implementation that can return a nullable byte[] result:

private static final PreparedStatementCallback<byte @Nullable []> DATA_CALLBACK = new PreparedStatementCallback<>() {

	@Override
	public byte @Nullable [] doInPreparedStatement(PreparedStatement ps)
			throws SQLException, DataAccessException {
		if ( ps.execute() ) {
			try (ResultSet rs = ps.getResultSet()) {
				if ( rs.next() ) {
					return rs.getBytes(1);
				}
			}
		}
		return null;
	}
};

NullAway reports this as an error:

error: [NullAway] method returns @Nullable, but superclass method org.springframework.jdbc.core.PreparedStatementCallback.doInPreparedStatement(java.sql.PreparedStatement) returns @NonNull
                public byte @Nullable [] doInPreparedStatement(PreparedStatement ps)
                                         ^
      (see http://t.uber.com/nullaway )
  1 error

I do not get this type of error using a nullable Object result like PreparedStatementCallback<@Nullable String>. Am I writing the byte @Nullable [] version incorrectly somehow? I feel like it should not generate this error, but am not sure.

[msridhar]

This might be a bug with anonymous classes and the diamond operator. Can you try writing this for the first line instead:

private static final PreparedStatementCallback<byte @Nullable []> DATA_CALLBACK = new PreparedStatementCallback<byte @Nullable []>() {

Does that remove the warning?

[msqr]

Thanks for the suggestion @msridhar, but unfortunately I still get the same warning.

[msridhar]

Ok, I'm very busy for the next few days but I'll try to take a look after that.

As another possible workaround, could you try declaring a static inner class (class Foo extends PreparedStatementCallback<byte @Nullable []> { ... }) instead of an anonymous class?