I am looking at JOHNZON-177 and I have noticed that we seem to be mapping primitives inconsistently inside the same class. MappingParserImpl.toObject() and MappingParserImpl.readObject() both contain special logic for handling longs and ints. Why wouldn't we want to handle these types consistently regardless of where they're used (fields or the root type)?
I have included code to handle the overflow/underflow case with a nice error message and that works fine when I added tests to the DefaultMappingTest from the JSON-B module. However, I found that MapperTest seems to already have similar tests for invalid long/int values. As I trace the logic, it doesn't hit my new code.
