Kohsuke Kawaguchi wrote:
> Paul Sandoz wrote:
>
>> Since the JAXB FI serializer is very simple and inherits mostly from
>> the JAXB StAX serializer we could generate the documents using the
>> FI StAXDocumentSerializer and the JAXB marshaller. The chance of a bug
>> is small.
>
>
> OK. I'll test marshal -> FI -> unmarshal route then.
>
> Note that this won't test the case where FI gives us a base64 data
> whereas JAXB expects it to be text.
>
Yes, that is a little more tricky. Perhaps we can have a simple test
with two JAXB beans one that maps an element to byte[] and another that
maps to char[].
>> For belts and braces we could test that the infoset parsed from the
>> XML output and the FI output of JAXB marhsalling are equivalent. I
>> think this is important anyway in the future when the change the
>> serializer. This might also show up inconsistencies in the binary data
>> to character conversion between JAXB and FI.
>
>
> OK. JAXB uses DOM comparator to check the equivalence. So I'll check the
> equivalence of marshal -> UTF8 -> DOM and marshal -> FI -> DOM.
>
Ah interesting does the production of the DOM ensure that attributes
ordered etc. i.e. a normalized canonical DOM?
Is the code in the JAXB CVS? maybe it is something better to use than
what we currently have for FI.
Paul.
--
| ? + ? = To question
----------------\
Paul Sandoz
x38109
+33-4-76188109