attached mail follows:
Hi All!
I would like to summarize the issues that I have come across
so far regarding the resolution for Bug 6468446. The bug fix
that has been put in b13 currently requires code changes in the
"sunlib" module of the JAXRPC tests. The changes has to be
made either in the Impl classes or in the WSDL used
for the tests such that the parameter ordering in the Impl
matches the ordering that is generated in the SEI.
It looks like the wrong ordering seems to be generated for method
calls that have more than 4 or 5 parameters. The rest of the
methods which use lesser number of parameters are not
affected. It so happens that the majority of the test methods
in the JAXRPC test base have less than 5 parameter with the
exception of the sunlib tests that have up to 7 parameters.
So at this point it looks like only the sunlib module alone will
be needing changes.
Vivek has mentioned that the ordering is unpredictable when
using HashSet and will vary with the kind of jdk used. So there
could potentially be a large number of real world applications
at customer sites that are affected by this wrong parameter ordering
and would require significant effort in changing their code base
as soon as they migrate to jdk 1.6 or when they apply the bug
fix with jdk 1.5.
The SQE jaxrpc tests now work fine with the code changes to
the sunlib module's WSDL files in b13. The fix is not yet checked
in to 9.1.
thanks
Balaji