Egads!  You're right - so why the heck did my Rampart build fail then??

Thanks for the catch; I'm going to revert this out of the 1.5 branch
pending further discussion.

--Glen

Andreas Veithen wrote:
> Glen,
> 
> Note that you merged the change from trunk to the 1.5 branch in r727458...
> 
> Andreas
> 
> On Tue, Jan 6, 2009 at 15:48, Glen Daniels <[email protected]> wrote:
>> Hi Amila,
>>
>> The Rampart build was broken with the Axis2 1.5 branch, and Andreas
>> clued me in that the problem was that change 727413 (adding the -lcmn
>> option) hadn't been made on the branch.  After taking a look at that
>> change, I'm going to belatedly -1 it for Axis2.
>>
>> Here's my reasoning.  First and foremost, the change was implemented not
>> by changing the JavaUtils.xmlNameToJavaIdentifier() method (for instance
>> by adding another override with a flag), but by simply not calling it
>> unless the lcmn option was specified.  This is seriously broken - it
>> means that if we have a WSDL method name "Foo-Bar", we'll now by default
>> codegen this into a Java method "Foo-Bar()"... which of course WON'T
>> COMPILE.  Second, almost as important, by changing the /default/
>> behavior instead of adding an option to invoke different behavior, we've
>> broken all of our users' codegen builds if they happen to use XML
>> identifiers which are uppercase or contain Java-incompatible characters.
>>  The fact that we needed to change the Rampart build to avoid a new
>> breakage should have been a dead giveaway - there are times when
>> incompatible change is inevitable and ok, but this was not one of them.
>>
>> Please revert this change immediately, and let's reconsider what you
>> were trying to do with it on the list.
>>
>> Also, this points out the fact that nowhere in our tests do we have
>> anything that tries various combinations of XML-to-Java to confirm that
>> WSDL2Java does the right thing with XML names like "Foo-Bar", etc.  I'll
>> open a JIRA on that.
>>
>> Thanks,
>> --Glen
>>

Reply via email to