The production WAR would have to be built manually.
If you are building the WAR file manually, then I’m not sure why copying the appropriate deployment descriptor into WEB-INF temporarily wouldn’t be part of the same Ant task.
I thought you were trying to use our deployment tool, in which case I would say that your suggestion here:
So the best option might be to have them all copied in the WEB-INF (via an ANT-Task) with the file being flagged as “uncommittable” there (via the .cvsignore)
would probably work. Then the deployment tool would WAR your entire WebRoot recursively (including all the extra deployment descriptors in the WEB-INF dir) and deploy the WAR file to whichever app server you were executing the deploy for via the tool. Then the app server would uncompress the WAR file and deploy the app and only read the deployment descriptors it cared about.
While this sounds sloppy, its probably fine for development. You could use your existing Ant task to build a “clean” WAR file for production deployment that only included the descriptors for the appropriate app server you were deploying too.
Am I understanding your request correctly? If not, please let me know.