Proxy config not quite right if mixing typeNames and types in proxyhint

[aclement]

The order of types when creating a JDK Proxy matters, and it would normally respect the order in a ProxyHint expressed in either types= or typeNames=. But if you mix those and use a combination of typeNames= or types= the ordering is not as easily controllable in the resultant proxy-config.json.

[sdeleuze]

Maybe for proxies we should enforce and document types or typeNames not both?

[aclement]

We could enforce sticking to one or the other but it feels nasty when you have the .class reference but are forced to write it as a string. After sleeping on it I had thought a simple order=N attribute in the hint annotation would work too.

[sdeleuze]

As discussed, I am more in favor of enforcing and documenting types or typeNames, and raising an error message if both are defined. For TypeHint this would have been blocking, but for ProxyHint I think this is not common enough to deserve a more complex and not so obvious from a programming model perspective mechanism.

[aclement]

There is a predictable ordering but it will be subtle without hiding something in the javadocs that no-one will ever read. I've done the quick thing, we have other battles to fight. You'll get an early failure if mixing the styles in a proxy hint. As a bonus it will even ll tell you which type contained the hint ;)