Java generic inference is unsound

[Andrei Alexandrescu]

Depending how you look at this, it is either a success or a failure of 
formal methods in designing programming languages.

http://twofoos.org/content/java-type-system-holes/

I didn't know about this. Loopholes in Java have come up now and then, 
but I thought they'd fixed them all. An older one (2001) was this:

http://fpl.cs.depaul.edu/Problem.java

Igarishi, Pierce and Wadler had shown (OOPSLA 1999) that Java's core 
type checking is correct, but have not included a host of generics in 
the proof. Apparently that's still a problem.


Andrei