Confusing disparity in behaviour from javac

[Alan Bateman]

On 26/10/2016 12:35, Andrew Dinn wrote:

> :
>
> If the problem is that this package has already been bagged by the JDK
> then should javac not report that?
When you compile code in the unnamed module then it reads java.base but 
java.base doesn't export jdk.internal to anyone so there isn't an issue 
at compile-time. However, at run-time you have a split package issue by 
way of deploying jdk.internal.Test on the class path (for reliability 
reasons, the JDK disallows splitting a package between the class path 
and the modules in the boot layer, even if the module in the boot layer 
is not defined to the application class loader). If jdk.internal.Test is 
in a module in a custom layer or even loaded by a custom class loader 
(unnamed module) then there won't be an issue. Looking at it now then we 
should minimally improve the launcher handling for the corner case that 
is the main class on the class path in the same package as a package in 
a JDK module.

-Alan