8156989: Trailing empty element in classpath ignored

[Mandy Chung]

> On May 16, 2016, at 11:53 AM, Alan Bateman <alan.bateman at oracle.com> wrote:
> 
> 
> Martin's MineField test has been excluded for some time, one reason is that it exercised -Xbootclasspath/p and so doesn't work with JDK 9. Jon is cleaning up this test via JDK-8156989 and lo behold, it finds a corner case.
> 
> The corner case is where the last element of the class is the empty string. Long standing behavior is to treat empty elements on the class path as the working directory. The issue crept in because the implementation uses String.split and is easily fixed:
>  http://cr.openjdk.java.net/~alanb/8156989/webrev/
> 

This looks fine.

> For tests then Jon will be pushing the updated MineField to jdk9/dev soon. We need to go over the new options and get javac and the runtime consistent. I think we want it so that empty elements in the new options are ignored, leaving class path for legacy reasons.

Yes we should add new tests to ensure the javac and runtime consistency of the new options.
Mandy