jtreg classpath setup inconsistency

[Jonathan Gibbons]

Chris,

See comment in bug.

-- Jon

On 10/03/2016 08:28 AM, Chris Plummer wrote:
> Hello,
>
> Can a jtreg expert please look at 
> https://bugs.openjdk.java.net/browse/JDK-8167001. I'd like to know why 
> invoking a jtreg test from the command line is executed differently 
> than doing so from test/Makefile. This results in different handling 
> of the classpath passed to the test, and is causing a test to fail as 
> a result. I have a solution to fix the test (pass "true" as the first 
> argument to ProcessTools.createJavaProcessBuilder), but I want to 
> better understand why the classpath isn't consistently set up.
>
> In summary, invoking from the command line will cause the classpath to 
> be setup be exporting CLASSPATH before running the test. This results 
> in the classpath being setup for java subprocesses that are created. 
> When invoking from test/Makefile, the classpath is setup by running 
> with -classpath. In this case the classpath is not implicitly passed 
> to the subprocesses, but can be explicitly passed by passing true for 
> the first argument to ProcessTools.createJavaProcessBuilder (which 
> defaults to true for hotspot jtreg tests, but false for jdk jtreg tests).
>
> thanks,
>
> Chris