RFR 8181080/10, Refactor several sun/net shell tests to plain java tests

[Felix Yang]

Alan and Chris,


On 2017/6/6 18:12, Alan Bateman wrote:
>
>
> On 06/06/2017 10:27, Felix Yang wrote:
>> Addressed all comments. Please review the updated patch:
>>
>> http://cr.openjdk.java.net/~xiaofeya/8181080/webrev.01/
>>
> Can you run the tests in sun/net/www/protocol/jrt with 
> `-vmoption:--limit-modules=java.desktop` to see how it behaves? I 
> think, but not certain, that this is the reason why we introduced a 
> shell to run that test.
All 3 tests will fail with error missing java.compiler module.
I also attempted with --limit-modules=java.se. Then other resources will 
fail with "java.lang.RuntimeException: Need to run with --limit-modules 
java.base". vmoptions will be added  to @run entry by jtreg, then lead 
the confusion.

SO IMO, the first proposal to fork a java process with constant 
arguments is probably better than using @run entry.
http://cr.openjdk.java.net/~xiaofeya/8181080/webrev.00/test/sun/net/www/protocol/jrt/OtherResourcesTest.java.html

Do you agree?

Thanks,
Felix
>
> -Alan