RFR: Fix shared tests to run without Shenandoah

[Roman Kennke]

Hi Aleksey,

> On 11/28/18 6:51 PM, Roman Kennke wrote:
>>> *) CriticalNativeStress.java: addition of "compact" + "-degen" does not make sense. Did you mean
>>> "passive" + "+degen"? It probably makes sense to split out for a separate backport.
>>
>> No, we already have passive+degen, I meand compact-degen. Also fixed a
>> number of other places like the above (added @test names, etc):
> 
> I don't understand. Every time we do "passive", we do "passive" with degen and without degen. Your
> change replaces one of the "passive" configs with "compact". Either add another "compact" @run line,
> or leave the whole thing alone? :)

Ugh. I thought it's 2x the same line. Zhengyu suggested it might have
intended to be compact. Let's simply revert this part.

>> http://cr.openjdk.java.net/~rkennke/fix-shared-tests/webrev.01/
> 
> Okay, what's up with TestFullGCCount test? Does it not run without Shenandoah built? It should. The
> @requires line there only protects from running the test with Shenandoah in explicit-concurrent
> mode, right? (While you are at it, you can probably change "vm.gc == "Shenandoah"" to
> "vm.gc.Shenandoah")

The way it was written was very confusing and it was wrong too. All the
@requires together would not select any test at all, except default GC.
The way I've split it up now should be good:

http://cr.openjdk.java.net/~rkennke/fix-shared-tests/webrev.02/

WDYT?

Roman