RFR: 8253750: use build-stable default seed for Utils.RANDOM_GENERATOR

[Joe Darcy]

I agree with Roger that this change should *not* go forward since it 
would have the effect of reducing test coverage.

Regression tests that use randomness should using the "randomness" jtreg 
label and should output the seed value used so a failing result can be 
replicated.

Thanks,

-Joe

On 10/2/2020 10:11 AM, Roger Riggs wrote:
> On Fri, 2 Oct 2020 15:39:01 GMT, Igor Ignatyev <iignatyev at openjdk.org> wrote:
>
>>> Hi all,
>>>
>>> could you please review the patch which updates `jdk.test.lib.Utils` to use md5 hash-sum of `java.vm.version` property
>>> as default seed for `Utils.RANDOM_GENERATOR`?
>>> from JBS:
>>>> using the same seed for all runs of a build will make it possible (easier) to compare results from different test runs
>>>> (e.g. on different platforms, w/ different flags) and consequently will make test results analysis easier.
>>> the patch also updates `RandomGeneratorTest` test, so it expects now that the same values are generated if no seed is
>>> provided.
>>> testing: ✅ tier1
>> ping?
> Is this really a good idea?  The purpose of using random numbers is to get broader coverage on multiple runs.
> If the seed only changes once per version (6 months), that reduces test coverage.
> At least for dev submitted runs, I would like to be different for every build (unless overridden).
>
> -------------
>
> PR: https://git.openjdk.java.net/jdk/pull/391