Regression test java/util/jar/JarFile/TurkCert.java incorrectly changes Locale settings

[Kelly O'Hair]

Jonathan Gibbons wrote:
> Daniel D. Daugherty wrote:
>>
>>
>> Jonathan Gibbons wrote:
>>> Alan Bateman wrote:
>>>> Pavel Tisnovsky wrote:
>>>>> Hi,
>>>>>
>>>>> the first statement (Locale.setDefault(new Locale("TR", "tr"));) in 
>>>>> regression test
>>>>> java/util/jar/JarFile/TurkCert.java
>>>>>
>>>>> changes locale settings on global level (i.e. in currently running 
>>>>> JTreg/JVM) which causes that other test fails, for example 
>>>>> sun/java2d/cmm/ColorConvertOp/ConstructorsNullTest/ConstructorsNullTest 
>>>>>
>>>>>
>>>>> I'm able to do patch for this test. Should I post webrev here or to 
>>>>> another OpenJDK mailing list?
>>>>>
>>>>> regards
>>>>> Pavel Tisnovsky
>>>> core-libs-dev is probably the best place. It's only recently that 
>>>> we've been running these tests in samevm mode and this is one that 
>>>> should probably run in othervm mode.
>>>>
>>>> -Alan.
>>> Alan,
>>>
>>> An alternative solution is to remember the locale before changing it 
>>> and resetting it when done.
>>>
>>> -- Jon
>>
>> Does that work when executing tests in parallel?
>>
>> Dan
>>
> jtreg does not support running tests in parallel for exactly this sort 
> of reason -- certainly not in samevm mode, where stdout/stderr are 
> already multiplexed between tests.
> 
> -- Jon

But will changing the locale impact the samevm testing?
Should any testcase that sets the locale be othervm?

I assume that if completely separate jtreg processes are running at
the same time that this locale setting has no impact. It's not
a system wide locale change is it?

-kto