RFR(XS): 8211287: ClassPathTests.java fails due to "Unable to map MiscData shared space at required address."

Calvin Cheung calvin.cheung at oracle.com
Tue Oct 2 21:52:15 UTC 2018


Hi Ioi,

Thanks for your suggestion.

Updated webrev:
     http://cr.openjdk.java.net/~ccheung/8211287/webrev.01/

thanks,
Calvin

On 10/2/18, 2:38 PM, Ioi Lam wrote:
> I think we should do this
>
>
> 98 try {
>   99                     m.invoke(tests);
> 100 } catch (InvocationTargetException ite) {
> 101 if (ite.getCause() instanceof SkippedException) {
> throw ite.getCause();
> 103 } else {
> 104 throw ite;
> 105 }
> 106 }
>
>
> That way the SkippedException will be recognized by the jtreg 
> framework and the test will be tagged as "skipped" (instead of "passed").
>
> Thanks
> - Ioi
>
> On 10/02/2018 02:27 PM, Calvin Cheung wrote:
>> Hi Jiangli,
>>
>> Thanks for taking a look.
>>
>> On 10/2/18, 1:42 PM, Jiangli Zhou wrote:
>>> Hi Calvin,
>>>
>>> With the change, would mach5 still be able to recolonize the test as 
>>> being skipped when testing doesn't perform due to mapping failure?
>> The test will still be run but it will be considered "passed" because 
>> of the SkippedException which is being thrown from CDSTestUtils.java.
>> The change is a standard practice on inspecting the checked exception 
>> from invoking a method reflectively.
>>
>> thanks,
>> Calvin
>>>
>>> Thanks,
>>>
>>> Jiangli
>>>
>>>
>>> On 10/2/18 1:09 PM, Calvin Cheung wrote:
>>>> bug: https://bugs.openjdk.java.net/browse/JDK-8211287
>>>>
>>>> webrev: http://cr.openjdk.java.net/~ccheung/8211287/webrev.00/
>>>>
>>>> The method to be invoked sometimes throws the SkippedException 
>>>> usually in case the archive cannot be mapped at the required 
>>>> address successfully. The fix is to catch the 
>>>> InvocationTargetException and inspecting its cause.  If the cause 
>>>> is SkippedException, consider the test has passed and output some 
>>>> message. Otherwise, rethrow the exception.
>>>>
>>>> Testing:
>>>>     hs-tier1,2,3
>>>>     Also ran the test 40 times on the windows host from which the 
>>>> failure was observed.
>>>>
>>>> thanks,
>>>> Calvin
>>>
>


More information about the hotspot-runtime-dev mailing list