RFR(xs) 8231097: [TESTBUG] runtime/cds/appcds/DirClasspathTest.java can fail with a mapping error

[Calvin Cheung]

Thanks David.

Calvin

On 9/24/19 3:57 PM, David Holmes wrote:
> Looks good to me too.
>
> Thanks,
> David
>
> On 25/09/2019 6:40 am, Calvin Cheung wrote:
>> Hi Ioi,
>>
>> Thanks for your review.
>>
>> Those two instances of the "isUnableToMap" check are in the 
>> TestCommon.testDump() methods.
>>
>> I've removed them and will do more testing before push.
>>
>> thanks,
>>
>> Calvin
>>
>> On 9/24/19 12:20 PM, Ioi Lam wrote:
>>> Hi Calvin,
>>>
>>> Looks good.
>>>
>>> I think we can also remove two instances of this check in the 
>>> TestCommon.dump(...) methods.
>>>
>>>             if (isUnableToMap(output)) {
>>>                 throw new SkippedException(UnableToMapMsg);
>>>             }
>>>
>>> No need for a new webrev.
>>>
>>> Thanks
>>> - Ioi
>>>
>>> On 9/24/19 10:15 AM, Calvin Cheung wrote:
>>>> bug: https://bugs.openjdk.java.net/browse/JDK-8231097
>>>>
>>>> webrev: http://cr.openjdk.java.net/~ccheung/8231097/webrev.00/
>>>>
>>>> Several CDS tests in tier-4 failed when running with 
>>>> -Dtest.dynamic.cds.archive=true due to mapping of the base archive 
>>>> has failed. The tests should be more tolerant on mapping failure by 
>>>> throwing a SkippedException.
>>>>
>>>> The failures were observed on windows and are hard to reproduce. 
>>>> The change was tested by simulating the mapping failure in 
>>>> FileMapInfo::map_region by temporary adding the following code at 
>>>> the beginning:
>>>>
>>>> if (1) {
>>>>     fail_continue("Unable to map ReadOnly shared space at required 
>>>> address");
>>>>     _memory_mapping_failed = true;
>>>>     return NULL;
>>>> }
>>>>
>>>> Also ran mach5 tier4 tests.
>>>>
>>>> thanks,
>>>>
>>>> Calvin
>>>>
>>>