Serguei, The test fails rarely and I'm not able to reproduce the issue.
JDWP call JvmtiGetLoadedClasses::getLoadedClasses then store received classes locally. So nothing prevent these classes from being unloaded and I think it's worth to fix. -Dmitry On 2016-04-19 19:49, [email protected] wrote: > On 4/19/16 09:42, [email protected] wrote: >> Hi Dmitry, >> >> 144 // Clazz become invalid since the time we get the class list >> 145 // Skip this entry >> 146 if (error == JVMTI_ERROR_INVALID_CLASS) { >> 147 continue; >> 148 } >> 149 >> I'd like to better understand the root cause of this issue. >> Could you, please, give an idea why a class in this test becomes invalid? >> Do you think there is a potential in this test for a class to get >> unloaded? >> Or is it something about anonymous classes? > > One more question: > Any ideas on why these two tests started failing in November 2015 and > did not fail before? > > > Thanks, > Serguei > > >> >> Thanks, >> Serguei >> >> >> On 4/19/16 06:56, Dmitry Samersoff wrote: >>> Any reviewers? >>> >>> On 2016-04-13 19:24, Dmitry Samersoff wrote: >>>> Everybody. >>>> >>>> Please review a small fix. >>>> >>>> http://cr.openjdk.java.net/~dsamersoff/JDK-8143921/webrev.01/ >>>> >>>> I don't have a reproducer, so the fix is based on coredump analyzes. >>>> >>>> Tested with rbt >>>> >>>> -Dmitry >>>> >>>> >> > -- Dmitry Samersoff Oracle Java development team, Saint Petersburg, Russia * I would love to change the world, but they won't give me the sources.
