Open pshipton opened 1 month ago
Issue Number: 20307 Status: Open Recommended Components: comp:test, comp:vm, comp:gc Recommended Assignees: pshipton, jasonfengj9, chengjin01
@hzongaro @r30shah fyi. I haven't seen this failure before (although I don't always do the best job of triaging harmony failures since there are a bunch of machine issues).
I see this one failing 1/50 times (build_info.php?build_id=79356), need to investigate this (May be 31-bit z/OS issue only??), I will do initial triage before getting one of us to work on it.
Internal build
[zOS S390 64bit Compressed Pointers] 80 Load_Level_2.harmony.5mins.Mode610 -Xcompressedrefs -Xjit -Xgcpolicy:gencon
fyrec510
It's not 32-bit only and it failed without -Xjit:count=0 so I've added it to the next milestone.
Thanks Peter for sharing, I think this will help nailing down the cause. @matthewhall2 Let's look into this together.
Internal build [zOS S390 64bit Compressed Pointers] 80 Load_Level_2.harmony.5mins.Mode610 fyrec60h
Internal build [Linux S390 64bit] 80 Load_Level_2.harmony.5mins.Mode109
Internal build [zOS S390] 80 Load_Level_2.harmony.5mins.Mode187
Internal build [zOS S390] 80 Load_Level_2.harmony.5mins.Mode112
Internal build [zOS S390] 80 Load_Level_2.harmony.5mins.Mode110
Internal build [zOS S390] 80 Load_Level_2.harmony.5mins.Mode187
Looking into this, given the assertion failure, to effectively investigate, I need to reproduce this reliably. I did picked up the last failing build from previous comment and launched grinder of 100x I can not reproduce this in 100x. Rerunning the s390x Linux with latest build 100x internal link . This one does not fail as well. I will check if I can reproduce this on specific set of machines (Though I checked couple of last reported failures - I see the grinder did launch some jobs on those machines so that seems unlikely).
It does fail in the builds fairly frequently. If it can't be reproduced via grinder, is there anything that can be done to the test to gather more information when it does fail?
It does fail in the builds fairly frequently. If it can't be reproduced via grinder, is there anything that can be done to the test to gather more information when it does fail?
I am looking into that direction. In past, I have dealt with such intermittent issue by changing the test to invoke a debug agent that can keep running the test to help identify the method causing the issue.
http://vmfarm.rtp.raleigh.ibm.com/job_output.php?id=96186470 [zOS S390] 80 Load_Level_2.harmony.5mins.Mode112
Internal build [zOS S390] 80 Load_Level_2.harmony.5mins.Mode112
-Xgcpolicy:gencon -Xjit:count=0 -Xnocompressedrefs
fyrec50yFailed 1/30 in a grinder (fyrec50z)