[
https://issues.apache.org/jira/browse/HBASE-18712?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16144747#comment-16144747
]
stack commented on HBASE-18712:
-------------------------------
bq. I don't remember which branch #833 was for.
precommit is master; i.e. jdk8 so permgen config does nothing.
The post you cite for the UseLoopPredicate flag has zero upvotes. It is last in
a long line of possible reasons for surefire timeout (i.e. anything that could
crash a JVM). The post begins: "This *could* also happen due to a totally
different issue.."... and continues.... " I found out that this *could* be
happening because of a JDK bug which *could* have caused this regression...."
[emphasis mine]. The flag looks like a work around for a bug introduced long
ago in early jdk7 long since solved. The bug report cited in the url you add is
from 2015.
I'm not against trying stuff but this seems like a long-shot w/ little by way
of substantiation. We probably want to avoid adding odd-looking configs just
getting our basic test suite passing.
> Specify MaxPermSize for surefire tests
> --------------------------------------
>
> Key: HBASE-18712
> URL: https://issues.apache.org/jira/browse/HBASE-18712
> Project: HBase
> Issue Type: Bug
> Reporter: Ted Yu
> Assignee: Ted Yu
> Attachments: 18712.v1.txt, 18712.v2.txt
>
>
> Currently hbase-surefire.argLine doesn't specify MaxPermSize for the test
> run(s).
> This sometimes resulted in mvn build prematurely exiting, leaving some large
> tests behind.
> The tests would be deemed timed out.
> As indicated by the following post:
> https://stackoverflow.com/questions/23260057/the-forked-vm-terminated-without-saying-properly-goodbye-vm-crash-or-system-exi
> We should specify large enough MaxPermSize so that mvn build doesn't end
> prematurely.
--
This message was sent by Atlassian JIRA
(v6.4.14#64029)
