[ 
https://issues.apache.org/jira/browse/SUREFIRE-1255?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15445169#comment-15445169
 ] 

Tibor Digana commented on SUREFIRE-1255:
----------------------------------------

The main difference between Versions {{2.18.1}} and {{2.19}} is ping feature 
and sending NOP command to the forked process which was not possible in 
{{2.18.1}} and earlier. The point is that the forked VM kills itself as soon as 
two commands NOP are not received because this would mean the main Maven 
process was shutdown via {{CTRL-C}} or killed.
And so if you override the communication channel in forked JVM, the VM 
activates this mechanism and kills itself which means the main process is 
waiting for command {{BOOTERCODE_NEXT_TEST}} which will never arive and there 
the main process hangs.

> Surefire 2.19.1 hangs before starting test execution
> ----------------------------------------------------
>
>                 Key: SUREFIRE-1255
>                 URL: https://issues.apache.org/jira/browse/SUREFIRE-1255
>             Project: Maven Surefire
>          Issue Type: Bug
>          Components: Maven Surefire Plugin, process forking
>    Affects Versions: 2.19.1
>            Reporter: Maximilian Michels
>            Assignee: Tibor Digana
>            Priority: Critical
>
> Seeing the same error as SUREFIRE-1193 in Apache Flink builds after upgrading 
> from 2.18.1 to 2.19.1. No errors, builds just gets stuck at the beginning of 
> tests with no log output from the test itself, e.g. 
> https://s3.amazonaws.com/archive.travis-ci.org/jobs/137118454/log.txt
> After a couple of minutes Surefire reports 
> {noformat}
> [ERROR] Failed to execute goal 
> org.apache.maven.plugins:maven-surefire-plugin:2.19.1:test 
> (integration-tests) on project flink-scala-shell_2.10: ExecutionException The 
> forked VM terminated without properly saying goodbye. VM crash or System.exit 
> called?
> [ERROR] Command was /bin/sh -c cd 
> /home/travis/build/mxm/flink/flink-scala-shell/target && 
> /usr/lib/jvm/java-8-oracle/jre/bin/java -Xms256m -Xmx800m -Dmvn.forkNumber=1 
> -XX:-UseGCOverheadLimit -jar 
> /home/travis/build/mxm/flink/flink-scala-shell/target/surefire/surefirebooter372520169616801479.jar
>  
> /home/travis/build/mxm/flink/flink-scala-shell/target/surefire/surefire8229439069544382018tmp
>  
> /home/travis/build/mxm/flink/flink-scala-shell/target/surefire/surefire_26373613144387982724tmp
> [ERROR] -> [Help 1]
> {noformat}
> We have a a couple of test classes that suffer from this problem. Tests don't 
> read from STDIN or replace it. Switching back to 2.18.1 eliminates the 
> problem (Ran over 100 builds). It seems to be a race condition because it 
> only occurs every once in a while, i.e. ~ 2 out of 10 builds on Travis. I 
> haven't been able to reproduce the problem locally.
> More logs: 
> https://s3.amazonaws.com/flink-logs-us/travis-artifacts/mxm/flink/849/849.1.tar.gz



--
This message was sent by Atlassian JIRA
(v6.3.4#6332)

Reply via email to