Sometimes qulice-checkstyle tests hang with 100% cpu usage

[krzyk]

Sometimes when I run qulice-checkstyle it stops working here:

[INFO] 
[INFO] --- maven-surefire-plugin:2.16:test (default-test) @ qulice-checkstyle ---
[INFO] Surefire report directory: /home/sg0891848/dev/tpc2/qulice/qulice/qulice-checkstyle/target/surefire-reports
[INFO] parallel='all', perCoreThreadCount=true, threadCount=4, useUnlimitedThreads=false, threadCountSuites=0, threadCountClasses=0, threadCountMethods=0

-------------------------------------------------------
 T E S T S
-------------------------------------------------------

And CPU usage jumps to 100%. I assume something is wrong with running tests in parallel.

[davvd]

we'll find someone to do this task, soon

[davvd]

thanks for the ticket, your account was topped for 15 mins, payment 43563700

[davvd]

@longtimeago the task is yours,please go ahead

[longtimeago]

@yegor256 sorry, can't reproduce and have no idea why it happens. Please, assign someone else.

[yegor256]

@davvd we need to assign someone else here

[davvd]

@davvd we need to assign someone else here

@yegor256 got it, someone else will be assigned soon

[davvd]

@palyuga please go ahead, it's your task now, keep this in mind, and don't hesitate to ask any technical questions you may have

Budget here is 30 mins (keep this principle in mind)

[palyuga]

@krzyk could you please capture a thread dump when it happens again?

[krzyk]

@palyuga sure

[palyuga]

@krzyk The issue is reproduced on my Windows machine once, a thread dump pointed to this JVM problem: http://bugs.java.com/bugdatabase/view_bug.do?bug_id=7183450
If you have Windows too, the solution is to add "-Djava.net.preferIPv4Stack=true" to your JAVA_OPTS
But I can't check it, as I can't reproduce the issue again (miracle)

[krzyk]

@palyuga I'm using Linux most of the time (unless there is a Windows bug that needs fixing), and this bug was reported under Linux for sure

[palyuga]

@krzyk in this case i'm still waiting for a thread dump from you

[krzyk]

@palyuga Here it is https://gist.github.com/krzyk/a5d71e7ce7a0435d8be2

I recreated it without proxy (I did about 14 builds and it happened)

[palyuga]

@krzyk As we can see there are no deadlocks and the first two threads seems to be just hanging in java.io.FileInputStream.readBytes(Native Method) (the similar IO problem was in my case, when InputStream.available() returns false). Doesn't look like a problem with quilice code. I think I can't help here/ Of course you could report an issue to: http://jira.codehaus.org/browse/MSHARED about checking availableness of an InputStream before reading. We can close it or I'll ask to assign to someone else, as you wish.

[krzyk]

@palyuga OK, so in that case please report this issuue to MSHARED and add a comment about this problem in our project. I'm not sure where, because it is not test related @yegor256 do you have some suggestions?

[yegor256]

@palyuga I think @krzyk is right, you should report this bug to MSHARED and put the link to our source code. I think the best place is the README.md file, in the "how to contribute" section. Just say there that due to a bug in MSHARED (here is the link) you may experience build hanging, sometimes, etc. That will be it for this task.

[palyuga]

@krzyk I've took another look to the problem. I was wrong, theese two processes are waiting for output from STDERR and STDOUT of a child process spawned by Runtime.exec(), where maven-surefire-plugin tries to execute unit tests. But the process hangs with no output.

I've found a discussion about such issue with a similar thread dump: http://comments.gmane.org/gmane.comp.jakarta.turbine.maven.user/135026
So it looks like migration to maven-surefire-plugin v. 2.18 could shed light on roots of this problem. At least we should receive an exception instead of silent hanging.

I've created an issue in jcabi-parent (jcabi/jcabi-parent#22), so lets wait for migration.

[davvd]

@palyuga this task is not yours any more (took too long), please stop. I'll give you something else... -60 to your rating, your total score is +90

[palyuga]

@davvd we are waiting for jcabi/jcabi-parent#22 here

[palyuga]

@krzyk now surefire-plugin 2.18.1 is used. Could you check again, does the problem occur now?

[krzyk]

@palyuga has the new parent been released and added to qulice pom?

[palyuga]

@krzyk oops, i've rushed with it. Lets wait for release)

[krzyk]

@palyuga thanks, I haven't seen this problem in a while