[ http://jira.codehaus.org/browse/CONTINUUM-667?page=comments#action_74334 
] 
            
Kenney Westerhof commented on CONTINUUM-667:
--------------------------------------------

The tests fail on windows because the process (the forked 'mvn') is not 
terminated due to window's bad process handling.
The batch file is terminated, but it's child processes aren't.

2 solutions:

* brett proposed we add something like -Dcontinuum.buildId=X to the commandline 
and grep the output
  using pstools or some other windows 'ps' and 'kill' capable tool to terminate 
the jvm started from batch files.

* we don't use the batchfiles but call java directly. All we need for this is 
the location of maven (using $M2_HOME)
  or grab the output of 'mvn --print-commandline <normal commandline>' which 
would print the commandline calling
  java; then we run that directly. That process is easily killed.

> need a way to limit the mount of time a unit test runs for
> ----------------------------------------------------------
>
>                 Key: CONTINUUM-667
>                 URL: http://jira.codehaus.org/browse/CONTINUUM-667
>             Project: Continuum
>          Issue Type: Bug
>          Components: Core system, Web interface
>    Affects Versions: 1.0.3
>            Reporter: james strachan
>         Assigned To: Kenney Westerhof
>            Priority: Blocker
>             Fix For: 1.1
>
>
> We get lots of builds hanging in CI due to test cases with timing issues on 
> different platforms. (You'd be amazed how easy that is to do). We need a way 
> to put in a maximum amount of time (say 5 minutes) for a single unit test - 
> or a max time per entire test run (say 1 hour) so that we can be alerted if a 
> test hangs.
> If a test hangs it normally hoses the entire CI server. The worst thing is 
> this makes the entire CI useless since it does no more builds and we don't 
> know its failed.
> I understand there are complications in killing stuff in Java. (Though 
> java.lang.Process.destroy() works for most things). One option is to just 
> have a status XML file we can poll for on a canoncial URL
> http;//serverhost/module/status.xml
> then we can look to see if any of the builds are blocked and if they are we 
> can get nagios or something to reboot the box / restart the CI server etc.

-- 
This message is automatically generated by JIRA.
-
If you think it was sent incorrectly contact one of the administrators: 
http://jira.codehaus.org/secure/Administrators.jspa
-
For more information on JIRA, see: http://www.atlassian.com/software/jira

        

Reply via email to