Hi,
On Sunday, 2013-03-03, Chao Shi wrote:
> Sorry for the break.
> I think this is related with my patch (which changes the synchronization
> mechanism when killing a pipeline). The cause is a racing problem on
> CrunchJobControl#runnerState. I will file a JIRA ticket to fix it.
> I'm interested in how Jenkins find it. I run "mvn install" to run
> intergration tests and do observe the stacktrace, but it doesn't cause
> CancelJobsIT to fail. Does anyone know how Jenkins do it?
There may be a race condition in Crunch that should be fixed, but
judging from past data, most broken builds are due to Apache's shaky
infrastructure or bugs in Jenkins itself (yes, this is a problem that
makes CI almost pointless). I have re-triggered the build to see if
the problem persists.
The build job doesn't do anything exciting. Basically, it fetches
changes from our git repo, cleans the workspace using "git clean -fdx"
and then runs "mvn clean install". I think it also uses a fresh Maven
repository for each build.
Regards,
Matthias
|