I think it's the other way around. Someone used a custom workspace option to force them into the SAME workspace, when the default is the other way around. I could get some karma and fix this, I guess.
On Tue, Jul 26, 2011 at 5:42 PM, Dennis Lundberg <denn...@apache.org> wrote: > Hi > > I've investigated some odd build failures on the Solaris slaves for the > builds of maven-site-plugin. There are two builds of the same source > code, but with different versions of Maven. Here are the jobs: > > https://builds.apache.org/view/M-R/view/Maven/job/maven-site-plugin-3.x/ > > https://builds.apache.org/view/M-R/view/Maven/job/maven-site-plugin-3.x-m2/ > > These jobs are set up to build on 4 generic OSes, including the generic > Solaris label. After analyzing the logs I found that both jobs start by > doing this: > > Deleting > /zonestorage/hudson/home/hudson/hudson-slave/maven-site-plugin-3.x/workspace/jdk/JDK > 1.5 (latest)/label/Solaris/target > > So when an SVN change triggers a build, or rather two builds in this > case, the first job is abruptly interrupted by the second one when the > second one simply deletes the entire workspace of the first. > > However the job "maven-site-plugin-3.x-m2" should have it's own > workspace, called "maven-site-plugin-3.x-m2", on the Solaris slave > shouldn't it? > > Is this something that can be configured or is automagical? > > -- > Dennis Lundberg >
