When rebuilding a single module of a multi-module project, I have an environment script which happens to fail (because $WORKSPACE is different from a normal top-level build, but any failure should do). When this environment script fails, the job naturally fails too, but the Xvfb plugin fails to terminate the Xvfb process it started, leading to problems in subsequent builds.
I'm having trouble coming up with an easy minimal reproducer, because I think it only happens when a multi-module build is involved. I suspect the failure has to happen in a Jenkins plugin (like Environment Script) too: a normal build failure is not enough.
This is the log from the rebuild which failed:
Building remotely on jenkins-slave-f20-kvm-7 (amd64-Fedora 20 amd64-Fedora-20 Fedora-20 Fedora kvm amd64) in workspace /home/fedora/jenkins/workspace/zanata-server-github-pull-requests/functional-test
Xvfb starting$ Xvfb :1 -screen 0 1024x768x24 -fbdir /home/fedora/jenkins/xvfb-2015-04-28_11-25-15-3566362744314223079.fbdir
[functional-test] $ /bin/sh -xe /tmp/org.zanata:functional-test6279959920485140886.sh
/tmp/org.zanata:functional-test6279959920485140886.sh: line 1: etc/scripts/allocate-jboss-ports: No such file or directory
FATAL: Unable to execute script, return code 1
Started calculate disk usage of build
Finished Calculation of disk usage of build in 0 seconds
Started calculate disk usage of workspace
Finished Calculation of disk usage of workspace in 0 seconds