Jenkins Pipeline job can't find script due to @tmp path being created

Question

I am writing a pipeline job that will call another script to execute. The Jenkinsfile and script exist in the same directory and yet the job fails to find the script to run.

This is the relevant bit of script;

stage ('Update') {
    try {
        dir('jenkins/pipeline/update-jenkins-plugins-ppln') {
            sh 'ls -l'
            sh 'update-plugins.sh'
        }
}

which returns the following error;

[update-jenkins-plugins-ppln] Running shell script
+ ls -l
total 8
-rw-r--r-- 1 jenkins jenkins 2441 Dec 20 09:34 Jenkinsfile
-rwxr-xr-x 1 jenkins jenkins  506 Dec 19 14:06 update-plugins.sh
[Pipeline] sh
[update-jenkins-plugins-ppln] Running shell script
+ update-plugins.sh
/var/lib/jenkins/workspace/update-jenkins-plugins-ppln/jenkins/pipeline/update-jenkins-plugins-ppln@tmp/durable-11cefdd0/script.sh: 2: /var/lib/jenkins/workspace/update-jenkins-plugins-ppln/jenkins/pipeline/update-jenkins-plugins-ppln@tmp/durable-11cefdd0/script.sh: update-plugins.sh: not found

As you can see, the pathing I'm using is correct because according to the ls the file I need update-plugins.sh is in the directory I've pathed to. For some reason though, when actually searching for the script Jenkins is adding @tmp/durable-8d48734f/script.sh onto the path.

Various troubleshooting:

• I read that you have to checkout the branch again even if you're already checking it out to get the Jenkinsfile, so I am.
• I have ssh'd into the Jenkins box to check and yes, the script is there.

Why is Jenkins adding the @tmp bit, and is there a way to prevent this behavior?

Solution 1

I guess your pwd is not in PATH so you have to call it like this: sh './update-plugins.sh'

Solution 2

Have you tried using the jenkins workspace environment variable WORKSPACE (absolute path of the workspace)? With that your line would look something like this:

sh '${WORKSPACE}/jenkins/pipeline/update-jenkins-plugins-ppln/update-plugins.sh'

Solution 3

I was having the same issue. I think @Oren technically answered your question about why Jenkins creates this tmp space, but I can share some info about how I solved it.

Basically, my Jenkins host symlinks bin/sh to dash; not bash. So, using a POSIX-compliant shell script solved the issue for me.

For example, I was trying to use shopt -s extglob to do some pattern matching:

stage {
    def shellCommand = $/ "rm -rf ! (_data|_includes|_plugins|Gemfile|_config.yml|page-builder)"/$
    sh(returnStdout: true, script: shellCommand).trim()
}

Since dash doesn't support extglob, replacing that with a POSIX-compliant find command worked:

stage {
    sh('find . -regextype posix-extended -not -regex ".*includes.*|.*data.*|.*plugins.*|.*config.yml|.*Gemfile.*"')
} 

Comments

[MisterStrickland]

Hi. Just wondering if you found a solution to your problem.

[Alex]

@Phil Nope, it's still an issue

[Alex]

Tried that already; same error + ./update-plugins.sh /var/lib/jenkins/workspace/update-jenkins-plugins-ppln@tmp/d‌​urable-88ee0ed4/scri‌​pt.sh: 2: /var/lib/jenkins/workspace/update-jenkins-plugins-ppln@tmp/d‌​urable-88ee0ed4/scri‌​pt.sh: ./update-plugins.sh: not found

[izzekil]

Then that's strange. I can only suggest you to try sh ' bash update-plugins.sh' or to calculate full path using pipeline func pwd(), concatenate and call the script by absolute path.

[Markus Weninger]

For me it worked using './' in the beginning of the script's name.

[overexchange]

Yes, /bin/sh -> /bin/dash in my case

[Alexander Samoylov]

For we adding of './' also worked. I think it may depend on the version of plugins and/or Jenkins. I use Jenkins 2.164.3, Pipeline Plugin 2.6, Durable Task Plugin 1.29.