I am currently enhancing the release process of our projects on Jenkins (1.430).
Current release jobs
Today, for one specific project, we have one job dedicated to the Release process.
The complete procedure is the following:
- The developer who is in charge of the release changes manually the version of all the pom.xml files (in fact using
mvn versions:set -DnewVersion=2.0
) to get rid of the -SNAPSHOT
.
- Then, he creates a tag in SVN (http://my-svn-repo/project/tags/V_2_0 for example).
- Once this tag has been created, he logs on our Jenkins server, and starts a Release build.
- This build will ask him which tag he wants to use for the build. The job is configured as a Parameterized build, with the parameter List Subversion tags.
- Jenkins will then build the artifacts from this tag, and deploy them on our Nexus instance.
- Once this is done, the developer set the pom.xml versions to the new development version (i.e.
2.1-SNAPSHOT
).
The advantage of this method is that I have only Jenkins job, as the build will rely only on a tag.
However, this procedure involves too many human interventiosns (changes of the pom.xml, commits, tags, etc.).
New release jobs
Now, I use the Maven release plugin.
I've created a job that asks three information to the user who launches the build:
- the version of the release (parameter
releaseVersion
of the release plugin);
- the version of development, after the release (parameter
developmentVersion
of the release plugin);
- the name of the tag (parameter
tag
of the release plugin).
This job works fine, except for one point: the job is based on the trunk or on a branch in SVN.
This means that if I have 2 branches (in addition to the trunk), I will need to create 3 release jobs: one per branch.
One idea to keep the best of the two worlds (i.e. using mvn release, but keeping 1 release job) it to add a build parameter that will ask the user for the path of the trunk / branch.
So instead of setting http://my-svn-repo/project/trunk
(or http://my-svn-repo/project/branches/BRANCH_V1
) in the job configuration, I will set http://my-svn-repo/project/$FROM_BRANCH
, and ask the user to input the FROM_BRANCH
parameter.
The problem with this solution is that the user will have to input either trunk
or branches/BRANCH_Vx
, which may lead to errors.
Ideally, I would love to have a build parameter that let me the choice of the branch (including trunk), as the parameter List Subversion tags exist for the choice of tags...
So my question: is there a better way to configure one Jenkins job that can work on all the branches?
Thanks.
Edit: I found the Validating String Jenkins plugin that can be interesting to ensure that the value defined by the user respects some regular expression. That is helpful in my case...
See Question&Answers more detail:
os 与恶龙缠斗过久,自身亦成为恶龙;凝视深渊过久,深渊将回以凝视…