I'm building a Maven project which has half a dozen modules.
I'm fine with importing it myself using either Maven or Ivy, but other teams would like to use those jars as well, but their practice is to commit the jars and source jars to version control.
I'd like to generate a zip/tar assembly of all modules and their sources which they can use however they like.
I've read Maven Assembly Plugin: Including Module Binaries but I'm shy of using it because:
The linked FAQ entry returns a 404;
I need to manually specify all modules.
Is there an alternative?
Update: I've tried using the built-in assembly descriptors
mvn assembly:assembly -DprojectModulesOnly=true
mvn assembly:assembly
and both failed with
[INFO] ------------------------------------------------------------------------
[ERROR] BUILD ERROR
[INFO] ------------------------------------------------------------------------
[INFO] Failed to create assembly: Error creating assembly archive bin: You must set at least one file.
right after all the module builds have run.
I think you're on the right lines, the moduleSets options of the assembly plugin handles what you're after.
If you're looking for some useful documentation, the Module Selection section of the Maven book covers it quite thoroughly, including how to configure includes and excludes, handle binaries and sources, and exclude external dependencies.
I had this problem, for me, the solution was NOT put / at the beginning of your <fileset><directory>
If you do that will work on Windows, not on Unix/Linux!
<fileSet>
<directory>src/main/</directory>
<outputDirectory></outputDirectory>
<includes>
<include>VERSION</include>
</includes>
</fileSet>
works whereas
<fileSet>
<directory>/src/main/</directory>
<outputDirectory></outputDirectory>
<includes>
<include>VERSION</include>
</includes>
</fileSet>
causes
[ERROR] Failed to execute goal org.apache.maven.plugins:maven-assembly-plugin:2.3:single (execution-pluggin-assembly) on project test3: Failed to create assembly: Error creating assembly archive assembly: You must set at least one file. -> [Help 1]
Have a look at the How to use assembly:assembly using predefined descriptor ids. I think the bin and src pre-defined descriptor files are what you need.
Sounds like you need a build-server of some kind. I was at JavaZone 2009 this week and looked at Hudson CI http://hudson-ci.org/
The server will create the artifacts you or other teams can use/download.
Related
I am having an issue running tests using tycho due to an incorrect dependency resolution that, somehow, is placing the the old Google Collections .jar on the classpath and not the Guava one, despite the fact that at no point in any of my poms do I specify a dependency on collections (only guava).
My unit tests fail due to things like NoSuchMethodError (ImmutableList.copyOf), NoClassDefFoundError (Joiner), which I pretty much narrowed down to 'finding the wrong jar'. These same tests pass when ran manually in Eclipse.
Here is the relevant part of the pom:
<dependencies>
<dependency>
<groupId>com.google.guava</groupId>
<artifactId>guava</artifactId>
<version>14.0.1</version>
</dependency>
...
</dependencies>
The phrase 'google collections' appears no where. The only other repository I specify is:
<repositories>
<repository>
<id>helios</id>
<layout>p2</layout>
<url>http://download.eclipse.org/releases/helios</url>
</repository>
</repositories>
My plugin imports 'com.google.common.base' and 'com.google.common.collect' as imported packages. I have my own bundled version of Guava 14 in my workspace for debugging, but in the POM I elect to not use my local module.
I followed Sean Patrick Floyd's answer on this question (JUnit throws java.lang.NoSuchMethodError For com.google.common.collect.Iterables.tryFind), and had my test throw an exception with the location of the .jar that the Iterables class was loaded from. It spat back out:
java.lang.IllegalArgumentException: file:/C:/Documents and Settings/Erika Redmark/.m2/repository/p2/osgi/bundle/com.google.collect/0.8.0.v201102150722/com.google.collect-0.8.0.v201102150722.jar
This is where I am now stuck. This google-collections jar is coming seemingly out of no where, and I don't know how to stop it. As long as it is being resolved, my unit tests will fail. How can I stop Tycho from trying to get the old Google Collections?
Just to clarify, this has not stopped building and deployment; the plugin update site is on an CI platform and we have been able to install the plugin on different Eclipse IDEs, so this issue is only affecting the tests.
Please let me know if additional information is needed.
The plug-in com.google.collect 0.8.0.v201102150722 is part of the Helios p2 repository that you have configured in your POM. This means that this plug-in is part of the target platform and so may be used to resolve dependencies.
If you want to ensure that the bundle is not used, make sure that it is not part of the target platform. In your case, the easiest way to do this is to explicitly remove the plug-in from the target platform:
<plugin>
<groupId>org.eclipse.tycho</groupId>
<artifactId>target-platform-configuration</artifactId>
<version>${tycho-version}</version>
<configuration>
<filters>
<filter>
<type>eclipse-plugin</type>
<id>com.google.collect</id>
<removeAll />
</filter>
</filters>
</configuration>
</plugin>
Next, you need to make sure that the guava plug-in is part of the target platform. You can add an artifact from a Maven repository to the target platform in the following way:
Declare a Maven dependency to the artifact in the dependencies section of the POM. You already have done this correctly.
Set the configuration parameter <pomDependencies> to consider on Tycho's target-platform-configuration plug-in.
Note that this will generally only work if the referenced artifact is already an OSGi bundle. This is the case here: com.google.guava:guava:14.0.1 seems to have all manifest headers needed by OSGi.
This should give you the result you wanted: In the test runtime, guava should now be used to match your com.google.common.* package imports.
And another general remark on declaring dependencies in Tycho: In Tycho, you can only declare dependencies in the PDE source files META-INF/MANIFEST.MF, feature.xml, etc.
The normal Maven-style dependencies declared in the POM do not add dependencies to the project. As explained above, the POM dependencies may only add artifacts to the target platform, i.e. the set of artifacts that may be used by Tycho to resolve the dependencies declared in the PDE source files. So in the end, the POM dependency may become part of the resolved dependencies, but only if the dependency resolver picks it for matching one of the declared dependencies.
by default, tycho will add any p2 artifacts you installed in your local maven repo to the target platform. If bundle com.google.collect exports the package which you import, it may be wired.
To stop tycho from including any locally installed artifacts, you can use -Dtycho.localArtifacts=ignore (or, remove the unwanted bundle from your local maven repo)
See http://wiki.eclipse.org/Tycho/Release_Notes/0.16#Improvements_and_Fixes
I created a simple Maven project (Packaging type - Jar) that has dependencies on Spring and My Sql library (mysql-connector). When I package this project with $mvn package I do get a jar file after successful execution of this command.
I was also trying to include all the dependencies in the output jar file, so I added a 'jar-with-dependencies' assembly descriptor, but as the documentation says:
The `jar-with-dependencies` descriptor builds a JAR archive with the contents of the main project jar along with the unpacked contents of all the project’s runtime dependencies.
I want to include the dependencies in JAR form, not the unpacked way. How can I do this?
Java cannot load classes from jars in your jar out of the box. You must configure it with project like one-jar and maven plugin.
Maybe this can help Maven Shade Plugin:Maven Excecutable Jar
Olivier
Vicky is trying to aggregate the jars and this is known as "building-an-aggregate-jar" in Maven's terms. You can find more in Robert's[1] and Artifact Technical Notes'[2] blog.
One discussion[3] that can help as well.
Regards
[1] http://rombertw.wordpress.com/2010/05/14/maven-recipe-building-an-aggregate-jar/
[2] http://blog.artifact-software.com/tech/?p=121
[3] http://maven.40175.n5.nabble.com/Aggregate-POM-for-thirdparty-package-dependencies-not-downloaded-td113927.html
At the end of my ant build id like it to call the equivalent of the command line call
mvn install:install-file -Dfile=my.jar -DgroupId=com.company.project -DartifactId=my_project -Dversion=1.0 -Dpackaging=jar -DgeneratePom=true
so that it will add the newly built jar to a maven repository which another project will rely on.
Ive tried using the maven-ant-task and have added the maven-ant-task jar to the ant built project and the following code to the build.xml:
<target name ="minstall" depends="jar">
<artifact:pom id="maven_install" file="maven_install.xml" />
<artifact:install file="${out.dir}/my_project.jar">
<pom refid="maven_install"/>
</artifact:install>
</target>
but seem to be missing something as it wont work for me. To begin with i get the error in the build.xml (ant build file) saying
The prefix "artifact" for element "artifact:pom" is not bound.
What am I doing wrong. I am fairly new to ant?
On a realted question what is the purpose of the associated POM file? I would not normally have a POM in this project as it is an ant build
Perhaps maven-ant-task jar is not installed, i.e. not in your ant CLASSPATH. You can follow this instruction for this.
As mentioned previously, you need to make sure the tasks are defined in your ant script, and the artifact namespace is understood.
The POM file is used (in this case) to tell the Maven repositories the dependencies of the JAR you are putting in the repository. The POM should also specify the JAR's identification information (groupId, artifactId, version number, license, etc.).
Strictly speaking, you do not need an external POM, you could define the information in your build.xml file as follows:
<!-- Assuming tasks defined, and 'artifact' namespace exists -->
<artifact:pom id="maven_install" groupId="com.whatever" artifactId="some-jar"
version="1.0" packaging="jar">
<dependency groupId="..." artifactId="..." version="..."/>
<dependency groupId="..." artifactId="..." version="..."/>
<license name="apache" url="http://www.apache.org"/> <!-- can be omitted -->
</artifact:pom>
<target name ="minstall" depends="jar">
<artifact:install file="${out.dir}/my_project.jar" pomRefId="maven_install"/>
</target>
When you install the JAR in the 'minstall' task, the POM should be generated with the appropriate dependencies in the local Repository.
That message means you are missing an xmlns:artifact attribute in your build.xml. Have a look at the installation page in the docs for an example.
As to the purpose of the POM file, it's mostly metadata so that maven can figure out dependencies properly. In a real maven build it also describes how to build, test and package. But in your case all that is done by ant instead.
I think that it makes no sense to put such commands in Ant's build.xml. If you want to have your jar file installed in your maven repo just use mvn install command.
Besides that, I guess that you are somehow confusing the purpose of Maven and Ant tools in your project. What I'd suggest is to use Maven as your main build tool. You can configure invokation of Ant targets in your POM file if you really need that. Personally, I think it is the best solution to have Ant called by Maven. Maven goals (such as clean, test, package, install and so on) are very simple to use and powerful (I guess that you can read it in every Maven tutorial).
We have our project build using maven. We try to run our unit test cases in maven build itself and for doing that we need to add DB2 driver jar in the dependency of all the sub projects.
Instead of doing that, we need a solution to specify the absolute path of the jar file as a mvn command line argument to use it in the running of unit test cases.
This is because the driver jar is available in our app server lib folder and we don't want to specify it in the dependencies of our projects.
Couldn't find a suitable solution googling it, hence requesting for an expert solution here.
Any workaround would be of greater help.
Thanks in advance.
The usual way would be to add a dependency to the database driver and limit the dependency to testing (test scope). So the library is available for unit tests but will not deployed and jar'ed.
Practically spoken, I'd create a maven artifact for this driver (just a basic POM file) and place it on the build servers maven repository (or the nexus, if you use it for the projects).
I'm using a dependency with scope set to 'system' to reference a jar that is available in the container but not in any maven repository. In this case the jar is put in a folder named 'lib' in the project like this, :
<dependency>
<groupId>groupId</groupId>
<artifactId>artifactId</artifactId>
<version>version</version>
<scope>system</scope>
<systemPath>${project.basedir}/lib/library.jar</systemPath>
</dependency>
The groupId, artifactId and version can be set to any value you want, the trick was that system dependencies have to be given with an absolute path, which is worked around by using the project.basedir property. It should also be possible to specify the complete path as a property.
We have our project build using maven. We try to run our unit test cases in maven build itself and for doing that we need to add DB2 driver jar in the dependency of all the sub projects.
Well, the maven way would be to declare the DB2 driver as dependency with a test scope in a parent project.
Instead of doing that, we need a solution to specify the absolute path of the jar file as a mvn command line argument to use it in the running of unit test cases.
You could use the additionalClasspathElement in the plugin configuration to pass the path to the driver:
<plugin>
<groupId>org.apache.maven.plugins</groupId>
<artifactId>maven-surefire-plugin</artifactId>
<configuration>
<additionalClasspathElements>
<additionalClasspathElement>path/to/additional/resources</additionalClasspathElement>
</additionalClasspathElements>
</configuration>
</plugin>
If you variablelize it, you could pass the value on the command line.
But to be honest, I can't understand why you don't install the driver in a corporate repository and declare it as dependency. And if you don't have a corporate repository, use a file based repo as described in this previous answer (please, don't use the system scope bad practice). There is no good reason to go the hacky way.
I'm new to maven, I keep running in to the following syntax:
<include>**/*</include>
Im not sure how to interpret **/*, is this some Java or maven convention?
It is more related to <fileset> Ant convention upon which Maven is built.
**/* : all files in all subdirectories: see Ant Patterns.
When ** is used as the name of a directory in the pattern, it matches zero or more directories.
For example: /test/** matches all files/directories under /test/, such as /test/x.java, or /test/foo/bar/xyz.html, but not /xyz.xml.