Compile Maven Module with Different Java Version

There are a number of hacks out there for compiling source code with a different version of the JDK than you are using to run Maven, for example you can use something like

<project>
  [...]
  <build>
    [...]
    <plugins>
      <plugin>
        <groupId>org.apache.maven.plugins</groupId>
        <artifactId>maven-compiler-plugin</artifactId>
        <version>2.5.1</version>
        <configuration>
          <executable><!-- path-to-javac --></executable>
        </configuration>
      </plugin>
    </plugins>
    [...]
  </build>
  [...] 
</project>

The issue with this approach is that you now have hard-coded the path to the JDK into your POM. Everything will work just fine on your machine but when you have to rebuild your machine because the HDD failed, or when you want to build on a different machine, you will be stuck as the path will most likely not match up.

The correct best practice way to handle this is via Toolchains. This will see you creating a ~/.m2/toolchains.xml file that describes where each of the different toolchains in your system are. Then the version of the JDK can be applied by the Maven Toolchains Plugin, e.g.

<plugins>
 ...
  <plugin>
    <groupId>org.apache.maven.plugins</groupId>
    <artifactId>maven-toolchains-plugin</artifactId>
    <version>1.0</version>
    <executions> 
      <execution>
        <phase>validate</phase>
        <goals>
          <goal>toolchain</goal>
        </goals>
      </execution>
    </executions>
    <configuration>
      <toolchains>
        <jdk>
          <version>1.6</version>
        </jdk>
      </toolchains>
    </configuration>
  </plugin>
  ...
</plugins>

The next thing is that you don’t need this as often as you would think. For example by using the source and target values you can generate the correct bytecode for the JRE that you are targeting… the only issue that you will then hit is the use of methods that are new in JRE 1.7… which is where Mojo’s Animal Sniffer Plugin comes in. Animal Sniffer can be used to ensure that you only use the methods of the JRE that you are targeting. The general community consensus is that the use of source and target configuration options in the Maven Compiler Plugin configuration coupled with the use of Mojo’s Animal Sniffer virtually eliminates the need for toolchains on the Compiler end of things…. on the Surefire end of things there is still need for toolchains… and I have a few edge cases that I need to update the compiler plugin and the toolchains plugins for to handle but, realistically you will not hit those edge cases 😉

Just to be sure that your original question is completely answered (since the above answers the question you wanted to ask – as opposed to the one you asked)

At present you are compiling with JDK 1.7 however depending on the version of the Maven Compiler Plugin you are using, you may be compiling with either <source>1.4</source><target>1.4</target> or <source>1.5</source><target>1.5</target> unless you have changed the configuration of the Maven Compiler Plugin in your pom.xml. That will dictate which language features are available to you, but not which classes… so you would be generating code that will work on JRE 1.7 and provided you have not used any new classes/methods introduced since 1.4/1.5 (Such as String.isEmpty()) should also work on JRE 1.4/1.5… the only way to be sure if it works on such an old JVM is to either: run it on the old JVM OR use Animal Sniffer.