What is the difference between the <pathelement> attributes 'path' and 'location' in Ant?

Question

I was running Selenium unit tests in TestNG with the Ant Java task like so:

<java classpathref="runtime.classpath"
    classname="org.testng.TestNG"
    failonerror="false">
    <arg value="-d" />
    <arg value="${grid.location}/target/reports" />
    <arg value="${lib.location}/testng.xml"/>   
</java>

runtime.classpath is a pathlike structure that included <pathelement path="${basedir}/target/classes/" />, which I thought was needed to let TestNG know which classes to run.

<path id="runtime.classpath">
        ...
        <!-- Target classes -->
        <pathelement path="${basedir}/target/classes/" />
</path>

However, I kept seeing in the log that TestNG found 0 applicable classes.

I eventually got some help from a colleague and it appears this was the key change:

<path id="runtime.classpath">
        ...
        <!-- path attribute changed to location -->
        <pathelement location="${basedir}/target/classes/" />
</path>

This also pulls in the test classes correctly:

   <java classpathref="runtime.classpath"
       classname="org.testng.TestNG"
       failonerror="false">
       <arg value="-d" />
       <arg value="${grid.location}/target/reports" />
       <arg value="${lib.location}/testng.xml"/>
       <classpath>
           <pathelement location="${basedir}/target/classes/" />
       </classpath> 
   </java>

What is the difference between the path and location attributes? I've looked at Writing a Simple Buildfile (specifically the Path-like Structures section), but in that manual it looks to me like location is more specific than path. That doesn't appear to be the case empirically, but I can't quite figure out why.

Answer 1

It looks like the difference between path and location is many entries vs one. A location is a file or directory, a path can be a list.

From the manual

The location attribute specifies a single file or directory relative to the project's base directory (or an absolute filename), while the path attribute accepts colon- or semicolon-separated lists of locations. The path attribute is intended to be used with predefined paths - in any other case, multiple elements with location attributes should be preferred.

Note that the JVM used by ant has just about no relation to the JVM used by the java task. By default the environment of ant isn't the same as that of things started with the java task via ant. This is actually helpful when you want to use a different JVM from the one ant wants to use and makes things explicit, helping avoid surprises later on.

Check out the docs for the java task, particularly clonevm

clonevm: If set to true, then all system properties and the bootclasspath of the forked Java Virtual Machine will be the same as those of the Java VM running Ant. Default is "false" (ignored if fork is disabled). since Ant 1.7