(I'm using 1.5.0 as a placeholder or Wikisearch's master branch)
Wikisearch uses google collections 1.0, which has since been replaced by Guava.
Hadoop 2 relies on Guava for its internals and includes it in the classpath. The Google Collections decommissioning notice warns
Guava is a fully compatible superset of the old Google Collections Library. It also includes years worth of bug fixes, doc fixes, and performance fixes, so we strongly encourage you to use Guava instead of Google Collections.
If your application ever ends up with both google-collect-1.0.jar and guava-r##.jar on the same classpath, simply remove the Google Collections JAR and you should be fine. If you don't remove it, it will either cause terrible problems (if it comes before guava in the classpath) or will never even be seen anyway (if it comes after).
Under the Hadoop 2 profile, google collections comes first and a bunch of tests fail as a result.