Instead of using the make-distribution.sh script from Spark, you can use Maven directly to compile the sources. For instance, if you wanted to build the default version of Spark, you could simply type (from the _spark_dir folder):

./build/mvn clean package

This would default to Hadoop 2.6. If your version of Hadoop was 2.7.2 and was deployed over YARN, you can do the following:

./build/mvn -Pyarn -Phadoop-2.7 -Dhadoop.version=2.7.2 -DskipTests clean package

You can also use Scala to build Spark:

./build/sbt package