The problem building Hadoop 0.20.205.0 is a Java 7 problem. Here's the details (with a patched version):
http://svn.apache.org/viewvc?view=revision&revision=1349441 I wanted to add code to my install script to drop this patch into place before building Hadoop, but I ran into another little thing. In the TUTORIAL.sh script that does the Hadoop build, it checks to see if the directory is already populated and doesn't do the "untar" step if the directory already exists — this is just what I want to be able to slip in my update. However, the Makefile.am file for Hadoop always removes the directory before starting — making it impossible to have the patch survive when you do "make hadoop-0.20.205.0" It seems that the right thing to do would be remove the "rm" from the Makefile.am. Thanks, Jim Donahue
