Travis support for new profiles, readme, comments in script

.travis.yml

@@ -4,9 +4,9 @@ jdk:
   - openjdk7
   - openjdk6
 env:
-  - PROFILE=hadoop_v1
-  - PROFILE=hadoop_yarn
+  - PROFILE=
+  - PROFILE="-Dhadoop.profile=2"
 
 install: true
 
-script: "mvn -P$PROFILE clean verify"
+script: "mvn $PROFILE clean verify"

README.md

@@ -35,7 +35,33 @@ If you’re a Debian/Ubuntu user, you’ll find a .deb package. We will continue
 
 	cd stratosphere-dist/target/stratosphere-dist-0.2-ozone-bin/stratosphere-0.2-ozone/
 
-Note: The directory structure here looks like the contents of the official release distribution.
+The directory structure here looks like the contents of the official release distribution.
+
+#### Build for different Hadoop Versions
+This section is for advanced users that want to build Stratosphere for a different Hadoop version, for example for Hadoop Yarn support.
+
+We use the profile activation via properties (-D).
+
+##### Build hadoop v1 (default)
+Build the default (currently hadoop 1.2.1)
+```mvn clean package```
+
+Build for a specific hadoop v1 version
+```mvn -Dhadoop-one.version=1.1.2 clean package```
+
+##### Build hadoop v2 (yarn)
+
+Build the yarn using the default version defined in the pom
+```mvn -Dhadoop.profile=2 clean package```
+
+Build for a specific hadoop v1 version
+```mvn -Dhadoop.profile=2 -Dhadoop-two.version=2.1.0-beta clean package```
+
+It is necessary to generate separate POMs if you want to deploy to your local repository (`mvn install`) or somewhere else.
+We have a script in `/tools` that generates POMs for the profiles. Use 
+```mvn -f pom.hadoop2.xml clean install -DskipTests```
+to put a POM file with the right dependencies into your local repository.
+
 
 ### Run your first program
 

tools/generate_specific_pom.sh

@@ -57,17 +57,15 @@ nupom="pom.${hadoop_version}.xml"
 poms=`find $ozone_home -name pom.xml`
 for p in $poms; do
   nuname="`dirname $p`/${nupom}"
-  # Now we do search and replace of explicit strings.  The best
-  # way of seeing what the below does is by doing a diff between
-  # the original pom and the generated pom (pom.xml.hadoop1 or
-  # pom.xml.hadoop2). We replace the compat.module variable with
-  # either hbase-hadoop1-compat or hbase-hadoop2-compat, we
-  # replace the version string in all poms, we change modules
-  # to include reference to the non-standard pom name, we
-  # adjust relative paths so child modules can find the parent pom,
-  # and we enable/disable hadoop 1 and hadoop 2 profiles as
-  # appropriate removing a comment string too.  We output the
-  # new pom beside the original.
+  # Now we do search and replace of explicit strings.  The best way of
+  # seeing what the below does is by doing a diff between the original
+  # pom and the generated pom (pom.hadoop1.xml or pom.hadoop2.xml). We
+  # replace the version string in all poms, we change modules to
+  # include reference to the non- standard pom name, we adjust
+  # relative paths so child modules can find the parent pom, and we
+  # enable/disable hadoop 1 and hadoop 2 profiles as appropriate
+  # removing a comment string too. We output the new pom beside the
+  # original.
   sed -e "s/${old_ozone_version}/${new_ozone_version}/" \
     -e "s/\(<module>[^<]*\)/\1\/${nupom}/" \
     -e "s/\(relativePath>\.\.\)/\1\/${nupom}/" \