Releasing OpenJPA

• Cleanup JIRA so the Fix Version in issues resolved since the last release includes this release version correctly.
• Update the files in openjpa-project -
  • Update the CHANGES.txt and RELEASE_NOTES.html based on the Text and HTML release reports from JIRA.
  • Review and update BUILDING.txt if needed.
• Stage any Roadmap or Release landing pages on the wiki.
• Use "mvn rat:check" to verify the source has the required headers before trying to release.
• Perform a full build (IANAL + tests) and TCK runs before trying to release.

• set MAVEN_OPTS=-Xmx1024m -XX:MaxPermSize=256m (especially for Java SE 6)
• Use Sun JDK (if possible)
  • If using IBM JDK, then also set the system property test.jvm.arguments to -Xdump:none (Avoids OOM heap dumps on selected tests. These extra files throw off the rat:check processing.) For example,

    mvn -Dtest.jvm.arguments="-Xdump:none" ...
    

• On Windows, use a cygwin window (vs a standard DOS prompt window) for the Release processing.

• Do not use Eclipse to do the checkout. The extra dot (.) files created by Eclipse throws off the rat:check processing.

   $ svn checkout https://svn.apache.org/repos/asf/openjpa/branches/1.2.x 1.2.x
  $ cd 1.2.x 

1. The new release number (default 1.2.0)
2. The SCM tag (default apache-openjpa-1.2.0) *this should be changed to just 1.2.0
3. The new development version (default 1.2.1)
4. optional if you have not specified a GPG passphrase in settings.xml you will be prompted for it.

   $ mvn -Prelease,docbook-profile release:prepare -DdryRun=true  
       <snip>
       [INFO] [release:prepare]
       [INFO] Verifying that there are no local modifications...
       [INFO] Executing: svn --non-interactive status
       [INFO] Working directory: /home/mikedd/workspaces/temp/jpadev/branches/1.0.x
       [INFO] Checking dependencies and plugins for snapshots ...
       What is the release version for "OpenJPA"? (org.apache.openjpa:openjpa-parent) 1.2.0: : 
       What is SCM release tag or label for "OpenJPA"? (org.apache.openjpa:openjpa-parent) openjpa-parent-1.2.0: : 1.2.0
       What is the new development version for "OpenJPA"? (org.apache.openjpa:openjpa-parent) 1.2.1-SNAPSHOT: : 
       

• verify signatures Verifying release signatures
• run examples. Running OpenJPA Examples
• if you have access to the TCK, run it Running the TCK

 $ mvn -Dmaven.test.execute=false -Prelease release:prepare 

For the 2.0.x code, different steps are required as there are problems with the maven-jar-plugin and maven-release-plugin when using the test-jar goal. See http://jira.codehaus.org/browse/MJAR-68 and http://jira.codehaus.org/browse/MRELEASE-285.

 $ mvn -DskipTests=true -Prelease release:prepare -DpreparationGoals="clean install"
 $ mvn clean install -Dtest=false -DfailIfNoTests=false -Dmaven.test.execute=false 
 $ mvn -DskipTests=true -Prelease release:prepare -DpreparationGoals="clean install" 

$ mvn -Prelease,docbook-profile release:perform 

For the 2.0.x code, you'll need additional parameters, due to the test-jar issues.

$ mvn -Prelease,docbook-profile,test-derby release:perform 

If your local OS userid doesn't match your Apache userid, then you'll have to set -Duser.name=<your_apache_id> on the cmdline and/or update your release profile in settings.xml

# ssh to people.apache.org 
$ cd ~/public_html/openjpa/${pom.version}/staging-site
$ unzip -qq apache-openjpa/downloads/apache-openjpa-${pom.version}-binary.zip 

Verify that the HTML links in staging-site/index.html are correct (usually only the downloads and docs links need to be updated to be prefixed with apache-openjpa/)

For the 2.x releases, not all of the *tests.jar are being signed (only 1 of 9) so you will need to manually sign and upload the signatures to the staging-repo. See http://people.apache.org/~henkp/repo/faq.html and http://www.apache.org/dev/release-signing.html#openpgp-ascii-detach-sig

 $ mvn stage:copy -Dsource="http://people.apache.org/~mikedd/openjpa/1.2.0/staging-repo" \
  -Dtarget="scp://people.apache.org/www/people.apache.org/repo/m2-ibiblio-rsync-repository" \
  -Dversion=1.2.0 \
  -DtargetRepositoryId=apache.releases

After the stage plugin completes it's a good idea to check the permissions on the file system.

$ ssh people.apache.org
$ cd /www/people.apache.org/repo/m2-ibiblio-rsync-repository/org/apache/openjpa/
$ ls -l openjpa-examples/1.2.0

# if you see something like the following you probably need to change the permissions.
# *-rw-r--r--*  1 mikedd  apcvs  59162 Jul 23 09:34 openjpa-examples-1.2.0-javadoc.jar

$ for file in `find /www/people.apache.org/repo/m2-ibiblio-rsync-repository/org/apache/openjpa/ -type d -iname '1.2.0'`\
  do \
  chmod -R g+w ${file} \ 
  done

$ ls -l openjpa-examples/1.2.0

# Now it should look something like this
# *-rw-rw--r--*  1 mikedd  apcvs  59162 Jul 23 09:34 openjpa-examples-1.2.0-javadoc.jar

Follow the latest steps and guides on the ASF website at http://www.apache.org/dev/openpgp.html#generate-key as you need to disable using SHA1 and new keys should be 4096 bits.

$ ssh-keygen -t dsa

• Program defaults should be fine. No passphrase is required for the ssh key generation. The keys will be saved in ~/.ssh/id_dsa (private) and ~/.ssh/id_dsa.pub (public).

  See Authenticating By Public Key (OpenSSH) for a good description on why and how to perform this task.

$ cd ~
$ ssh <your userid>@people.apache.org

• At this point, you will still be prompted for your password.

$ gpg --gen-key

• The program's default values should be fine. For the "Real Name" enter your full name (ie. Stan Programmer). For the "e-mail address" enter your apache address (ie. sprogrammer@apache.org). You will also be required to enter a "passphrase" for the GPG key generation. Keep track of this as you will need this for the Release processing.

  The generated keys are stored in $HOME/.gnupg or %HOME%\Application Data\gnupg subdirectory. Save the content in this subdirectory to a safe media. This contains your private key used to sign all the OpenJPA release materials.

}}. See the commands describe at the beginning of this KEYS file to perform this task. The gpg key-pair is used to sign the published artifacts for the OpenJPA releases.

$ gpg --list-sigs <Real Name> && gpg --armor -- export <Real Name>

The https://svn.apache.org/repos/asf/openjpa/site/docs/KEYS file is updated via normal svn commit procedures. How the http://www.apache.org/dist/openjpa/KEYS file gets updated is still a mystery to me...

• Use absolute path in the find command's root directory.
  Problem symptom:

  $ find . -name m2-repository -not -path "*openjpa-project*" \
      -exec mvn -f "c:/tmp/maven-stage-plugin/pom.xml" stage:copy \
      -Dsource=file://{} -Dtarget=scp://allee8285@people.apache.org/www/people.apache.org/repo/m2-ibiblio-rsync-repository \
      -Dversion=1.0.1 \;
  [INFO] Scanning for projects...
  [INFO] Searching repository for plugin with prefix: 'stage'.
  ........
  [INFO] Downloading file from the source repository:
  [INFO] ------------------------------------------------------------------------
  [ERROR] BUILD ERROR
  [INFO] ------------------------------------------------------------------------
  [INFO] Error copying repository from file://./openjpa-all/target/site/m2-repository to \
         scp://allee8285@people.apache.org/www/people.apache.org/repo/m2-ibiblio-rsync-repository
  

• Quote and use the drive name in the path.
  Problem symptom:

  $ find /cygdrive/c/OpenJPA.1.0.1.Release/1.0.1 -name m2-repository -not -path "*openjpa-project*" \
      -exec mvn -f /cygwin/c/tmp/maven-stage-plugin/pom.xml stage:copy \
      -Dsource=file://{} -Dtarget=scp://allee8285@people.apache.org/www/people.apache.org/repo/m2-ibiblio-rsync-repository \
      -Dversion=1.0.1 \;
  [INFO] Scanning for projects...
  [INFO] Searching repository for plugin with prefix: 'stage'.
  ........
  [INFO] Downloading file from the source repository:
  [INFO] ------------------------------------------------------------------------
  [ERROR] BUILD ERROR
  [INFO] ------------------------------------------------------------------------
  [INFO] Error copying repository from file:///cygdrive/c/OpenJPA.1.0.1.Release/1.0.1/openjpa-all/target/site/m2-repository to
         scp://allee8285@people.apache.org/www/people.apache.org/repo/m2-ibiblio-rsync-repository
  
  Embedded error: Could not read from file: c:\cygdrive\c\OpenJPA.1.0.1.Release\1.0.1\openjpa-all\target\site\m2-repository
  \cygdrive\c\OpenJPA.1.0.1.Release\1.0.1\openjpa-all\target\site\m2-repository (Access is denied.)
  

• Specify the people.apache.org user id in the target property.
  Problem symptom:

  $ find "c:/OpenJPA.1.0.1.Release/1.0.1" -name m2-repository -not -path "*openjpa-project*" \
     -exec mvn -f "c:/tmp/maven-stage-plugin/pom.xml" stage:copy 
     -Dsource=file://{} -Dtarget=scp://people.apache.org/www/people.apache.org/repo/m2-ibiblio-rsync-repository \
     -Dversion=1.0.1 \;
  [INFO] Scanning for projects...
  [INFO] Searching repository for plugin with prefix: 'stage'.
  [INFO] ----------------------------------------------------------------------------
  ........
  [INFO] Downloading file from the source repository: /org/apache/openjpa/openjpa/maven-metadata.xml.sha1
  [INFO] Downloading metadata from the target repository.
  Password:: *********
  ........
  Password:: *********
  [INFO] ------------------------------------------------------------------------
  [ERROR] BUILD ERROR
  [INFO] ------------------------------------------------------------------------
  [INFO] Error copying repository from file://c:/OpenJPA.1.0.1.Release/1.0.1/openjpa-all/target/site/m2-repository to \
         scp://people.apache.org/www/people.apache.org/repo/m2-ibiblio-rsync-repository
  

• Continue to use the Windows form of path name, e.g. C:\a\b\c
• Use '/' instead of '\' character as file separator, e.g. C:/a/b/c
• Quote all path name using '"' character, e.g. "C:/a/b/c"
• Avoid using space characters in path name, e.g. "C:/OpenJPA.Release/1.0.1"

1. Eclipse control files, .classpath, .project
2. User created log files

1. Don't use Eclipse's svn plugin to "Check out" files to a Eclipse project. Simply use the svn command, as described in the release instructions.
2. Don't create, directly or indirectly, any files under the release build tree.
3. Don't use the IBM JDK. By default, the IBM JDK will create heap dumps when hitting the OOM condition in some of our tests. These extra files throw off the rat:check processing. If you must use the IBM JDK, then also set MAVEN_OPTS=-Xdump:none.