Check if someone else has already begun work on the change you have in mind in the issue tracker
- If not, create a ticket describing the change you're proposing in the issue tracker
- Check out the latest version of the source code
git clone http://git-wip-us.apache.org/repos/asf/cassandra.git cassandra-trunk
- Modify the source to include the improvement/bugfix
Verify that you follow Cassandra's CodeStyle.
- Verify that your change works by adding a unit test.
- Make sure all tests pass by running "ant test" in the project directory.
You can run specific tests like so: ant test -Dtest.name=SSTableReaderTest`
For testing multi-node behavior, https://github.com/pcmanus/ccm is useful
- When you're happy with the result create a patch:
git add <any new or modified file>
git diff --cached > branchname-issue.txt (e.g. trunk-123.txt, cassandra-0.6-123.txt)
- Attach the newly generated patch to the issue and click "Submit patch" in the left side of the JIRA page
- Wait for other developers or committers to review it and hopefully +1 the ticket
- Wait for a committer to commit it.
Testing and Coverage
Setting up and running system tests:
Running the functional tests for Thrift RPC
Install CQL: svn checkout https://svn.apache.org/repos/asf/cassandra/drivers; cd drivers/py; python setup.py build; sudo python setup.py install.
Install the nose test runner (aptitude install python-nose, easy_install nose, etc).
Install the Thrift compiler (see InstallThrift) and Python libraries (cd thrift/lib/py && python setup.py install).
Generate Cassandra's Python code using ant gen-thrift-py.
Build the source ant clean build.
Run nosetests test/system/ from the top-level source directory.
If you need to modify the system tests, you probably only need to care about test/system/test_thrift_server.py. (test/system/init.py takes care of spawning new cassandra instances for each test and cleaning up afterwards so they are isolated.)
Running the code coverage task
ant codecoverage -Dcobertura.dir=/path/to/cobertura
/path/to/cobertura/cobertura-report.sh --destination build/cobertura/html source code src/java
Buildbot runs the Cassandra tests continuously: http://ci.apache.org/builders/cassandra-trunk. (Builders for stable branches also exist.)
- IntelliJ Project Settings:
Cassandra uses ivy [http://ant.apache.org/ivy/] to fetch compile-time dependencies. Ivy needs to be able to access the web servers that host the dependencies (typically maven repositories). If your internet access is proxied, ivy (ant, really) needs to know about it. There are two ways to accomplish this:
Include -Dhttp.proxyHost=myproxy -Dhttp.proxyPort=3128 in the ANT_OPTS environment variable. See the ant wiki [http://wiki.apache.org/ant/TheElementsOfAntStyle] for more information about ANT_OPTS.
Specify these values on the command line: ant clean build -Dhttp.proxyHost=myproxy -Dhttp.proxyPort=3128
Using git to manage patches during reviews: http://spyced.blogspot.com/2009/06/patch-oriented-development-made-sane.html
Uploading and applying patches from JIRA automagically: GitAndJIRA
Branch-specific Git repo
Sometimes it's necessary to create an entirely new repository instance to work on a branch (for example, if you need to work in a separate IDE configuration). This is particularly common in the long-term support of stable releases. The following recipe can be used to create a Git repository for tracking/committing to/from a single branch.
mkdir cassandra-0.7 cd cassandra-0.7 git init git remote add -f -t cassandra-0.7 -m master origin git://github.com/apache/cassandra.git git remote set-head origin cassandra-0.7 git merge origin # Git-svn setup cd .git; wget http://git.apache.org/authors.txt; cd .. git config svn.authorsfile ".git/authors.txt" git svn init --prefix=origin/ --branches=branches https://svn.apache.org/repos/asf/cassandra
Got commit access? Outstanding! Here are the conventions we follow.
Commit messages take the form of
<explanation> patch by <author>; reviewed by <committer> for CASSANDRA-<ticket>
When committing to multiple branches, start with the most-stable and merge forwards. For instance, if you had a fix to apply to 1.1, 1.2, and trunk, you would first commit to 1.1, and push changes. Then, switch to your 1.2 branch by doing
git checkout cassandra-1.2
- and run
git merge cassandra-1.2
If there are conflicts, resolve them and commit, followed by a push. Finally, switch to trunk by doing
git checkout trunk
git merge cassandra-1.2
again resolve conflicts if the exist, and commit and push.
See http://www.youtube.com/watch?v=AJ-CpGsCpM0 for an in-depth explanation of why fixes should be merged forwards from more-stable branches, rather than backported from trunk.
This workflow also makes it so git knows what commits have been made to earlier branches but not to trunk: if you forget to merge a fix immediately, the next time someone goes to merge from the branch, git will incorporate the forgotten ones too.