Each commit to the repository must have a non-empty log message. The message format should start with a one-line short summary prefixed , with the Jira issue number. If you do not have a Jira issue number, create one before you commit. a capitalized first letter of the first word. The short summary need not be a complete sentence with proper punctuation and grammar since it is best to keep it to 72 50 characters or less.
After the short summary line, you may continue the message with one or more paragraphs separated by an empty line, each line about 72 characters or less. The empty line after the short summary is necessary if you want to add more text since source control tools like git may send the log message as an email and use the first line up to an empty line as the email subject. When committing contributions from non-committers, please attribute the original author in the commit messages.
Since we support many operating systems, it would be best to list which one the commit was tested on to let the community know what has been covered.
When submitting a commit with significant changes authored by another contributor, be sure to attribute such changes to the author in your commit comment.
Short, TS-####one shortline summary (7250 chars or less), not a sentence Tested: Ubuntu-9.0.4,OSX-10.5,FreeBSD-7.2,OpenSolaris-2009.06 Author: John Doe <firstname.lastname@example.org> paragraph More detailed explanatory text as necessary, patch attributions, and what platforms the commit has been tested on. Wrap it to about 72 characters or so. In some contexts, the first line is treated as the subject of an email and the rest of the text as the body. The blank line separating the summary from the body is critical (unless you omit the body entirely); tools like rebase can get confused if you run the two together. This is an example of a second paragraph. The 'Author:' line is required if the commit represents code submitted by somebody who does not have commit permissions. The person doing the commit is implicitly the reviewer of the submitted code. The email address is optional but it is nice to have so the original author can be uniquely identified. It is OK for the committer to modify or cleanup the code and if done so, should be noted in the commit log.
In the future, I'd like to see the need for listing the operating systems tested on go away when we implement a testing environment that can accept pre-commit patch and verify that patch passes a series of smoke tests.
Commits that resolve a Jira ticket should also include an update to the CHANGES file - just add the full commit 'short summary' line and an attribution if relevant (to make it easier to view version content). Commits to this file should preserve double blank lines where they exist.
Trunk is CTR, Commit-Then-Review, policy. This was discussed and voted on by the community a long time ago. This does grant a lot of freedom to committing experimental or unstable code to the trunk. However, with such powers, comes some responsibilities:
- If the trunk is broken, i.e. it doesn't build and/or run even without your changes, please don't commit! No more fuel on the fire please. It should be relatively easy to track down which change set broke the build from the buildbot dashboard and inform the committer or community.
- You are responsible for latest trunk to build and at least start up on one of our supported platforms with your changes before you push a commit. This typically implies that you should to do a `git pull` to your tree and build before committing. Running the regression tests (trunk/ci/regression) is also highly recommended.
- At a minimum the the build should not break on the current primary buildbot slave which is Ubuntu Linux. If the primary build bot is broken, it is the commiter's responsibility to get the build working or find someone to help them.
- If your changes result in a non-functioning Traffic Server (e.g. start fails, simple forward proxy browsing is broken, etc) please consider backing out your changes and submitting the changes for review under a TS Jira ticket and asking for reviewers.
The master branch
Changes to the master branch are governed by a Review-Then-Commit policy that has been modified in the following way: Instead of a Consensus Approval, commits require at least one +1 Vote from another committer.
The master branch is expected to be stable, and changes should be tested and reviewed.
1. We only schedule one main release / year. Each main release branch is 2 years LTS (Long-Term Support).
2. A release branch is cut off master once a year. For example, 7.x.
3. The community can agree to make a minor release, e.g. v7.2.0 from a main release branch at any given time during the 7.x 2-year life cycle. Such a release is still branched off the corresponding main release branch. (Minor releases should not be branched from master.) Any such minor release is also LTS, that is, its support term is the same as that of the main release branch upon which it is based.
4. The Release Manager and community should agree on which commits from master should go into any minor release. This would be for example a new feature, or a new plugin etc. Critical crashes / security fixes is dealt with as normal with a patch release.
5. Critical fixes (e.g. security fixes) on an LTS branch must be applied to every LTS minor release on that branch. As an example, this could mean making a v7.1.1 release as well as a v7.2.1 release.