Further clarifications: - Creating the fork https://help.github.com/articles/fork-a-repo/
- Creating a branch for every commit (and creating the pull request) https://help.github.com/articles/creating-a-pull-request-from-a-fork/ - How to keep your fork in-sync with the upstream repository https://help.github.com/articles/syncing-a-fork/ - How long-lived Apache feature branches will work in this model. In this case, we'd still need a branch off of the feature branch for every commit from the fork. In this workflow, a feature branch is no different than any other branch. If you want a commit to land in a branch, then you create a new branch based off that branch. You create multiple branches if you want your commits to land in multiple branches. It might sound like a daunting task initially, but believe me, its very easy and straightforward to create a branch and open pull requests for review. And once a pull request is opened, you can make changes by simply pushing commits to the same branch. - How to merge long-lived feature branches into Apache Feature branches or any other branch could be merged with trunk or any branch by creating a new pull request. A new pull request could be opened by selecting two branches - a base branch and a head branch. In this case, if you want to merge a feature branch with trunk, then you select feature branch as base branch and trunk as head branch. I have attached a screen-shot for reference. I agree with you on creating a wiki page to cover all the scenarios. Thanks, Vivek Ratnavel On Thu, Jan 4, 2018 at 1:05 PM, Jonathan Hurley <[email protected]> wrote: > Thanks for the clarifications. This sounds like the "Forking Workflow" as > opposed to the "Feature Branch Workflow". I'm fine with that since it lets > non-commiters help. > > We should try to capture all of these scenarios in a wiki page which we > can then all agree upon. Things which we need to cover are: > > - Creating the fork > - Creating a branch for every commit (and creating the pull request) > - How to keep your fork in-sync with the upstream repository > - How long-lived Apache feature branches will work in this model. In this > case, we'd still need a branch off of the feature branch for every commit > from the fork. > - How to merge long-lived feature branches into Apache > > A few of the items above haven't been specified yet - like keeping the > forked repo in sync and how to manage long-lived feature branches in Apache. > > I still do not think we need [component-1][component-2] in the commit > message. We can use the fields in Apache Jira for this. It makes our commit > messages long, hard to read, and ugly. > > > On Jan 4, 2018, at 12:57 PM, Vivek Ratnavel <[email protected]> > wrote: > > > > Let me clarify a few things here. > > > > > > - Before opening any pull requests, one needs to fork > > https://github.com/apache/ambari. This is a one time process. > > - Before working on any JIRA, lets say AMBARI-12345, one needs to > create > > a branch from their own fork. Everyone can have their own naming > > conventions to name this branch since this is not going to affect the > > public repository in any way. > > - To answer Nate's question, if a JIRA has to be committed to > branch-2.6 > > and trunk, one needs to create two branches from their own fork - a > branch > > based on branch-2.6 and another branch based on trunk. Let's name them > > AMBARI-12345-branch-2.6 and AMBARI-12345-trunk. Again this could be > > anything as long as you can differentiate. > > - After committing patches to both the newly created branches, you need > > to open two pull requests against two public branches - branch2.6 and > > trunk. This link should help - > > https://help.github.com/articles/creating-a-pull-request-from-a-fork/ > > - If there is no conflict, github offers "squash and merge" option > which > > will let you remove unnecessary commit messages and merge any number of > > commits as one commit. For more info - > > https://help.github.com/articles/about-pull-request-merges > > > > Hope this clarifies the flow. > > > > To clarify Jonathan's suggestion > > > > * I do not think that adding a [COMPONENT] tag is useful. Many commits > span > > ambari-server and ambari-agent, and a good number also span ambari-web > and > > ambari-server. I also think that we should have the title of the JIra > match > > the commit exactly as we do today. > > > > If a commit spans multiple components, lets say ambari-server and > > ambari-web, the PR title should be [AMBARI-12345][ambari-server][ > ambari-web] > > Title. This is especially useful to categorize the open pull requests > based > > on their components, so that other folks working in those components can > > work on clearing those open pull requests. > > > > Please let me know if you need more clarification on anything discussed > > here. > > > > Thanks, > > Vivek Ratnavel > > > > On Thu, Jan 4, 2018 at 8:45 AM, Nate Cole <[email protected]> wrote: > > > >> Please also clarify the following scenario: > >> > >> I’m working on a fix for branch-2.6, and when I’m done, I need to merge > to > >> trunk. > >> > >> What is the flow? > >> - Create a fork > >> - Commit to branch-2.6 (on my fork) > >> - Commit to trunk (on my fork) > >> - Create pull request to bring changes to both branches? > >> Or > >> - Create a fork > >> - Commit to branch-2.6 (on my fork) > >> - Create pull request > >> - Commit to trunk (on my fork) > >> - Create pull request > >> > >> This is exposing my git n00bness > >> > >> > >> > >> > >> > >> On 1/4/18, 11:32 AM, "Attila Doroszlai" <[email protected]> > >> wrote: > >> > >>> * Since this new flow model requires a branch for a commit, we > >> should enforce a naming strategy. These short-lived feature branches for > >> commits must be easy to find and remove. We should also make the > community > >> aware that once you have had your pull request merged, you should get > rid > >> of your branch. As for branch naming conventions, I haven't thought > through > >> it very much, but perhaps simply the name of the associated JIRA, such > as > >> AMBARI-12345. > >> > >> Correct me if I'm wrong, but the branch to be merged should be > created > >> in your own fork, not in the apache/ambari repo. Otherwise > non-committers > >> would not be able to create pull requests. I think this eliminates the > >> need to coordinate branch naming, although some convention or pattern > would > >> be helpful anyway. > >> > >> -Attila > >> > >> > >> > >
