Brock, 

Yes, I think we should branch off from Flume-1.3.0, since trunk has some local 
commits(non-RTC - though the code is the same) in addition to the merge 
commits. Please branch off from flume-1.3.0 when you are ready.

Thanks
Hari

-- 
Hari Shreedharan


On Tuesday, October 16, 2012 at 12:48 PM, Brock Noland wrote:

> Hi,
> 
> OK, I am working on the RC and I had this question, should the
> flume-1.4.0 branch be based off trunk or 1.3.0? Normally I would say
> trunk, but it was my understanding we didn't want the release branches
> to have merge commits, which trunk has.
> 
> Brock
> 
> On Fri, Aug 24, 2012 at 3:57 PM, Hari Shreedharan
> <[email protected] (mailto:[email protected])> wrote:
> > To be clear, all commits should go to flume-1.3.0(till of course, 
> > flume-1.3.0 is finalized) as of now. By current release branch, I meant the 
> > branch which will be released (as opposed to the ones already released, 
> > which should not be touched).
> > 
> > Hari
> > 
> > --
> > Hari Shreedharan
> > 
> > 
> > On Friday, August 24, 2012 at 1:27 PM, Hari Shreedharan wrote:
> > 
> > > Please read this page for some more details:
> > > http://flume.apache.org/source.html
> > > 
> > > 
> > > Thanks
> > > Hari
> > > 
> > > --
> > > Hari Shreedharan
> > > 
> > > 
> > > On Friday, August 24, 2012 at 12:52 PM, Hari Shreedharan wrote:
> > > 
> > > > Also note that I have cherry picked all commits till this morning. Any 
> > > > further commits will have to be handled by the committer who commits it 
> > > > to trunk.
> > > > 
> > > > Thanks
> > > > Hari
> > > > 
> > > > --
> > > > Hari Shreedharan
> > > > 
> > > > 
> > > > On Friday, August 24, 2012 at 12:50 PM, Hari Shreedharan wrote:
> > > > 
> > > > > Fellow developers,
> > > > > 
> > > > > The Flume PMC has decided to keep an active release branch along with 
> > > > > trunk. All commits that go into trunk will also have to be committed 
> > > > > using git cherry-pick and pushed to the current release branch.
> > > > > 
> > > > > This does not change anything for contributors, but changes the 
> > > > > workflow for committers.
> > > > > 
> > > > > Committers should first commit the patch to trunk on your local repo:
> > > > > git commit (Enter commit message).
> > > > > git log
> > > > > 
> > > > > Copy the commit hash of the commit you just made. Then:
> > > > > git checkout flume-1.2.0 (or the current release branch)
> > > > > git cherry-pick <commit hash of the commit you made> (This should get 
> > > > > committed immediately).
> > > > > 
> > > > > Now push to both trunk and the release branch:
> > > > > git push -u origin trunk
> > > > > git push -u origin flume-1.2.0
> > > > > 
> > > > > 
> > > > > 
> > > > > 
> > > > > The committer should make sure the commits are pushed to both 
> > > > > branches. Please make sure all commits to the release branch are fast 
> > > > > forward commits and there are no merge commits on the release branch.
> > > > > 
> > > > > This process requires a little more work, but this guarantees that 
> > > > > our release tags will not have accidental and local commits in its 
> > > > > history, as we can force push to the release branches to remove these 
> > > > > from history. Ideally we should try to keep the history on release 
> > > > > branches linear, but if at some point we decide to start using 
> > > > > feature branches, we might end up having merge commits on these 
> > > > > branches too, but that is expected and required - since that would 
> > > > > represent the list of commits for that feature.
> > > > > 
> > > > > I will update the website soon with this information. Please let me 
> > > > > know if you have any questions - which I expect many would have.
> > > > > 
> > > > > 
> > > > > Thanks,
> > > > > Hari
> > > > > 
> > > > > 
> > > > > --
> > > > > Hari Shreedharan
> > > > > 
> > > > 
> > > > 
> > > 
> > > 
> > 
> > 
> 
> 
> 
> 
> -- 
> Apache MRUnit - Unit testing MapReduce - http://incubator.apache.org/mrunit/
> 
> 


Reply via email to