After Falcon is donated I expect to be making lot of changes to complete its 
MXML support. I strongly prefer to be able to work in SVN and not to have to 
learn Yet Another Source Code Control System. Is the idea that SVN would stay 
the primary repository and Git might become any option for development branches?

- Gordon

-----Original Message-----
From: Carol Frampton [mailto:cfram...@adobe.com] 
Sent: Friday, August 10, 2012 7:31 AM
To: flex-dev@incubator.apache.org
Subject: Re: What would it take to move to Git?



On 8/10/12 5 :00AM, "Carlos Rovira" <carlos.rov...@codeoscopic.com> wrote:

>We must take into account that low activity was due to donation and 
>create a first incubation SDK. In the next months more people should 
>come to the project as it helps to develop some patch, bug or modification.
>
>The infrastructure should support huge changes and must be set up 
>before those changes start to be developed. Changes to UIComponent, 
>HTML5 port, and so on are huge and must live with the actual evolution 
>of other minor things.
>
>So taking into account that there are aggresive modification plans in 
>mind, I think we should go with the best branching model and not the 
>simplest.
>If
>there was no plans, maybe the later will be best.
>
>
>>Agreed, but considering the relatively low commit activity ATM (I'm
>listed 7th at [1] with 32 commits since January, which have nothing to 
>do with Flex code for example)., I'm wondering if it would make more 
>sense to start with the *simplest* branching model as opposed to the
>*best* one, for now. You might need the latter downstream, but probably 
>not right now.


Personally I'd like to come up with a relatively simple scheme for branching 
using SVN and then get to work, especially since Apache infra still considers 
Git experimental.  I'm guessing we are one of the larger projects at Apache in 
terms of code size and I want us to be able to get infra support.  I know we've 
used their services quite a few times already.

We've spent 8 months pushing existing code around and not making any forward 
progress with Flex.  I understand that some people prefer Git but at this point 
everyone knows SVN to some level and not everyone knows Git.

Assuming we will release from trunk, I don't think everyone working in trunk 
can work, even aside from the stability issues.  If we all worked for the same 
company, and all signed up to get our features done by date X so we could 
release on date Y it might be possible.  In the Apache model people work on 
things when they can, so there might be features/components that aren't done 
when someone decides to do a release.

I think Alex is proposing that we all work in the dev branch (aka
unstable) and when it is time to release we cherry pick changes to move back to 
trunk and then release from trunk.  We would not be merging on a daily or even 
weekly basis but at release time (perhaps too corporate but could also have 
base levels and merge then).  I think Justin raises some legitimate concerns 
about multiple changes being made to the same file but I think we can come up 
with a solution for that. Perhaps changes don't get even get committed to the 
dev branch until you think they are worthy of release but there are other 
alternatives as well.

I think you could also flip this and do dev work in trunk and then create a 
branch for the release that was based on the last release tag and then you move 
changes from trunk over to the release branch.

I'm having a hard time following the Alex/Justin discussion but I have read 
some misinformation about SVN in there.  You can do an svn merge and then if 
there are conflicts, go back and resolve those conflicts and mark the files 
resolved with svn resolve.  Then you svn commit the results of the merge.  You 
can also block merges with svn --record-only merge.

Carol

Reply via email to