[Dune-devel] Write access for Andreas

Martin Nolte nolte at mathematik.uni-freiburg.de
Wed May 8 09:24:35 CEST 2013


Hi Christian, hi Andreas, hi all,

maybe I do not understand the experiment. From what I gather this is a plain 
branch and - when finished - merge concept. In my opinion, this should be the 
default development model, anyway.

Moreover, I do not understand why a private branch (on dune-project.org) is 
preferable to
(a) a feature branch
(b) a private branch hosted somewhere else
In case (b), I would advise to push the branch as a feature branch before 
merging, so that everyone can have a look. This is the true power of git.

As Markus pointed out, my private existed purely due to Subversion requiring a 
central repository. Now that we have git, I migrated it our git server in 
Freiburg. All developments are first pushed to this server, can then be 
cherry-picked into a (still private) feature branch, which in turn can be 
pushed to dune-project.org. When discussion is over, it can be merged into the 
master branch. Of course, in case of a single bug-fix patch (most frequent 
case) I do directly cherry-pick it into the master branch on dune-project.org.

Finally, I would like to point out a "downside" (it is an upside really) of 
git: We have to talk to each other. Simply expressing our opinion on an annual 
developer meeting will (in my opinion) be insufficient in the future. The more 
branches we have (and each clone can be considered one), the more they can 
diverge. Moreover, I think a merge of feature branches should be voted on. 
 From my experience (even with Subversion), lacking discussion is at the root 
of our problem.

Best,

Martin


PS: I warmly welcome git because I can develop first and discuss later, i.e., 
proposals have been tested before discussion.



On 05/07/2013 10:49 PM, Dedner, Andreas wrote:
> Hi.
> Is there a good reason not to try the experiment - we do not have to make this the
> default (yet), we can decide now try this for now only in this case.
> Christian and Andreas can report back how its working and if they see some advantage, that have
> been overlooked so far.
> Andreas
> ________________________________________
> From: dune-devel-bounces+a.s.dedner=warwick.ac.uk at dune-project.org [dune-devel-bounces+a.s.dedner=warwick.ac.uk at dune-project.org] on behalf of Christian Engwer [christian.engwer at uni-muenster.de]
> Sent: 07 May 2013 21:37
> To: Martin Nolte
> Cc: dune-devel at dune-project.org
> Subject: Re: [Dune-devel] Write access for Andreas
>
> On Tue, May 07, 2013 at 09:19:21PM +0200, Martin Nolte wrote:
>> No objections against full write access. That's the way we always
>> handled this stuff.
>
> With Andreas I had the idea to try the fork/join model, that's why we
> wanted to do this and in this case full write access is just not
> necessary. But as I said, it will not chnage anything for us ;-)
>
> Christian
>
> _______________________________________________
> Dune-devel mailing list
> Dune-devel at dune-project.org
> http://lists.dune-project.org/mailman/listinfo/dune-devel
>
>

-- 
Dr. Martin Nolte <nolte at mathematik.uni-freiburg.de>

Universität Freiburg                                   phone: +49-761-203-5630
Abteilung für angewandte Mathematik                    fax:   +49-761-203-5632
Hermann-Herder-Straße 10
79104 Freiburg, Germany




More information about the Dune-devel mailing list