Re: GNOME git repositories? (was Re: GNOME subversion migration)



I'm going to describe the system we have at work, because I personally
think that it rocks and that something like this in Gnome would be
really awesome.  We use SFEE for bug tracking (it sort of sucks, but
that's unrelated) and we use bzr for version control.  I think a system
like this could work for git too, or probably Mercurial.  I don't know
enough about the details of these individual SCMs right now to say
"let's use bzr".. I'm just interested in the workflow we get from our
system.

We have a mainline that stores the official product, and we can create
branches either on our server via sftp or create them locally and then
push them onto the server at any point.  Nobody ever commits directly
into mainline.. everything happens through the bug tracker, basically.
We write the code, push it to the server, and then set the URL of the
branch in the bug entry.  Once the branch is "ready" we mark is
NEEDREVIEW and there is a script that goes through and auto-builds it
and keeps it up-to-date against mainline until it gets approved.  When
two people approve it, (assuming it builds correctly) it gets merged
into mainline and marks the bug RESOLVED.. if there's a problem, the
script marks it NEEDREVIEW and tells you that it couldn't merge.  That
doesn't happen often at all though.

Maybe Gnome is too large of a project for all the auto-building, I don't
know.  Maybe not.  It's a really nice system though, and I think it
makes it much easier to manage things.  While we're talking about
upgrading version control systems, we might try to see if we can take
the opportunity to come up with something that really improves the
workflow of the projects rather than just some incremental improvements
over what cvs provides.  I think what I just described, or something
like it, would be a huge improvement.  It would take some real work to
develop it and integrate it with Bugzilla, though.

/ Cody

On Tue, 2006-12-26 at 14:22 -0800, Sriram Ramkrishna wrote:
> On Tue, Dec 26, 2006 at 06:44:58PM -0300, Germ�Po�ama�rote:
> > That is a big misunderstanding about how it works.  Using a distributed
> > source system doesn't mean that doesn't exist any central ('main')
> > repository.
> > 
> > Moreover, it can works the same way it has been working until now.
> > The big difference is any contributor can have his or her own copy
> > of the repository (as usual) but the whole history.
> 
> Even better, you don't have to give a cvs/svn account to every
> contributor.  Allowing the barrier of entry to be a lot easier.
> 
> > One best practice is 'commit early, commit often and merge with main
> > regularly'.
> 
> Indeed.
> 
> > Your concern could happen even today if the developer doesn't commit
> > his or her code from his or her copy.  Nobody cares, because 'main'
> > rules.
> 
> The cycle of maintenance is a lot easier I think.
> 
> sri
> 




[Date Prev][Date Next]   [Thread Prev][Thread Next]   [Thread Index] [Date Index] [Author Index]