[DRE-maint] Experiments with git

Lucas Nussbaum lucas at lucas-nussbaum.net
Mon Jan 17 19:41:15 UTC 2011


On 17/01/11 at 20:55 +0200, Antono Vasiljev wrote:
> Excerpts from Lucas Nussbaum's message of Mon Jan 17 20:30:56 +0200 2011:
> 
> > > > svn-buildpackage wouldn't work with git.
> > > > That's a -1000 considering that our current workflow deeply relies on
> > > > svn-bp.
> > > Ok. Let's think positive. What required in order to replace
> > > svn-buildpackage with $OUR_SCRIPT_NAME.
> > > Maybe git-buildpackage may be used?
> 
> > I think so, yes. All the docs refer to it, the new processes need to be
> > defined, etc, etc. For what benefit? OK, git is faster, and I prefer git
> > for new projects too.
> 
> It's easier to contribute and live in modern ruby world with git.
> Almost all big ruby project use git. Why should we use svn? :)

... except the ruby interpreter. ;)

> Almost every ruby developer (potential package maintainer) is
> already git user.
> 
> If we decide some day to keep upstream sources in tree it will
> be much more easy with git.
> 
> As for me svn is pain in the ass.

Don't get me wrong. I like git a lot, and use it preferably to SVN.
However, I don't think that SVN has any real problem *for our use*.

> > However, it is something that can be done at the same time as switching
> > to gem2deb. So help make this happen, and we can do the switch to git at
> > the same time, since gem2deb will force us to redefine all our processes
> > anyway (with clear benefits).
> 
> I can make work on this. I need some mentoring from more
> experienced debian developer. What parts of workflow are broken
> without svn-buildpackage? What should be implemented?
> 
> I saw some checks made by svn-inject when I imported my
> packages. I suppose this checks are first point in TODO.

So, one of the first TODO is probably to see how one could use git +
gem2deb to maintain Debian packages, and document the process. For
example, how would one handle new upstream releases?

Then, another question is how we would be able to maintain 100s of
packages with git. Should we do one-repo-per-package? How could we make
that easier? I think that git has sub-modules. Could we leverage that?

Finally, it would be great to have more participation in the -ruby@
thread. When everybody is quiet, it's always a bit uncomfortable because
we can't know if it's because everybody agrees, or nobody cares.

- Lucas



More information about the Pkg-ruby-extras-maintainers mailing list