[DRE-maint] Experiments with git

Gunnar Wolf gwolf at gwolf.org
Tue Jan 18 18:01:43 UTC 2011


Antono Vasiljev dijo [Mon, Jan 17, 2011 at 01:53:55PM +0200]:
> Hello, All.
> 
> I cloned svn repo with git-svn and experimented a little bit.
> (...)
> 
> Almost every big ruby project use git today. Why should we use SVN?
> 
> So. I would like to know your cons/pros about switching to git.
> 
> Its my '+1' for git for pkg-ruby-extras.

Hi,

This discussion has been had before. The main reason we have stuck
with SVN -besides having everything done and documented this way- is
that SVN (and svn-buildpackage) allows us to have -and track- one
single repository for our whole flow. Yes, there is a git-buildpackage
(and I use it on other packages), but its logic is quite different. 

git-buildpackage Is not well suited for multi-package
repositories. Not only because it likes working at the root of the Git
tree (which would of course be fixable), but because of the
branches/tags it generates would clash. This is, as SVN tracks a
repository from the current directory downwards, it is easy to come up
with simple schemes for setting (copying) the branches/tags to
../../tags/$version; were we using Git, the tag would just be a label
($version).

Of course, we could come up with a naming logic here, even if it led
to very long identifiers. But there is another argument: Our SVN tree
is mainly used to track common advance of the history of each of the
projects - Not to do development on it. Storing what amounts to only
the diffs to the orig.tar.gz means our repository is quite light to
work with, but useless without the "real" thing. This means - We are
not using SVN for tracking development, but for sharing state, for a
very simple level of collaboration.



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