[ros-dev] GIT mirror

[Alexander Potashev]

On 02:49 Fri 09 Jan     , Samuel serapion wrote:
> On Thu, Jan 8, 2009 at 11:47 PM, Steven Edwards <winehacker at gmail.com>wrote:
> > That being said as a (sometime) developer on Wine and a (even less
> > active) ReactOS developer, I think that having a git gateway is a good
> > idea. I like git, I most likely will use the git mirror for my stuff.
> > Even the Linux kernel has CVS and SVN gateways and everyone knows how
> > much Linus hates them. If nothing else a git gateway will enable those
> > developers that wont to take advantage of the branching support and
> > are comfortable with git to develop in the way that works best for
> > them. If others want to setup an Hg mirror then more power to them
> > too! I am all for the Hg and Git mirrors but...
> >
> 
> Yeah, lets try to spin this little debate into a new direction. We could
> find a way to convert commits from git to svn or have some simple diff
> mechanism that would sync git-tree into the main svn tree. Dunno... sounds
> messy, maybe a bad idea. How do the linux kernel nuts maintain these
> gateways consistent ?

I've never heard about cvs and svn gateways for the Linux kernel
sources, those gateways are probably used to just grab the sources but
not to send changes upstream. In the Linux kernel development 'git pull'
is used only between the Git tree of "high-level" maintainers. Ordinary
developers just send patches to LKML (or to another mailing list).

Yes, we should probably setup a backward gateway to svn. Usually
(without a Git mirror) people imports an svn repo to their local Git
repo a then 'git svn dcommit' back to svn repo.
This won't work if you clone the mirror at
git://git.reactos.org/reactos.git , because now you're dealing with a
normal Git repo. But we can configure (somehow) the mirror to dcommit
the changes somebody has pushed.

> -- 
> encoded
> Samuel Serapión Vega
> Computer Science, Polytechnic University of Puerto Rico.