[GRASS-PSC] GRASS 7 Migration from CVS to SVN

[Markus Neteler]

Hello Paul,

On 9/14/07, Paul Kelly <paul-grass at stjohnspoint.co.uk> wrote:
> Hello Markus
>
> On Fri, 14 Sep 2007, Markus Neteler wrote:
>
> > Dear PSC,
> >
> > we should decide (or give a recommendation to the
> > developers team) about the migration from CVS to SVN.
> > This is currently holding the start of GRASS 7 development.
>
> I would say not necessarily - a release branch could be created for the
> next stable release and then the major changes for 7 made in the CVS HEAD.

Well, the expected (heavy) moving around of files is rather a pain
in CVS while easy in SVN. The risk of messing everything up in
SVN is low (while no guarantee in CVS, we already have some odd
files there which Martin discovered in his attempts to convert to SVN).

> But I think what we more or less decided in the last discussion was that
> we would use the move to SVN as a "crutch" to aid the 6.x/7.x development
> split. So in that sense you're right. As long as we all understand the
> implications of that (especially with regard to where to make bug fixes
> and where to add new development) then I think it will actually be a good
> help.

Additionally, the user management is way easier. We could grant
restricted "translator" access and so forth while there is only global
access for CVS. Just check our previous discussions about CVS vs
SVN.


> > Essentially, all want to migrate to SVN for various
> > advantages already discussed. But we have to define
> > where the hosting will take place. Currently there
> > are two offers:
> >
> > - SVN with GForge at Intevation in Germany
> > - SVN with Trac and Wiki at OSGeo.org
> (out of interest - where is the OSGeo SVN server physically located and
> who is the sysadmin for it?)
>
> I think (as Helena initially suggested in the last discussion) we should
> move to the OSGeo infrastructure, primarily for the social/shared support
> benefits Frank outlines. I just think we have to be clear that we're not
> proposing leaving Intevation because of any deficiencies with the
> infrastructure there - I can only say it's been amazingly reliable for
> years with almost no downtime at all and to my knowledge never a bug or an
> issue to report. Bernhard has always been there when needed to sort out
> any queries with access to the CVS server. He's been about for a while
> now and as far as I can make out from the mailing list archives was one
> of the major influences in actually having GRASS released under the GPL,
> and I think we owe him a lot of thanks for all his support of GRASS.

Absolutely no doubts about that!
Intevation - Bernhard - Jan - the team there have greatly supported GRASS
and luckily much influenced the decision to go for GPL and CVS!
We are all most grateful to them.

> I wonder is it reasonable to expect Intevation to continue to support our
> old CVS repositories (grass, grass6, web, libgrass - are there any others)
> when we move the latest developments to a new infrastructure - perhaps we
> should simply move the whole thing to SVN?

That is not so easy once it comes to branches. With Martin I discussed
what about moving only HEAD to SVN and keeping all other in CVS
(hoping that Intevation is keeping support for us). The GRASS 6.x would
then continue there (basically bug fix releases which we'll have for longer
time).
Concering Web, we'll have to see if to change (since I am doing most of
it alone and did not manage to get much folks involved over the years).
But that's a different discussion.

> I don't know if they have any
> other projects using their CVS server - they might wish to close it down
> once we leave, to reduce the maintenance/support effort of maintaining it.

I only know about
- grass (which is GRASS 5)
- grass6 (which wil hopefully GRASS 7 one day)
- web (us again)
- libgrass (Frank, but dead since there is the plugin)
- newsletter/ (meanwhile OSGeo Journal, hosted there)
- grass_doc/ (dead, don't even remembered that is was there)
- progmangrass50/ (all doxygen now)

That's all (see http://freegis.org/cgi-bin/viewcvs.cgi/ ).

Migration candidate is only grass6/ there, and probably only for HEAD.
It would be important to keep all other repositories alive there.
(There is another Wiki document discussing the technical issues
of CVS -> SVN migration).

> This is where I have slight reservations about the idea of using the move
> to SVN as a split-point in 6.x/7.x development: are we going to simply
> copy the grass7 repository in its entirety and rename it something like
> grass7? Or are we going to subtly modify the files and then reimport them
> into grass7. I'm not quite sure what's the plan and this certainly needs
> to be ironed out.

There was some discussion on it recently:
See last paragraphs of:
http://grass.itc.it/pipermail/grass-dev/2007-August/032599.html
http://grass.itc.it/pipermail/grass-dev/2007-August/032612.html

> As to the other benefits of the OSGeo infrastructure (apart from
> community/social integration with the other projects), if the bug tracker
> / repository integration works out well then that would be a nice added
> bonus. I have to say the way you couldn't Cc e-mail discussions into the
> Gforge bug tracker like you could with the old RT one was a big downside
> to using it more for me - I wonder does Trac have that capability?

No idea. But GForge has some more limitations (see Wiki page and GForge
reports itself) - we observe that the old RT was way more used by developers
than GForge. There must be reasons for that.

Markus