[Pgpool-hackers] Exodus from pgfoundry

[Guillaume Lelarge]

On Wed, 2011-06-01 at 18:47 +0200, Jehan-Guillaume (ioguix) de Rorthais
wrote:
> On 01/06/2011 18:38, Guillaume Lelarge wrote:
> > On Thu, 2011-06-02 at 00:31 +0900, Tatsuo Ishii wrote:
> >>> On 01/06/2011 11:14, Tatsuo Ishii wrote:
> >>>>>> 4) Prepare everything necessary for the new server
> >>>>>
> >>>>> Well, my only concern here would be to have yet another login credential.
> >>>>
> >>>> No way:-)
> >>>>
> >>>>>>    - Source repository (probably starting with CVS, then migrate to
> >>>>>>      git.hub?)
> >>>>>
> >>>>> github would be perfect for 3.1. Do you have an account ? You can create
> >>>>> the pgpool organization here, but all pgpool commiters need all to have
> >>>>> a github account then.
> >>
> >> It seems git.postgresql.org is another choice...
> >>
> > 
> > Yes. We could do something like pgadmin: have the main repository on
> > git.postgresql.org, have another one on github that developers could
> > easily fork (the one on github would be updated once in a while,
> > automatically, just like the pgadmin one).
> > 
> >>>> My concern is, I would like to bring CVS histories to git. I know
> >>>> PostgreSQL developers worked very hard and took long time to do that.
> >>>> If it takes so long time, we need to continue to use CVS.
> >>>
> >>> AFAIK, Guillaume Lelarge is working on this.
> >>>
> >>> In my opinion, as soon as 3.1 is out, we should freeze the development
> >>> and focus on cvs -> git conversation whatever it takes. We already
> >>> discussed the problem inherited from cvs while thinking about releasing
> >>> 3.1 sooner, we should not use CVS for 3.2.
> >>
> >> I expect git conversion finishes within 1 or 2 days at most. I don't
> >> want to be blocked by the conversion work.
> >>
> > 
> > Git conversion takes 2 minutes on my laptop.
> > 
> > Here are some informations.
> > 
> > We first need to patch a few files. Patch is available on
> > cvs_pgpool.patch file. It fixes some issues on encoding.
> > 
> > default.options is a default config file that cvs2git will read in order
> > to do the conversion. If you want to play with it, you need to
> > change /home/guillaume/cvs_pgpool with the full path to the CVS
> > repository (not the one you get by "cvs checkout" but the real one with
> > *, files, which you'll find on pgfoundry.org:/cvsroot/pgpool).
> > 
> > Once you changed that, you can use the convert.sh script file. It will
> > create a git repository for each CVS repository (pgpool, pgpoolAdmin,
> > pgpool-ha, pgpool-II, pgpool-web).
> > 
> > Once you executed the script file, it will create a git directory with a
> > subdirectory for each repository. If you want to use one, you first need
> > to clone it, this way:
> > 
> > [guillaume at laptop git_pgpool]$ mkdir git2
> > [guillaume at laptop git_pgpool]$ cd git2
> > [guillaume at laptop git2]$ git clone ../git/pgpool-II.git/
> > Cloning into pgpool-II...
> > done.
> > 
> > Then, you can enter the pgpool-II directory, and use any git commands
> > you can think of.
> > 
> > BTW, something that first surprises me is that cvs2git doesn't convert
> > each cvs branch in a git branch. It does that because the cvs branch was
> > not used, hence it is not needed. We could create the branch afterwards
> > if we need it.
> > 
> > What we still need to do:
> > 
> >  * if you took a look at the patch, I fixed a japanese comment with
> >    "(something)"... this needs to be fixed with the real comment in
> >    latin1
> 
> As far as I understand it, git support UTF8 messages for commits
> internally. So maybe we could keep those commit message in japanese if
> cvs related files are in UTF-8.
> See man page of git-commit, section "DISCUSSION" about commit messages
> in UTF-8 and some more options about how to set it explicitly in another
> encoding.
> 

You can't have it both ways. All ",v" files are in latin1 (at least, the
log sections). So we need to keep that this way. I would better fix a
string than fix all the other ones :)

> >  * default.options has the list of all the commiters, but some
> >    name/email pairs are probably obsolete... if you could give me the
> >    informations to fix this, that would be great
> > 
> > 
> 
> 


-- 
Guillaume
  http://blog.guillaume.lelarge.info
  http://www.dalibo.com