[bt-devel] working with Git / SVN

[Olaf Radicke]

Am Samstag, den 24.07.2010, 09:30 +0300 schrieb Jaak Ristioja:
> -----BEGIN PGP SIGNED MESSAGE-----
> Hash: SHA1
> 
> On 23.07.2010 23:45, Olaf Radicke wrote:
> > I see you concern, but mixing is a foul way.
> > 
> > o = Olaf
> > bs = Bibletime-SF.net
> > bg = Bibletime-gitorious
> > 
> >    patch-in-
> >             \
> > bs-bs---bs--bs--bs----->
> >   |     \         \
> >   \bg---bg--------bg---->
> >       |             \
> >       \o---o--------krash!--o-->
> >             \
> >              -patch-out
> > 
> > If i send you a patch and you commit this in sf.net, the  history is lost for git.
> > If the code come back to me, over git-mirror, i become a conflict. Easier to handle
> > is, when i pull and you pull with '--no-commit'. So you have a checkup befor commit.
> > And the history is not lost.
> 
> Well, if that commit would be the new GUI, it would only be the initial commit for the
> feature. I don't think the git history up to that point would matter much. I'm not exactly
> sure how --no-commit would help with SVN, which has quite a linear commit history.

With '--no-commit' you can pitch on specific paches and condemned the
leftover or commit later or commit in different brunch.

Yes, the GUI it's a big bite, but you can commit later with...

git checkout master
git checkout alter-gui-branch /src/alter
git add /src/alter
git checkout alter-gui-branch /src/main.cpp
git commit -a -m 'I add the new alter GUI in master-branch now.'

Can you picture that? Be that as it may, SVN is a hitch in the work
flow. 

> >> I think
> >> its risky because the code quality of the backend still needs a lot of refactoring and
> >> some architectural changes. 
> > 
> > We can us proxy-classes for stabile interfaces. It's standard-tool of refactoring - i think.
> 
> I don't think we have enough developers for this right now.

I think, i dont need this, bud it's a obtion if we/i need this.