[Kde-scm-interest] Re: kdebase-workspace and kdebase-runtime git repos to validate

[Ian Monroe]

On Tue, Dec 14, 2010 at 02:26, Oswald Buddenhagen <ossi at kde.org> wrote:
> On Mon, Dec 13, 2010 at 06:22:38PM -0600, Ian Monroe wrote:
>> The branches (especially the early CVS ones) are a bit confused, this
>> is a work-in-progress.
>
>> cvs2svn has some known issues I think,
>>
> that's to put it mildly. one could also say that it monumentally screwed
> up entire branches, probably most of them (i wonder how you approached
> that?).

Well its all in the kde-ruleset repo.

> a more reliable way would be probably doing a "naive" cvs2svn conversion
> (which does not consider server-side moves) with a current version and
> then fix up the history (remove the ghost files) with svn2git min
> revision rules.
> of course that would mean that all rules before the the old revision
> would have to be re-adjusted ... but well, i'm not saying anything new
> here, so whatever.

With the kdebase-* repos I don't even bother with tags and repos <4
since those repositories didn't really exist before then. And if some
of the old kdelibs branches are useless I guess it might make sense to
not even keep them.

Not that regardless we need to keep the SVN repo around if we really
want any historical accuracy at all. Unlike the cvs->svn transition,
we are splitting up the repo and KDE's interdependent code will lose
its greater context in git.

My main goal with the old code is to make the 'master' history
reasonable, so that if someone want to follow the progress of a
function they can do that. And like I said, if anything our git
history will be better at SVN in doing this.

Ian